Association of Adverse Events With Antibiotic Use
in Hospitalized Patients
Pranita D. Tamma, MD, MHS; Edina Avdic, PharmD, MBA; David X. Li, BS;
Kathryn Dzintars, PharmD; Sara E. Cosgrove, MD, MS

A ntibiotic use is common in the inpatient setting.Approximately 50% of hospitalized patients receive atleast 1 antibiotic during their hospital stay,1 with an es-
timated 20% to 30% of inpatient days of antibiotic therapy con-
sidered unnecessary.2-6 The reasons for antibiotic overuse are
myriad, including administration of antibiotics for nonbacte-
rial or noninfectious syndromes, treatment of conditions
caused by colonizing or contaminating organisms, and dura-
tions of therapy that are longer than indicated. Unnecessary
use of antibiotics is particularly concerning because antibiot-
ics may be associated with a number of adverse drug events
(ADEs), including allergic reactions, end-organ toxic effects,

subsequent infection with antibiotic-resistant organisms, and
Clostridium difficile infections (CDIs).7-12

Estimates of the incidence of antibiotic-associated ADEs
in hospitalized patients are generally unavailable. Previ-
ously, Shehab and colleagues13 conducted a retrospective analy-
sis of ADEs among patients presenting to emergency depart-
ments and found that antibiotics were implicated in 19% of all
emergency department visits for ADEs. It is unclear whether
these data are generalizable to hospitalized patients for a num-
ber of reasons: (1) acutely ill hospitalized patients may be pre-
disposed to certain ADEs, such as antibiotic-associated neph-
rotoxic effects, particularly those admitted with acute renal

IMPORTANCE Estimates of the incidence of overall antibiotic-associated adverse drug events
(ADEs) in hospitalized patients are generally unavailable.

OBJECTIVE To describe the incidence of antibiotic-associated ADEs for adult inpatients
receiving systemic antibiotic therapy.

DESIGN, SETTING, AND PARTICIPANTS Retrospective cohort of adult inpatients admitted to
general medicine wards at an academic medical center.

EXPOSURES At least 24 hours of any parenteral or oral antibiotic therapy.

MAIN OUTCOMES AND MEASURES Medical records of 1488 patients were examined for 30
days after antibiotic initiation for the development of the following antibiotic-associated
ADEs: gastrointestinal, dermatologic, musculoskeletal, hematologic, hepatobiliary, renal,
cardiac, and neurologic; and 90 days for the development of Clostridium difficile infection or
incident multidrug-resistant organism infection, based on adjudication by 2 infectious
diseases trained clinicians.

RESULTS In 1488 patients, the median age was 59 years (interquartile range, 49-69 years),
and 758 (51%) participants were female. A total of 298 (20%) patients experienced at least
1 antibiotic-associated ADE. Furthermore, 56 (20%) non–clinically indicated antibiotic
regimens were associated with an ADE, including 7 cases of C difficile infection. Every
additional 10 days of antibiotic therapy conferred a 3% increased risk of an ADE. The most
common ADEs were gastrointestinal, renal, and hematologic abnormalities, accounting for 78
(42%), 45 (24%), and 28 (15%) 30-day ADEs, respectively. Notable differences were
identified between the incidence of ADEs associated with specific antibiotics.

CONCLUSIONS AND RELEVANCE Although antibiotics may play a critical role when used
appropriately, our findings underscore the importance of judicious antibiotic prescribing to
reduce the harm that can result from antibiotic-associated ADEs.

JAMA Intern Med. 2017;177(9):1308-1315. doi:10.1001/jamainternmed.2017.1938
Published online June 12, 2017.

Author Affiliations: Division of
Pediatric Infectious Diseases,
Department of Pediatrics, Johns
Hopkins University School of
Medicine, Baltimore, Maryland
(Tamma); Department of Pharmacy,
Johns Hopkins Hospital, Baltimore,
Maryland (Avdic, Dzintars); Division
of Infectious Diseases, Department of
Medicine, Johns Hopkins University
School of Medicine, Baltimore,
Maryland (Li, Cosgrove).

Corresponding Author: Pranita D.
Tamma, MD, MHS, Division of
Pediatric Infectious Diseases,
Department of Pediatrics,
Johns Hopkins University
School of Medicine,
200 N Wolfe St, Ste 3149,
Baltimore, MD 21287
([email protected]).

Research

JAMA Internal Medicine | Original Investigation

1308 JAMA Internal Medicine September 2017 Volume 177, Number 9 (Reprinted) jamainternalmedicine.com

© 2017 American Medical Association. All rights reserved.

Downloaded From: https://jamanetwork.com/ on 08/01/2022

failure for non–antibiotic-related reasons; (2) hospitalized pa-
tients are frequently administered intravenous antibiotic
therapy, often at high doses, which may have different ad-
verse event profiles than the oral regimens more commonly
prescribed in the outpatient setting14; (3) hospitalized pa-
tients are commonly administered multiple medications con-
currently, causing a potentially synergistic increase in the risk
of ADE development15; and (4) hospitalized patients are more
likely to be elderly or have multiple medical conditions, re-
sulting in impaired drug elimination and an increased risk of
ADE development.16,17 Previous studies evaluating antibiotic-
associated ADEs in the inpatient setting have used adminis-
trative databases and have not accounted for antibiotic-
associated ADEs that occurred after hospital discharge.18,19

Additionally, they have limited their evaluation of ADEs to
single antibiotic classes or single infectious syndromes.18-21 A
comparative analysis of the incidence of ADEs across all classes
of antibiotics has yet to be performed. Therefore, in the pre-
sent study, we sought to describe the incidence of antibiotic-
associated ADEs for adult inpatients receiving systemic anti-
biotic therapy while hospitalized in general medicine wards.

Methods
Setting and Patients
This study was conducted at the Johns Hopkins Hospital, a 1194-
bed tertiary care facility in Baltimore, Maryland. This study was
approved by the Johns Hopkins University School of Medicine
Institutional Review Board, with a waiver of informed consent
due to the retrospective nature of the study. The data were ret-
rospectively collected on patients 18 years and older admitted
to 4 general medicine services between September 2013 and
June 2014.6 All patients who received antibiotics for at least 24
hours were included. Exclusion criteria included prophylactic
antibiotic use with no clear stop dates, antibiotics used for non-
infectious indications (eg, rifaximin for hepatic encephalopa-
thy, erythromycin for intestinal motility), topical or inhaled
antibiotics, and antituberculosis regimens.

Data Collection and Definitions
Demographic data, preexisting medical conditions, antibi-
otic regimens, and ADEs were collected via patient medical rec-
ord review. Both inpatient and outpatient medical records were
reviewed to obtain follow-up data for patients in the Johns
Hopkins Health System. In addition, the Epic Care Everywhere
Network, a secure health information exchange, was ac-
cessed to view patient data from a large number of health care
facilities throughout the United States.22 This enabled the iden-
tification of patients presenting to outside emergency depart-
ments, hospitals, or primary care clinics with antibiotic-
associated ADEs, if these facilities were in the Epic system.

All antibiotic regimens were adjudicated for appropriate-
ness and associated ADEs by at least 2 infectious diseases phy-
sicians or pharmacists (P.D.T., E.A., K.D., and S.E.C.). Days of
therapy (DOTs) were defined as the number of days from
antibiotic initiation until the completion of antibiotic courses.
A single DOT was recorded for each individual antibiotic

administered to a patient on a given calendar day. Unneces-
sary antibiotic days were defined as DOTs that were not clini-
cally indicated based on recommendations in the Johns
Hopkins Hospital Antibiotic Guidelines.23 For calculations of
overall rates of ADEs, the denominator included all patients
receiving antibiotics (n = 1488). For calculations involving a
single antibiotic, the denominator included only patients
receiving that particular antibiotic.

Avoidable ADEs were defined as the proportion of overall
ADEs that occurred in patients for whom antibiotic therapy was
considered not indicated. Nonindicated antibiotic regimens did
not include patients with prolonged durations of therapy be-
cause our goal was to determine the incidence of adverse re-
actions for patients for whom no antibiotic therapy was nec-
essary. For example, if a patient received ciprofloxacin for 15
days for pyelonephritis when 7 days would have been suffi-
cient and the patient developed tendinitis on day 16, one would
be unable to attribute the adverse event to the 7 indicated days
of ciprofloxacin use or the additional 8 days of unnecessary
ciprofloxacin use. We also did not consider overly broad spec-
trum antibiotic therapy prescribed for valid indications as not
indicated because of the impossibility of knowing whether the
patient would or would not have developed an ADE with a nar-
rower choice, particularly in the same class of antibiotics.

Criteria used to define antibiotic-associated ADEs are sum-
marized in Table 1. These definitions were derived from avail-
able literature, package inserts, and/or consensus opinions prior
to any data collection related to the present work. Patients were
observed for 30 days from the date of antibiotic initiation for
most ADEs (gastrointestinal, dermatologic, musculoskeletal,
hematologic, hepatobiliary, renal, cardiac, and neurologic
events) and for 90 days from the date of antibiotic initiation
for CDI and the development of multidrug-resistant organ-
ism (MDRO) infections not previously identified. All ADEs other
than CDI or incident MDRO infections were censored at 30 days
due to concerns for underestimating the incidence if a longer
evaluation period was used because these ADEs generally
occur during exposure to particular antibiotics or shortly there-
after. In contrast, data suggest that CDI and the emergence of
MDRO infections can become clinically apparent several weeks
to months after discontinuing antibiotic therapy.26,27

Key Points
Question What is the likelihood of developing antibiotic-
associated adverse drug events (ADEs) for hospitalized patients
receiving antibiotic therapy?

Findings In this cohort study, medical records of 1488 adult
inpatients were examined for 30 days after antibiotic initiation for
the development of the following antibiotic-associated ADEs:
gastrointestinal, dermatologic, musculoskeletal, hematologic,
hepatobiliary, renal, cardiac, and neurologic; and 90 days for the
development of Clostridium difficile infection or incident
multidrug-resistant organism infection. Twenty percent of patients
experienced at least 1 antibiotic-associated ADE.

Meaning These findings underscore the importance of judicious
antibiotic prescribing to reduce the harm that can result from
antibiotic-associated ADEs.

Adverse Events and Antibiotic Use in Hospitalized Patients Original Investigation Research

jamainternalmedicine.com (Reprinted) JAMA Internal Medicine September 2017 Volume 177, Number 9 1309

© 2017 American Medical Association. All rights reserved.

Downloaded From: https://jamanetwork.com/ on 08/01/2022

All potential ADEs were adjudicated in the context of the pa-
tient’s medical history and clinical course to ensure that each
event was likely to have been antibiotic associated, both to rule
out alternative explanations and to appropriately categorize
ADEs. Each ADE was then attributed to a single antibiotic, based
on the likelihood of that antibiotic causing the specific ADE and
the temporal relationship of the antibiotic’s administration to the
ADE. For example, acute kidney injury in a patient receiving van-
comycin and cefepime would have been attributed to vancomy-
cin use only. This step was performed to avoid overestimating
the incidence of ADEs because most patients in our cohort re-
ceived multiple antibiotics during their hospital stays. However,
because virtually all antibiotics can cause CDI or the emergence
of MDRO infections, the development of either of these 90-day
ADEs was attributed to all preceding antibiotic used.

Statistical Analysis
Rates per 10 000 person-days and 95% confidence intervals
were calculated for each ADE and antibiotic class. For 30-day
ADEs, the numerator was the number of ADEs attributed to
each antibiotic or class of antibiotics. The denominator was the
person-time at risk for all patients who received that particu-
lar antibiotic or class of antibiotics, computed as the time, in
days, from antibiotic initiation to the ADE for patients who ex-
perienced the ADE, with censoring at 30 days for patients who
did not experience the ADE. The proportion of 30-day anti-
biotic-associated ADEs per antibiotic or antibiotic class and the
proportion of patients receiving a particular antibiotic or
antibiotic class who developed a 30-day ADE were also calcu-

lated. For 90-day ADEs, the numerator accounted for all pre-
ceding antibiotics rather than only a single antibiotic. The de-
nominator was the person-time at risk for all patients who
received antibiotics, computed as the time, in days, from
antibiotic initiation to ADE onset, with censoring at 90 days.
Hazard ratios were calculated to identify the incremental risk
of an ADE conferred by each additional day of antibiotic use.
All analyses were performed using Stata 13 (StataCorp).

Results
Antibiotic Regimens
Of the 5579 patients admitted to the 4 included medicine wards
during the study period, 1488 (27%) patients received anti-
biotics for at least 24 hours and were included in the analysis.
Previous work describes the demographic data, preexisting
medical conditions, sources of infection, and “appropriate-
ness” of antibiotic use of the included population in more
detail.6 In brief, the median age was 59 years (interquartile
range [IQR], 49-69 years) and 758 (51%) participants were
female. The most common underlying medical conditions were
diabetes (491 [33%]), structural lung disease (327 [22%]), and
congestive heart failure with an ejection fraction of less than
40% (178 [12%]). The median length of hospital stay was 4 days
(IQR, 2-9 days). The most common indications for antibiotic
therapy were urinary tract infections (179 [12%]), skin and
soft-tissue infections (119 [8%]), and community-acquired
pneumonia (104 [7%]).

Table 1. Criteria Used for Antibiotic-Associated Adverse Drug Events

Adverse Drug Event Definition
Within 30 d of Antibiotic Initiation

Non–Clostridium
difficile–associated diarrhea

>3 Loose stools per day associated with antibiotic administration and documented as “diarrhea” in the medical record,
in the absence of laxative use or preexisting enteritis. Patients with a positive C difficile PCR test result were excluded
from this category

Nausea and vomiting Nausea and vomiting associated with antibiotic administration, in the absence of an alternate explanation

Hematologic Anemia (hemoglobin level <10 g/dL), leukopenia (white blood cell count <4500 cells/μL), or thrombocytopenia (platelet
count <150 × 103/μL) with levels below patient’s baseline and in the absence of bleeding or myelosuppressive therapies

Hepatobiliary Cholestasis (total bilirubin level >3 mg/dL) or transaminitis (aspartate transaminase or alanine transaminase level
>3 times patient’s baseline) in the absence of existing hepatobiliary disease or recent biliary instrumentation

Renal Increase in serum creatinine level >1.5 times patient’s baseline in the absence of precipitating factors for acute kidney
injury such as sepsis or the receipt of intravenous contrast or other nephrotoxic agents24

Neurologic Altered mental status, peripheral neuropathy, or seizures in the absence of preexisting neurologic conditions,
substance-related toxic effects, or infectious syndromes

Dermatologic Rash, including hives, nonhives rashes, and red man syndrome, temporally associated with antibiotic administration
with resolution on antibiotic discontinuation; excluding vancomycin-associated red man syndrome

Cardiac QTc >440 ms in males or >460 ms in females in the absence of preexisting arrhythmias, based on ≥2 electrocardiograms

Anaphylaxis Acute onset of respiratory compromise, hypotension, or end-organ dysfunction within minutes after initiation
of antibiotic administration, in the absence of an alternative explanation

Myositis Increase in creatine phosphokinase level >5 times patient’s baseline, in the absence of existing myopathy or statin use

Within 90 d of Antibiotic Initiation

C difficile infection Clinical signs and symptoms consistent with C difficile infection in the setting of a positive C difficile PCR test result
and the absence of laxative use

Infection with
MDR organism25

Infection with any of the following organisms, in a patient without a history of colonization or infection with the same
organism: methicillin-resistant Staphylococcus aureus; vancomycin-resistant enterococci; carbapenem-resistant
Enterobacteriaceae; MDR Acinetobacter; MDR Pseudomonas; or a gram-negative organism with a greater than 2-fold
increase in the minimum inhibitory concentration of an antibiotic compared with the initial infection

Abbreviations: MDR, multidrug-resistant; PCR, polymerase chain reaction.

SI conversion factors: To convert hemoglobin to grams per liter, multiply by 10.0; to convert white blood cell count to ×109 per liter, multiply by 0.001; to convert
platelet count to ×109 per liter, multiply by 1.0; to convert bilirubin to micromoles per liter, multiply by 17.104.

Research Original Investigation Adverse Events and Antibiotic Use in Hospitalized Patients

1310 JAMA Internal Medicine September 2017 Volume 177, Number 9 (Reprinted) jamainternalmedicine.com

© 2017 American Medical Association. All rights reserved.

Downloaded From: https://jamanetwork.com/ on 08/01/2022

The most frequently prescribed antibiotics were third-
generation cephalosporins (607 [41%] regimens), parenteral
vancomycin (544 [37%] regimens), and cefepime (414 [28%]
regimens) (Table 2). The majority of patients (1176 [79%]) re-
ceived more than 1 antibiotic during the hospitalization. The
median DOTs per patient was 7 days (IQR, 4-14 days). A total
of 324 unique ADEs occurred; 298 (20%) patients experi-
enced at least 1 antibiotic-associated ADE. The overall rate of
antibiotic-associated ADEs was 22.9 per 10 000 person-days.

Every additional 10 antibiotic DOTs conferred a 3%
increased risk of an ADE. A total of 236 (73%) antibiotic-
associated ADEs occurred during hospitalization and the re-
maining 88 (27%) occurred after hospital discharge including
33 (18%) 30-day ADEs, 11 (20%) CDIs, and 44 (52%) MDRO in-
fections. The study investigators determined that 287 (19%)
of antibiotic regimens were not clinically indicated, most com-
monly because of treatment of asymptomatic bacteriuria or
treatment of noninfectious lower respiratory tract conditions
(eg, aspiration pneumonitis, congestive heart failure).6 Of the
287 nonindicated antibiotic regimens, 56 (20%) were associ-
ated with an ADE.

30-Day ADEs
Of the 324 overall ADEs, 186 (57%) were 30-day ADEs. The me-
dian time to development of a 30-day ADE was 5 days (IQR,
3-8 days). The median times to 30-day ADEs for the various
organ systems were as follows: cardiac, 11 days (IQR, 4-18 days);
gastrointestinal, 5 days (IQR, 2-9 days); hematologic, 12 days
(IQR, 6-24 days); hepatobiliary, 8 days (IQR, 4-12 days); renal,
5 days (IQR, 2-10 days); and neurologic, 3 days (IQR, 2-4 days).
The most common ADEs were gastrointestinal, renal, and he-
matologic abnormalities, accounting for 78 (42%), 45 (24%),
and 28 (15%) 30-day ADEs, respectively (Table 2). Table 3 and
Table 4 outline the proportions of 30-day ADEs attributable
to specific antibiotics or antibiotic classes and the proportion
of patients receiving a specific antibiotic or antibiotic class who
developed 30-day ADEs, respectively.

Aminoglycosides, parenteral vancomycin, and trimetho-
prim-sulfamethoxazole were associated with the highest rates
of nephrotoxic effects at 21.2 (95% CI, 12.5-66.0), 12.1 (95% CI,
7.7-19.0), and 13.2 (95% CI, 5.9-29.3) episodes per 10 000 per-
son-days, respectively (Table 2). Two patients experienced QTc
prolongation—1 receiving azithromycin and 1 receiving cipro-
floxacin after 4 and 18 days of therapy, respectively. Seven pa-
tients (6.7 [95% CI, 2.7-12.0] episodes per 10 000 person-
days) receiving cefepime developed neurotoxic effects,
including encephalopathy or seizures. Less frequent 30-day
ADEs, all occurring in single patients, included cefepime-
associated anaphylaxis, piperacillin-tazobactam–associated
drug fever, daptomycin-associated myositis, ciprofloxacin-
associated tendinitis, trimethoprim-sulfamethoxazole–
associated pancreatitis, linezolid-associated peripheral
neuropathy, vancomycin-associated hives, and a trimethoprim-
sulfamethoxazole–associated nonhives rash.

90-Day ADEs
There were 138 ADEs occurring within 90 days, accounting for
43% of all ADEs. Of these 138 ADEs, 54 (39%) were CDI and 84

(61%) were MDRO infections. The median time to develop-
ment of a 90-day ADE was 15 days (IQR, 4-34 days). The rate
of CDI was 3.9 (95% CI, 3.0-5.2) per 10 000 person-days for pa-
tients receiving antibiotics, corresponding to 54 (4%) study pa-
tients developing CDI within 90 days of antibiotic initiation.
The antibiotics most frequently associated with CDI were third-
generation cephalosporins (present in 28 [52%] regimens pre-
ceding CDI), cefepime (26 [48%] regimens), and fluoroquino-
lones (19 [35%] regimens).

The rate of emergence of incident MDRO infections was
6.1 (95% CI, 4.9-7.6) per 10 000 person-days, corresponding
to 84 [6%] study patients developing an infection with a new
MDRO within 90 days of antibiotic initiation. Subsequent gram-
positive resistance was observed in 60 (4%) patients, at a rate
of 4.8 (95% CI, 3.7-6.1) cases per 10 000 person-days. Forty
(67%) of the MDRO c ases were related to vancomyc in-
resistant enterococci infections. Gram-negative resistance
occurred less frequently at a rate of 1.7 (95% CI, 1.2-2.6) cases
per 10 000 person-days, or in 30 (2%) patients, with extended-
spectrum β-lactamase production being the most common
resistance mechanism identified.

Clinically Significant ADEs
Antibiotic-associated ADEs were then categorized into clini-
cally significant and non–clinically significant categories. Only
1 category was selected per patient, with the more severe cat-
egory selected when multiple categories were met. A total of
314 (97%) of the 324 antibiotic-associated ADEs were consid-
ered clinically significant because of the following reasons: new
hospitalization(s) (n = 10 [3%]), prolonged hospitalization
(n = 77 [24%]), additional clinic or emergency department vis-
its (n = 29 [9%]), and additional laboratory tests, electrocar-
diograms, or imaging (n = 198 [61%]). There were no deaths
attributable to any antibiotic-associated ADE.

Discussion
We found that 20% of hospitalized patients receiving at least
24 hours of antibiotic therapy developed an antibiotic-
associated ADE. Moreover, 20% of ADEs were attributable to
antibiotics prescribed for conditions for which antibiotics were
not indicated. Every 10 DOTs conferred an additional 3% risk
of an ADE. Our findings underscore the importance of avoid-
ing unnecessary antibiotic prescribing to reduce the harm that
can result from antibiotic-associated ADEs.

