Infectious Diseases Now 53 (2023) 104751

Available online at

www.sciencedirect.com

Review

Diagnostic uncertainty in infectious diseases: Advocacy for a nosological

framework
Pierre-Marie Roger a,b,⇑, Olivia Keïta-Perse c, Jean-Luc Mainardi d,e
a
Infectiologie, Centre Hospitalier Universitaire de Guadeloupe, France
b
Faculté de Médecine, Université des Antilles, France
c
Epidémiologie et Hygiène Hospitalière, Centre Hospitalier Princesse Grace, 98000, Monaco
d
Service de Microbiologie, Hôpital Européen Georges Pompidou, AP-HP Centre, 75015 Paris, France
e
Université Paris Cité, Paris, France

a r t i c l e i n f o a b s t r a c t

Article history: Diagnostic uncertainty (DU) is frequent in infectious diseases (ID), being recorded in 10% to over 50% of
Received 17 February 2023 patients. Herein, we show that in several fields of clinical practice, high rates of DU are constant over
Revised 22 June 2023 time. DUs are not taken into account in guidelines, as therapeutic propositions are based on an estab-
Accepted 3 July 2023
lished diagnosis. Moreover, while other guidelines underline the need for rapid broad-spectrum antibi-
Available online 6 July 2023
otic therapy for patients with sepsis, many clinical conditions mimic sepsis and lead to unnecessary
antibiotic therapy. Considering DU, many studies have been carried out to look for relevant biomarkers
Keywords:
of infections, which also attest to non-infectious diseases mimicking infections. Therefore, diagnosis is
Diagnostic uncertainty
Antibiotic use
often primarily a hypothesis, and empirical antibiotic therapy should be reassessed when microbiological
Guidelines data are available. However, other than for urinary tract infections or unexpected primary bacteremia,
Biomarkers the high frequency of sterile microbiological samples implies that DU remains central in follow-up, which
Antimicrobial stewardship does not facilitate clinical management or antibiotic optimization. The main way to resolve the therapeu-
tic challenge of DU could be to precisely describe the latter through a consensual definition that would
facilitate consideration of DU and its mandatory therapeutic implications. A consensual definition of
DU would also clarify responsibility and accountability for physicians in the antimicrobial approval pro-
cess and l provide an opportunity to instruct their students in this large field of medical practices and to
productively conduct relevant research.

1. Introduction mendations are made for a well-established diagnosis, and difficul-

ties in diagnosis are generally not discussed. As a result, a list of
Antimicrobial stewardship (ASM) was initially aimed at redu- etiologies that could explain therapeutic failure is very rarely pro-
cing antibiotic consumption in order to limit the emergence of vided. As an example, in 2007 the American guidelines for
multi-drug resistant (MDR) bacteria [1]. After several years of suc- community-acquired pneumonia enumerated the etiologies of
cessive approaches, ASM now also seeks to improve antibiotic use antibiotic failures, which involved non-infectious diseases such
for the benefit of patients [2]. To achieve these targets, antimicro- as vasculitis, pulmonary embolism, and cancer [7]. One might con-
bial stewardship programs (ASPs) include a list of actions, such as sider antibiotic failure in case of pulmonary vasculitis as a diagno-
proposed therapeutic guidelines and audits with feedback [2]. sis error associated with antibiotic misuse, and not as an antibiotic
The benefit of ASM and guidelines to improve antibiotic use in failure. Moreover, when associated with different therapeutic
daily practice has been demonstrated, as they reduce the morbidity options according to different guidelines, the absence of differen-
and mortality associated with infectious diseases and/or antibiotic tial diagnosis, can be confusing for clinicians, and become a factor
use and the emergence of MDR bacteria [3–6]. of non-adherence [11].
Institutional guidelines are based on international and/or That said, it is indeed difficult to correctly diagnose infectious
national recommendations, the latter beginning with description diseases, and a recent prospective multicenter audit showed that
of the infectious diseases under consideration [7–11]. The Recom- unnecessary antibiotic use accounted for 169/453 (37%) of pre-
scriptions, and that 106/169 (63%) of unnecessary prescriptions
concerned non-infectious syndromes [12]. The high level of unnec-
⇑ Corresponding author at: Infectiologie, Centre Hospitalier Universitaire de essary antibiotic prescription constitutes a vicious circle insofar as
Guadeloupe, France.
a significant proportion of patients have a favorable outcome,
E-mail address: pierre-marie.roger@chu-guadeloupe.fr (P.-M. Roger).

https://doi.org/10.1016/j.idnow.2023.104751
2666-9919/Ó 2023 Elsevier Masson SAS. All rights reserved.
P.-M. Roger, O. Keïta-Perse and J.-L. Mainardi Infectious Diseases Now 53 (2023) 104751

thereby bolstering prescriber confidence. Even in patients with note, primary bacteremia represent 10% to 12% of all bacteremia
severe diseases mimicking septic shock, non-infectious inflamma- in modern hospitals [29], and as shown in Table 1, the rates of
tory diseases are frequent (>10%), including those associated with unknown sources of bacteremia, i.e. instances of DU, have not been
adverse effects of drugs, vasculitis, and cancers [13]. Of note, a sim- reduced over the last twenty years.
ilar rate of sepsis of unknown origin has been reported in patients
with cancer [14]. Concerning infectious diseases, as an antibiotic 2.3. Antimicrobial stewardship and diagnostic uncertainty
compound rarely has a low spectrum of activity, it is possible to
observe a favorable outcome despite an incorrect diagnosis. This It is now well-established that the main tools of ASP are audits
could at least partially explain the high frequency of antibiotic with feedback, guidelines, and pre-prescription authorization [2].
combinations observed in daily practice, 30% to 45% of which are Auditing of antibiotic therapy is a key means of DU observation.
not justified [15]. In 2003, we showed that in an emergency ward, elderly patients
The aim of the present review is to raise awareness on the fre- with fever and respiratory signs were most of the time misdiag-
quency of DU, its impact on antibiotic use and patient outcomes, nosed with community-acquired pneumonia rather than bronchi-
and to consider how a nosological framework for DUs could yield tis or cardiac failure with pulmonary involvement, and 50% of
therapeutic improvement. We have focused on in-hospital care, antibiotic treatments were stopped, without any negative impact
as DUs in family practices are beyond the scope of this paper. on outcomes [30]. In another study, involving multiple wards in
various medical specialties, over one third of the patients had been
2. Awareness on the frequency of diagnostic uncertainty misdiagnosed [31]. Of note, it was shown with a before/after
approach that guideline implementation in emergency depart-
Four categories of works have considered DUs: those on the epi- ments led to a reduction of 17% to 25% of unnecessary antibiotic
demiology of sepsis, studies on bacteremia, those focusing on prescriptions, a finding suggesting misdiagnosis [32]. In-depth
antimicrobial stewardship programs (ASPs), and a fourth category, analysis of specific studies showed that ASP in emergency depart-
less common, which includes studies aimed at quantifying DUs and ments required improved diagnosis as the first step toward better
at determining risk factors. As these groups of studies define DU antimicrobial use [20,21].
differently, we shall indicate their results successively.
2.4. Specifically designed studies on diagnostic uncertainty and risk
2.1. Epidemiology of sepsis and diagnostic uncertainty factors