Previous studies on antibiotic-associated ADEs in the in-
patient setting have largely been limited to single infectious
syndromes or single antibiotic classes.18-21,28 For example, Lin
and colleagues18 evaluated the incidence of antibiotic-
associated ADEs using an administrative database of hospi-
talized patients with pneumonia. They found that even though
less than 1% of patients developed ADEs, the presence of an
antibiotic-associated ADE was an independent predictor of
prolonged hospital lengths of stay and total hospital charges.
Werner et al20 evaluated the frequency of adverse events re-
lated to unnecessary fluoroquinolone use in hospitalized
patients based on medical record review. They found that

Adverse Events and Antibiotic Use in Hospitalized Patients Original Investigation Research

jamainternalmedicine.com (Reprinted) JAMA Internal Medicine September 2017 Volume 177, Number 9 1311

© 2017 American Medical Association. All rights reserved.

Downloaded From: https://jamanetwork.com/ on 08/01/2022

Ta
bl

e
2.

R
at

es
of

30
-D

ay
A

nt
ib

io
ti

c-
A

ss
oc

ia
te

d
A

dv
er

se
D

ru
g

Ev
en

ts
in

14
8

8
H

os
pi

ta
liz

ed
Pa

ti
en

ts
R

ec
ei

vi
ng

A
nt

ib
io

ti
cs

pe
r1

0
0

0
0

Pe
rs

on
-d

ay
s

(P
D

)a

A
nt

ib
io

ti
c

A
ge

nt

N
o.

of
Pa

ti
en

ts
R

ec
ei

vi
ng

A
ge

nt

Ca
rd

ia
c

G
as

tr
oi

nt
es

ti
na

lb
H

em
at

ol
og

ic
H

ep
at

ob
ili

ar
y

R
en

al
N

eu
ro

lo
gi

c
O

th
er

Ev
en

ts
c

N
o.

R
at

e
pe

r
10

00
0

PD
(9

5%
CI

)
N

o.

R
at

e
pe

r
10

00
0

PD
(9

5%
CI

)
N

o.

R
at

e
pe

r
10

00
0

PD
(9

5%
CI

)
N

o.

R
at

e
pe

r
10

00
0

PD
(9

5%
CI

)
N

o.

R
at

e
pe

r
10

00
0

PD
(9

5%
CI

)
N

o.

R
at

e
pe

r
10

00
0

PD
(9

5%
CI

)
N

o.

R
at

e
pe

r
10

00
0

PD
(9

5%
CI

)
β-

La
ct

am
d

11
87

0
…

59
17

.4
(1

3.
5-

22
.4

)
27

8.
7

(5
.3

-1
1.

3)
6

3.
4

(3
.1

-7
.9

)
17

6.
9

(3
.1

-1
2.

9)
10

3.
8

(1
.5

-5
.3

)
2

0.
6

(0
.1

-2
.2

)

A
m

pi
ci

lli
n

63
0

…
2

11
.6

(2
.9

-4
6.

2)
1

5.
6

(0
.8

-3
9.

6)
0

…
1

5.
6

(0
.8

-3
9.

6)
0

…
0

…

A
m

ox
ic

ill
in

–
cl

av
ul

an
at

e
10

2
0

…
3

12
.6

(4
.5

-2
5.

2)
0

…
0

…
0

…
0

…
0

…

A
m

pi
ci

lli
n-

su
lb

ac
ta

m
52

0
…

1
7.

2
(1

.0
-5

1.
2)

0
…

0
…

2
14

.2
(1

3.
9-

49
.6

)
0

…
0

…

O
xa

ci
lli

n
33

0
…

4
37

.1
(1

2.
0-

10
5.

0)
1

10
.8

(1
.5

-7
6.

4)
2

21
.6

(5
.4

-8
6.

6)
0

…
0

…
0

…

Pi
pe

ra
ci

lli
n-

ta
zo

ba
ct

am
31

5
0

…
16

14
.8

(1
3.

1-
23

.0
)

4
4.

3
(1

.6
-1

1.
4)

1
1.

1
(0

.2
-7

.9
)

1
1.

1
(0

.2
-7

.9
)

1
1.

1
(0

.2
-7

.9
)

1
1.

1
(0

.2
-7

.9
)

Ce
fa

zo
lin

79
0

…
0

…
1

4.
4

(0
.6

-3
1.

4)
0

…
2

8.
2

(1
.6

-2
4.

8)
0

…
0

…

Ce
ft

ri
ax

on
e

60
7

0
…

14
8.

0
(4

.7
-1

3.
5)

11
6.

2
(3

.4
-1

1.
3)

3
2.

1
(1

.4
-1

2.
3)

5
2.

8
(1

.2
-6

.8
)

1
0.

6
(0

.1
-3

.9
)

0
…

Ce
fp

od
ox

im
e

89
0

…
2

7.
7

(1
.9

-3
0.

9)
0

…
0

…
0

…
0

…
0

…

Ce
fe

pi
m

e
41

4
0

…
10

8.
5

(4
.6

-1
5.

8)
6

5.
0

(2
.2

-1
1.

1)
0

…
6

5.
0

(2
.2

-1
1.

1)
7

6.
7

(2
.7

-1
2.

0)
1

0.
8

(0
.1

-5
.8

)

Er
ta

pe
ne

m
85

0
…

3
12

.1
(3

.9
-3

7.
6)

0
…

0
…

0
…

0
…

0
…

M
er

op
en

em
80

0
…

4
18

.0
(6

.8
-4

8.
0)

3
12

.9
(4

.2
-4

0.
1)

0
…

0
…

1
4.

4
(0

.8
-2

9.
4)

0
…

N
on

–β
-l

ac
ta

m
s

A
m

in
og

ly
co

si
de

s
32

0
…

0
…

0
…

0
…

2
21

.2
(1

2.
5-

66
.0

)
0

…
0

…

A
zi

th
ro

m
yc

in
40

0
1

0.
8

(0
.1

-5
.9

)
1

0.
8

(0
.1

-5
.9

)
0

…
4

3.
4

(1
.3

-9
.0

)
0

…
0

…
0

…

Cl
in

da
m

yc
in

19
3

0
…

3
5.

4
(1

.8
-1

6.
8)

0
…

0
…

0
…

0
…

0
…

D
ap

to
m

yc
in

8
0

…
0

…
0

…
0

…
0

…
0

…
1

44
.8

(6
.3

-3
18

.3
)

D
ox

yc
yc

lin
e

57
0

…
2

12
.4

(3
.1

-4
9.

7)
0

…
0

…
0

…
0

…
0

…

Fl
uo

ro
qu

in
ol

on
es

39
4

1
0.

9
(0

.1
-6

.2
)

5
4.

4
(1

.8
-1

0.
6)

1
0.

9
(0

.1
-6

.2
)

3
2.

6
(0

.8
-8

.0
)

1
0.

9
(0

.1
-6

.2
)

1
0.

9
(0

.1
-6

.2
)

1
0.

9
(0

.1
-6

.2
)

Li
ne

zo
lid

23
0

…
0

…
0

…
0

…
0

…
1

15
.8

(2
.2

-1
12

.3
)

0
…

M
et

ro
ni

da
zo

le
17

5
0

…
1

2.
0

(0
.3

-1
4.

2)
0

…
0

…
0

…
1

2.
0

(0
.3

-1
4.

2)
0

…

Tr
im

et
ho

pr
im

–
su

lf
am

et
ho

xa
zo

le
15

5
0

…
5

11
.2

(4
.7

-2
6.

9)
0

…
0

…
6

13
.2

(5
.9

-2
9.

3)
0

…
1

2.
1

(0
.3

-1
5.

1)

In
tr

av
en

ou
s

va
nc

om
yc

in
54

4
0

…
2

1.
3

(0
.3

-5
.2

)
0

…
0

…
19

12
.1

(7
.7

-1
9.

0)
0

…
2

1.
3

(0
.3

-5
.2

)

O
ve

ra
ll

ra
te

s
14

88
e

2
0.

4
(0

.1
-1

.8
)

78
18

.2
(1

4.
6-

22
.8

)
28

6.
4

(4
.4

-9
.2

)
13

2.
9

(1
.7

-5
.0

)
45

10
.6

(7
.9

-1
4.

2)
13

2.
9

(1
.7

-5
.0

)
7

1.
6

(0
.8

-3
.3

)
a

Th
e

fo
llo

w
in

g
re

gi
m

en
s

ar
e

in
cl

ud
ed

in
th

e
ov

er
al

lr
at

es
an

d
re

su
lte

d
in

no
30

-d
ad

ve
rs

e
dr

ug
ev

en
ts

:p
en

ic
ill

in
(2

1)
,a

m
ox

ic
ill

in
(4

7)
,d

ic
lo

xa
ci

lli
n

(1
),

ce
ph

al
ex

in
(4

4)
,s

ec
on

d-
ge

ne
ra

tio
n

ce
ph

al
os

po
rin

s
(3

8
),

ce
ft

az
id

im
e

(6
),

ce
ft

ar
ol

in
e

(8
),

az
tr

eo
na

m
(2

2)
,f

os
fo

m
yc

in
(1

0
),

ni
tr

of
ur

an
to

in
(2

6
),

tig
ec

yc
lin

e
(3

),
or

al
va

nc
om

yc
in

(8
4)

.
b

In
cl

ud
es

na
us

ea
,e

m
es

is
,n

on
–C

lo
st

rid
iu

m
di

ff
ic

ile
–a

ss
oc

ia
te

d
di

ar
rh

ea
.

c
O

th
er

ad
ve

rs
e

dr
ug

ev
en

ts
in

cl
ud

e
ce

fe
pi

m
e-

as
so

ci
at

ed
an

ap
hy

la
xi

s
(1

),
pi

pe
ra

ci
lli

n-
ta

zo
ba

ct
am

–a
ss

oc
ia

te
d

dr
ug

fe
ve

r(
1)

,c
ip

ro
flo

xa
ci

n-
as

so
ci

at
ed

te
nd

in
iti

s
(1

),
da

pt
om

yc
in

-a
ss

oc
ia

te
d

m
yo

si
tis

(1
),

tr
im

et
ho

pr
im

–
su

lfa
m

et
ho

xa
zo

le
–a

ss
oc

ia
te

d
pa

nc
re

at
iti

s
(1

),
va

nc
om

yc
in

-a
ss

oc
ia

te
d

hi
ve

s
(1

),
an

d
tr

im
et

ho
pr

im
–

su
lfa

m
et

ho
xa

zo
le

-r
el

at
ed

no
nh

iv
es

ra
sh

(1
).

d
So

m
e

pa
tie

nt
s

re
ce

iv
ed

m
or

e
th

an
1β

-la
ct

am
an

tib
io

tic
.

e
M

os
tp

at
ie

nt
s

(1
17

6
[7

9
%

])
re

ce
iv

ed
m

or
e

th
an

1a
nt

ib
io

tic
.

Research Original Investigation Adverse Events and Antibiotic Use in Hospitalized Patients

1312 JAMA Internal Medicine September 2017 Volume 177, Number 9 (Reprinted) jamainternalmedicine.com

© 2017 American Medical Association. All rights reserved.

Downloaded From: https://jamanetwork.com/ on 08/01/2022

approximately 40% of days of fluoroquinolone therapy were
unnecessary and 27% of regimens were associated with ad-
verse events including gastrointestinal events (14%), MDRO
colonization (8%), and CDI (4%). Finally, Macy and Contreras19

evaluated the incidence of cephalosporin-associated ADEs
using an administrative database and found that the most
frequently reported serious ADEs were CDI, occurring in ap-
proximately 1% of patients.

We believe that our study enhances these investigations
in a number of ways. First, unlike previous studies, we evalu-
ated antibiotic-associated ADEs that occurred in both the in-
patient setting as well as the outpatient setting after hospital
discharge, enabling us to produce a more global picture of the
overall incidence of antibiotic-associated ADEs.13,18,19,29 Our
previous work suggests that approximately 40% of antibiot-
ics prescribed for hospitalized patients represent antibiotics
prescribed at the time of hospital discharge that are to be con-
tinued after leaving the hospital.6 We believe that it is impor-

tant to include these antibiotic days in estimates of antibiotic-
associated adverse events for hospitalized patients. Second,
in our cohort, infectious diseases physicians and pharma-
cists reviewed all patient medical records to identify ADEs and
to determine whether they were most likely attributable to re-
cent or current antibiotic use using strict, predefined criteria.
In contrast, previous studies have generally used administra-
tive databases, in which relevant events are commonly mis-
coded and through which attributable risk cannot always be
assigned.13,18 Furthermore, we did not limit our evaluation to
specific antibiotic classes but, rather, included all antibiotic
classes.

Limitations
Our study has a number of limitations. This was a single-
center study at an academic hospital with a medically com-
plex patient population. Replication of our results at other in-
stitutions and in other patient populations is necessary to

Table 3. Proportion of 30-Day Antibiotic-Associated Adverse Drug Events in 1488 Hospitalized Patients Receiving Systemic Antibiotic Therapya

Antibiotic Agent

No. of
Patients
Receiving
Agent

No. (%)

Cardiac
Gastro-
intestinalb Hematologic

Hepato-
biliary Renal Neurologic

Other
Eventsc

β-Lactamsd 1187 0 59 (5.0) 27 (2.3) 6 (0.5) 17 (1.4) 10 (0.8) 2 (0.2)

Ampicillin 63 0 2 (3.2) 1 (1.6) 1 (1.6) 1 (1.6) 0 0

Amoxicillin-
clavulanate

102 0 3 (2.9) 0 0 0 0 0

Ampicillin-
sulbactam

52 0 1 (1.9) 0 0 2 (3.8) 0 0

Oxacillin 33 0 4 (12.1) 1 (3.0) 2 (6.0) 0 0 0

Piperacillin-
tazobactam

315 0 16 (5.1) 4 (1.3) 1 (0.3) 1 (0.3) 1 (0.3) 1 (0.3)

Cefazolin 79 0 0 1 (1.3) 0 2 (2.5) 0 0

Ceftriaxone 607 0 14 (2.3) 11 (1.8) 3 (0.5) 5 (0.8) 1 (0.2) 0

Cefpodoxime 89 0 2 (2.2) 0 0 0 0 0

Cefepime 414 0 10 (2.4) 6 (1.4) 0 6 (1.4) 7 (1.7) 1 (0.2)

Ertapenem 85 0 3 (3.5) 0 0 0 0 0

Meropenem 80 0 4 (5.0) 3 (3.8) 0 0 1 (1.3) 0

Non–β-lactams

Aminoglycosides 32 0 0 0 0 2 (6.3) 0 0

Azithromycin 400 1 (0.3) 1 (0.3) 0 4 (1.0) 0 0 0

Clindamycin 193 0 3 (1.6) 0 0 0 0 0

Daptomycin 8 0 0 0 0 0 0 1 (12.5)

Doxycycline 57 0 2 (3.5) 0 0 0 0 0

Fluoroquinolones 394 1 (0.3) 5 (1.3) 1 (0.3) 3 (0.8) 1 (0.3) 1 (0.3) 1 (0.3)

Linezolid 23 0 0 0 0 0 1 (4.3) 0

Metronidazole 175 0 1 (0.6) 0 0 0 1 (0.6) 0

Trimethoprim-
sulfamethoxazole

155 0 5 (3.2) 0 0 6 (3.9) 0 1 (0.6)

Intravenous
vancomycin

544 0 2 (0.4) 0 0 19 (3.5) 0 2 (0.4)

Any antibiotics 1488e 2 (0.1) 78 (5.2) 28 (1.9) 13 (0.9) 45 (3.0) 13 (0.9) 7 (0.5)
a The following regimens are included in the overall rates and resulted in no

30-d adverse drug events: penicillin (21), amoxicillin (47), dicloxacillin (1),
cephalexin (44), second-generation cephalosporins (38), ceftazidime (6),
ceftaroline (8), aztreonam (22), fosfomycin (10), nitrofurantoin (26),
tigecycline (3), oral vancomycin (84).

b Includes nausea, emesis, non–Clostridium difficile–associated diarrhea.
c Other adverse drug events include cefepime-associated anaphylaxis (1),

piperacillin-tazobactam–associated drug fever (1), ciprofloxacin-associated
tendinitis (1), daptomycin-associated myositis (1), trimethoprim-
sulfamethoxazole–associated pancreatitis (1), vancomycin-associated hives (1),
and trimethoprim-sulfamethoxazole-associated nonhives rash (1).

d Some patients received more than 1 β-lactam antibiotic.
e Most patients (1176 [79%]) received more than 1 antibiotic.

Adverse Events and Antibiotic Use in Hospitalized Patients Original Investigation Research

jamainternalmedicine.com (Reprinted) JAMA Internal Medicine September 2017 Volume 177, Number 9 1313

© 2017 American Medical Association. All rights reserved.

Downloaded From: https://jamanetwork.com/ on 08/01/2022

enhance the generalizability of our findings. This would also
allow for ADE estimates for antibiotic agents not included on
our hospital formulary. Furthermore, because prescriptions of
some antibiotics were so infrequent (eg, penicillin, ceftaro-
line fosamil, tigecycline), accurate estimates of some drug-
specific ADEs could not be calculated. Our approximations of
antibiotic-associated ADEs are likely underestimations for a
number of reasons. First, our hospital has had a robust anti-
biotic stewardship program since 2002 that remained active
during the study period, likely reducing overall antibiotic pre-
scriptions, durations of antibiotic therapy, and consequently
antibiotic-associated ADEs. Second, we were unable to evalu-
ate data from patients who had follow-up medical care out-
side the Epic Care Everywhere network, for example those who
presented to primary care clinicians, emergency depart-
ments, or urgent care centers not using the Epic electronic
medical record system.22 Of note, only 119 (8%) patients were
considered lost to follow-up with no subsequent inpatient or

outpatient visits documented in the Epic Care Everywhere net-
work. Additionally, it is plausible that a portion of patients in
this cohort may have previously experienced serious antibiotic-
associated ADEs, leading to future avoidance of these agents
(eg, hives from penicillin use as a child), also potentially
underestimating the incidence of antibiotic-associated ADEs.
Finally, we did not include excessively prolonged durations of
antibiotic therapy or inappropriately broad antibiotic use
toward our calculation of avoidable antibiotic-associated ADEs,
likely underestimating this value.

Conclusions
In summary, antibiotic-associated ADEs are common among
inpatients receiving antibiotics, some of which may be avoid-
able with more judicious use of antibiotics. The frequency of
antibiotic-associated ADEs may not be recognized by clini-

Table 4. Proportion of 1488 Patients Receiving Systemic Antibiotic Therapy Who Developed Adverse Drug Events (ADEs) Within 30 Daysa

Antibiotic Agents

No. (%)

Total ADEs Cardiac
Gastro-
intestinalb Hematologic

Hepato-
biliary Renal Neurologic

Other
Eventsc

Any β-lactamd 121 (65.1) 0 59 (75.6) 27 (96.4) 6 (46.2) 17 (37.8) 10 (76.9) 2 (28.6)

Ampicillin 4 (2.2) 0 2 (2.6) 1 (3.6) 0 1 (2.2) 0 0

Amoxicilin-
clavulanate

3 (1.6) 0 3 (3.8) 0 0 0 0 0

Ampicillin-
sulbactam

3 (1.6) 0 1 (1.3) 0 0 2 (4.4) 0 0

Oxacillin 7 (3.8) 0 4 (5.1) 1 (3.6) 2 (15.4) 0 0 0

Piperacillin-
tazobactam

24 (12.9) 0 16 (20.5) 4 (14.3) 1 (7.7) 1 (2.2) 1 (7.7) 1 (14.3)

Cefazolin 3 (1.6) 0 0 1 (3.6) 0 2 (4.4) 0 0

Ceftriaxone 34 (18.3) 0 14 (17.9) 11 (39.3) 3 (23.1) 5 (11.1) 1 (7.7) 0

Cefpodoxime 2 (1.1) 0 2 (2.6) 0 0 0 0 0

Cefepime 30 (16.1) 0 10 (12.8) 6 (21.4) 0 6 (13.3) 7 (53.8) 1 (14.3)

Ertapenem 3 (1.6) 0 3 (3.8) 0 0 0 0 0

Meropenem 8 (4.3) 0 4 (5.1) 3 (10.7) 0 0 1 (7.7) 0

Non–β-lactams

Aminoglycosides 2 (1.1) 0 0 0 0 2 (4.4) 0 0

Azithromycin 6 (3.2) 1 (50.0) 1 (1.3) 0 4 (30.8) 0 0 0

Clindamycin 3 (1.6) 0 3 (3.8) 0 0 0 0 0

Daptomycin 1 (0.5) 0 0 0 0 0 0 1 (14.3)

Doxycycline 2 (1.1) 0 2 (2.6) 0 0 0 0 0

Fluoroquinolones 13 (7.0) 1 (50.0) 5 (6.4) 1 (3.6) 3 (23.1) 1 (2.2) 1 (7.7) 1 (14.3)

Linezolid 1 (0.5) 0 0 0 0 0 1 (7.7) 0

Metronidazole 2 (1.1) 0 1 (1.3) 0 0 0 1 (7.7) 0

Trimethoprim-
sulfamethoxazole

12 (6.5) 0 5 (6.4) 0 0 6 (13.3) 0 1 (14.3)

Intravenous
vancomycin

23 (12.4) 0 2 (2.6) 0 0 19 (42.2) 0 2 (28.6)

All antibioticse 186 (100) 2 (100) 78 (100) 28 (100) 13 (100) 45 (100) 13 (100) 7 (100)
a The following regimens are included in the overall rates and resulted in no

30-d adverse drug events: penicillin (21), amoxicillin (47), dicloxacillin (1),
cephalexin (44), second-generation cephalosporins (38), ceftazidime (6),
ceftaroline (8), aztreonam (22), fosfomycin (10), nitrofurantoin (26),
tigecycline (3), oral vancomycin (84).

b Includes nausea, emesis, non-Clostridium difficile–associated diarrhea.
c Other ADEs include cefepime-associated anaphylaxis (1), piperacillin-

tazobactam–associated drug fever (1), ciprofloxacin-associated tendinitis (1),
daptomycin-associated myositis (1), trimethoprim-sulfamethoxazole–
associated pancreatitis (1), vancomycin-associated hives (1), and
vancomycin-associated nonhives, non–red man syndrome rash (1).

d Some patients received more than 1 β-lactam antibiotic.
e Most patients (1176 [79%]) received more than 1 antibiotic.

Research Original Investigation Adverse Events and Antibiotic Use in Hospitalized Patients

1314 JAMA Internal Medicine September 2017 Volume 177, Number 9 (Reprinted) jamainternalmedicine.com

© 2017 American Medical Association. All rights reserved.

Downloaded From: https://jamanetwork.com/ on 08/01/2022

cians because ADEs have varied manifestations, clinicians may
be unaware of the risks associated with specific antibiotic
agents, or because they may occur after patients are dis-
charged from the hospital. Our findings provide quantitative

data about the risk of ADEs that clinicians should consider
when weighing decisions to initiate or discontinue antibiotic
therapy and lend further credence to the importance of anti-
biotic stewardship to optimize patient safety.

ARTICLE INFORMATION

Accepted for Publication: April 6, 2017.

Published Online: June 12, 2017.
doi:10.1001/jamainternmed.2017.1938

Author Contributions: Dr Tamma had full access to
all of the data in the study and takes responsibility
for the integrity of the data and the accuracy of the
data analysis.
Study concept and design: Tamma, Avdic, Li,
Cosgrove.
Acquisition, analysis, or interpretation of data: All
authors.
Drafting of the manuscript: Tamma, Avdic, Li,
Dzintars.
Critical revision of the manuscript for important
intellectual content: Avdic, Li, Dzintars, Cosgrove.
Statistical analysis: Tamma, Li.
Obtained funding: Tamma, Avdic, Cosgrove.
Administrative, technical, or material support: Avdic,
Dzintars.
Supervision: Dzintars, Cosgrove.

Conflict of Interest Disclosures: None reported.

Funding/Support: This study was made possible
by an investigator-initiated grant from Pfizer
Independent Grants for Learning and Change and
The Joint Commission.

Role of the Funder/Sponsor: The funders had no
role in the design and conduct of the study;
collection, management, analysis, and
interpretation of the data; preparation, review, or
approval of the manuscript; and decision to submit
the manuscript for publication.

Additional Contributions: We thank Yuan Zhao,
MPH, Johns Hopkins University, and John Keenan,
MD, Johns Hopkins University, for their assistance
with data collection. Dr Keenan received a portion
of his salary from Pfizer/The Joint Commission.

REFERENCES

1. Magill SS, Edwards JR, Beldavs ZG, et al;
Emerging Infections Program Healthcare-
Associated Infections and Antimicrobial Use
Prevalence Survey Team. Prevalence of
antimicrobial use in US acute care hospitals,
May-September 2011. JAMA. 2014;312(14):1438-1446.

2. Hecker MT, Aron DC, Patel NP, Lehmann MK,
Donskey CJ. Unnecessary use of antimicrobials in
hospitalized patients: current patterns of misuse
with an emphasis on the antianaerobic spectrum of
activity. Arch Intern Med. 2003;163(8):972-978.

3. Camins BC, King MD, Wells JB, et al. Impact of an
antimicrobial utilization program on antimicrobial
use at a large teaching hospital: a randomized
controlled trial. Infect Control Hosp Epidemiol.
2009;30(10):931-938.

4. Ingram PR, Seet JM, Budgeon CA, Murray R.
Point-prevalence study of inappropriate antibiotic
use at a tertiary Australian hospital. Intern Med J.
2012;42(6):719-721.

5. Cosgrove SE, Seo SK, Bolon MK, et al; CDC
Prevention Epicenter Program. Evaluation of
postprescription review and feedback as a method
of promoting rational antimicrobial use:
a multicenter intervention. Infect Control Hosp
Epidemiol. 2012;33(4):374-380.

6. Tamma PD, Avdic E, Keenan JF, et al. What is the
more effective antibiotic stewardship intervention:
pre-prescription authorization or post-prescription
review with feedback? Clin Infect Dis. 2017;64(5):
537-543.

7. Bell BG, Schellevis F, Stobberingh E, Goossens H,
Pringle M. A systematic review and meta-analysis of
the effects of antibiotic consumption on antibiotic
resistance. BMC Infect Dis. 2014;14:13.

8. Hensgens MP, Goorhuis A, Dekkers OM, Kuijper
EJ. Time interval of increased risk for Clostridium
difficile infection after exposure to antibiotics.
J Antimicrob Chemother. 2012;67(3):742-748.

9. Liu NW, Shatagopam K, Monn MF, et al. Risk for
Clostridium difficile infection after radical
cystectomy for bladder cancer: analysis of a
contemporary series. Urol Oncol. 2015;33(12):
503.e17-503.e22.

10. Kuster SP, Rudnick W, Shigayeva A, et al;
Toronto Invasive Bacterial Diseases Network.
Previous antibiotic exposure and antimicrobial
resistance in invasive pneumococcal disease:
results from prospective surveillance. Clin Infect Dis.
2014;59(7):944-952.

11. Alshammari TM, Larrat EP, Morrill HJ, Caffrey
AR, Quilliam BJ, LaPlante KL. Risk of hepatotoxicity
associated with fluoroquinolones: a national
case-control safety study. Am J Health Syst Pharm.
2014;71(1):37-43.

12. Torres MJ, Blanca M, Fernandez J, et al; ENDA;
EAACI Interest Group on Drug Hypersensitivity.
Diagnosis of immediate allergic reactions to
beta-lactam antibiotics. Allergy. 2003;58(10):
961-972.

13. Shehab N, Patel PR, Srinivasan A, Budnitz DS.
Emergency department visits for
antibiotic-associated adverse events. Clin Infect Dis.
2008;47(6):735-743.

14. File TM Jr, Segreti J, Dunbar L, et al.
A multicenter, randomized study comparing the
efficacy and safety of intravenous and/or oral
levofloxacin versus ceftriaxone and/or cefuroxime
axetil in treatment of adults with community-
acquired pneumonia. Antimicrob Agents Chemother.
1997;41(9):1965-1972.

15. Hammond DA, Smith MN, Li C, Hayes SM,
Lusardi K, Bookstaver PB. Systematic review and
meta-analysis of acute kidney injury associated
with concomitant vancomycin and
piperacillin/tazobactam. Clin Infect Dis. 2017;64(5):
666-674.

16. Martin RM, Biswas PN, Freemantle SN, Pearce
GL, Mann RD. Age and sex distribution of suspected

adverse drug reactions to newly marketed drugs in
general practice in England: analysis of 48 cohort
studies. Br J Clin Pharmacol. 1998;46(5):505-511.

17. Pretorius RW, Gataric G, Swedlund SK, Miller JR.
Reducing the risk of adverse drug events in older
adults. Am Fam Physician. 2013;87(5):331-336.

18. Lin RY, Nuruzzaman F, Shah SN. Incidence and
impact of adverse effects to antibiotics in
hospitalized adults with pneumonia. J Hosp Med.
2009;4(2):E7-E15.

19. Macy E, Contreras R. Adverse reactions
associated with oral and parenteral use of
cephalosporins: a retrospective population-based
analysis. J Allergy Clin Immunol. 2015;135(3):
745-52.e5.

20. Werner NL, Hecker MT, Sethi AK, Donskey CJ.
Unnecessary use of fluoroquinolone antibiotics in
hospitalized patients. BMC Infect Dis. 2011;11:187.

21. Blumenthal KG, Kuhlen JL Jr, Weil AA, et al.
Adverse drug reactions associated with ceftaroline
use: a 2-center retrospective cohort. J Allergy Clin
Immunol Pract. 2016;4(4):740-746.

22. Epic Care Everywhere Network.
https://www.epic.com/CareEverywhere/. Accessed
December 6, 2016.

23. Johns Hopkins Medicine Antibiotic Guidelines
2015-2016. http://www.hopkinsmedicine.org/amp.
Accessed February 5, 2017.

24. Goldstein SL, Kirkendall E, Nguyen H, et al.
Electronic health record identification of
nephrotoxin exposure and associated acute kidney
injury. Pediatrics. 2013;132(3):e756-e767.

25. Centers for Disease Control and Prevention.
Antimicrobial resistant phenotype definitions.
https://www.cdc.gov/nhsn/pdfs/ps-analysis
-resources/phenotype_definitions.pdf. Accessed
January 26, 2017.

26. Anand A, Bashey B, Mir T, Glatt AE.
Epidemiology, clinical manifestations, and outcome
of Clostridium difficile-associated diarrhea. Am J
Gastroenterol. 1994;89(4):519-523.

27. Gerding DN, Olson MM, Peterson LR, et al.
Clostridium difficile-associated diarrhea and colitis
in adults: a prospective case-controlled
epidemiologic study. Arch Intern Med. 1986;146(1):
95-100.

28. Johannes CB, Ziyadeh N, Seeger JD, Tucker E,
Reiter C, Faich G. Incidence of allergic reactions
associated with antibacterial use in a large,
managed care organisation. Drug Saf. 2007;30(8):
705-713.

29. Meropol SB, Localio AR, Metlay JP. Risks and
benefits associated with antibiotic use for acute
respiratory infections: a cohort study. Ann Fam Med.
2013;11(2):165-172.

Adverse Events and Antibiotic Use in Hospitalized Patients Original Investigation Research

jamainternalmedicine.com (Reprinted) JAMA Internal Medicine September 2017 Volume 177, Number 9 1315

© 2017 American Medical Association. All rights reserved.

Downloaded From: https://jamanetwork.com/ on 08/01/2022

Lyons I, et al. BMJ Qual Saf 2018;27:892–901. doi:10.1136/bmjqs-2017-007476892

Original research

► Additional material is
published online only. To view
please visit the journal online
(http:// dx. doi. org/ 10. 1136/
bmjqs- 2017- 007476).

For numbered affiliations see
end of article.

Correspondence to
Professor Bryony Dean Franklin,
Centre for Medication Safety
and Service Quality, Imperial
College Healthcare NHS Trust/
UCL School of Pharmacy,
London, W6 8RF, UK;
bryony. [email protected] nhs. net

Received 9 October 2017
Revised 12 March 2018
Accepted 24 March 2018
Published Online First
7 April 2018

To cite: Lyons I, Furniss D,
Blandford A, et al.
BMJ Qual Saf
2018;27:892–901.

Errors and discrepancies in the
administration of intravenous
infusions: a mixed methods
multihospital observational study

imogen lyons,1 Dominic Furniss,1 ann Blandford,1 gillian chumbley,2
ioanna iacovides,3 li Wei,4 anna cox,1 astrid Mayer,5,6 Jolien Vos,1
galal h galal-edeen,1,7 Kumiko O schnock,8,9 Patricia c Dykes,9,10
David W Bates,8,9 Bryony Dean Franklin4,11

ABSTRACT
Introduction Intravenous medication administration
has traditionally been regarded as error prone, with high
potential for harm. A recent US multisite study revealed few
potentially harmful errors despite a high overall error rate.
However, there is limited evidence about infusion practices
in England and how they relate to prevalence and types of
error.
Objectives To determine the prevalence, types and severity
of errors and discrepancies in infusion administration in
English hospitals, and to explore sources of variation,
including the contribution of smart pumps.
Methods We conducted an observational point prevalence
study of intravenous infusions in 16 National Health
Service hospital trusts. Observers compared each infusion
against the medication order and local policy. Deviations
were classified as errors or discrepancies based on their
potential for patient harm. Contextual issues and reasons
for deviations were explored qualitatively during observer
debriefs.
Results Data were collected from 1326 patients and
2008 infusions. Errors were observed in 231 infusions
(11.5%, 95% CI 10.2% to 13.0%). Discrepancies were
observed in 1065 infusions (53.0%, 95% CI 50.8% to
55.2%). Twenty-three errors (1.1% of all infusions) were
considered potentially harmful; none were judged likely to
prolong hospital stay or result in long-term harm. Types and
prevalence of errors and discrepancies varied widely among
trusts, as did local policies. Deviations from medication
orders and local policies were sometimes made for efficiency
or patient need. Smart pumps, as currently implemented,
had little effect, with similar error rates observed in infusions
delivered with and without a smart pump (10.3% vs 10.8%,
p=0.8).
Conclusion Errors and discrepancies are relatively
common in everyday infusion administrations but most
have low potential for patient harm. Better understanding
of performance variability to strategically manage risk
may be a more helpful tactic than striving to eliminate all
deviations.

InTRoduCTIon
Intravenous medication administration
is complex, and data suggest that errors

are common. For example, a systematic
review of nine studies across various
stages of intravenous medication prepa-
ration and administration reported errors
in 73% of intravenous doses.1 However,
published error rates vary widely, from
18% to 173% of intravenous doses in
studies using structured observation of
medication administration.2

Amidst concerns over safety, technol-
ogies such as ‘smart pumps’ have been
advocated. These incorporate dose error
reduction software to check programmed
infusion rates against preset limits within a
customisable drug library. However, dose
limits can be over-ridden, and evidence
regarding their impact is mixed.3 4 While
unintended infusion overdoses repre-
sent a major safety concern, there are
many factors that affect infusion admin-
istration, and smart pumps are just one
possible solution.

A recent multisite US study using
structured observation reported a high
prevalence of intravenous infusion admin-
istration errors and procedural failures,
even with the use of smart pumps, yet few
potentially harmful errors.4 Building on
this and an earlier US study,5 we therefore
wanted to conduct a similar study in the
UK with a larger sample size6 to confirm
or refute these findings in a different
context in which smart pumps are less
common. In contrast to previous studies,
we also wanted to incorporate a Safety II
approach to interpret our findings.7 This
approach moves away from the tradi-
tional focus of classifying all deviations as
errors and blaming the human for unre-
liable processes. Instead it encourages

o
n
A

u
g

u
st 1

, 2
0
2

2
b

y g
u

e
st. P

ro
te

cte
d

b
y co

p
yrig

h
t.

h
ttp

://q
u
a
litysa

fe
ty.b

m
j.co

m
/

B
M

J Q
u

a
l S

a
f: first p

u
b

lish
e

d
a

s 1
0

.1
1

3
6

/b
m

jq
s-2

0
1

7
-0

0
7

4
7
6
o

n
7

A
p
ril 2

0
1
8
. D

o
w

n
lo

a
d
e
d
fro

m

893Lyons I, et al. BMJ Qual Saf 2018;27:892–901. doi:10.1136/bmjqs-2017-007476

Original research

consideration of deviations in terms of performance
variability, how to understand and manage this vari-
ability, and that the human component can make posi-
tive contributions to safety.7 8 Our objectives were to
determine the prevalence, types and severity of errors
and discrepancies in intravenous infusions in England
and to explore sources of variation, including the
potential contribution of smart pumps, using a Safety
II approach.

MeThodS
Study design
We used a point prevalence observational study of intra-
venous infusions in a sample of hospitals, followed
by debriefs with staff at each site to gather additional
context. Although we built on previous studies,4 5 we did
not consider all deviations from the medication order
or local policy to be errors: minor or intentional devi-
ations were classed as discrepancies. The study protocol
was published previously6 and the study was approved
by a National Health Service (NHS) Research Ethics
Committee (14/SC/0290).

Study setting and sample
We used a purposive sampling strategy to select 16
NHS trusts in England, aiming for a diverse range of
organisations in terms of type, size, location, patient
safety metrics and use of infusion devices and smart
pump technology.6 Online supplementary appendices
1 and 2 summarise the recruitment process and char-
acteristics of each participating trust. We conducted
observations in three clinical areas (general medicine,
general surgery and critical care) in 13 trusts; in eight
of these we also conducted observations in paediatrics
and oncology day care. Two specialist children’s hospi-
tals collected paediatric data only. One further trust
collected oncology day care data at three hospital sites.
We aimed to include a sample of 2100 infusions across
all participating sites to give a CI around a 10% error
rate of 8.7%–11.3%.6

data collection
Data were collected between April 2015 and December
2016. At each trust, two observers (usually a nurse and a
pharmacist) employed in the organisation were trained
by the research team to collect data. This training
included highlighting the types of deviations to look
for, conducting observations in the presence of the
research team where possible and using sample cases to
facilitate discussion about classification of deviations
identified. Observers were also requested to identify
and familiarise themselves with relevant local policies
and guidelines prior to data collection. Observers then
spent 1 weekday or equivalent collecting data in each
clinical area. One clinical area could comprise one or
more wards. Observers aimed to collect data on all
intravenous infusions being administered at the time

of data collection, including drugs, fluids, blood prod-
ucts and nutrition. Bolus doses were excluded, except
where a prescribed bolus was given as an infusion, or
vice versa. Completed infusions were excluded even if
still attached to the patient. Patients were not observed
if they were in isolation due to infection risks, were
receiving care that would have required interruption
or were off the ward.

Observers compared each medication being admin-
istered against the prescription and local policies/
guidance,6 and consulted clinical staff if needed to
understand any deviations. Data were recorded using
a standardised paper form and subsequently uploaded
to a secure web-based tool.9 No patient identifiable
data were recorded. Suspected errors were raised with
clinical staff so they could be corrected if needed; local
reporting practices were then followed.

Identifying and assessing deviations
We recorded any deviations from a prescriber’s
written or electronic medication order, the hospital’s
intravenous policy and guidelines, or the manufactur-
er’s instructions. We included the administration of
medication to which the patient had a documented
allergy or sensitivity, but did not assess other aspects of
the clinical appropriateness of the medication order.
We also collected data on policy violations and proce-
dural or documentation deviations that may increase
the likelihood of medication administration errors
occurring. These included patients not wearing an
identification wristband with the correct information,
medication or infusion administration sets not being
labelled in accordance with hospital policy and failure
to document the administration of medication in line
with policy. Finally, we encouraged observers to record
any other irregularities, anomalies or workarounds
related to the administration. Some of these were
grouped together for analysis and formed new catego-
ries. Online supplementary appendix 3 presents defi-
nitions of deviation types.

Local observers rated each deviation using an adap-
tation of the National Coordinating Council for Medi-
cation Error Reporting and Prevention (NCCMERP)
severity index.10 Ratings were based on the likelihood
of the deviation resulting in patient harm if it had not
been intercepted, and were used to classify the devia-
tions as discrepancies (rated A1 or A2) or errors (rated
from Cto I) (online supplementary appendix 4).6 Based
on these ratings we developed and clarified our clas-
sifications, recognising that deviations could be either
errors or discrepancies, either in medication adminis-
tration or in the associated procedural and documen-
tation requirements (figure 1). We report separately
on a comparison between the NCCMERP ratings and
an alternative severity classification method based on
expert judgement.11

Observers at each trust documented brief descrip-
tions of any deviations identified and provided further

o
n
A

u
g

u
st 1

, 2
0
2

2
b

y g
u

e
st. P

ro
te

cte
d

b
y co

p
yrig

h
t.

h
ttp

://q
u
a
litysa

fe
ty.b

m
j.co

m
/

B
M

J Q
u

a
l S

a
f: first p

u
b

lish
e

d
a

s 1
0

.1
1

3
6

/b
m

jq
s-2

0
1

7
-0

0
7

4
7
6
o

n
7

A
p
ril 2

0
1
8
. D

o
w

n
lo

a
d
e
d
fro

m

894 Lyons I, et al. BMJ Qual Saf 2018;27:892–901. doi:10.1136/bmjqs-2017-007476

Original research

qualitative insights during semistructured debriefs
once data collection was complete.

data management and analysis
Clinicians within the research team reviewed devia-
tions that had uncertain classifications, for example,
where local observers highlighted that they found
categorisation difficult and where observers had clas-
sified similar deviations differently. Clinicians within
the research team also reviewed each error rated cate-
gory D (‘likely to have required increased monitoring
and/or intervention to preclude harm’) and above.
Minor changes were made to classifications of type
and severity of error as needed.

Error and discrepancy rates were calculated as the
proportion of infusions with at least one error or
discrepancy using total opportunities for error (total
number of doses administered, plus any omitted doses)
as the denominator. Results are presented according
to overall error and discrepancy rates, and individual
types of errors and discrepancies, grouped into medi-
cation administration deviations, and procedural and
documentation deviations. Variations in deviation
rates between clinical areas, delivery modes and infu-
sion types were explored descriptively with their 95%
CIs, and Χ2 tests where appropriate. Qualitative data
were analysed inductively.

ReSulTS
Overall, 6491 patients were present in the clinical
areas observed, of whom 1545 (23.8%) were receiving
and/or prescribed an intravenous infusion at the time
of data collection. Data were collected from 1326

(85.8%) patients, who were administered and/or
prescribed 2008 infusions.

Frequency, types and potential severity of errors and
discrepancies
Overall, 240 errors and 1491 discrepancies were
identified across 2008 intravenous infusions. Table 1
presents the numbers and percentages of infusions and
patients affected. Table 2 shows the types of devia-
tions observed and their likely harm. Ninety per cent
of observed errors were considered unlikely to cause
harm despite reaching the patient (NCCMERP cate-
gory C). Twenty-two errors (9.5%) were category
D, and one (0.4%) category E; these 23 potentially
harmful errors represent 1.1% of infusions. Examples
in each severity category are presented in table 3.

Medication administration deviations
Overall, 427 (21.3%) infusions involved at least one
medication administration error (n=211) or discrep-
ancy (n=257). The most frequent types of deviation
concerned rates and unauthorised medications.