Table 1 lists successive published works over 20 years showing In dedicated studies, the rates of DU were consistently high,
variable rates of DU, ranging from 3.3% to 35%. As stated above, this generally exceeding 30%. In a prospective audit of the antibiotic
large variation in the rate of DUs is at least in partially related to therapies prescribed in a French academic emergency department
variable definitions, the most common of which is the absence of in 2002, the rate of DU, which had been defined by four indepen-
a recognized site of infection [16–19]. In a nationwide USA study dent experts, was 33% [33]. Several years later, other studies using
enrolling more than 6 million patients with sepsis in 2001, 10% comparable methods found similar rates of DU in wards as well as
of the cases were defined as being of unknown origin [16]. Sepsis institutions, suggesting a lack of improvement in clinical practices
of unknown origin was not specifically defined and was finally despite concerted concomitant efforts to develop new biomarkers
interpreted as an undetermined port of entry. Interestingly, this in daily management (see Table 1) [34–38]. In a recent study focus-
significant rate of DU did not generate any comments and maynot ing on urinary tract infections, following admission [only 40% of
even have been septic. The variations in DU rates reported in other the patients treated in the emergency department had a confirmed
epidemiological studies are partly related to the department hav- diagnosis 39].
ing accommodated them. Of note, while the accuracy of diagnosis To date, few studies have sought out risk factors for DU [33,35].
in the emergency room has been questioned in several works, a In one study, inexperienced internists and inadequate interpreta-
review of antimicrobial stewardship tools in emergency wards tion of radiological findings were found in a multivariate analysis
(43 studies) found that measurement of the diagnostic accuracy to lead to diagnostic errors, which were associated with inade-
of infectious diseases had yet to be investigated [20,21]. quate antibiotic therapy (odds ratios of 5.50 and 4.52 respectively)
[33]. In another study, the diagnosis was more likely correct when
the initial condition was related to the reason for hospital admis-
2.2. Bacteremia and diagnostic uncertainty
sion [35]. Finally, in another multivariate analysis, unnecessary
antibiotic therapy was associated with the absence of blood cul-
Bacteremia without a port of entry is category of DU at least
tures (AOR, 5.26) [12].
until the microbial result is available. Accordingly, significant dif-
ferences in the rates of bacteremia without port of entry, i.e. DU,
have been observed in both community-acquired and nosocomial 3. Diagnostic uncertainty occurs in a well-described nosological
infections, ranging from 12.7% to 79.0% [22–28]. Once again, vari- framework
ations in the definition of DU may explain this broad range, and
the classification of catheter-related bacteremia in unknown foci Studies have shown that high fever (38.5 °C) in a suspected
of infection has been assessed in various settings (see Table 1). In upper respiratory infection, or pain acute otitis media are among
a Canadian study, among 480 patients discharged from an emer- the risk factors for antibiotic prescriptions [40]. In contrast,
gency department with bacteremia, higher risks of readmission absence of fever is a risk factor for an unfavorable outcome, as clin-
were associated with ‘‘not yet diagnosed” bacteremia [24]. In a ret- icians do not suspect ongoing infection. In a cohort study of 2605
rospective multicenter study (9 countries, 25 centers) of 767P. bacteremia infections in elderly patients, the absence of fever
aeruginosa bacteremia, more than 70% of which were associated was associated with a worse outcome (OR 1.83) [41]; in adult
with care, DU defined as an unknown source of bacteremia patients, the absence of fever was observed in 23% of 696 cases
accounted for 22.4% [28]. Finally, the concept of primary bac- of meningitis, which was a source of delay in the diagnosis [42].
teremia, defined as bacteremia of unknown origin despite careful Thereby, fever is a cardinal sign of an ongoing infectious disease
clinical assessment and other explorations, characterizes DU. Of and is associated with antibiotic prescriptions, even if the initial
2
 Table 1

P.-M. Roger, O. Keïta-Perse and J.-L. Mainardi

Studies reporting diagnostic uncertainty rates over the past twenty years. Some of the studies included herein date back to the beginning of the century, suggesting that these rates have not been reduced by emerging methods or
technologies.

References Method Setting n patients or antibiotic lines Definition of Rate of DU