Rate deviations
Overall, 152 infusions (7.6%) were being adminis-
tered at a different rate from that prescribed; 77 were
classified as errors (rated ≥C) and 75 as discrepancies
(rated A1 or A2). A large proportion involved order
changes that had not been correctly documented
and infusions titrated based on the patient’s clinical
need or fluid allowance without such titration being
prescribed. Three deviations involved prescribed

Figure 1 Classification of deviations, errors and discrepancies.

o
n
A

u
g

u
st 1

, 2
0
2

2
b

y g
u

e
st. P

ro
te

cte
d

b
y co

p
yrig

h
t.

h
ttp

://q
u
a
litysa

fe
ty.b

m
j.co

m
/

B
M

J Q
u

a
l S

a
f: first p

u
b

lish
e

d
a

s 1
0

.1
1

3
6

/b
m

jq
s-2

0
1

7
-0

0
7

4
7
6
o

n
7

A
p
ril 2

0
1
8
. D

o
w

n
lo

a
d
e
d
fro

m

895Lyons I, et al. BMJ Qual Saf 2018;27:892–901. doi:10.1136/bmjqs-2017-007476

Original research

boluses administered as infusions, and one was a
prescribed infusion given as a bolus.

About 31% of rate errors occurred in infusions
delivered via gravity (without using a pump), despite
accounting for just 8% of infusions. Of the 12 most
serious rate errors (rated D), eight were administered
via gravity; these included red blood cells, vancomycin,
paracetamol and piperacillin/tazobactam. Many medi-
cation orders specified a duration rather than rate (eg,
over 8 hours). In one case an infusion was observed
running at a very high rate to ‘catch up’—1 L of Plas-
malyte 148 had been prescribed over 24 hours; at the
time of observation, 27 hours after the start time, the
rate was set at 500 mL/hour.

Unauthorised medication
Eighty-nine infusions did not have a corresponding
medication order. Thirteen were flushes that did
not require a medication order according to local
policy. Therefore, 76 infusions (3.8%; 75 errors, one
discrepancy) were judged to be unauthorised. Almost
half were fluids used to flush the line, commonly in
oncology settings, including sodium chloride 0.9%
(n=29), dextrose (1), Plasmalyte (2) and heparin
(3). A further seven infusions were sodium chloride
0.9% administered at low rates to keep the vein
open. Twenty infusions were unauthorised repeats of
previously prescribed maintenance fluids. Four were
administered based on verbal orders that had not
been documented at the time of observation. Of the
remaining 10 unauthorised infusions, seven involved
maintenance fluids and three were drugs (calcium foli-
nate, remifentanil, insulin). The remifentanil infusion
had been prescribed and subsequently discontinued,
but not represcribed after a decision to resedate the
patient.

Procedural and documentation deviations
Overall, 961 infusions (47.9%) had at least one proce-
dural or documentation error (n=24) or discrepancy
(n=1219). Table 2 shows the frequency and severity of
different types of procedural and documentation devi-
ations. Non-compliance with hospital requirements
for labelling infusion administration sets was most
common. Procedural or documentation errors mostly
involved unlabelled syringes, or infusions where the
label was significantly inaccurate. For example, a
patient prescribed 60 mg pamidronate was being
administered an infusion labelled as 30 mg, but staff
confirmed the patient had received the correct dose.

While some of the discrepancies identified in our
study were deviations from protocols that may have
been intentional workarounds, this was not always
the case. Some were minor, non-clinically significant
variations from what was prescribed that did not meet
our definition of a medication administration error
(eg, small deviations in flow rate or concentration,
or minor delays to maintenance fluids’ start or finish T

ab
le

1

N
um

be
r a

nd
p

ro
po

rt
io

n
of

in
fu

si
on

s
an

d
pa

tie
nt

s
w

ith
a

t l
ea

st
o

ne
e

rr
or

o
r d

is
cr

ep
an

cy

In
fu

si
o

ns
(

n=
20

08
)

Pa
ti

en
ts

(
n=

13
26

)

O
ne

o
r

m
o

re
e

rr
o

rs
(

ie
, C

to
I

s
ev

er
it

y
ra

ti
ng

s)
n

(%
; 9

5%
C

I)

O
ne

o
r

m
o

re
d

is
cr

ep
an

ci
es

(
ie

, A
1

an
d

A

2
se

ve
ri

ty
r

at
in

g
s)

n
(%

; 9
5%

C
I)

O
ne

o
r

m
o

r e
e

rr
o

rs
(

ie
, C

to
I

s
ev

er
it

y
ra

ti
ng

s)
n

(%
; 9

5%
C

I)

O
ne

o
r

m
o

r e
d

is
cr

ep
an

ci
es

(
ie

, A
1

an
d

A

2
se

ve
ri

ty
r

at
in

g
s)

n
(%

; 9
5%

C
I)

M
ed

ic
at

io
n

ad
m

in
is

tr
at

io
n

de
vi

at
io

ns
20

7
(1

0.
3;

9
.1

%
to

1
1.

7%
)

24
1

(1
2.

0;
1

0.
6%

to
1

3.
5%

)
19

7
(1

4.
9;

1
3.

0%
to

1
6.

9%
)

22
2

(1
6.

7;
1

4.
8%

to
1

8.
8%

)
Pr

oc
ed

ur
al

o
r d

oc
um

en
ta

tio
n

de
vi

at
io

ns
24

(1
.2

; 0
.8

%
to

1
.8

%
)

95
3

(4
7.

5;
4

5.
3%

to
4

9.
7%

)
23

(1
.7

; 1
.1

%
to

2
.6

%
)

69
4

(5
2.

3;
4

9.
6%

to
5

5.
1%

)
M

is
ce

lla
ne

ou
s

5
(0

.2
; 0

.1
%

to
0

.6
%

)
13

(0
.6

; 0
.3

%
to

1
.1

%
)

5
(0

.4
; 0

.2
%

to
0

.9
%

)
13

(1
.0

; 0
.6

%
to

1
.7

%
)

A
ll

de
vi

at
io

ns
*

23
1

(1
1.

5;
1

0.
2%

to
1

3.
0%

)
10

65
(5

3.
0;

5
0.

8%
to

5
5.

2%
)

21
9

(1
6.

5;
1

4.
6%

to
1

8.
6%

)
78

1
(5

8.
9;

5
6.

2%
to

6
1.

6%
)

*S
om

e
in

fu
si

on
s

w
er

e
af

fe
ct

ed
b

y
m

or
e

th
an

o
ne

ty
pe

o
f d

is
cr

ep
an

cy
; t

he
re

fo
re

th
e

nu
m

be
r a

nd
p

er
ce

nt
ag

e
of

in
fu

si
on

s
af

fe
ct

ed
b

y
at

le
as

t o
ne

e
rr

or
o

r d
is

cr
ep

an
cy

o
f a

ny
ty

pe
is

n
ot

th
e

su
m

o
f e

ac
h

de
vi

at
io

n
ty

pe
.

o
n
A

u
g

u
st 1

, 2
0
2

2
b

y g
u

e
st. P

ro
te

cte
d

b
y co

p
yrig

h
t.

h
ttp

://q
u
a
litysa

fe
ty.b

m
j.co

m
/

B
M

J Q
u

a
l S

a
f: first p

u
b

lish
e

d
a

s 1
0

.1
1

3
6

/b
m

jq
s-2

0
1

7
-0

0
7

4
7
6
o

n
7

A
p
ril 2

0
1
8
. D

o
w

n
lo

a
d
e
d
fro

m

896 Lyons I, et al. BMJ Qual Saf 2018;27:892–901. doi:10.1136/bmjqs-2017-007476

Original research

times due to being interrupted to administer intrave-
nous antibiotics), and some were minor documenta-
tion discrepancies.

Sources of variation in error and discrepancy rates
Error rates among trusts ranged from 2.7% to 24.4%,
and discrepancy rates from 13.5% to 100% of infu-
sions, with no evidence of a relationship between
error and discrepancy rates (figure 2). Procedural or
documentation deviations ranged from 9.9% to 100%
of infusions across trusts, reflecting wide variation in
hospital policies and how they were enacted in prac-
tice. Some trusts had stringent policy requirements (eg,
trust K) whereas others did not (eg, trust J); some had
requirements that staff were unaware of in practice
(eg, trusts D and P).

Variation was also evident among clinical areas
and different infusion types (online supplementary
appendix 5). Infusions observed in critical care had a
lower error rate (7.0%); the error rate for paediatric
areas was similar to that for adult non-critical care

areas. Patient-controlled analgesia pumps and syringe
drivers had the lowest error rates at 6.4% and 5.1%,
respectively, with infusions delivered via gravity the
highest (21.5% of 163 infusions). Error rates also
varied by type of medication; maintenance fluids (eg,
sodium chloride 0.9%) had a high error rate (18.5%)
compared with drugs (6.9%), blood products (9.1%)
and parenteral nutrition (2.9%).

Eleven of 16 hospitals (69%) used smart pumps
(ie, an infusion pump with a drug library and/or dose
error reduction software enabled) in at least one clin-
ical area. However, just 640 (32%) infusions were
administered using a smart pump (online supplemen-
tary appendix 5). Infusions delivered using smart
pumps had similar error rates to those using other
pumps (10.3% vs 10.8%; p=0.8). No appropriate
entry was available in the drug library for one-third
of infusions administered using a smart pump. Of 424
infusions with a library entry available, the library
was used in 356 (84%) cases. There was no significant
difference in error rates for doses given via a drug

Table 2 Number, frequency and potential severity of each type of deviation

Type of deviation

Errors Discrepancies

NCCMERP severity rating n (% of 2008
infusions)

NCCMERP severity
rating n (% of 2008

infusions)C D E A1 A2

Medication administration deviations
Rate deviation 65 12 – 77 (3.8) 48 27 75 (3.7)
Unauthorised medication 72 3 – 75 (3.7) – 1 1 (0.0)
Administration start time discrepancy 13 – – 13 (0.6) 31 8 39 (1.9)
Incomplete or delayed completion 10 – – 10 (0.5) 4 27 31 (1.5)
Expired drug 11 – – 11 (0.5) 1 1 2 (0.1)
Dose discrepancy 5 2 – 7 (0.3) 6 6 12 (0.6)
Wrong drug/fluid/diluent 11 – – 11 (0.5) 1 1 2 (0.1)
Omitted medications (not administered at time of

data collection)
2 3 – 5 (0.2) 1 6 7 (0.3)

Roller clamp positioned incorrectly or inappropriately 1 – – 1 (0.0) – 10 10 (0.5)
Concentration discrepancy – – 1 1 (0.0) 7 2 9 (0.4)
Drug library not used or incorrectly used (in the case

of smart pumps)
– – – – – 67 67 (3.3)

Allergy oversight – – – – 2 – 2 (0.1)
All medication administration deviations 190 20 1 211 101 156 257
Procedure or documentation deviations
Infusion administration set not tagged/labelled

correctly
– – – – – 537 537 (26.8)

Documentation of the administration 1 – – 1 (0.0) – 334 334 (16.6)
Additive label missing or incorrect 16 1 – 17 (0.8) 2 200 202 (10.1)
Patient identification* 6 – – 6 (0.3) – 110 110 (5.5)
Documentation of the medication order – – – – 7 31 36 (1.8)
All procedure or documentation deviations 23 1 – 24 9 1212 1219
Miscellaneous 4 1 – 5 (0.2) 4 9 13 (0.6)
All deviations 217 22 1 240 114 1377 1491
*Deviations are counted per infusion; this figure includes patient identification deviations (ie, no name band) applied to all infusions for those patients.
There were 88 patient identification discrepancies, counting each once per patient.
NCCMERP, National Coordinating Council for Medication Error Reporting and Prevention.

o
n
A

u
g

u
st 1

, 2
0
2

2
b

y g
u

e
st. P

ro
te

cte
d

b
y co

p
yrig

h
t.

h
ttp

://q
u
a
litysa

fe
ty.b

m
j.co

m
/

B
M

J Q
u

a
l S

a
f: first p

u
b

lish
e

d
a

s 1
0

.1
1

3
6

/b
m

jq
s-2

0
1

7
-0

0
7

4
7
6
o

n
7

A
p
ril 2

0
1
8
. D

o
w

n
lo

a
d
e
d
fro

m

897Lyons I, et al. BMJ Qual Saf 2018;27:892–901. doi:10.1136/bmjqs-2017-007476

Original research

library versus those given without (online supple-
mentary appendix 5). Discrepancy rates were higher
in infusions delivered using smart pumps (61.7% of
640 infusions) compared with those without smart
features (46.4% of 1202 infusions, p<0.001). Sixty-
seven discrepancies were identified in the use of a
smart pump drug library: 61 where the drug library
was bypassed completely and six where the wrong
entry was selected. However, differences in discrep-
ancy rates were more commonly linked with policy
requirements for labelling infusions and administra-
tion sets at different sites; when discrepancies related
to use of a drug library are excluded, the discrepancy
rate remains higher in infusions delivered via smart
pump (59.2% of 640 infusions).

Qualitative insights
Information provided by observers revealed some
reasons for deviations. Some were simple slips or
lapses such as confusing diluents and forgetting
to open roller clamps to start the infusion; others
involved lack of knowledge of policy requirements.
Staff also reported deliberate deviations that would
benefit patients but conflicted with official rules
and formal procedures, for example, giving patients
fluids that had not yet been prescribed when a doctor
was unavailable (unauthorised fluids) and keeping
lines patent by switching to a low infusion rate in
anticipation of another infusion being needed (rate
deviation). There were several instances of inaccu-
rate prescriptions that were ‘corrected’ and admin-
istered by nurses without getting the order changed
prior to administration. However, in one case the
administering nurse incorrectly assumed an unusual
prescription was wrong (table 3—piperacillin/tazo-
bactam).

In some instances, nursing staff actively tried to
balance risk and efficiency rather than follow proce-
dures mechanistically. For example, staff reported
stopping infusions (delay in completion) when patients
left the ward for investigations so a nurse did not have
to accompany the patient when staffing resources
were stretched. In addition, some nurses objected to
spending time labelling administration sets and writing
batch numbers on additive labels for short infusions
that would soon be discarded.

Observers at some trusts reported that collecting
the study data provided insights into the reasons for
some deviations and helped them identify solutions.
For example, at one site where poor compliance with
documentation of medication administration was
recorded, the trust subsequently purchased handheld
computers to allow staff to access electronic records in
closer proximity to patients.

dISCuSSIon
We found that 1 in 10 intravenous infusions involved
an error, and one in two involved a discrepancy.
However, few were considered likely to cause patient
harm. There was considerable variability in errors,
discrepancies, policies and practices among trusts. Our
mixed methods approach offers insights into some
reasons for this variability. Nurses can be a source of
resilience, compensating for deficiencies and vulner-
abilities in the system; however, this same adaptive
capacity can also lead to unsatisfactory outcomes.12
Informed by Safety II, 7 8 our findings suggest the need
to question traditional notions of ‘error’ and the goal
of eliminating all errors and discrepancies. Instead we
reflect on a broader notion of deviations, highlight
positive contributions to efficiency and safety that go
beyond compliance and explore strategic interventions
to manage performance variability.

Table 3 Examples of observed deviations in the administration
of intravenous infusions

Severity
category Examples

E ► Patient was administered 2 g vancomycin diluted in 250 mL of
sodium chloride 0.9%. The drug should have been diluted in
500 mL of sodium chloride 0.9% (concentration error: severity
category E) and administered over at least 240 min. The drug
was observed running too fast via gravity feed (rate error: D).
The chart had not been signed to confirm the administration had
been double-checked as required (documentation discrepancy:
A2). The patient suffered from pain and red lumps along arm.

D ► Piperacillin/tazobactam was prescribed to be given over 3 hours.
However, it was given as a bolus over 3–5 min, which is the
most common way to administer this antibiotic. The nurses
presumed the doctors had made a mistake and corrected it.
However, this had been prescribed intentionally after discussions
with the consultant, with microbiology, with pharmacy and the
drug manufacturer due to the patient’s poor renal function. This
clinical decision was recorded in the patient’s notes but nursing
staff had not reviewed these.

► 40 mmol of potassium chloride rather than the prescribed
20 mmol was administered together with 10 mmol magnesium
sulfate in sodium chloride 0.9% at 1000 mL/hour.

C ► 1 L sodium chloride 0.9% with potassium chloride 0.15% was
prescribed over 12 hours. The documented start time was 23:25.
When observed at 13:00 the following day the infusion was not
running and approximately 150 mL remained. The infusion should
have been complete but the pump was not plugged in and the
battery was empty.

► A medication order for 20 mcg fentanyl stated diluent as
dextrose 5%, however the drug was prepared and administered
in sodium chloride 0.9%.

A2 ► Electronic prescription specified 1 L of sodium chloride 0.9%
over 8 hours. Started at 02:00 thus due to finish 10:00 but at
09:25 there was still 500 mL to run. The infusion was paused
at the time of observation as the patient was receiving an
intermittent amoxicillin infusion.

► Hartmann’s solution had been selected in the smart pump’s drug
library but the infusion being administered was sodium chloride
0.9% (at the correct rate prescribed).

A1 ► The prescribed rate was 250 mL/hour for 123 mg paclitaxel in
250 mL sodium chloride 0.9%. However, the final reconstituted
volume was 290.5 mL, which was being infused at 290 mL/hour
to give the same rate of administration as prescribed.

► Administration of piperacillin/tazobactam was delayed by
approximately 30 min.

o
n
A

u
g

u
st 1

, 2
0
2

2
b

y g
u

e
st. P

ro
te

cte
d

b
y co

p
yrig

h
t.

h
ttp

://q
u
a
litysa

fe
ty.b

m
j.co

m
/

B
M

J Q
u

a
l S

a
f: first p

u
b

lish
e

d
a

s 1
0

.1
1

3
6

/b
m

jq
s-2

0
1

7
-0

0
7

4
7
6
o

n
7

A
p
ril 2

0
1
8
. D

o
w

n
lo

a
d
e
d
fro

m

898 Lyons I, et al. BMJ Qual Saf 2018;27:892–901. doi:10.1136/bmjqs-2017-007476

Original research

disentangling errors, discrepancies and harm
Overall, we found a lower error rate (11.5%) than
that reported in much research into intravenous
medication error (range 35%–85.9%).13–15 Some of
this difference can be explained by methodological
differences, for example, inclusion of bolus doses and
preparation errors in other studies. The difficulties in
comparing error rates between studies using different
methods and definitions, in different contexts, have
been well documented.15 16 Comparing studies using
similar methods,4 5 we found broadly comparable rates
of potentially harmful errors, with errors rated D or
above in 1.1% of infusions in our study, and 0.4%4
and 3.8%5 elsewhere. We also identified similar error
types, with the most common medication adminis-
tration errors being rate deviations and unauthor-
ised medications, and the most common procedural
and documentation deviations concerning labelling
of medication and administration sets. However, our
overall error rate remains lower than in these studies,
probably due to our more nuanced distinction between
errors and discrepancies.

While several studies consider the potential harm
associated with errors and some distinguish between

medication administration errors and procedural
failures or policy violations,4 13 we are not aware of
previous studies that sought to understand the context
of the deviation by distinguishing errors and discrepan-
cies. Researchers and practitioners may have differing
views on what constitutes an error,17 with a range of
situations identified that clinicians may not consider
errors.18 19 These judgements are largely ignored in
definitions of errors adopted in most previous studies.
Separation of discrepancies and errors in our study
allowed us to better capture the complexities of current
intravenous practices, and may be more acceptable to
clinicians who feel that the realities of practice mean
that policies cannot always be adhered to.

Previous studies have highlighted the importance
of procedural failures and policy violations in identi-
fying system weaknesses that may create latent condi-
tions for patient harm.5 In this study, we recognise
that both medication administration and procedural/
documentation deviations occur on a spectrum from
minor discrepancies to serious errors with potential
for harm. While severe errors naturally attract greater
attention, and are often the focus for intervention, a

Figure 2 Variation in error and discrepancy rates between National Health Service (NHS) trusts.

o
n
A

u
g

u
st 1

, 2
0
2

2
b

y g
u

e
st. P

ro
te

cte
d

b
y co

p
yrig

h
t.

h
ttp

://q
u
a
litysa

fe
ty.b

m
j.co

m
/

B
M

J Q
u

a
l S

a
f: first p

u
b

lish
e

d
a

s 1
0

.1
1

3
6

/b
m

jq
s-2

0
1

7
-0

0
7

4
7
6
o

n
7

A
p
ril 2

0
1
8
. D

o
w

n
lo

a
d
e
d
fro

m

899Lyons I, et al. BMJ Qual Saf 2018;27:892–901. doi:10.1136/bmjqs-2017-007476

Original research

Safety II perspective encourages us to look at ‘normal’
discrepancies to identify potential system weaknesses.
According to Safety II, people make adjustments to
respond to the demands of the situation and compen-
sate for system weaknesses. We identified several
cases where these adjustments avoided or mitigated
potential harm. However, these same adaptive mech-
anisms can also lead to unsatisfactory outcomes, as
identified in one instance in this study. A challenge
for safety management is that everyday discrepancies
appear trivial but can contribute to rarer and more
serious incidents.20 Our approach to distinguishing
discrepancies and errors may help clinicians to reflect
on different kinds of deviations, consider which are
important and identify discrepancy patterns that may
be concerning.

Policy and practice gaps
Much of the variability among trusts related to gaps
between policy and practice. Better understanding of
the reasons behind such performance variability is neces-
sary to target interventions that improve safety. Proce-
dural and documentation deviations may not always
represent poor practice but rather a poor fit between
official policy and everyday practice due to situational
constraints. In some cases, policies that better reflect
existing practice may be more beneficial in managing
risk to both patients and staff than enforcing compli-
ance with existing policy. For example, policies allowing
administration of flushes without a medication order in
specific circumstances or for specific patient groups,
already in place in many trusts, could be introduced at
hospitals where unprescribed flushes are accepted local
practice by clinical staff but are technically unauthorised.
National standardisation may be helpful for whether or
not small volume flushes need to be prescribed and if
so how, labelling requirements for intravenous infusions
and giving sets, and requirements for double-checking.