Diagnostic Uncertainty (%)
Part 1: Diagnostic uncertainties reported in epidemiological studies of sepsis
Angus et al., Prospective observational study 847 US facilities 6,621,559 Sepsis of 13.0
2001 [16] Unknown Origin
Vincent et al., Prospective cohort observational study 198 ICUs in 24 European 1177 Infection without recognized port of entry 5.0
2006 [17] for 15 days countries
Roger et al., Prospective cohort observational study 1 ID department in a tertiary 12,597 By exclusion of recognized sources of infections and other non- 14.9
2017 [18] care hospital in France infectious diagnosis
Gouel-Cheron et al., Retrospective cohort study Patients for at least 150,948 Absence of site-specific infection diagnosis code 29.0
2022 [19] 3 days in ICU
in 85 US hospitals
Part 2: Diagnostic uncertainties in bacteremia
Pedersen et al., Population-based cohort study Nine community, non-academic Population of Denmark Unknown port of entry of the bacteremia or  2 probable sources 21.0
2003 [22] hospitals in the US
Anderson et al., Retrospective multicenter cohort study Nine community, non-academic 1470 Unknown port of entry of the bacteremia mixed with 79.0
2014 [23] hospitals in the US those due to central venous catheter1
Chan et al., Retrospective single cohort study ID department in a tertiary care 480 Fever as ‘‘not yet diagnosed” 12.7
2015 [24] teaching hospital in Canada
Mehl et al., Prospective cohort study 1 emergency ward in one 1995 None of the criteria for ascertaining a focus From 0.5 to
2017 [25] 2002 to 2013 general hospital in Norway 20.9*
Aillet et al., Retrospective multicenter observational 5 emergency wards in general 169 Undetermined diagnosis (i.e., not written in patient’s chart) 22.4
2018 [26] study hospitals in France
Courjon et al., Cohort study from 2005 to 2018 1 ID department in a tertiary 1034 Discrepancy between diagnosis at admission and final diagnosis at 35
2019 [27] care hospital in France discharge
Babich et al., Retrospective multicenter observational 9 countries, 767 Unknown source of bacteremia 22.4
2020 [28] study 25 centers
3

Part 3: Antibiotic audit and diagnostic uncertainties (numbers indicate antibiotic prescriptions)
Roger et al., Retrospective cohort study 1 emergency ward in a tertiary 169 elderly patients and Diagnostic errors according to ID diagnostic and therapeutic advice 66.0
2003 [30] over 3 months hospital in France respiratory signs
Pulcini et al., Prospective observational study for 8 wards in a teaching hospital in 112 Unnecessary antibiotic treatment due to misdiagnosis 34.0
2007 [31] 9 weeks France
Dinh et al., Single-center 1 emergency department in a 769/580 No written diagnosis of infection recorded in the patient’s file, or 25.6/17.4
2017 [32] before/after study tertiary care hospital in France antibiotics not indicated according to the prescribing guidelines
Roger et al., Prospective multicenter observational 17 private hospitals in France 453 absence of an identified diagnosis of ID in the patient’s electronic 23.0
2019 [5] study report
Part 4: Studies aimed at measuring diagnostic uncertainty
Roger et al., Prospective 1 emergency department in a 117 questionnaires to Diagnostic errors according to four experts from different specialties 33.0
2002 [33] 6-month survey on diagnosis and tertiary hospital in France emergency doctors
antibiotic prescribing after their own prescriptions
Klouwenberg et al., Prospective observational study 2 ICUs 206 Interobserver disagreement 21.0
2013 [34] in the Netherlands
Filice et al., Retrospective cohort study 1 University Hospital in the US 500 Diagnostic accuracy assessed by four ID physicians 31.0

Infectious Diseases Now 53 (2023) 104751

2015 [35]
Caterino et al., Prospective observational study in 1 tertiary care center in the US 424 Under- and overdiagnosis Diagnosis as assessed by two experts 33–27**
2017 [36] elderly patients
Lopansri et al., Prospective multicenter observational 7 ICUs in the US 249 Diagnostic agreement between attending physicians 29.7
2019 [37] study
Coon et al., Retrospective cohort study over two 1 children’s hospital in the US 181 Diagnostic discordance between ED physicians and pediatric 62.0
2020 [38] seasons hospitalists
1
. This definition was inadequate, with a clearly overestimated rate of DU. 2. IDS = infectious disease specialists. * Depending on the bacteria involved in bacteremia ** Underdiagnosis of infection compared to gold-standard: 33%,
and 27% overdiagnosis.
P.-M. Roger, O. Keïta-Perse and J.-L. Mainardi Infectious Diseases Now 53 (2023) 104751