Implications for practice: strategic interventions and
smart pumps
Appreciating the nuances of frequency, types and
severity of deviations occurring in different contexts
moves us beyond interventions focused on improving
compliance and eliminating error, towards more stra-
tegic interventions to proactively manage risk. Care is
rarely delivered in ideal circumstances and a more prag-
matic and practical approach, incorporating a wider
range of strategies, is needed.8 Strategic decisions to
live with certain deviations might be made if efforts to
resolve them are likely to distract from other aspects
of patient care, or not translate into gains for patient
safety. More work is needed to understand if and how
routine performance variability in intravenous infusions
can spiral into rare and unsatisfactory outcomes, what
conditions contribute to poor outcomes and which
interventions should be prioritised to prevent harm
rather than only reducing discrepancies.

Smart pumps are one possible intervention to improve
safety in intravenous infusion administration. Similar
to previous US studies,4 5 we found that smart pumps,
as currently implemented in English hospitals, do not
seem to reduce the risk of error in everyday practice.
Although smart pumps may have a role in preventing
more severe and rare errors, our relatively limited
observation periods did not identify these. In addition,
greater attention to the configuration and usability of
pumps is required: a third of smart pumps used in our
study offered no advantage over standard pumps due
to incomplete drug libraries. Using smart pumps as part
of an integrated system with bar code scanning and
interfacing with electronic systems could guard against
a broader range of deviations. Although the costs
and benefits of implementing such a system have not
yet been established,4 5 such approaches have become
standard practice in the USA as both were included in
the government’s Meaningful Use programme, which
provides financial incentives to promote the use of
health information technologies to improve quality.
Such configurations are rare in English hospitals; no
participating trusts used bar code administration, and
only a minority had trust-wide electronic prescribing
and medication administration records. The high error
rate associated with infusions delivered without a pump
in our study suggests that efforts to reduce reliance on
gravity feed, where it is difficult to control the delivery
rate, may be a more immediate and achievable priority
than the expansion of smart pump technology.

Strengths and limitations
This was a large multisite study, incorporating hospi-
tals with widely differing medication processes and
systems, reflecting the diversity of intravenous infusion
practices within the English NHS. Adopting a mixed
methods approach provided a rich understanding of
intravenous medication errors and the contexts in
which they occur. There are advantages and disadvan-
tages of using local observers versus observers from
a research team. Employing local data collectors may
have allowed less conspicuous observation and reduced
the likelihood of nurses modifying their behaviour on
observation days. However, using local staff may have
resulted in some interobserver variability or institu-
tional blindness to local poor practice. Variability was
minimised as much as possible by using two observers
from different professional backgrounds at each site
where possible, providing training, and subsequent
review of data by the multidisciplinary research team.
Resource limitations and confidentiality agreements
precluded measurement of interobserver reliability
across sites.

Other limitations are acknowledged. The timing
of data collection at each trust depended on local
approvals and staff availability; both daily and
seasonal variation in staffing levels and workload
may have affected deviation rates. We focused on

o
n
A

u
g

u
st 1

, 2
0
2

2
b

y g
u

e
st. P

ro
te

cte
d

b
y co

p
yrig

h
t.

h
ttp

://q
u
a
litysa

fe
ty.b

m
j.co

m
/

B
M

J Q
u

a
l S

a
f: first p

u
b

lish
e

d
a

s 1
0

.1
1

3
6

/b
m

jq
s-2

0
1

7
-0

0
7

4
7
6
o

n
7

A
p
ril 2

0
1
8
. D

o
w

n
lo

a
d
e
d
fro

m

900 Lyons I, et al. BMJ Qual Saf 2018;27:892–901. doi:10.1136/bmjqs-2017-007476

Original research

infusions running at the time of observation and will
therefore have underestimated the overall medication
administration error rate; observation of prescribing,
dispensing, preparation and setting up infusions is
likely to have revealed further errors.21 Errors already
identified and corrected by smart pumps or a double
check by another staff member prior to our obser-
vations would also not be captured using our meth-
odology. Ward managers were aware the study was
investigating medication administration errors and
discrepancies, so it is possible that nurses changed
their behaviour on observation days. However, obser-
vation dates were not publicised in advance and nurses
were not directly observed, thus the impact is likely
minimal. Finally, our study was not powered to test
for associations between pump and infusion types and
error rates; our findings instead highlight areas for
further investigation.

ConCluSIon
Overall, we identified errors in 1 in 10 infusions, but
very few were likely to result in patient harm. Smart
pumps, as currently implemented, seemed to have little
effect, with similar error rates observed in infusions
delivered with and without a smart pump. Measuring
the prevalence, types and severity of errors and discrep-
ancies can provide valuable insights for reflection.
However, this needs to be coupled to causal accounts
and contextual understanding of local hospital poli-
cies, cultures, customs and practices. Not all deviations
from medication order or policy are bad; many arise
as nurses actively manage safety and productivity pres-
sures. This study suggests there is a need to shift the
focus away from the goal of eliminating deviations to
enable strategic intervention to manage infusion risk
in the context of everyday performance variability
and working conditions. Future work should explore
where efforts should be targeted to prevent harm
rather than only reducing discrepancies.

Author affiliations
1UCL Interaction Centre, University College London, London, UK
2Pain Management Centre, Imperial College Healthcare NHS Trust, London, UK
3Institute of Educational Technology, Open University, Milton Keynes, UK
4Research Department of Practice and Policy, UCL School of Pharmacy, London,
UK
5UCL Medical School, University College London, London, UK
6Royal Free London NHS Foundation Trust, London, UK
7Faculty of Computers and Information, Cairo University, Cairo, Egypt
8Brigham and Women’s Hospital, Boston, Massachusetts, USA
9Harvard Medical School, Boston, Massachusetts, USA
10Department of Medicine, Brigham and Women’s Hospital, Boston,
Massachusetts, USA
11Centre for Medication Safety and Service Quality, Imperial College Healthcare
NHS Trust, London, UK

Contributors As per previous submission.

Funding This work is supported by the National Institute
for Health Research (NIHR) grant [12/209/27], from the
Health Services and Delivery Research (HS&DR) stream. The
research is also supported by the NIHR Imperial Patient Safety
Translational Research Centre. The views expressed are those

of the authors and not necessarily those of the NHS, the NIHR
or the Department of Health.

Competing interests None declared.

Patient consent Not required.

Ethics approval NHS Research Ethics Committee (14/SC/0290)

Provenance and peer review Not commissioned; externally
peer reviewed.

Open access This is an open access article distributed
in accordance with the terms of the Creative Commons
Attribution (CC BY 4.0) license, which permits others to
distribute, remix, adapt and build upon this work, for
commercial use, provided the original work is properly cited.
See: http:// creativecommons. org/ licenses/ by/ 4. 0/

© Article author(s) (or their employer(s) unless otherwise
stated in the text of the article) 2018. All rights reserved.
No commercial use is permitted unless otherwise expressly
granted.

ReFeRenCeS
1 McDowell SE, Mt-Isa S, Ashby D, et al. Where errors occur in

the preparation and administration of intravenous medicines:
a systematic review and Bayesian analysis. Qual Saf Healthcare
2010;19:341–5.

2 Franklin BD, Taxis K, Barber N. Parenteral drug errors.
Reported error rates are likely to be underestimation. BMJ
2009;338:b1814.

3 Ohashi K, Dalleur O, Dykes PC, et al. Benefits and risks
of using smart pumps to reduce medication error rates: a
systematic review. Drug Saf 2014;37:1011–20.

4 Schnock KO, Dykes PC, Albert J, et al. The frequency of
intravenous medication administration errors related to smart
infusion pumps: a multihospital observational study. BMJ Qual
Saf 2017;26:131–40.

5 Husch M, Sullivan C, Rooney D, et al. Insights from the
sharp end of intravenous medication errors: implications
for infusion pump technology. Qual Saf Health Care
2005;14:80–6.

6 Blandford A, Furniss D, Lyons I, et al. Exploring the Current
Landscape of Intravenous Infusion Practices and Errors
(ECLIPSE): protocol for a mixed-methods observational study.
BMJ Open 2016;6:e009777.

7 Hollnagel E, Wears RL, Braithwaite J. From Safety-I to
Safety-II: a white paper. The resilient health care net. Published
simultaneously by the University of Southern Denmark, USA:
University of Florida and Australia: Macquarie University,
2015.

8 Vincent C, Amalberti R. Safer healthcare. Cham: Springer
International Publishing, 2016.

9 Harris PA, Taylor R, Thielke R, et al. Research electronic
data capture (REDCap)–a metadata-driven methodology
and workflow process for providing translational research
informatics support. J Biomed Inform 2009;42:377–81.

10 Hartwig SC, Denger SD, Schneider PJ. Severity-indexed,
incident report-based medication error-reporting program. Am
J Hosp Pharm 1991;48:2611–6.

11 Fahmy S, Garfield S, Furniss D, et al. A comparison of two
methods of assessing the potential clinical importance of
medication errors. Safety in Health 2018;4.

12 Hollnagel E. FRAM, the functional resonance analysis method:
modelling complex socio-technical systems. Farnham: Ashgate
Publishing Ltd, 2012.

13 Westbrook JI, Rob MI, Woods A, et al. Errors in the
administration of intravenous medications in hospital and the

o
n
A

u
g

u
st 1

, 2
0
2

2
b

y g
u

e
st. P

ro
te

cte
d

b
y co

p
yrig

h
t.

h
ttp

://q
u
a
litysa

fe
ty.b

m
j.co

m
/

B
M

J Q
u

a
l S

a
f: first p

u
b

lish
e

d
a

s 1
0

.1
1

3
6

/b
m

jq
s-2

0
1

7
-0

0
7

4
7
6
o

n
7

A
p
ril 2

0
1
8
. D

o
w

n
lo

a
d
e
d
fro

m

901Lyons I, et al. BMJ Qual Saf 2018;27:892–901. doi:10.1136/bmjqs-2017-007476

Original research

role of correct procedures and nurse experience. BMJ Qual Saf
2011;20:1027–34.

14 Keers RN, Williams SD, Cooke J, et al. Prevalence and nature
of medication administration errors in health care settings:
a systematic review of direct observational evidence. Ann
Pharmacother 2013;47:237–56.

15 McLeod MC, Barber N, Franklin BD. Methodological
variations and their effects on reported medication
administration error rates. BMJ Qual Saf 2013;22:278–89.

16 Lisby M, Nielsen LP, Brock B, et al. How are medication errors
defined? A systematic literature review of definitions and
characteristics. Int J Qual Health Care 2010;22:507–18.

17 Franklin BD. ‘Smart’ intravenous pumps: how smart are they?
BMJ Qual Saf 2017;26:93–4.

18 Baker HM. Rules outside the rules for administration of
medication: a study in New South Wales, Australia. Image J
Nurs Sch 1997;29:155–8.

19 Tamuz M, Thomas EJ, Franchois KE. Defining
and classifying medical error: lessons for patient safety
reporting systems. Qual Saf Health Care
2004;13:13–20.

20 Macrae C. Close calls: managing risk and resilience in airline
flight safety. UK: Palgrave Macmillan UK, 2014.

21 Taxis K, Barber N. Ethnographic study of incidence and
severity of intravenous drug errors. BMJ 2003;326:684.

o
n
A

u
g

u
st 1

, 2
0
2

2
b

y g
u

e
st. P

ro
te

cte
d

b
y co

p
yrig

h
t.

h
ttp

://q
u
a
litysa

fe
ty.b

m
j.co

m
/

B
M

J Q
u

a
l S

a
f: first p

u
b

lish
e

d
a

s 1
0

.1
1

3
6

/b
m

jq
s-2

0
1

7
-0

0
7

4
7
6
o

n
7

A
p
ril 2

0
1
8
. D

o
w

n
lo

a
d
e
d
fro

m

Journal of

Clinical Medicine

Article

Determinants of In-Hospital Mortality in Elderly Patients Aged
80 Years or above with Acute Heart Failure: A Retrospective
Cohort Study at a Single Rural Hospital

Yusuke Watanabe 1,*, Kazuko Tajiri 2 , Hiroyuki Nagata 1 and Masayuki Kojima 3

����������
�������

Citation: Watanabe, Y.; Tajiri, K.;

Nagata, H.; Kojima, M. Determinants

of In-Hospital Mortality in Elderly

Patients Aged 80 Years or above with

Acute Heart Failure: A Retrospective

Cohort Study at a Single Rural

Hospital. J. Clin. Med. 2021, 10, 1468.

https://doi.org/10.3390/jcm10071468

Academic Editor: Nuria Farre

Received: 11 February 2021

Accepted: 23 March 2021

Published: 2 April 2021

Publisher’s Note: MDPI stays neutral

with regard to jurisdictional claims in

published maps and institutional affil-

iations.

Copyright: © 2021 by the authors.

Licensee MDPI, Basel, Switzerland.

This article is an open access article

distributed under the terms and

conditions of the Creative Commons

Attribution (CC BY) license (https://

creativecommons.org/licenses/by/

4.0/).

1 Department of Internal Medicine, Hitachiomiya Saiseikai Hospital, 3033-3 Tagouchichou, Hitachiomiya,
Ibaraki 319-2601, Japan; [email protected]

2 Department of Cardiology, Faculty of Medicine, University of Tsukuba, Tsukuba 305-8577, Japan;
[email protected]

3 Department of Surgery, Hitachiomiya Saiseikai Hospital, Hitachiomiya 319-2256, Japan;
[email protected]

* Correspondence: [email protected]; Tel.: +81-295-52-5151; Fax: +81-295-52-5725

Abstract: Heart failure is one of the leading causes of mortality worldwide. Several predictive risk
scores and factors associated with in-hospital mortality have been reported for acute heart failure.
However, only a few studies have examined the predictors in elderly patients. This study investigated
determinants of in-hospital mortality in elderly patients with acute heart failure, aged 80 years or
above, by evaluating the serum sodium, blood urea nitrogen, age and serum albumin, systolic blood
pressure and natriuretic peptide levels (SOB-ASAP) score. We reviewed the medical records of
106 consecutive patients retrospectively and classified them into the survivor group (n = 83) and the
non-survivor group (n = 23) based on the in-hospital mortality. Patient characteristics at admission
and during hospitalization were compared between the two groups. Multivariate stepwise regression
analysis was used to evaluate the in-hospital mortality. The SOB-ASAP score was significantly better
in the survivor group than in the non-survivor group. Multivariate stepwise regression analysis
revealed that a poor SOB-ASAP score, oral phosphodiesterase 3 inhibitor use, and requirement of
early intravenous antibiotic administration were associated with in-hospital mortality in very elderly
patients with acute heart failure. Severe clinical status might predict outcomes in very elderly patients.

Keywords: acute heart failure; elderly patients; SOB-ASAP score; phosphodiesterase 3 inhibitor; antibiotics

1. Introduction

Heart failure is one of the most common diseases worldwide and generally affects the
elderly [1–3]. Elderly patients with heart failure are likely to have many comorbidities [2].
The prevalence of heart failure is on the rise due to the rapidly aging population [3,4]; in
2019, the estimated number of individuals aged 65 years or above in Japan was approx-
imately 36 million (28.4% of the total population). Heart failure is the leading cause of
mortality in Japan. Although several novel medications have been developed [5], thera-
peutic strategies for improving patients’ prognoses are yet to be identified.

Previous studies have suggested that risk score systems are useful for predicting
prognosis in inpatients and outpatients with heart failure [6–8]. Other studies have reported
factors leading to in-hospital mortality, such as acute kidney injury, new-onset atrial
fibrillation, and nutritional index [9–11]. However, most of these studies have focused
on relatively younger populations than the Japanese elderly population, and very few
studies have focused on the very elderly population [12,13]. The population of hospitalized
patients with acute heart failure is aging, even in rural areas. Recently, a novel scoring
system, the serum sodium, blood urea nitrogen, age and serum albumin, systolic blood
pressure and natriuretic peptide level (SOB-ASAP) score, was developed in Japanese

J. Clin. Med. 2021, 10, 1468. https://doi.org/10.3390/jcm10071468 https://www.mdpi.com/journal/jcm

J. Clin. Med. 2021, 10, 1468 2 of 10

registries [14]. The SOB-ASAP score ranges from 0 to 14; the highest score indicates a high
in-hospital mortality rate [14].

This study aimed to reveal other determinants of in-hospital mortality in very el-
derly patients (aged 80 years or older) with acute heart failure by evaluating their SOB-
ASAP score.

2. Materials and Methods

This single-hospital retrospective observational study was conducted at the Hita-
chiomiya Saiseikai Hospital, Japan. We reviewed the medical records of all consecutive
patients with heart failure admitted to our hospital between January 2017 and December
2019. The inclusion criteria were as follows: (1) hospitalized patients aged ≥80 years and
diagnosed with heart failure, (2) clinical status corresponding to heart failure, according to
the Framingham criteria [15], and (3) left ventricular function assessed with echocardiog-
raphy, at least. The exclusion criteria were as follows: (1) brain natriuretic peptide (BNP)
level <100 pg/mL or unknown, (2) readmission of the same patient with acute heart failure,
(3) no diagnosis of acute heart failure, and (4) requiring transfer to a tertiary hospital. We
divided the participants into two groups based on the prevalence of in-hospital mortality:
the survivor group and the non-survivor group.

This study was approved by our institutional review board (ID 20-06) and was con-
ducted in accordance with the Declaration of Helsinki for experiments involving humans.
The requirement for written informed consent was waived by our institutional review
board due to the retrospective nature of the study.

In echocardiography, data acquisition was performed by an expert. The variables mea-
sured and derived using echocardiography were determined as follows: two-dimensional
left ventricular ejection fraction (LVEF) was computed from the calculated left ventricular
end-diastolic and end-systolic volumes. Valvular heart disease was defined as moderate to
severe valvular disease (according to current guidelines [16–18]) or a history of valvular
surgery or cardiac surgery. Wall motion abnormality was defined as localized abnormal
wall motion, such as akinesis, hypokinesis, and dyskinesis.

Hypertension was defined as the use of medication for hypertension and/or a
history of hypertension before admission. Dyslipidemia was defined as a triglyceride
level ≥150 mg/dL, low-density lipoprotein cholesterol level ≥140 mg/dL, high-density
lipoprotein cholesterol level ≤40 mg/dL, the use of medication for dyslipidemia, or a
history of dyslipidemia. Diabetes mellitus was defined as a hemoglobin A1c level ≥6.5%
(National Glycohemoglobin Standardization Program value), the use of medication for
diabetes mellitus, or a history of diabetes mellitus. Chronic obstructive pulmonary
disease (COPD) was defined as the use of medical treatment for COPD and/or a history
of COPD before admission. We calculated the estimated glomerular filtration rate (eGFR)
from the serum creatinine levels, age, weight, and sex using the following formula:
eGFR (mL/min/1.73 m2) = 194 × s-Cr (−1.094) × age (−0.287) × 0.739 (if female) [19].
Worsening renal function was defined as an increase in the serum creatinine level
to >0.3 mg/dL [20]. The SOB-ASAP score was calculated according to the previously
published formula [14].

Statistical Analyses

All statistical analyses were performed using SPSS 26.0 for Windows (SPSS, Chicago,
IL, USA).

Continuous data are expressed as mean ± standard deviation (SD). Normality was
tested using the Shapiro–Wilk test. Normally distributed continuous variables were com-
pared between the two groups using the unpaired Student’s t-test. Continuous variables
were compared using the Mann–Whitney U-test. Categorical variables were expressed as
numbers and percentages and were compared using the Pearson’s χ2 test or the Fisher’s
exact test. Multivariate stepwise regression analysis was used to evaluate the in-hospital
mortality in elderly patients (aged ≥ 80 years) with acute heart failure; the included vari-

J. Clin. Med. 2021, 10, 1468 3 of 10

ables were found to be significant (p < 0.1) using a univariate logistic regression analysis.
We analyzed the relationship between LVEF and the in-hospital mortality by constructing a
receiver operating characteristics (ROC) curve and calculating the area under the curve. The
area under the curve was 0.66 (95% confidence interval (CI): 0.53–0.79, p = 0.018). The sensi-
tivity and specificity for an LVEF of 49.6% were 60.2% and 39.1%, respectively. Therefore, an
LVEF ≥ 50% was analyzed in the univariate analysis. Meanwhile, the variables included in
the SOB-ASAP scoring system, such as the systolic blood pressure and BNP, were excluded
from the multivariate model. A p-value < 0.05 was considered statistically significant.

3. Results

A total of 106 patients (36.8% men, mean age: 89.8 ± 4.5 years) were included in the
study (survivor group: n = 83; non-survivor group: n = 23) (Figure 1).

Figure 1. Study flow chart. BNP: brain natriuretic peptide.

Table 1 shows the comparison of the baseline characteristics and medication usage at
admission between the two groups. The systolic blood pressure was significantly higher in
the survivor group than in the non-survivor group (139 ± 33 mmHg vs. 119 ± 20 mmHg,
p = 0.011). A systolic blood pressure of ≤100 mmHg, indicating an unstable hemody-
namic status, was more prevalent in the non-survivor group than in the survivor group
(26.1% vs. 8.4%, p = 0.022). The SOB-ASAP score was significantly better in the survivor
group than in the non-survivor group (4.3 ± 2.3 vs. 6.8 ± 2.7, p < 0.001). Mineralocorti-
coid receptor antagonist usage was significantly lower in the survivor group than in the
non-survivor group (10.8% vs. 30.4%, p = 0.028).

J. Clin. Med. 2021, 10, 1468 4 of 10

Table 1. Comparison of the baseline characteristics and medication usage at admission between the survivor and non-
survivor groups.