diagnosis is unclear. Therefore, it is no surprise that attendance by involving 26 institutions we included the item ‘‘diagnosis written
ID specialists in medical wards is associated with more diagnostic in the patient’s chart” in our definition of optimal antibiotic ther-
testing (39% vs 82%), reduction of fever of unknown origin (24% vs apy (OAT) [5]. We found that the rates of OAT were low, i.e.,
36%) and of community-acquired febrile syndromes, and with <15%. Considering the many tools of ASP already implemented in
more appropriate antibiotic prescriptions (55% vs 43%) [43]. these institutions at the time of antibiotic audits, the low rates of
When fever and unclear diagnosis are persistent over one week OAT could be surprising, but the item ‘‘identifiable diagnosis”
despite adequate clinical and successive investigations, patients had yet to contribute to OAT definition. Additionally, DU was asso-
are categorized as having fevers of unknown origin (FUO). A recent ciated with the absence of effectively simplified antibiotic
review revealed that 11% to 55% of FUOs were related to an infec- reassessment [5].
tious disease [44]. Accordingly, using a medical table for twelve In clinical practice, antibiotic prescriptions outside guidelines
years in an Infectious Diseases Department in a teaching hospital, are still frequently given by medical and surgical specialists. To
we showed that 14.9% of patients had inflammatory non-infectious increase adherence to guidelines, quality indicators have been
diagnoses, and that 3.6% had self-limiting fever of unknown origin devised, but the initial diagnosis was still not questioned [52]. Of
[18]. Thereby, the creation of a framework for diagnostic uncer- note, we have reported that accompanied self-antibiotic reassess-
tainty in infectious diseases, before FUO categorization, will be a ment (ASAR), i.e., adversarial discussions with the prescribers with
pathway for care improvement, including better antibiotic use emphasis on diagnostic accuracy regarding their patients and their
(see below). antibiotic prescriptions, was associated with higher adherence to
guidelines, ranging from 60% to 71% among 1221 antibiotic pre-
scriptions from 51 senior doctors [53]. In fact, ASP encompasses
4. Do biomarkers help in case of diagnostic uncertainty?
a bundle of interventions, at best as structured clinical pathways
for most common infections and their differential diagnoses, which
There have been numerous studies on biomarkers in the man-
should be developed in accordance with local or regional human
agement of infectious diseases, illustrating both the aims in terms
and technical resources. Finally, the significant antibiotic overuse
of clinical improvement, and failure to achieve sought-after
and misuse associated with DU is a major reason to create the
improvement. There currently exists evidence of the usefulness
nosological framework ‘‘Diagnostic uncertainty in infectious dis-
of biomarkers such as procalcitonin (PCT) to evaluate the prognosis
eases”, as it could be associated with better therapeutic use.
of infectious diseases and to guide antibiotic treatment duration
[45]. Similarly, numerous studies have shown that C-reactive pro-
tein (CRP) and PCT help to distinguish bacterial from non-bacterial 6. Impact of diagnostic uncertainty on outcome
infections, even though the diagnostic accuracy of these markers
varies substantially (see Fig. 1A), while sensitivity and specificity To the best of our knowledge, while information has been
depend on threshold values [46]. However, to the best of our extracted from the studies listed above, no studies to date have
knowledge, no study to date has reported improved diagnostic specifically focused on this subject. In a recent audit conducted
accuracy in the field of infectious diseases using such biomarkers in 26 institutions, optimal antibiotic therapy (OAT), defined as an
[47]. As shown in Fig. 1B, most biomarkers fail to suspect bac- established diagnosis and appropriate antibiotic use, including
teremia in the case of a patient with fever but without any port empirical treatment and reassessment, was associated in a multi-
of entry. variate analysis with less adverse outcomes compared to non-
Regarding clinical follow-up after antibiotic prescriptions, we optimal antibiotic therapy (2 vs. 11%, adjusted odds ratio = 0.17
have reported that the kinetics of fever provided similar informa- [0.06–0.49]) [5]. In another study aiming to determine the reasons
tion as CRP in 83% of 392 infections [48]. Importantly, we had that had led to unnecessary antibiotic therapies, the univariate
selected infected patients who did not require surgical treatment, analysis showed them to be strongly associated with the absence
as they were associated with biomarker modifications. Of note, of diagnosis (p < 0.001) and with unfavorable outcomes
without the use of any biomarkers, more than 20% of the antibiotic (p < 0.001) [12]. In a multivariate Cox model by Contou et al., inten-
treatments in two intensive care units were stopped without neg- sive care unit patients with a shock of unknown origin had a higher
ative impact on the patients (mean SAPS II score 43.5) [49]. These risk of mortality compared to those for whom the type of shock
results were obtained through constructive discussion on diagnos- had been established (hazard ratio = 1.75 [1.07–2.88]) [13]. In a
tic accuracy between ID and intensive care bedside specialists, large-scale European study involving 198 intensive care units, pri-
with 33% of disagreement. Indeed, the question ‘‘biomarkers, mary bacteremia, i.e., without a recognized port of entry, was asso-
friends or foes to clinicians?” is regularly put forward in the liter- ciated with a worse outcome compared to other diagnoses (odds
ature [47]. ratio 2.0 [1.4–2.6]) [17]. In a Danish population-based study on
bacteremia, an undetermined source of infection was associated
with the highest case fatality rate (OR 3.2 [2.2–4.7]) [22]. In a study
5. Diagnostic uncertainty and antimicrobial misuse
of bacteremia in five emergency wards in France including 169
patients, the absence of a diagnosis in the patient’s chart was the
Several studies reported herein showed the link between DU
highest risk factor for an unfavorable outcome (adjusted
and inadequate use of antimicrobials. In a study from 131 US hos-
OR = 9.34 [2.21–39.48]) [26]. However, other studies did not report
pitals including 26 036 bacteremia infections, 17% of the patients
statistically significant associations between DUs and unfavorable
did not receive antimicrobials immediately after admission (a form
outcomes [24,25], or they did not search for such a relationship
of misuse), and 19% received inappropriate treatment due to resis-
[28,29,31,32]. All in all, a large number of studies have indicated
tant strains [50]. The frequently unknown source of bacteremia
a more unfavorable prognosis in case of DU.
(39.2%) explained the high rates of antibiotic misuse. In another
study from a teaching hospital in France, 50% of the patients with
bacteremia had inadequate antimicrobial prescriptions [51]. 7. Possible definitions of diagnostic uncertainty to replace
While several recommendations have focused on how to orga- imprecise terminology
nize ASP and which tools should be used, they did not focus on
how to deal with DU [1,2]. Considering the importance of diagnosis Despite improved sepsis definitions over recent decades,
in antibiotic prescription, in a multicenter prospective study clinicians still encounter difficulties in identifying really infected
4
P.-M. Roger, O. Keïta-Perse and J.-L. Mainardi Infectious Diseases Now 53 (2023) 104751

Fig. 1. Biomarkers do not help to reduce diagnostic uncertainties. Considerable overlapping values limited the diagnostic pertinence of procalcitonin (PCT) and C-reactive
protein (CRP) 1.A. Lower respiratory tract infections (n = 330) and biomarkers. Dedicated unit (24 beds) for elderly patients, mean age (std dev) 83 ± 8 years, with flu-like
syndrome during the winter period from 2000 to 2005. Diagnoses at discharge were community-acquired pneumonia (n = 154, 47%), bronchitis (n = 151, 45%), or chronic
obstructive pulmonary disease exacerbation (n = 25, 8%). Proven microbial diseases were defined by positive laboratory analysis (either sputum, blood cultures, urinary
antigens, or serology). Patients with presumed viral infections received no antimicrobials, while those with presumed bacterial infections did receive an antibiotic therapy. C-
reactive protein (CRP) was determined at hospital admission and leucocytes were determined at the beginning of the first day after admission, early in the morning before
breakfast, walking, or smoking to avoid physiological variations. The rate of patients with biomarker values suggesting viral infections was 34%. Unpublished data. 1.B.
Primary bacteremia (n = 96) and biomarkers. Primary bacteremia was defined by an unknown origin despite clinical assessment and other explorations performed in a
specialized ward. The rate of patients with normal values for both biomarkers was 22%. Unpublished data from [29].