Survivor Group
(n = 83)

Non-Survivor Group
(n = 23)

p-Value

Age, years 89.5 ± 4.6 91.0 ± 4.2 0.17
Male sex, n (%) 31 (37.3) 8 (34.8) 0.82

Height, cm 147 ± 10 149 ± 9 0.34
Body weight, kg 49.8 ± 10.6 46.4 ± 11.0 0.087

Systolic blood pressure, mmHg (n, %) 139 ± 33 119 ± 20 0.011
Diastolic blood pressure, mmHg (n, %) 78.5 ± 23.1 72.0 ± 16.0 0.35

Heart rate, beats/minute (n, %) 89 ± 26 93 ± 31 0.41
Respiratory rate, breaths/minute (n, %) 21 ± 6 19 ± 4 0.13

Systolic blood pressure ≤ 100 mmHg, n (%) 7 (8.4) 6 (26.1) 0.022
Hypertension, n (%) 76 (91.6) 21 (91.3) 0.62
Dyslipidemia, n (%) 21 (25.3) 5 (21.7) 0.48

Diabetes mellitus, n (%) 25 (30.1) 6 (26.1) 0.71
Atrial fibrillation/atrial flutter, n (%) 46 (55.4) 16 (69.6) 0.22

Pacemaker implantation, n (%) 13 (15.7) 1 (4.3) 0.14
Chronic obstructive pulmonary disease, n (%) 7 (8.4) 2 (8.7) 0.62

eGFR < 60 mL/min/1.73 m2, n (%) 61 (73.5) 16 (69.6) 0.71
Ambulance transport to emergency department,

n (%)
16 (19.3) 6 (26.1) 0.33

NYHA functional classification at admission 0.32
3, n (%) 27 (32.5) 5 (21.7)
4, n (%) 56 (67.5) 18 (78.3)

SOB-ASAP score, (n, %) 4.3 ± 2.3 6.8 ± 2.7 <0.001
Medication usage at admission

ACE-I and/or ARB, n (%) 47 (56.6) 9 (39.1) 0.14
β blockers, n (%) 27 (32.5) 7 (30.4) 0.85

Calcium channel blockers, n (%) 35 (42.2) 7 (30.4) 0.31
Loop diuretics, n (%) 47 (56.6) 17 (73.9) 0.13

Mineralocorticoid receptor antagonists, n (%) 9 (10.8) 7 (30.4) 0.028
Thiazides, n (%) 2 (2.4) 2 (8.7) 0.21
Tolvaptan, n (%) 6 (7.2) 3 (13.0) 0.30
Digitalis, n (%) 4 (4.8) 1 (4.3) 0.70

PDE3-inhibitor, n (%) 1 (1.2) 3 (13.0) 0.031
Statins, n (%) 12 (14.5) 3 (13.0) 0.58

Oral anti-diabetes mellitus agents, n (%) 12 (14.5) 2 (8.7) 0.37
Anti-platelets, n (%) 18 (21.7) 8 (34.8) 0.20

Anti-coagulants, n (%) 27 (32.5) 5 (21.7) 0.32

Categorical variables are expressed as numbers and percentages. Continuous variables are expressed as mean ± SD. ACE-I: angiotensin-
converting enzyme inhibitor; ARB: angiotensin II receptor blocker; eGFR: estimated glomerular filtration rate; GWTG-HF: Get with the
Guideline-Heart Failure; NYHA: New York Heart Association, PDE3: phosphodiesterase 3; SOB-ASAP: SO, serum sodium; B, blood urea
nitrogen; A, age and serum albumin; S, systolic blood pressure; and P, natriuretic peptide levels.

Table 2 shows the differences in the baseline laboratory data and echocardiographic
parameters between the two groups. The serum sodium levels were significantly higher in the
survivor group than in the non-survivor group (139 ± 5 mEq/L vs. 136 ± 5 mEq/L, p = 0.025).
The C-reactive protein level was significantly lower in the survivor group than in the non-
survivor group (1.5 ± 2.3 mg/dL vs. 5.1 ± 7.2 mg/dL, p = 0.0073). The BNP level was signifi-
cantly lower in the survivor group than in the non-survivor group (646.9 ± 586.9 pg/mL vs.
1170.7 ± 1018.8 pg/mL, p = 0.0033). The LVEF value was significantly better in the survivor
group than in the non-survivor group (52.8 ± 17.5% vs. 42.1 ± 19.9%, p = 0.018).

J. Clin. Med. 2021, 10, 1468 5 of 10

Table 2. The comparison of baseline laboratory data and echocardiographic parameters between the survivor and non-
survivor groups.

Survivor Group
(n = 83)

Non-Survivor Group
(n = 23)

p-Value

Laboratory data

Total protein, g/dL (n, %) 6.6 ± 0.7 (80, 96.4) 6.4 ± 0.8 (22, 95.7) 0.22
Serum albumin, g/dL (n, %) 3.4 ± 0.5 3.1 ± 0.7 0.10
Total bilirubin, g/dL (n, %) 0.79 ± 0.51 (82, 98.8) 0.76 ± 0.40 (22, 95.7) 0.10

Aspartate aminotransferase, U/L (n, %) 37 ± 32 83 ± 152 0.70
Alanine aminotransferase, U/L (n, %) 23 ± 17 51 ± 93 0.99

Serum sodium, mEq/L (n, %) 139 ± 5 136 ± 5 0.025
Serum potassium, mEq/L (n, %) 4.2 ± 0.7 4.4 ± 1.0 0.17

Blood glucose, mg/dL (n, %) 139 ± 44 (81, 90.4) 151 ± 55 (21, 91.3) 0.12
Blood urea nitrogen, mg/dL (n, %) 26.5 ± 16.5 29.1 ± 16.6 0.29

Serum creatinine, mg/dL (n, %) 1.24 ± 0.76 1.25 ± 0.60 0.53
Estimated glomerular filtration rate,

mL/min/1.73 m2 (n, %)
47.2 ± 23.3 43.4 ± 20.5 0.59

C-reactive protein, mg/dL (n, %) 1.5 ± 2.3 5.1 ± 7.2 0.0073
Hemoglobin, g/dL (n, %) 10.9 ± 2.0 11.0 ± 1.9 0.81

Brain natriuretic peptide, pg/mL (n, %) 646.9 ± 586.9 1170.7 ± 1018.8 0.0033
Echocardiography results

Left ventricular ejection fraction, (%) 52.9 ± 17.5 42.1 ± 19.9 0.018
Interventricular septal thickness, mm 9.7 ± 1.7 (82, 98.8) 9.6 ± 2.9 (22, 95.7) 0.36

Left ventricular end-diastolic diameter, mm 44.9 ± 8.8 45.6 ± 11.6 0.95
Left ventricular end-systolic diameter, mm 32.4 ± 9.4 36.3 ± 12.6 0.30

Posterior left ventricular wall thickness, mm 9.8 ± 1.9 (88, 98.8) 10.0 ± 2.2 (22, 95.7) 0.71
Left ventricular end-diastolic volume, mL 95.9 ± 44.3 103.7 ± 60.9 0.98
Left ventricular end-systolic volume, mL 47.3 ± 32.4 65.6 ± 51.2 0.28

Left atrial diameter, mm 43.1 ± 9.4 (82, 98.8) 41.8 ± 8.2 0.56
Aortic diameter, mm 29.7 ± 4.2 (80, 96.4) 31.8 ± 5.2 0.078

Valvular heart disease, n (%) 74 (89.2) 22 (95.7) 0.31
Wall motion abnormality, n (%) 9 (10.8) 4 (17.4) 0.30

Categorical variables are expressed as numbers and percentages. Continuous variables are expressed as mean ± standard deviation (SD).

Table 3 shows the comparison of treatment strategies and subsequent outcomes during
hospitalization between the two groups. The prevalence of intravenous administration of
diuretics within 48 h of hospitalization was significantly higher in the survivor group than
in the non-survivor group (80.7% vs. 56.5%, p = 0.017). The prevalence of intravenous cate-
cholamine support requirement and intravenous administration of antibiotics within 48 h
of hospitalization was significantly lower in the survivor group than in the non-survivor
group (1.2% vs. 13.0%, p = 0.031; 14.5% vs. 34.8%, p = 0.033, respectively). Intravenous
catecholamine support requirement, intravenous administration of antibiotics, and non-
invasive positive pressure ventilation support during the entire period of hospitalization
were more prevalent in the non-survivor group than in the survivor group (3.4% vs. 30.4%,
p < 0.001; 18.0% vs. 52.2%, p < 0.001; and 5.6% vs. 21.7%, p = 0.029, respectively). Worsening
of renal function was less frequent in the survivor group than in the non-survivor group
(21.3% vs. 60.9%, p < 0.001).

J. Clin. Med. 2021, 10, 1468 6 of 10

Table 3. Comparison of treatment and outcomes during hospitalization between the survivor and non-survivor groups.

Survivor Group
(n = 83)

Non-Survivor Group
(n = 23)

p-Value

Details of treatment within 48 h of hospitalization

Intravenous diuretic administration, n (%) 67 (80.7) 13 (56.5) 0.017
Intravenous carperitide administration, n (%) 1 (1.2) 0 (0.0) 0.78

Tolvaptan introduction, n (%) 5 (6.0) 0 (0.0) 0.29
Intravenous nitric acid administration, n (%) 11 (13.0) 1 (4.3) 0.21

Digoxin administration, n (%) 3 (3.6) 2 (8.7) 0.30
Intravenous catecholamine support requirement, n (%) 1 (1.2) 3 (13.0) 0.031

Oral PDE3-inhibitor and/or catecholamine addition n (%) 0 (0.0) 1 (4.3) 0.22
Intravenous antibiotic administration, n (%) 12 (14.5) 8 (34.8) 0.033

NPPV support requirement, n (%) 4 (4.8) 3 (13.0) 0.17
Morphine use, n (%) 2 (2.4) 0 (0.0) 0.61

Details of treatment and results during the entire period of hospitalization

Intravenous diuretic administration, n (%) 74 (83.1) 19 (82.6) 0.58
Intravenous carperitide administration, n (%) 1 (1.1) 0 (0.0) 0.80

Tolvaptan introduction, n (%) 11 (12.4) 6 (26.1) 0.099
Intravenous nitric acid administration, n (%) 13 (14.6) 1 (4.3) 0.17

Digoxin administration, n (%) 5 (5.6) 2 (8.7) 0.44
Intravenous catecholamine support requirement, n (%) 3 (3.4) 7 (30.4) <0.001

Oral PDE3-inhibitor and/or catecholamine
administration, n (%)

3 (3.4) 2 (8.7) 0.27

Intravenous antibiotic administration, n (%) 16 (18.0) 12 (52.2) <0.001
NPPV support requirement, n (%) 5 (5.6) 5 (21.7) 0.029

Morphine use, n (%) 1 (1.1) 2 (8.7) 0.11
Maximum serum creatinine during hospitalization,

mg/dL (n, %)
1.47 ± 0.96 2.03 ± 1.10 0.0089

Worsening renal function, n (%) 19 (21.3) 14 (60.9) <0.001

Categorical variables are expressed as numbers and percentages. Continuous variables are expressed as mean ± SD. NPPV: noninvasive
positive pressure ventilation; PDE3: phosphodiesterase 3.

Table 4 shows the results of the univariate logistic regression analysis and the multi-
variate stepwise regression analysis for predicting in-hospital mortality. Only variables
that had significant differences (p-value < 0.1), except for age and male sex, are shown
in the results of the univariate logistic regression analysis (Supplemental Table S1). The
multivariate stepwise regression analysis model showed that a poor SOB-ASAP score (per
point increase; odds ratio (OR): 1.449, 95% CI: 1.159–1.812, p = 0.0010), oral phosphodi-
esterase 3 inhibitor usage at admission (OR: 14.276, 95% CI: 1.119–182.170, p = 0.041), and
intravenous antibiotic administration within 48 h of hospitalization (OR: 3.887, 95% CI:
1.142–13.224, p = 0.030) were significantly associated with in-hospital mortality.

Table 4. Results of the univariate logistic regression analysis and the multivariate stepwise regression analysis for predicting
in-hospital mortality.

Univariate Analysis OR 95% CI p-Value

Age (per year increase) 1.077 0.969–1.197 0.17
Male sex 0.897 0.340–2.352 0.82

Systolic blood pressure (per mmHg increase) 0.975 0.956–0.993 0.0086
SOB-ASAP score (per point increase) 1.455 1.194–1.774 <0.001

Mineralocorticoid receptor antagonist use at admission 3.597 1.167–11.090 0.026
Phosphodiesterase 3 inhibitor use at admission 12.300 1.214–124.581 0.034

Serum albumin (per g/dL increase) 0.462 0.204–1.044 0.063
Aspartate aminotransferase (per U/L increase) 1.007 0.999–1.015 0.079
Alanine aminotransferase (per U/L increase) 1.012 0.998–1.025 0.083

Serum sodium (per mEq/L increase) 0.917 0.838–1.004 0.060

J. Clin. Med. 2021, 10, 1468 7 of 10

Table 4. Cont.

Univariate Analysis OR 95% CI p-Value

Serum potassium (per mEq/L increase) 1.716 0.926–3.182 0.086
C-reactive protein (per mg/dL increase) 1.267 1.087–1.486 0.0036

Brain natriuretic peptide (per pg/mL increase) 1.001 1.000–1.001 0.0080
Left ventricular end-systolic volume (per mL increase) 1.012 1.000–1.023 0.046

Left ventricular ejection fraction ≥50% 0.446 0.173–1.147 0.094
Aortic diameter (per mm increase) 1.101 0.997–1.216 0.058

Intravenous diuretic administration within 48 h of hospitalization 0.310 0.116–0.834 0.020
Catecholamine support requirement within 48 h of hospitalization 12.300 1.214–124.581 0.034
Intravenous antibiotic administration within 48 h of hospitalization 3.156 1.100–9.052 0.033

Multivariate Analysis OR 95% CI p-Value

SOB-ASAP score (per point increase) 1.449 1.159–1.812 0.0010
Phosphodiesterase 3 inhibitor use at admission 14.276 1.119–182.170 0.041

Intravenous antibiotic administration within 48 h of hospitalization 3.887 1.142–13.224 0.030

CI: confidence interval; GWTG-HF: Get with the Guideline-Heart Failure; NPPV: noninvasive positive pressure ventilation; NYHA: New
York Heart Association; OR: odds ratio; SOB-ASAP: SO, serum sodium; B, blood urea nitrogen; A, age and serum albumin; S, systolic blood
pressure; and P, natriuretic peptide levels.

4. Discussion

The three main findings of the present study are as follows: (1) the SOB-ASAP score
can predict in-hospital mortality even in very elderly patients with acute heart failure,
and (2) in addition to the SOB-ASAP score, use of oral phosphodiesterase 3 inhibitors
at admission, and requirement of intravenous antibiotic administration within 48 h of
hospitalization were important factors for predicting in-hospital mortality secondary to
acute heart failure.

The SOB-ASAP score, which can predict the clinical outcomes of patients with acute
heart failure, was found to be useful and practical. Several previous studies have performed
risk assessments for patients with heart failure, including assessments with the Get with
the Guideline-Heart Failure (GWTG-HF) risk score that was adopted globally to anticipate
the outcomes of acute heart failure [21]. Although the SOB-ASAP score was validated in
accordance with previous risk scores such as the GWTG-HF risk score, the SOB-ASAP score
includes novel serum parameters such as BNP and N-terminal pro-BNP (NT-pro BNP),
which were not considered in the previous studies [14]. Therefore, the SOB-ASAP score may
predict the outcomes of hospitalized patients with acute heart failure. Furthermore, one of
the greatest advantages of the SOB-ASAP score is that it predicts the clinical outcomes of
patients with acute heart failure during the very acute phase of hospitalization [14]. The
present study suggests the usefulness of a novel risk scoring system even in very elderly
patients with acute heart failure, which could be valuable for physicians treating patients
with heart failure.

The present study also underscores the clinical impact of treatment with oral phospho-
diesterase 3 inhibitors at admission. Phosphodiesterase 3 inhibitors are often required in
patients with heart failure having a severe clinical status [22]. One study showed that despite
their hemodynamically beneficial effects, long-term therapy with oral phosphodiesterase
3 inhibitors could increase the morbidity and mortality of patients with severe chronic heart
failure [23]. In other words, patients treated with oral phosphodiesterase 3 inhibitors are in
a worse clinical situation. Therefore, in-hospital mortality could be frequently observed in
patients with acute heart failure requiring oral phosphodiesterase 3 inhibitors.

The association between intravenous antibiotic administration within 48 h of hospi-
talization and worse clinical outcomes should be discussed further. The present study
showed that patients who required intravenous antibiotic administration in the early phase
of hospitalization were often suspected of being affected with pneumonia, because pa-
tients with worse clinical outcomes had higher serum C-reactive protein levels as well as
symptoms similar to that of pneumonia. Furthermore, a previous study showed that the

J. Clin. Med. 2021, 10, 1468 8 of 10

coexistence of comorbidities, such as pneumonia, could increase the risk of mortality in the
elderly [24,25]. The requirement of intravenous antibiotic administration in the early phase
of hospitalization could lead to acute infections such as pneumonia, and could therefore
affect the clinical outcomes of elderly patients with heart failure.

Our study has several limitations. First, because this was a single-center retrospective
observational study, there is a risk of selection bias. Second, although both left and
right cardiac function may affect the clinical prognosis [26,27], the present study did not
comprehensively assess the cardiac function. Moreover, the present study did not show the
etiology of heart failure. Furthermore, the prognosis of patients with heart failure having a
preserved ejection fraction (HFpEF) was as bad as that of patients with heart failure having
a reduced ejection fraction (HFrEF). Additionally, the prevalence of HFpEF was high in
the elderly patients with heart failure [28,29]. Therefore, the relationship between LVEF
and prognosis should be carefully interpreted. Third, the frailty and nutritional status
of patients could affect their prognosis [30,31]; however, we could not obtain sufficient
information on the baseline frailty and nutritional status due to the severity of acute heart
failure and its emergent clinical setting. Fourth, the severity of acute heart failure itself
might affect the selection of treatment. For example, in some cases, physicians may hesitate
to administer intravenous diuretics due to unstable hemodynamics. Finally, the present
study was observational in nature and had a relatively small study population; therefore,
it can be considered as a pilot study whose results need to be confirmed prospectively in
further extensive multicenter studies.

In conclusion, a poor SOB-ASAP score, oral phosphodiesterase 3 inhibitor use at ad-
mission, and requirement of early intravenous antibiotic administration were significantly
associated with in-hospital mortality in elderly patients (aged ≥ 80 years) with acute heart
failure. Recognizing patients with severe disease and high SOB-ASAP scores, which is a
novel risk scoring system, could help physicians to treat patients with heart failure.

Supplementary Materials: The following are available online at https://www.mdpi.com/article/10
.3390/jcm10071468/s1, Table S1: Univariate logistic regression analysis to predict in-hospital mortality.

Author Contributions: Conceptualization, Methodology, Original draft preparation, Data curation,
and writing: Y.W. Editing, Supervision, Reviewing Writing, and Investigation: K.T. Supervision and
Editing: H.N. Supervision and Reviewing: M.K. All authors have read and agreed to the published
version of the manuscript.

Funding: This research received no external funding.

Institutional Review Board Statement: The study was conducted according to the guidelines of the
Declaration of Helsinki and was approved by the Institutional Review Board of the Hitachiomiya
Hospital (ID 20-06).

Informed Consent Statement: Patient consent was waived due to the retrospective nature of
the study.

Data Availability Statement: The datasets generated and/or analyzed during the current study are
not publicly available because the study dataset contains potentially identifying clinical information,
but are available from the corresponding author upon reasonable request.

Acknowledgments: We are grateful to the staff of the clinical laboratory department at the Hita-
chiomiya Saiseikai Hospital for their support.

Conflicts of Interest: The authors declare no conflict of interest. The funders had no role in the design
of the study; in the collection, analyses, or interpretation of data; in the writing of the manuscript, or
in the decision to publish the results.