patients [30,31,39,41,47,49–54]. In daily practice, most physicians definitions as sepsis in 42% and as septic shock in 14%; the other
in hospitals refer to sepsis or septic shock, without considering a clinical presentations were not specified, suggesting that the term
more precise clinical diagnosis, which could lead to more accurate ‘‘sepsis” did not adequately characterize the other patients [50].
identification of the pathogens involved, anda better choice of The sepsis concept could be used if it led to similar levels of quality
antimicrobials. In a North American multicenter study on management whatever the infectious diseases under considera-
bacteremia, patients were described in accordance with Sepsis-3 tion, but it does not. For example, antibiotic reassessment is easier
5
P.-M. Roger, O. Keïta-Perse and J.-L. Mainardi Infectious Diseases Now 53 (2023) 104751

in UTI compared to CAP, with microbiological data often failing to To combat this unfounded fear, and given that guidelines
discover an etiologic agent for the latter. Moreover, as sepsis is a should lead to better care, we need to establish a definition of
syndrome encompassing numerous entities, it frequently yields DU through which it would evolve from a subliminal notion to
antibiotic prescription, even in the absence of any sign of severity. an established concept. In the current literature, DU goes by a mul-
In real life, the term sepsis is too generic, and when non- titude of names, which preclude correct understanding of the mag-
specialized doctors encounter cases of sepsis-like syndrome, a nitude of the problem (see Tables). Therefore, application of the
vicious circle sets in with the prescription of broad-spectrum PRISMA method to review DU is inefficient insofar as the concept
antibiotic combinations in a global environment characterized by and associated terms are not consensual.
increasing prevalence of MDR-bacteria [12–14,49,54]. Accordingly, Several advantages are expected for ‘‘diagnostic uncertainty”
in an observational cohort study in six US hospitals, 30% of the patients: the designation will compel the medical team to work
antibiotic therapies were prescribed to afebrile patients with nor- together, to remain in on alert, and to follow clinical progression
mal white cell counts [55]; by Day 5, 66% of them had not under- and microbial results, potentially with improved antibiotic
gone reassessment of their broad-spectrum empirical therapy. reassessment. In other words, a consensual definition of DU would
Prescription of unnecessary antibiotics or lack of antimicrobial help to maintain professional relationships, strengthening interdis-
reassessment, is the rule insofar as the term ‘‘sepsis” is overused, ciplinary discussions and possibly modifying physician behavior.
particularly in emergency, intensive care units and surgical depart- Furthermore, recognition of DUs will be conducive to their epi-
ments [12,13,49,53,54]. In this context, fear, rather than substan- demiological description and related microbial results. In a previ-
dard knowledge, could be the main obstacle to prevention of ous study [27], we defined DU as a discrepancy between the
antibiotic overuse. Accordingly, one could conclude that physician diagnosis suspected at admission and the final diagnosis at dis-
doubt, i.e. DU, is not considered in the traditional model of knowl- charge from an ID ward, with a focus on patients with
edge in hospitals, which focuses on monitoring and regulatory community-acquired bacteremia. The spectrum of responsible bac-
approaches [2]. teria was different from those in patients without DU, with more
polymicrobial bacteremia and more strains resistant to third- gen-
eration cephalosporins as compared to amoxicillin/clavulanic acid.
Table 2
Attempts to define diagnostic uncertainty (DU). In the current literature, DUs are Thus, defining Dus Precise definition of DU could help to optimize
assigned different names: inter-observer disagreement, diagnostic discordance, empirical antibiotic therapy and its reassessment, which is all but
diagnostic inaccuracy, misdiagnosis, or names reflecting expected consequences: impossible without specific studies. As for FUO, the definition
indiscriminate antibiotic prescription, antibiotic misuse or antibiotic overuse. Blood could be time-dependent, in conjunction with laboratory and/or
cultures should always be performed when DU is perceived.
radiological findings [44]. Table 2 indicates the advantages and
Definition (1) No diagnostic hypothesis despite limitations of attempts at such definitions, all of them being
After admission thorough analysis of medical history, time-dependent in accordance with the availability of biological
symptoms, vital signs, clinical
examination findings, and after a
and/or radiological results. Before further studies, the consensual
second medical opinion or more than definition of DU and its therapeutic consequence will need to be
two diagnoses after this careful tested in real-life situations.
analysis A consensual definition of DU should improve all therapeutic
Advantages Limitations
guidelines by means of the quantification associated with each
- no resource consumption - too imprecise and thus too many
- easy to reproduce patients concerned specific focus of infection, with subsequent details on the other
- reduction of diagnostic errors* - does not include microbiological diagnoses accounting for DU.
analysis
Definition (2) No diagnostic hypothesis after
After a first round of biological thorough analysis of medical history,
and radiological means symptoms, vital signs, clinical
8. Conclusion
examination findings, plus basic
laboratory and radiological results, Over the past twenty years, DU rates have remained suggesting
excluding CT scan that concomitant improvement in microbial and radiological tech-
Advantages Limitations
niques has been associated with limited benefits in clinical prac-
- in line with in-hospital care - resource consumption
- inclusion of rapid - time-dependence tice. Combinations of microbial and inflammatory biomarkers in
microbiological tests point-of-care equipment could potentially help physicians.
- without the presence of vital During the same period, ASP has been implemented, including
signs and without continuing medical education, but DU is still insufficiently
agranulocytosis, no antimicrobials
necessary
acknowledged [1,2]. Moreover, the frequent use of imprecise ter-
Definition (3) No diagnostic hypothesis after minology such as sepsis, for which the definition has varied over
By the end of Day-2 thorough analysis of medical history, time, limits the use of a proper lexicon [11–14,17,54]. Accordingly,
symptoms, vital signs, clinical through a systematic review of the literature designed to deter-
examination findings, plus non-
mine what could be a responsible use of antibiotics, the authors
contributory laboratory results
including sterile urine and blood found 22 key elements to be included in the definition, but clinical
cultures and radiological results diagnosis and uncertainties were not considered [56]. If DU
including CT scan remains unclassified, it will limit the development of research in
Advantages Limitations this field and patient care. Of note, a recent review on machine
- information enrichment based - resource consumption
learning in infectious diseases indicated that only 33% (20/60) of
on medical history, physical - ‘‘long-term” requirement
examination, and previous the support systems included diagnosis of infection [57]. More-
therapeutic means over, less than half of these diagnoses included vital signs, symp-
- without the presence of vital toms, and physical examination findings in the process, while
signs, no antimicrobials necessary
other major parameters (outcomes, effects of drugs. . .) were nearly
invisible. All things considered, a consensual definition of DU
*As reported for community-acquired pneumonia [40], feeling constrained to make should be established so as to reduce the complexity of hospital-
a rapid diagnosis may be associated with more diagnostic errors.
based ASP interventions, and related microbial results, therapeutic
6
P.-M. Roger, O. Keïta-Perse and J.-L. Mainardi Infectious Diseases Now 53 (2023) 104751