J. Clin. Med. 2021, 10, 1468 9 of 10

References
1. Virani, S.S.; Alonso, A.; Benjamin, E.J.; Bittencourt, M.S.; Callaway, C.W.; Carson, A.P.; Chamberlain, A.M.; Chang, A.R.; Cheng,

S.; Delling, F.N.; et al. Heart disease and stroke statistics-2020 update: A report from the American Heart Association. Circulation
2020, 141, e139–e596. [CrossRef] [PubMed]

2. Ambrosy, A.P.; Fonarow, G.C.; Butler, J.; Chioncel, O.; Greene, S.J.; Vaduganathan, M.; Nodari, S.; Lam, C.S.P.; Sato, N.;
Shah, A.N.; et al. The global health and economic burden of hospitalizations for heart failure: Lessons learned from hospitalized
heart failure registries. J. Am. Coll. Cardiol. 2014, 63, 1123–1133. [CrossRef] [PubMed]

3. Sidney, S.; Go, A.S.; Jaffe, M.G.; Solomon, M.D.; Ambrosy, A.P.; Rana, J.S. Association between aging of the US population and
heart disease mortality from 2011 to 2017. JAMA Cardiol. 2019, 4, 1280–1286. [CrossRef] [PubMed]

4. Takahashi, T.; Morisawa, T.; Saitoh, M.; Yokoyama, M.; Shimada, K.; Fujiwara, T.; Daida, H. Current status and future development
of acute and cardiac physiotherapies in Japan. Phys. Ther. Res. 2020, 23, 1–7. [CrossRef]

5. Yancy, C.W.; Jessup, M.; Bozkurt, B.; Butler, J.; Casey, D.E., Jr.; Drazner, M.H.; Fonarow, G.C.; Geraci, S.A.; Horwich, T.;
Januzzi, J.L.; et al. 2013 ACCF/AHA guideline for the management of heart failure: A report of the American College of Cardiology
Foundation/American Heart Association Task Force on practice guidelines. Circulation 2013, 128, e240–e327. [CrossRef]

6. Stiell, I.G.; Perry, J.J.; Clement, C.M.; Brison, R.J.; Rowe, B.H.; Aaron, S.D.; McRae, A.D.; Borgundvaag, B.; Calder, L.A.;
Forster, A.J.; et al. Prospective and explicit clinical validation of the Ottawa heart failure risk scale, with and without use of
quantitative NT-proBNP. Acad. Emerg. Med. 2017, 24, 316–327. [CrossRef]

7. Lee, D.S.; Stitt, A.; Austin, P.C.; Stukel, T.A.; Schull, M.J.; Chong, A.; Newton, G.E.; Lee, J.S.; Tu, J.V. Prediction of heart failure
mortality in emergent care: A cohort study. Ann. Intern. Med. 2012, 156, 767–775. [CrossRef] [PubMed]

8. Miró, Ó.; Pocock, S.J. Predicting 30-day mortality for patients with acute heart failure in the emergency department. Ann. Intern.
Med. 2018, 168, 899–900. [CrossRef]

9. Campanile, A.; Castellani, C.; Santucci, A.; Annunziata, R.; Tutarini, C.; Reccia, M.R.; Del Pinto, M.; Verdecchia, P.; Cavallini, C.
Predictors of in-hospital and long-term mortality in unselected patients admitted to a modern coronary care unit. J. Cardiovasc.
Med. 2019, 20, 327–334. [CrossRef] [PubMed]

10. Rivero-Ayerza, M.; Scholte Op Reimer, W.; Lenzen, M.; Theuns, D.A.; Jordaens, L.; Komajda, M.; Follath, F.; Swedberg, K.;
Cleland, J.G. New-onset atrial fibrillation is an independent predictor of in-hospital mortality in hospitalized heart failure patients:
Results of the EuroHeart Failure Survey. Eur. Heart J. 2008, 29, 1618–1624. [CrossRef]

11. Alataş, Ö.D.; Biteker, M.; Yildirim, B.; Acar, E.; Gökçek, K. Comparison of objective nutritional indexes for the prediction of
in-hospital mortality among elderly patients with acute heart failure. Eur. J. Emerg. Med. 2020, 27, 362–367. [CrossRef]

12. Flather, M.D.; Shibata, M.C.; Coats, A.J.; Van Veldhuisen, D.J.; Parkhomenko, A.; Borbola, J.; Cohen-Solal, A.; Dumitrascu, D.;
Ferrari, R.; Lechat, P.; et al. Randomized trial to determine the effect of nebivolol on mortality and cardiovascular hospital
admission in elderly patients with heart failure (SENIORS). Eur. Heart J. 2005, 26, 215–225. [CrossRef]

13. Düngen, H.D.; Apostolovic, S.; Inkrot, S.; Tahirovic, E.; Töpper, A.; Mehrhof, F.; Prettin, C.; Putnikovic, B.; Neskovic, A.N.;
Krotin, M.; et al. Titration to target dose of bisoprolol vs. carvedilol in elderly patients with heart failure: The CIBIS-ELD trial.
Eur. J. Heart Fail. 2011, 13, 670–680. [CrossRef] [PubMed]

14. Shiraishi, Y.; Kohsaka, S.; Abe, T.; Nagai, T.; Goda, A.; Nishihata, Y.; Nagatomo, Y.; Saji, M.; Toyosaki, Y.; Takei, M.; et al.
Derivation and validation of clinical prediction models for rapid risk stratification for time-sensitive management for acute heart
failure. J. Clin. Med. 2020, 9, 3394. [CrossRef]

15. McKee, P.A.; Castelli, W.P.; McNamara, P.M.; Kannel, W.B. The natural history of congestive heart failure: The Framingham study.
N. Engl. J. Med. 1971, 285, 1441–1446. [CrossRef]

16. Baumgartner, H.; Hung, J.; Bermejo, J.; Chambers, J.B.; Edvardsen, T.; Goldstein, S.; Lancellotti, P.; LeFevre, M.; Miller, F., Jr.; Otto,
C.M. Recommendations on the echocardiographic assessment of aortic valve stenosis: A focused update from the European
Association of Cardiovascular Imaging and the American Society of Echocardiography. J. Am. Soc. Echocardiogr. 2017, 30, 372–392.
[CrossRef] [PubMed]

17. Zoghbi, W.A.; Adams, D.; Bonow, R.O.; Enriquez-Sarano, M.; Foster, E.; Grayburn, P.A.; Hahn, R.T.; Han, Y.; Hung, J.;
Lang, R.M.; et al. Recommendations for noninvasive evaluation of native valvular regurgitation: A report from the Ameri-
can Society of Echocardiography Developed in Collaboration with the Society for Cardiovascular Magnetic Resonance. J. Am. Soc.
Echocardiogr. 2017, 30, 303–371. [CrossRef] [PubMed]

18. Baumgartner, H.; Hung, J.; Bermejo, J.; Chambers, J.B.; Evangelista, A.; Griffin, B.P.; Iung, B.; Otto, C.M.; Pellikka, P.A.;
Quiñones, M. Echocardiographic assessment of valve stenosis: EAE/ASE recommendations for clinical practice. J. Am. Soc.
Echocardiogr. 2009, 22, 1–23. [CrossRef]

19. Matsuo, S.; Imai, E.; Horio, M.; Yasuda, Y.; Tomita, K.; Nitta, K.; Yamagata, K.; Tomino, Y.; Yokoyama, H.; Hishida, A. Revised
equations for estimated GFR from serum creatinine in Japan. Am. J. Kidney Dis. 2009, 53, 982–992. [CrossRef]

20. Damman, K.; Valente, M.A.; Voors, A.A.; O’Connor, C.M.; van Veldhuisen, D.J.; Hillege, H.L. Renal impairment, worsening renal
function, and outcome in patients with heart failure: An updated meta-analysis. Eur. Heart J. 2014, 35, 455–469. [CrossRef]

21. Peterson, P.N.; Rumsfeld, J.S.; Liang, L.; Albert, N.M.; Hernandez, A.F.; Peterson, E.D.; Fonarow, G.C.; Masoudi, F.A. A validated
risk score for in-hospital mortality in patients with heart failure from the American Heart Association get with the guidelines
program. Circ. Cardiovasc. Qual. Outcomes 2010, 3, 25–32. [CrossRef]

J. Clin. Med. 2021, 10, 1468 10 of 10

22. Dobashi, S.; Watanabe, I.; Nakanishi, R.; Hisatake, S.; Kiuchi, S.; Kabuki, T.; Oka, T.; Fujii, T.; Ikeda, T. Comparing the effects of
milrinone and olprinone in patients with congestive heart failure. Heart Vessel. 2020, 35, 776–785. [CrossRef] [PubMed]

23. Packer, M.; Carver, J.R.; Rodeheffer, R.J.; Ivanhoe, R.J.; DiBianco, R.; Zeldis, S.M.; Hendrix, G.H.; Bommer, W.J.; Elkayam, U.;
Kukin, M.L.; et al. Effect of oral milrinone on mortality in severe chronic heart failure. The PROMISE Study Research Group. N.
Engl. J. Med. 1991, 325, 1468–1475. [CrossRef]

24. Bahadir, A.; Ortakoylu, M.G.; Iliaz, S.; Kanmaz, Z.D.; Bagci, B.A.; Iliaz, R.; Caglar, E. Prevalence and outcomes of comorbid
illnesses in elderly patients with respiratory diseases. Geriatr. Gerontol. Int. 2016, 16, 791–796. [CrossRef]

25. Nakagawa, N.; Saito, Y.; Sasaki, M.; Tsuda, Y.; Mochizuki, H.; Takahashi, H. Comparison of clinical profile in elderly patients
with nursing and healthcare-associated pneumonia, and those with community-acquired pneumonia. Geriatr. Gerontol. Int. 2014,
14, 362–371. [CrossRef]

26. Damy, T.; Viallet, C.; Lairez, O.; Deswarte, G.; Paulino, A.; Maison, P.; Vermes, E.; Gueret, P.; Adnot, S.; Dubois-Randé, J.L.; et al.
Comparison of four right ventricular systolic echocardiographic parameters to predict adverse outcomes in chronic heart failure.
Eur. J. Heart Fail. 2009, 11, 818–824. [CrossRef]

27. Gavazzoni, M.; Badano, L.P.; Vizzardi, E.; Raddino, R.; Genovese, D.; Taramasso, M.; Sciatti, E.; Palermo, C.; Metra, M.; Muraru, D.
Prognostic value of right ventricular free wall longitudinal strain in a large cohort of outpatients with left-side heart disease. Eur.
Heart J. Cardiovasc. Imaging 2020, 21, 1013–1021. [CrossRef] [PubMed]

28. The survival of patients with heart failure with preserved or reduced left ventricular ejection fraction: An individual patient data
meta-analysis. Eur. Heart J. 2012, 33, 1750–1757. [CrossRef]

29. Tsuchihashi-Makaya, M.; Hamaguchi, S.; Kinugawa, S.; Yokota, T.; Goto, D.; Yokoshiki, H.; Kato, N.; Takeshita, A.; Tsutsui, H.;
JCARE-CARD Investigators. Characteristics and outcomes of hospitalized patients with heart failure and reduced vs preserved
ejection fraction. Report from the Japanese Cardiac Registry of Heart Failure in Cardiology (JCARE-CARD). Circ. J. 2009,
73, 1893–1900. [CrossRef] [PubMed]

30. Nochioka, K.; Sakata, Y.; Takahashi, J.; Miyata, S.; Miura, M.; Takada, T.; Fukumoto, Y.; Shiba, N.; Shimokawa, H. Prognostic
impact of nutritional status in asymptomatic patients with cardiac diseases: A report from the CHART-2 Study. Circ. J. 2013,
77, 2318–2326. [CrossRef] [PubMed]

31. Narumi, T.; Watanabe, T.; Kadowaki, S.; Takahashi, T.; Yokoyama, M.; Kinoshita, D.; Honda, Y.; Funayama, A.; Nishiyama, S.;
Takahashi, H.; et al. Sarcopenia evaluated by fat-free mass index is an important prognostic factor in patients with chronic heart
failure. Eur. J. Intern. Med. 2015, 26, 118–122. [CrossRef] [PubMed]

Harbin et al. BMC Geriatrics (2022) 22:458
https://doi.org/10.1186/s12877-022-03161-w

R E S E A R C H

Barriers and facilitators of appropriate
antibiotic use in primary care institutions
after an antibiotic quality improvement
program – a nested qualitative study
Nicolay Jonassen Harbin1*, Morten Lindbæk1 and Maria Romøren2

Abstract
Background: Antibiotic prescribing by physicians in primary care institutions is common and affected by several fac-
tors. Diagnosis and treatment of infections in a nursing home (NH) resident is challenging, with the risk of both under-
and overtreatment. Identifying barriers and facilitators of appropriate antibiotic prescribing in NHs and municipal
acute care units (MACUs) is essential to ensure the most adequate antibiotic treatment possible and develop future
antibiotic stewardship programs.

Methods: After implementing a one-year antibiotic quality improvement program, we conducted six semi-struc-
tured focus group interviews with physicians (n = 11) and nurses (n = 14) in 10 NHs and 3 MACUs located in the
county of Østfold, Norway. We used a semi-structured interview guide covering multiple areas influencing antibiotic
use to identify persistent barriers and facilitators of appropriate antibiotic prescribing after the intervention. The inter-
views were audio-recorded and transcribed verbatim. The content analysis was performed following the six phases of
thematic analysis developed by Braun and Clarke.

Results: We identified thirteen themes containing barriers and facilitators of the appropriateness of antibiotic use
in primary care institutions. The themes were grouped into four main levels: Barriers and facilitators 1) at the clinical
level, 2) at the resident level, 3) at the next of kin level, and 4) at the organisational level. Unclear clinical presentation
of symptoms and lack of diagnostic possibilities were described as essential barriers to appropriate antibiotic use.
At the same time, increased availability of the permanent nursing home physician and early and frequent dialogue
with the residents’ next of kin were emphasized as facilitators of appropriate antibiotic use. The influence of nurses
in the decision-making process regarding infection diagnostics and treatment was by both professions described as
profound.

Conclusions: Our qualitative study identified four main levels containing several barriers and facilitators of appropri-
ate antibiotic prescribing in Norwegian NHs and MACUs. Diagnostic uncertainty, frequent dialogue with next of kin
and organisational factors should be targeted in future antibiotic stewardship programs in primary care institutions. In
addition, for such programs to be as effective as possible, nurses should be included on equal terms with physicians.

© The Author(s) 2022. Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which
permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the
original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or
other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line
to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory
regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this
licence, visit http:// creat iveco mmons. org/ licen ses/ by/4. 0/. The Creative Commons Public Domain Dedication waiver (http:// creat iveco
mmons. org/ publi cdoma in/ zero/1. 0/) applies to the data made available in this article, unless otherwise stated in a credit line to the data.

Open Access

*Correspondence: [email protected]

1 Antibiotic Center for Primary Care, Department of General Practice, Institute
of Health and Society, University of Oslo, Postboks 1130 Blindern, 0317 Oslo,
Norway
Full list of author information is available at the end of the article

Page 2 of 15Harbin et al. BMC Geriatrics (2022) 22:458

Background
Antimicrobial resistance is an increasing challenge
worldwide [1], and the call for prudent use of antibiotics
is in demand to slow down and reverse further resistance
development [2]. Previous studies have shown that 1/2 to
more than 3/4 of nursing home (NH) residents receive
one or more courses of antibiotics during a calendar year
[3–6]. Bacterial infections requiring antimicrobial treat-
ment has a high prevalence in long-term care residents
compared to elderly living at home [7], many of them
deemed inappropriate [8–10].

Prescribing antibiotics can be classified either as medi-
cally or ethically appropriate based on the circumstances
of the specific antibiotic-requiring infection. Medically
appropriate antibiotic use typically covers the clinical,
microbiological, pharmacokinetic and dynamic aspects
of the specific cases. Simultaneously, antibiotic treatment
among old and frail NH residents is primarily about pro-
longing life, which raises the question of whether anti-
biotic treatment is appropriate or inappropriate from
an ethical perspective. Ethical questions make infection
diagnostics and treatment a more significant challenge
for NH physicians and nurses, as multiple comorbidities,
polypharmacy, speech and hearing disabilities, cogni-
tive debilitation, and atypical symptom manifestation of
infections is more common in the NH population [11–
15]. Limited diagnostic on-site possibilities at the NHs
further complicate diagnosis and treatment of infections
[16]. Next of kin is usually more involved in promoting
NH residents’ interests and demands than in their earlier
adult life. Although the Norwegian health law demands
more next of kin involvement when residents cannot
consent, the providing physician has to make final deci-
sions in cases involving consent incompetence [17].
Besides apparent advantages and benefits of next of kin
involvement, this guardianship role may lead to tension,
disagreement and conflict between relatives and health
care professionals regarding infection treatment [18, 19].

Although the final decision to prescribe antibiot-
ics or not is taken by one or several medical physicians,
the decision-making of prescribing is multifactorial and
complex. Previous studies have identified several factors
influencing antibiotic prescribing in hospitals and gen-
eral practice, for example, physician-specific and patient-
related factors, availability of diagnostic tools and local
antibiotic resistance data, patient satisfaction and cul-
tural and organizational factors [20–31]. Many of the fac-
tors identified in general practice and hospitals may apply

for NHs, and some issues are specific to NHs affecting
antibiotic prescribing. Studies on antibiotic prescribing
decisions in NHs and older aged patients have identified
the availability of evidence-based guidelines, physicians’
habits, perceived risks of antibiotic prescribing, the influ-
ence of other health care professionals, and residents’
current clinical situation and medical history as impor-
tant factors influencing antibiotic treatment duration and
the decision-making whether or not to prescribe antibi-
otics [16, 32–36].

Responding to the emerging antimicrobial resist-
ance threat, the Norwegian Government published its
“National Action Plan against Antibiotic Resistance in
the Health Services” in 2016 [37]. As part of the plan, the
Antibiotic Centre for Primary Care launched the “RASK”
intervention in the county of Østfold, a quality improve-
ment programme aiming to optimize treatment of infec-
tions and improve antibiotic prescribing in Norwegian
NHs and municipal acute care units (MACUs) [38]. The
intervention aimed to increase the knowledge regarding
appropriate antibiotic treatment and increase the aware-
ness of the institutions’ antibiotic prescribing patterns.

Previous Norwegian studies have found a wide vari-
ation in total antibiotic use between Norwegian NHs,
indicating a potential for improving antibiotic prescrib-
ing in this sector [39, 40]. In addition, we have identified
only one previous Norwegian qualitative study on fac-
tors influencing antibiotic treatment in NH residents,
who primarily investigated the ethical problems related
to intravenous antibiotic administration perceived by
NH nurses [18]. Further studies, investigating the cur-
rent topic on a broader level is warranted. Therefore, this
focus group study aimed to in-depth explore both physi-
cians’ and nurses’ perceptions of persisting barriers and
facilitators of appropriate antibiotic use in Norwegian
NHs and MACUs after the implementation of a struc-
tured antibiotic improvement program.

Methods
This article conforms to the “Standards for Reporting
Qualitative Research (SRQR): 21-items checklist” [41].

Study setting
The current study was initiated after completing the
“RASK” intervention in Østfold county, located in the
South-Eastern part of Norway, which lasted from Octo-
ber 2016 to October 2017. In Norway, NHs may be clas-
sified as long-term, short-term or mixed (both long- and

Keywords: Nursing home, Municipal acute care unit, Antibiotic stewardship program, Barriers, Facilitators, Urinary
tract infection, Life-prolonging treatment

Page 3 of 15Harbin et al. BMC Geriatrics (2022) 22:458

short-term) based on the residency. In addition to ordi-
nary NHs, we have municipal acute care units (MACUs)
to alleviate the use of hospital services. Although classi-
fied as a part of the NH sector, MACUs differ from tra-
ditional NHs in that the patients live at home and are
admitted due to acute onset disease by general practi-
tioners during regular working hours and out-of-hours.
When the patients have been treated, they are usually
discharged home.

We invited all 37 NHs and MACUs in the county to the
intervention, resulting in 34 institutions participating in
the project. NH and MACU physicians, nurses and other
healthcare professionals from included institutions were
then invited to a one-day conference with professional
presentations and workshops on infections and appro-
priate use of antibiotics. All participating institutions
received a report presenting their antibiotic use based
on sales statistics from supplying pharmacies, compared
to other participating institutions in the county. After
the starting conference, participants were instructed to
arrange educational activities on the same topics for their
colleagues and set a goal for their institution during the
one-year project period. In addition, the institutions were
asked to register bi-monthly point prevalence surveys
on antibiotic use and indication, and tailor-made clini-
cal checklists were offered as tools to be used during the
intervention year. Follow-up conferences were held after
six and 12 months, and participating institutions received
new antibiotic reports for further academic audit and
feedback. All physicians and nurses working in the NHs
or MACUs that took part in the intervention were invited
orally at the final “RASK” conference to participate in
the current study, and in addition they were personally
invited by email or telephone. Willing informants were
included until we decided that a saturation point had
been reached. The saturation point was assessed continu-
ously by comparing the current interview with summa-
ries and transcripts of previous interviews, and decided
upon when no new themes and no additional information
on pre-existing themes occurred. We conducted six focus
group interviews between October 2017 and December
2018 with 11 physicians and 14 nurses from 13 institu-
tions. The participants did not work in the same institu-
tion, except from one interview where one physician and
three nurses were employed at the same MACU ward.
Nine physicians and six nurses had participated at one
or more conferences during the intervention year, and all
participants were familiar with the programme through
the antibiotic reports, educational material, and the use
of intervention tools at the institutions. Each focus group
consisted of three to six participants, a size range decided
upon to obtain interactive group discussions during
the interviews and to increase the involvement level for

each participant. Four of the interviews were conducted
with both physicians and nurses mixed to explore the
dynamics between the two occupational groups. The two
last interviews were conducted with only physicians in
one interview and only nurses in the other to see if we
received other information when the groups were inter-
viewed separately.

Researcher characteristics
NJH is a part time NH physician working at a short-term
NH in one of the municipalities that participated in the
“RASK” intervention. In addition, he was the respon-
sible coordinator for the “RASK” intervention in the
county of Østfold. NJH had no experience in qualitative
research prior to conducting the current study. ML is a
long-time GP and a researcher in the field of antibiotic
prescribing in general practice and NHs, and had the
overall leadership responsibility for the “RASK” interven-
tion in Østfold. MR is a GP and a long-time researcher
in the field of NH medicine and general practice. Both
ML and MR have extensive prior experience and train-
ing in conducting qualitative research. All authors share
a common interest in quality improvement in primary
care institutions, and factors affecting antibiotic prescrib-
ing in particular. Based on the authors background, one
could imagine that the informants would formulate their
answers based on what they thought was expected to be
answered. However, we perceived the discussions in the
interviews as open and rich, and that the health person-
nel talked uncensored about their thoughts, experiences
and dilemmas in their clinical work. The prior knowledge
of the organizational structure and clinical everyday life
rather was an advantage in penetrating and understand-
ing the informants’ stories and perceptions.

Data collection
The interview duration varied from 56 minutes to 87 min-
utes, with a mean overall duration of 75 minutes. NJH was
the main interviewer in all six interviews, while ML and
MR participated as co-interviewers in one and five inter-
views, respectively. We used a semi-structured interview
guide that NJH, MR, and ML developed. The interview
guide contained four main topics; 1) factors influencing
physicians’ antibiotic prescribing, 2) factors influencing
physicians’ choice to deviate from antibiotic guidelines,
3) influence of nurses on physicians’ antibiotic prescrib-
ing and 4) what ethical dilemmas physicians and nurses
experience regarding antibiotic treatment. The inform-
ants were provided with written information about the
main topics of the study by email in advance. All inter-
views were started by shortly describing the main top-
ics, followed by encouraging the informants to describe
two experienced cases where antibiotic treatment had

Page 4 of 15Harbin et al. BMC Geriatrics (2022) 22:458

been initiated; one where there had been no doubt and
the other where there had been hesitations regarding the
treatment. The main topics were then presented to the
informants step-wise through the interviews to initiate
discussions in the group. The informants were moderated
if they deviated greatly from the topics. When saturation
on the relevant topics became evident, the interviewers
moved on to the next topic in the interview guide. The
interviewers also engaged informants who were not as
involved in the discussions and complemented with in-
depth questions along the way as needed.