means, and outcomes should be examined to improve care for [17] Vincent JL, Sakr Y, Sprung CL, Ranieri VM, Reinhart K, Gerlach H, et al. Sepsis in
European intensive care units: results of the SOAP study. Crit Care Med
numerous patients. With this perspective, ASP metrics will come
2006;34:344–53. https://doi.org/10.1097/01.ccm.0000194725.48928.3a.
to conclude a fifth category, which would be the first in clinical [18] Roger PM, Demonchy E, Risso K, Courjon J, Leroux S, Leroux E, et al. Medical
practice: quality of the diagnosis, antimicrobial utilization and table: A major tool for antimicrobial stewardship policy. Med Mal Infect
costs, unintended consequences, patient outcomes, and process 2017;47:311–8. https://doi.org/10.1016/j.medmal.2017.03.005.
[19] Gouel-Cheron A, Swihart BJ, Warner S, Mathew L, Strich JR, Mancera A, et al.
appraisal [58]. Epidemiology of ICU-onset bloodstream infection: prevalence, pathogens, and
risk factors among 150,948 ICU patients at 85 U.S. hospitals. Crit Care Med
2022;50:1725–36. https://doi.org/10.1097/CCM.0000000000005662.
Author contributions [20] Pulia M, Redwood R. Empiric antibiotic prescribing for suspected sepsis: a
stewardship balancing Act. Am J Med Sci 2020;360:613–4. https://doi.org/
10.1016/j.amjms.2020.08.030.
All authors contributed significantly to the review, and all have [21] Losier M, Ramsey TD, Wilby KJ, Black EK. A systematic review of antimicrobial
read and consented to the article. Pierre-Marie Roger contributed stewardship interventions in the emergency department. Annal Pharmacother
2017;51:774–90. https://doi.org/10.1177/10600 28017709820.
to the design and wrote the article. Olivia Keïta-Perse and Jean-
[22] Pedersen G, Schønheyder HC, Sørensen HT. Source of infection and other
Luc Mainardi contributed to the writing of the article. factors associated with case fatality in community-acquired bacteremia-a
Danish population-based cohort study from 1992 to 1997. Clin Microbiol
Infect 2003;9:793–802. https://doi.org/10.1046/j.1469-0691.2003.00599.x.
References [23] Anderson DJ, Moehring RW, Sloane R, Schmader KE, Weber DJ, Fowler Jr VG,
et al. Bloodstream infections in community hospitals in the 21st century: a
[1] Dellit TH, Owens RC, McGowan JE, Gerding DN, Weinstein RA, Burke JP, et al. multicenter cohort study. PLoS One 2014;9:e91713.
Infectious diseases society of America and the society for healthcare [24] Chan J, Wong J, Saginur R, Forster AJ, van Walraven C. Epidemiology and
epidemiology of America guidelines for developing an institutional program outcomes of bloodstream infections in patients discharged from the
to enhance antimicrobial stewardship. Clin Infect Dis 2007;44:159–77. https:// emergency department. CJEM 2015;17:27–37. https://doi.org/10.2310/
doi.org/10.1086/510393. 8000.2013.131349.
[2] Mendelson M, Morris AM, Thursky K, Pulcini C. How to start an antimicrobial [25] Mehl A, Åsvold BO, Kümmel A, Lydersen S, Paulsen J, Haugan I, et al. Trends in
stewardship programme in a hospital. Clin Microbiol Infect 2020;26:447–53. antimicrobial resistance and empiric antibiotic therapy of bloodstream
https://doi.org/10.1016/j.cmi.2019.08.007. infections at a general hospital in Mid-Norway: a prospective observational
[3] White Jr AC, Atmar RL, Wilson J, Cate TR, Stager CE, Greenberg SB. Effects of study. BMC Infect Dis 2017;17:. https://doi.org/10.1186/s12879-017-2210-
requiring prior authorization for selected antimicrobials: expenditures, 6116.
susceptibilities, and clinical outcomes. Clin Infect Dis 1997;25:230–9. [26] Aillet C, Jammes D, Fribourg A, Léotard S, Pellat O, Etienne P, et al. Bacteraemia
https://doi.org/10.1086/514545. PMID: 9332517. in emergency departments: effective antibiotic reassessment is associated
[4] Grohs P, Kernéis S, Sabatier B, Lavollay M, Carbonnelle E, Rostane H, et al. with a better outcome. Eur J Clin Microbiol Infect Dis 2018;37:325–31. https://
Fighting the spread of AmpC-hyperproducing Enterobacteriaceae: beneficial doi.org/10.1007/s10096-017-3136-z.
effect of replacing ceftriaxone with cefotaxime. J Antimicrob Chemother [27] Courjon J, Chirio D, Demonchy E, Michelangeli C, Degand N, Roger PM.
2014;69:786–9. https://doi.org/10.1093/jac/dkt403. Amoxicillin/clavulanic acid + aminoglycoside as empirical antibiotic treatment
[5] Roger PM, Espinet A, Ravily D, Meyer MJ, Moll F, Montera E, et al. Simplified in severe community-acquired infections with diagnostic uncertainty. Eur J
therapeutic guidelines: the main tool of antimicrobial stewardship programs Clin Microb Infect Dis 2019;38:895–901. https://doi.org/10.1007/s10096-019-
associated with optimal antibiotic therapy. Eur J Clin Microbiol Infect Dis 03496-0.
2022;41:375–83. https://doi.org/10.1007/s10096-022-04406-7. [28] Babich T, Naucler P, Valik JK, Giske CG, Benito N, Cardona R, et al. Risk factors
[6] Davey P, Marwick CA, Scott CL, Charani E, McNeil K, Brown E, et al. for mortality among patients with Pseudomonas aeruginosa bacteraemia: a
Interventions to improve antibiotic prescribing practices for hospital retrospective multicentre study. Int J Antimicrob Agents 2020;55:. https://doi.
inpatients. Cochrane Database Syst Rev 2017(2):. https://doi.org/ org/10.1016/j.ijantimicag.2019.11.004105847.
10.1093/jacamr/dlac127CD003543. [29] Courjon J, Demonchy E, Degand N, Risso K, Ruimy R, Roger PM. Patients with
[7] Mandell LA, Wunderink RG, Anzueto A, Bartlett JG, Campbell GD, Dean NC, community-acquired bacteremia of unknown origin: clinical characteristics
et al. Infectious Diseases Society of America/American Thoracic Society and usefulness of microbiological results for therapeutic issues: a single-
consensus guidelines on the management of community-acquired center cohort study. Ann Clin Microbiol Antimicrob 2017;16:. https://doi.org/
pneumonia in adults. Clin Infect Dis 2007;44(Suppl. 2):S27–72. https://doi. 10.1186/s12941-017-0214-040.
org/10.1086/511159. [30] Roger PM, Farhad R, Pulcini C, Mariette A, Taurel M, Oualid H, et al. Elderly
[8] Nicolle LE, Gupta K, Bradley SF, Colgan R, DeMuri GP, Drekonja D, et al. Clinical patients presenting with fever and respiratory problems in an intensive care
practice guideline for the management of asymptomatic bacteriuria: 2019 unit. Diagnostic, therapeutic and prognostic impact of a systematic infectious
Update by the Infectious Diseases Society of America. Clin Infect Dis 2019;68: disease consultation. Presse Med 2003;32:1699–704.
e83–e175. https://doi.org/10.1093/cid/ciy1121. [31] Pulcini C, Cua E, Lieutier F, Landraud L, Dellamonica P, Roger PM. Antibiotic
[9] US Food and Drug Administration. Acute bacterial skin and skin structure misuse: a prospective clinical audit in a French university hospital. Eur J Clin
infections: developing drugs for treatment. Guidance for industry. 2013. Microbiol Infect Dis 2007;26:277–80. https://doi.org/10.1007/s10096-007-
https://www.fda.gov/downloads/Drugs/Guidance/UCM071185.pdf (accessed 0277-5.
12 Sep 2022). [32] Dinh A, Duran C, Davido B, Bouchand F, Deconinck L, Matt M, et al. Impact of an
[10] Bruna JL, Castan B, de Barbeyrac B, Cazanave C, Charvériat A, Faure K, et al. antimicrobial stewardship programme to optimize antimicrobial use for
Pelvic inflammatory diseases: updated french guidelines. J Gynecol Obstet outpatients at an emergency department. J Hosp Infect 2017;97:288–93.
Hum Reprod 2020;49:. https://doi.org/10.1016/j.jogoh.2020.101714101714. https://doi.org/10.1016/j.jhin.2017.07.005.
[11] Kwok M, McGeorge S, Mayer-Coverdale J, Graves B, Paterson DL, Harris PNA, [33] Roger PM, Martin C, Taurel M, Fournier JP, Nicole I, Carles M, et al. Motives for
et al. Guideline of guidelines: management of recurrent urinary tract the prescription of antibiotics in the emergency department of the University
infections in women. BJU Int 2022;130(Suppl 3):11–22. https://doi.org/ Hospital Center in Nice a prospective study. Press Med 2002;31:58–63.
10.1111/bju.15756. [34] Klein Klouwenberg PM, Ong DS, Bos LD, de Beer FM, van Hooijdonk RT, Huson
[12] Roger PM, Montera E, Lesselingue D, Troadec N, Charlot P, Simand A, et al. Risk MA, et al. Interobserver agreement of Centers for Disease Control and
factors for unnecessary antibiotic therapy: a major role for clinical management. Prevention criteria for classifying infections in critically ill patients. Crit Care
Clin Infect Dis 2019;69:466–72. https://doi.org/10.1093/cid/ciy921. Med 2013;41:2373–8. https://doi.org/10.1097/CCM.0b013e3182923712.
[13] Contou D, Roux D, Jochmans S, Coudroy R, Guérot E, Grimaldi D, et al. Septic [35] Filice GA, Drekonja DM, Thurn JR, Hamann GM, Masoud BT, Johnson JR.
shock with no diagnosis at 24 hours: a pragmatic multicenter prospective Diagnostic errors that lead to inappropriate antimicrobial use. Infect Control
cohort study. Crit Care 2016;20:. https://doi.org/10.1186/s13054-016-1537- Hosp Epidemiol 2015;36:949–56. https://doi.org/10.1017/ice.2015.113.
5360. [36] Caterino JM, Leininger R, Kline DM, Southerland LT, Khaliqdina S, Baugh CW,
[14] Gudiol C, Albasanz-Puig A, Cuervo G, Carratalà J. Understanding and managing et al. Accuracy of current diagnostic criteria for acute bacterial infection in
sepsis in patients with cancer in the era of antimicrobial resistance. Front Med older emergency department patients. J Am Geriatr Soc 2017;65:1802–9.
2021;8:. https://doi.org/10.3389/fmed.2021.636547636547. https://doi.org/10.1111/jgs.14912.
[15] Roger PM, Brofferio P, Labate C, Barrière JR, Minguet JM, Foulon P, et al. [37] Lopansri BK, Miller Iii RR, Burke JP, Levy M, Opal S, Rothman RE, et al. Physician
Prospective evaluation of antibiotic combinations in a general hospital. Med agreement on the diagnosis of sepsis in the intensive care unit: estimation of
Mal Infect 2010;40:165–71. https://doi.org/10.1016/j.medmal.2009.08.019. concordance and analysis of underlying factors in a multicenter cohort. J
[16] Angus DC, Linde-Zwirble WT, Lidicker J, Clermont G, Carcillo J, Pinsky MR. Intensive Care 2019;7:. https://doi.org/10.1186/s40560-019-0368-213.
Epidemiology of severe sepsis in the United States: analysis of incidence, [38] Coon ER, Maloney CG, Shen MW. Antibiotic and diagnostic discordance
outcome, and associated costs of care. Crit Care Med 2001;29:1303–10. between ED physicians and hospitalists for pediatric respiratory illness. Hosp
https://doi.org/10.1097/00003246-200107000-00002. Pediatr 2015;5:111–8. https://doi.org/10.1542/hpeds.2014-0110.