Data analysis
The interviewers discussed the interviews immedi-
ately after completion of each interview, and a sum-
mary for each interview was written and discussed
further by email. All interviews were audio-recorded
and transcribed verbatim. We based the analyses on the
six phases of thematic analysis developed by Braun and
Clarke [42], primarily with a inductive and semantic
approach: 1) familiarizing with the depth and breadth of
the content by reading repeatedly through the interviews
to gain a general impression, 2) generating initial codes
for the entire material using both theory and data-driven
approaches, 3) searching for themes and re-sorting the
initial codes into potential themes, 4) reviewing potential
themes at the level of the coded data extracts and creat-
ing a candidate thematic map and secondly considering
the validity of individual themes and the candidate the-
matic map in relation to the data set, 5) further defining,
refining and naming the themes and identifying barriers
and facilitators of appropriate antibiotic use, 6) identify-
ing, defining and naming overarching levels, and further
group the themes at the appropriate level, 7) produc-
tion of the article including illustrative extracts from the

material that captures the essence of the points demon-
strated. NJH transcribed the interviews and performed
the initial coding and analysis of the content. MR and ML
further evaluated the transcribed data material, as well
as the initial coding of the content. All authors partici-
pated substantially in the process of searching, defining,
reviewing and naming relevant themes and in the article
production. The qualitative data analysis software pro-
gram NVIVO 12 was used for analyses and data manage-
ment [43].

Ethical approval
All participants provided written consent prior to the
interviews. We replaced any names and places with num-
bers and characters in the transcribed text to protect the
anonymity of the participants. The Regional Commit-
tees for Medical and Health Research Ethics of South-
East Norway granted ethics approval for the study (ref.:
2017/1711), and the Norwegian Centre for Research Data
approved data protection (55,887 / 3 / LAR).

Results
The demographic characteristics of the informants are
presented in Table 1.

We identified thirteen themes grouped into four main
overarching levels affecting antibiotic use during the
analysis, ranging from individual to external and sys-
temic factors (Fig. 1). Most of the barriers and facilitators
described by the informants applied to both NHs and
MACUs. We have chosen to use the term NH further
in the article when discussing factors that apply to both
types of institutions, while we specify when factors were
applicable only to NHs or MACUs.

Table 1 Characteristics of the study informants

Demographics Physicians (n = 11) Nurses (n = 14) Overall (n = 25)

Sex Female 7 13 20

Male 4 1 5

Age Mean (range) 42 (32 – 64) 43 (25 – 62) 43 (25 – 64)

Years clinical experience Mean (range) 14 (4 – 37) 16 (2 – 41) 15 (2-41)

Type of facility Nursing home 7 10 17

Municipal acute care unit 4 4 8

Speciality Nursing home General practitioner (1) Geriatric and palliative medicine (1) –

Internal medicine (2) Rehabilitation medicine (1)

In specialisation (3) Registered nurse (8)

No specialisation (1)

Municipal acute care unit General practitioner (1) Acute geriatric medicine (3) –

In specialisation (3) Registered nurse (1)

Page 5 of 15Harbin et al. BMC Geriatrics (2022) 22:458

Barriers and facilitators at the clinical level
Unclear clinical presentation
Unclear clinical manifestations caused by infections,
especially in cognitively impaired residents, were
regarded as major contributors to diagnostic uncertainty
and difficult treatment decisions. Some physicians high-
lighted the difficulty of distinguishing viral from bacterial
respiratory tract infections, due to a perception that frail
and old residents with viral respiratory tract infections
often present with typical hallmarks of bacterial infec-
tions like fever, crackles over the lungs and increased
C-reactive protein concentrations. Identifying bacterial
aetiology was pointed out as a relevant challenge, espe-
cially during flu seasons, which often led to uncertainty
among physicians regarding the initiation of antibiot-
ics. In such cases, the resident’s general condition was
described as decisive for whether antibiotic treatment
should be initiated. Several physicians described that
they had a lower threshold for starting antibiotic treat-
ment in residents with a chronic obstructive pulmonary
disease when in doubt about the microbiological cause
of the respiratory infection. Diagnosis and treatment
of urinary tract infection (UTI) was perceived as chal-
lenging, especially in demented residents, due to non-
specific symptoms, poor anamnesis and high prevalence
of asymptomatic bacteriuria, leading to a high level of
uncertainty. Both physicians and nurses described sev-
eral approaches to uncertain UTI situations, including
watchful waiting, intravenous fluid therapy and antibiotic
prescribing to be on the safe side. With uncertain focus of

infections, several of the physicians and nurses described
that broad-spectrum antibiotics were often prescribed
to cover both the airways and urinary tract system. One
of the physicians claimed that in cases where residents
presented with new-onset non-specific symptoms, like
confusion and agitation, it was better to try a short-term
course of antibiotics than more side-effect burdened
medications.

Physician, male, 35 – 39 years: “They do not have the
clear urinary tract infection symptoms. They can be
agitated and have a positive urine stick. Instead of
starting up with haloperidol or something similar, it
is after all a bit better to give a course of pivmecilli-
nam to check if a urinary tract infection is the cause.
Therefore, you treat a little more on vague indica-
tions.”

Lack of diagnostic possibilities
Lack of diagnostic possibilities was described as a gen-
eral barrier for both the diagnostic process and for the
choice of antibiotic by both professions. On-site x-ray
was an opportunity they missed, mainly when dealing
with emerging cases of respiratory infections to avoid
unwanted and burdensome referrals to the local hospital.
One physician described a case that involved a resident
who had experienced several respiratory infections, lead-
ing to multiple antibiotic treatments. After referring the
resident to an x-ray investigation at the hospital, lung
cancer was identified. The informant emphasized that

Fig. 1 Overarching levels (in the circle) and associated themes affecting antibiotic prescribing by physicians nursing homes

Page 6 of 15Harbin et al. BMC Geriatrics (2022) 22:458

the resident could have been spared from many antibiotic
courses if inhabiting an on-site x-ray at the NH. The lack
of diagnostic possibilities were also described as a barrier
for narrow-spectrum antibiotics, as it often led to a “bet-
ter safe than sorry” approach.

Physician, male, 35 – 39 years: “We are first-line
service, so we do not have all the diagnostic tools.
When you get a case with no clear clinical focus, and
you have not performed a good advance care plan-
ning, I feel a bit trapped. I do not want to refer the
patient to the hospital, but still you want to feel that
what we do is justifiable. Then it happens that you
end up with broad spectrum (antibiotics), and it is
often with a bit of distaste, right?!”

Among the nurses, a primary concern was an often pro-
longed time from sampling to blood biochemistry and
bacteriological cultivation results. This was regarded as
a barrier to treating infections in accordance with the
National guidelines for antibiotic use and, potentially,
leading to decisions not to prescribe the recommended
first-line antibiotics.

Knowledge and awareness
The physicians and nurses pointed to a shortcoming of
knowledge, mainly amongst auxiliary nurses and regis-
tered nurses, regarding indication, sampling and inter-
pretation of point-of-care test results as a persistent
and major barrier to appropriate antibiotic treatment of
residents. Not interpreting the test results alongside the
clinical signs and symptoms was described as a recurrent
problem in the diagnostic process.

Physician, male, 40 – 44 years: “The next day the
C-reactive protein has risen, but the fever has gone
down and the patient is out of bed. Then there are
some who think that the antibiotics does not work
… and then… then you have to quickly sign up for a
course and start paying attention.”

Both professions emphasized increased knowledge
regarding the correct use of diagnostic tests as one of the
most important measures of the intervention. In particu-
lar, the indication and interpretation of urine dipstick and
C-reactive protein tests were mentioned.

To increase awareness among healthcare profession-
als, concerning own antibiotic prescribing practices and
the development of antimicrobial resistance (locally and
globally) was described as an additional intervention
measure to facilitate prudent antibiotic use.

Physician, female, 40 – 44 years: I gave the introduc-
tory antimicrobial resistance lecture, from the first
conference meeting, to my nurses. When they saw

the maps changing color throughout Europe, from
mainly green to mainly red countries, there was a
gasp from the assembly. Therefore, I think with good
and correct knowledge it at least makes it easier and
safer for me to explain and justify my choices to my
colleagues.

Strategies for coping with uncertainty
The physicians and nurses mentioned several strategies to
counteract diagnostic uncertainty and thereby facilitate
appropriate antibiotic use. Watchful waiting, often com-
bined with intravenous fluid treatment, was described as
a commonly used strategy, especially when dealing with
suspected but uncertain UTI cases. Utilization of a urine
culture to avoid unnecessary antibiotic treatment was
additionally lifted as a strategy when encountering non-
specific UTI symptoms.

Physician, male, 35 – 39 years: “When speaking of
UTIs, which is where perhaps the most disagreement
is, I think if I am in doubt “yes, we’ll send it for cul-
ture”. It takes a few days or a week until the culture
is ready, and by that time the resident has become
better or has developed more clear symptoms. Then
I gain some time on it, and can postpone or avoid
antibiotics.”

Some nurses and physicians described the clinical
checklist for suspected UTIs, offered to the participat-
ing institutions as part of the intervention, as a valuable
and effective tool in reducing the number of antibiotic
treatments in pending cases of UTIs. One perceived rea-
son for this was that the threshold for sampling urinary
dipstick tests increased amongst both auxiliary and reg-
istered nurses, leading to fewer tests being presented to
the physician for evaluation. Finally, a referral system to
the local hospital for diagnosis, treatment decision and
level of care, informally called a “diagnostic loop referral”,
was highlighted as an additional strategy when dealing
with diagnostic uncertainty. This system was mainly uti-
lized by the informants working at MACUs, and to some
extent by the NH informants.

Barriers and facilitators at the resident level
One of the physicians described the balancing act of anti-
biotic treatment in NH residents, and hence life-prolong-
ing treatment, as operating in “gray areas”.

Resident autonomy and consent competence
Most of the nurses and physicians described an increased
focus on assessing consent competence in NHs in
recent years. In addition, several of the physicians and
nurses emphasized the importance of the patients’

Page 7 of 15Harbin et al. BMC Geriatrics (2022) 22:458

voice regarding antibiotic treatment, even if cognitively
impaired and apparently consent incompetent.

Nurse, female, 35 – 39 years: “We have demented
residents who say: “No, I am now so ill that this is
not compatible with life”. And it’s a bit like, what do
you express in your own illness when you are there?
We have demented pneumonia patients who can
answer us: “Yes, thank you. It’s nice that the doctor
has been here, but I will not take any pills”. Then you
have those who want treatment for all it is worth.”

Both professions expressed that if a consent-compe-
tent resident, and to some extent consent-incompetent
residents, express a specific wish not to be treated with
antibiotics in case of a life-threatening infection, the
residents wish was usually respected. Nevertheless,
some nurses described that it was not unusual that phy-
sicians still initiated antibiotic treatments against the
residents wish not to receive treatment. Several factors
were perceived as barriers to good judgment and deci-
sions regarding antibiotic treatment in end-of-life situa-
tions. These included poor advance care planning, lack of
residents own voice, cognitive impairment, and residents
changing their minds regarding antibiotic treatment dur-
ing an ongoing infection. Nevertheless, according to both
nurses and physicians, such cases often culminated in
antibiotic treatments.

Physician, male, 35 – 39 years: “Then you have that
“fresh product” as you mentioned earlier, where you
have a patient who earlier said it does not want any
life-prolonging treatment. Then the patient gets an
infection and then they want treatment, and then
they get (antibiotic) treatment.”

Quality of life, frailty and short life expectancy
One decisive factor that was perceived as crucial regard-
ing life prolonging treatment or not, including antibiotic
treatment, was residents’ quality of life. Several different
factors influencing quality of life were mentioned, such as
familiar joys, a wish to experience important events (the
next Olympics, a wedding, etc.), or simply to enjoy good
food. When evaluating the quality of life, the physicians
and nurses expressed that this was an assessment prefera-
bly done by residents themselves. In cases with an absent
resident voice, both professions described that they usu-
ally involved next of kin to elicit information regarding
the resident’s earlier preferences. If the involvement of
the next of kin did not provide adequate information, the
choice of treatment was to be decided by the responsi-
ble physician. One physician described that degree of
cognitive impairment, pain and agitation were the most
essential prerequisites for assessing the quality of life in

residents unable to communicate themselves. There was
a a general agreement among both nurses and physicians
throughout the interviews that antibiotic treatment at the
expense of residents’ quality of life was considered uneth-
ical and inappropriate.

Physician, male, 35 – 39 years: “If the measure you
take to live longer takes away the quality of life….”

Physician, male, 40 – 44 years: “Then it may not be
the right measure.”

One physician shared that many antibiotic treatment
courses in NHs are both medically and ethically inappro-
priate, as it often prolongs residents’ suffering. Accord-
ing to the same physician, the reason for this, and hence
being a barrier to proper antibiotic treatment in these
situations, was that physicians refuse to decide to refrain
from treatment as it is perceived as unpleasant. Further-
more, that the prescriber should assess the level of frailty
and underlying disorders before initiating life-prolonging
antibiotic therapy and reflect on the life situation the
resident eventually would return to, given a successful
treatment. Achieving this, according to the same physi-
cian, would facilitate both the medically and ethically
appropriateness of antibiotic therapy in NH residents.
The same physician also applied this to residents experi-
encing recurring bacterial infections and in cases where
the preferred antibiotics lacked effect, leaving the ques-
tion of whether to change to broader spectrum antibiot-
ics or not.

Physician, male, 40 – 44 years: “You have the option
to start with penicillin, then it does not work so
you add gentamycin, and if that does not work you
switch to cefotaxim. If you choose that road, it is
clearly not the right way to go with someone that
frail in the first place. Although you might do it cor-
rect medically, you are ethically completely out of
your mind.”

Antibiotic treatment of palliative and pre-terminal
residents with short life expectancies generated mixed
responses. Several physicians expressed a tendency to
treat palliative care residents with antibiotics primarily
with a symptom-based, not life-prolonging, approach to
relieving pain and discomfort associated with the infec-
tion. On the other hand, the majority of both nurses and
physicians described a clear reluctance towards treating
pre-terminal patients with antibiotics, including treat-
ment to relieve symptoms.

Physician, female, 35 – 39 years: “It seems directly
unethical really. If you think they will die in a short
time, giving antibiotics, I do not know. There are also

Page 8 of 15Harbin et al. BMC Geriatrics (2022) 22:458

side effects. I cannot see many situations where it
can be justified.”

Barriers and facilitators at the next of kin level
Drivers for testing and treatment
Despite health personnel considering the treatment
medically and ethically inappropriate, influence and
pressure for antibiotic treatment from residents’ next
of kin were described as a persistent barrier towards
correct antibiotic use in NHs. Both the physicians and
nurses expressed that diagnostic and antibiotic treatment
should be based on residents’ wishes and medical deci-
sions made by physicians. However, several informants
from both professions were inclined to give in to non-
coherent wishes and pressure from the next of kin. Lack
of residents’ own voice, and cases where residents’ ear-
lier wishes prior to the incapacity were unknown, were
described as potential conflict-generating situations.
Some nurses even described cases where residents clearly
had expressed specific treatment wishes, but through
persuasion by next of kin had chosen to change their
position regarding treatment per the relatives’ wishes.

Nurse, female, 60 – 62 years: “This man, at that
time, he said clearly and unequivocally: “it is I who
decide”. Then the relatives persuaded him when we
were not present. Therefore, I think there is a lot of
hassle with family members.”

Disagreements and conflicts
Disagreement between health care professionals and
next of kin was widely discussed in several interviews
as a common challenge concerning antibiotic treat-
ment. Regarding frail elderly residents, the nurses and
physicians described relatives who exhibit unrealistic
expectations of the treatment effect of antibiotics and
do not understand why diagnostic testing needs a clini-
cal indication. Fear of negative coverage in the local
newspaper or filing complaints to the municipality’s
management department were reasons described for
giving in to pressure from next of kin. Another reason
for succumbing to the wishes of residents’ next of kin
was the experience that this approach generated less
work for the physician, as dealing with disagreements
and conflicts were time-consuming and tiring. When
exposed to strong disagreements regarding antibiotic
treatment, some expressed that they simply chose to
follow the wishes of next of kin to avoid conflicts. One
physician explained that in cases where family mem-
bers demanded antibiotic treatment in apparent uneth-
ical or medically futile situations, he often decided to
treat with a narrow-spectrum antibiotic knowing that

it would have no effect at all on the infection. Oth-
ers described compromise-based approaches, which
resulted in an antibiotic treatment attempt of two or
three days, with subsequent termination of the treat-
ment if the resident did not show signs of improvement.

Nurse, female, 60 – 64 years: “But the grandson of
the resident, who himself was a doctor, would not
give up. Therefore, we talked with our physician and
expressed that this was not correct. Then we decided
to prescribe antibiotics for two or three days and
then discontinue the treatment.”

Some of the nurses further expressed an impression
that residents’ relatives over the years increasingly have
gained power concerning diagnostic procedures and ini-
tiating antibiotic treatment.

Nurse, female, 45-49 years: “Whom are we actually
treating? Are we treating the resident or the rela-
tives?”

Dialogue and advance care planning
Early stage dialogue with residents’ next of kin, often con-
nected with advance care planning, was highlighted as
facilitators for achieving ethically and medically correct
antibiotic treatment of NH residents. Perceived benefits
of early dialogue included building trust relationships,
maturing and curbing relatives for future deteriorations
in health and deciding level of antibiotic treatment prior
to these events. Professional experience of healthcare
workers, continuity in terms of full-time employment of
physicians, and collaboration and agreement between
health care professionals on complex issues regarding
antibiotic treatment were described as important facili-
tators for trust building with next of kin. Although next
of kin was generally described as needing repeated real-
ity-oriented conversations concerning conflicting issues,
such dialogues were considered an essential measure for
ethically and medically correct antibiotic treatment.

Nurse, female, 60 – 64 years: “The patient has a seri-
ous illness not yet diagnosed, which the hospital has
chosen not to investigate any further. Therefore, if
we do not talk openly about this the relatives might
wonder why in the world we do not treat their dad,
right?! Such cases need clarification.”

Advance care plans, applying primarily for long-term-
care NHs, were generally seen as valuable, reassuring and
important by both physicians and nurses when dealing
with new-onset infectious conditions, mainly concerning
whether the resident should receive antibiotic treatment
or not.

Page 9 of 15Harbin et al. BMC Geriatrics (2022) 22:458

Barriers and facilitators at the organisational level
Barriers and facilitators in nursing homes
Deficient staffing resources, especially concerning physi-
cians, were described as an important barrier in terms of
optimizing diagnostics and antibiotic treatment of resi-
dents. By having frequent access to the permanent NH
physician, who inhabits knowledge of the residents and
their medical history, this was perceived as a benefit for
the residents themselves and facilitate both medically
and ethically appropriate use of antibiotics.

Physician, male, 40 – 44 years: “By being present
every day, I think you get to use antibiotics in a
much better way compared to arriving at a NH to
attend an ill resident you do not know from before.
Then it is much easier to think, “yes we’ll start anti-
biotic treatment because he is ill.”

Both professions described the collaboration between
nurses and physicians as non-problematic in terms of
antibiotic treatment. Some physicians emphasized that
due to the intervention, the collaboration worked better
because the nurses to a lesser extent conducted point-
of-care testing on their own initiative. Likewise, some
physicians also highlighted decreased pressure from the
nurses to initiate antibiotic treatment during and after
the intervention, further adding to better collaboration
between the two professions. Both professions pointed
to the crucial role of nurses regarding the diagnostic pro-
cess, where several of the physicians expressed that the
nurses literally acted as their “eyes and ears” in many
clinical decisions.

Physician, female, 40 – 44 years: “We are very
dependent on the nurses, it is therefore very impor-
tant that they have a competent clinical view and
that the collaboration works well.”

Similarly, the nurses perceived their role in clinical deci-
sion-making processes as significant, and thus having
a major impact on physicians’ clinical decisions. One
physician pointed out that if a nurse wanted a resident
treated with antibiotics, the nurse would have no prob-
lem convincing the physician into treating the resident.

Physician, male, 35 – 39 years: “Yes, so it is how they
(the nurses) describe it. They will get a cure for uri-
nary tract infection if they want. They can report
that the patient is more restless, has frequent urina-
tion and so on.”

Some physicians and nurses described two potential bar-
riers of appropriate antibiotic use regarding the nurse
role. First, different nurses may have consistently different
interpretations and reports of clinical observations. Sec-
ondly, nurses’ accuracy in relation to adherence regarding

dosing intervals of oral antibiotics was described as often
inaccurate, while with intravenous antibiotics the adher-
ence to the intervals was usually accurate. Some of the
nurses confirmed this, and described that one possible
reason may be that residents treated with oral antibiot-
ics are not considered as ill as those receiving intravenous
antibiotics.

Challenges specific for municipal acute care units
One issue that applied explicitly to MACUs was that the
diagnosis given in referral letters from general practition-
ers (GPs) and emergency physicians (OOHS) often was
perceived as deliberately incorrect. Several of the MACU
nurses and one MACU physician expressed a suspicion
that the referring physicians often used wrong referral
diagnosis as a shell hide for the real reason for admit-
tance. UTIs and dehydration were mentioned as fre-
quently used diagnoses to justify admissions to MACUs,
while the real reasons for referral often were perceived as
pressure from the patient’s relatives or the home nurse
service regarding inclining difficulties living at home
due to age, cognitive impairment and frailty. Referrals to
MACUs often include an antibiotic initiation plan when
the admission diagnosis is an infection and is usually not
re-evaluated until the MACU physician returns to the
ward. Several of the MACU nurses looked upon this as
a barrier of correct antibiotic treatment, as many of the
referred UTI cases were actual cases of asymptomatic
bacteriuria not requiring antibiotic treatment.

Nurse, female, 25 – 29 years: “We get patients
referred with a plan, like an antibiotic regimen. But
the real reason for admission is that relatives are
going on holiday … they take a urine sample, find
something on the dip stick and then they are admit-
ted to us.”

Treatment initiated by out‑of‑hours service
The nurses generally expressed that OOHS was some-
thing they strived to avoid using as far as possible, as
contacting the OOHS was tantamount to calling in a
treatment order.

Nurse, female, 35 – 39 years: “You call in an order,
and you get what you ask for.”

Regarding antibiotic treatment, several of the nurses
and physicians in the interviews shared the opinion that
OOHS physicians had a lower threshold for initiating
broad-spectrum antibiotics. Lack of knowledge con-
cerning resident history and lack of clinical examination
of residents as the treatment decision often happens by
telephone consultation were described as main barriers