7
P.-M. Roger, O. Keïta-Perse and J.-L. Mainardi Infectious Diseases Now 53 (2023) 104751

[39] Shallcross LJ, Rockenschaub P, Mc Nulty D, Freemantle N, Hayward A, Gill MJ. intensive care units. J Antimicrob Chemother 2006;57:546–50. https://doi.
Diagnostic uncertainty and urinary tract infection in the emergency org/10.1093/jac/dki483.
department: a cohort study from a UK hospital. BMC Emerg Med 2020;20:. [50] Kadri SS, Lai YL, Warner S, Strich JR, Babiker A, Ricotta EE, et al. Inappropriate
https://doi.org/10.1186/s12873-020-00333-y40. empiric antibiotic therapy in bloodstream infections at U.S. hospitals based on
[40] Donnelly JP, Baddley JW, Wang HE. Antibiotic utilization for acute respiratory discordant in vitro susceptibilities: a retrospective cohort analysis of
tract infections in U.S. emergency departments. Antimicrob Agent Chemother prevalence, predictors and mortality risk. Lancet Infect Dis 2021;21:241–51.
2014;58:1451–7. https://doi.org/10.1128/AAC.02039-13. https://doi.org/10.1016/S1473-3099(20)30477-1.
[41] Hernández C, Fehér C, Soriano A, Marco F, Almela M, Cobos-Trigueros N, et al. [51] Chirio D, Demonchy E, Le Marechal M, Gaudart A, Lotte R, Carles M, et al. 24/7
Clinical characteristics and outcome of elderly patients with community-onset workflow for bloodstream infection diagnostics in microbiology laboratories:
bacteremia. J Infect 2015;70:135–43. https://doi.org/10.1016/j. the first step to improve clinical management. Clin Chem Lab Med
jinf.2014.09.002. 2022;61:349–55. https://doi.org/10.1515/cclm-2022-0667.
[42] van de Beek D, de Gans J, Spanjaard L, Weisfelt M, Reitsma JB, Vermeulen M. [52] Berrevoets MAH, Ten Oever J, Hoogerwerf J, Kullberg BJ, Atsma F, Hulscher ME,
Clinical features and prognostic factors in adults with bacterial meningitis. N et al. Appropriate empirical antibiotic use in the emergency department: full
Engl J Med 2004;351:1849–59. https://doi.org/10.1056/NEJMoa040845. compliance matters! JAC Antimicrob Resist 2019;1:. https://doi.org/
[43] Borer A, Gilad J, Meydan N, Schlaeffer P, Riesenberg K, Schlaeffer F. Impact of 10.1093/jacamr/dlz061dlz061.
regular attendance by infectious disease specialists on the management of [53] Roger PM, Peyraud I, Vitris M, Romain V, Bestman L, Blondel L, et al. Impact of
hospitalised adults with community-acquired febrile syndromes. Clin simplified therapeutic guidelines on antibiotic prescriptions: a prospective
Microbiol Infect 2004;10:911–6. https://doi.org/10.1111/j.1469- multicentre comparative study. J Antimicrob Chemother 2020;75:747–1745.
0691.2004.00964.x. https://doi.org/10.1093/jac/dkz490.
[44] Wright WF, Auwaerter PG. Fever and fever of unknown origin: review, recent [54] Donnelly JP, Safford MM, Shapiro NI, Baddley JW, Wang HE. Application of the
advances, and lingering dogma. Open Forum. Infect Dis 2020;7:. https://doi. third international consensus definitions for Sepsis (Sepsis-3) classifications in
org/10.1093/ofid/ofaa132ofaa132. a population-based cohort: the reasons for geographic and racial differences in
[45] Wirz Y, Meier MA, Bouadma L, Luyt CE, Wolff M, Chastre J, et al. Effect of stroke study. Lancet Infect Dis 2017;17:661–70. https://doi.org/10.1016/
procalcitonin-guided antibiotic treatment on clinical outcomes in intensive S1473-3099(17)30117-2.
care unit patients with infection and sepsis patients: a patient-level meta- [55] Braykov NP, Morgan DJ, Schweizer ML, Uslan DZ, Kelesidis T, Weisenberg SA,
analysis of randomized trials. Crit Care 2018;22:. https://doi.org/10.1186/ et al. Assessment of empirical antibiotic therapy optimisation in six hospitals:
s13054-018-2125-7191. an observational cohort study. Lancet Infect Dis 2014;14:1220–7. https://doi.
[46] Kapasi AJ, Dittrich S, González IJ, Rodwell TC. Host biomarkers for org/10.1016/S1473-3099(14)70952-1.
distinguishing bacterial from non-bacterial causes of acute febrile illness: a [56] Monnier AA, Eisenstein BI, Hulscher ME, Gyssens IC, DRIVE-AB WP1 group..
comprehensive review. PLoS ONE 2016;11(8):e0160278. Towards a global definition of responsible antibiotic use: results of an
[47] Schuetza P, Aujeskyb D, Muellerc C, Muelle B. Biomarker-guided personalised international multidisciplinary consensus procedure. J Antimicrob
emergency medicine for all – hope for another hype? Swiss Med Wkly 2015:. Chemother 2018;73(Suppl 6):vi3–vi16. https://doi.org/10.1093/jac/dky114.
https://doi.org/10.4414/smw.2015.14079145. [57] Peiffer-Smadja N, Rawson TM, Ahmad R, Buchard A, Georgiou P, Lescure FX,
[48] Roger PM, Hung S, de Salvador F, Allieri-Rosenthal A, Farhad R, Pulcini C, et al. et al. Machine learning for clinical decision support in infectious diseases: a
Usefulness of C-reactive protein in the therapeutic follow-up of infected narrative review of current applications. Clin Microbiol Infect
patients. Med Mal Infect 2009;39:319–24. https://doi.org/10.1016/j. 2020;26:584–95. https://doi.org/10.1016/j.cmi.2019.09.009.
medmal.2009.03.001. [58] Dodds Ashley ES, Kaye KS, DePestel DD, Hermsen ED. Antimicrobial
[49] Pulcini C, Pradier C, Samat-Long C, Hyvernat H, Bernardin G, Ichai C, et al. stewardship: philosophy versus practice. Clin Infect Dis 2014;59(Suppl. 3):
Factors associated with adherence to infectious diseases advice in two S112–21. https://doi.org/10.1093/cid/ciu546.