Diagnostic error in the pediatric hospital: a narrative review

Abstract

Diagnostic error is a prevalent type of medical error that is associated with considerable patient harm and increased medical costs. The majority of literature guiding the current understanding of diagnostic error in the hospital setting is from adult studies. However, there is research to suggest this type of error is also prevalent in the pediatric specialty. Thus, we set out to define the current understanding of diagnostic error in the pediatric hospital through a structured literature review. We searched PubMed and identified studies focusing on three aspects of diagnostic error in pediatric hospitals: the incidence or prevalence, contributing factors, and related interventions. We used a tiered review, screened 59 abstracts, and included 23 full-text studies in the final review. Seventeen of the 23 studies focused on the incidence or prevalence, with only 3 studies investigating the utility of interventions. Most studies took place in an intensive care unit or emergency department with very few studies including only patients on the general wards. Overall, the prevalence of diagnostic error in pediatric hospitals varied greatly and depended on the measurement technique and specific hospital setting. Both healthcare system factors and individual cognitive factors were found to contribute to diagnostic error, with there being limited evidence to guide how best to mitigate the influence of these factors on the diagnostic process. The general knowledge of diagnostic errors in pediatric hospital settings is limited. Future work should incorporate structured frameworks to measure diagnostic errors and examine clinicians’ diagnostic processes in real-time to help guide effective hospital-wide interventions.

Introduction

The National Academy of Medicine (NAM) defines diagnostic error as the failure to (a) establish an accurate and timely explanation of a patient’s health problem(s) or (b) communicate that explanation to the patient [1]. Over the last several years, diagnostic errors have become a national priority in patient safety research [2–6]. Despite an encompassing definition and a call from national organizations to prioritize research in this domain, diagnostic error remains an understudied field [5,7,8].

From adult studies, each measuring diagnostic error slightly differently, it is estimated that this type of error occurs in 10% to 20% of patient encounters [9–11] and is associated with considerable patient harm and increased medical costs [1,9,12]. While most studies investigating diagnostic error have focused on adult patients, there is evidence that this type of error also occurs in the pediatric specialty. Similar to adult studies, diagnostic error is the most prevalent misadventure in pediatric malpractice claims and is associated with the highest median indemnity payment [13]. Studies looking at the prevalence of diagnostic error in pediatrics have found that 15% of pediatricians’ self-identified making at least one diagnostic error each month [14], and 45% identified making a diagnostic error that harmed patients at least once per year [15].

There is less information related to diagnostic error in pediatric hospital settings, but the limited evidence base suggests pediatric hospitals are not immune to this type of error [16]. Pediatric hospitals should be a focus for future study as diagnostic errors in the hospital are associated with increased patient harm compared to outpatient settings [17]. Prior reviews on diagnostic error in the pediatric hospital have focused on single locations, such as the pediatric intensive care unit (PICU), neonatal intensive care unit (NICU), or cardiac intensive care unit (CICU) [18–20]. However, to help guide hospital-wide safety initiatives, which in general have been associated with improved patient outcomes [21–23], there is a need to summarize the current understanding of diagnostic error in all pediatric hospital locations. Thus, we set out to accomplish three objectives: (1) identify mechanisms used to measure diagnostic error in the pediatric hospital and define the incidence and/or prevalence of diagnostic error in this setting, (2) determine what factors contribute to diagnostic error in pediatric hospital settings and (3) identify interventions that have been used to prevent, mitigate or reduce diagnostic error in the pediatric hospital. Through a structured literature review, we aimed to describe the current evidence base, identify gaps in knowledge, and apprise future research directions to help better understand, measure, and mitigate diagnostic errors in the pediatric hospital.

Methods

Search Strategy

We performed a structured review to summarize the current literature on diagnostic errors in pediatric hospital settings. We defined the hospital settings as general wards, emergency department (ED), pediatric intensive care unit (PICU), pediatric cardiac intensive care unit (CICU), and neonatal intensive care unit (NICU). With the help of a librarian with a Master’s degree in Information Science and trained in conducting literature searches and systematic reviews, we performed a focused search of the US National Library of Medicine’s PubMed database to identify candidate articles that focused on diagnostic error in the pediatric hospital. The initial search included the MESH terms “Hospitals, Pediatric”, “Intensive Care Units, Pediatric”, “Intensive Care, Neonatal”, “Pediatric Emergency Medicine”, and “Diagnostic Errors”. The complete list of search terms entered into the PubMed database is listed in the Supplemental Appendix. We limited our search to English-language articles with full abstracts published after the year 2000 to ensure the review was capturing only the most recent and relevant literature.

Selection Strategy

Given the heterogeneity of the definition of diagnostic error in the literature, we did not limit our article selection to any one definition. We included articles that focused on three distinct aspects of diagnostic error: (1) the incidence or prevalence, including methods to measure diagnostic error, (2) factors that contribute to diagnostic error, and (3) interventions to mitigate, reduce, or prevent diagnostic error. We determined that interventions could include actual interventions implemented and tested in practice or theoretical interventions that could be beneficial but have not yet been implemented or studied empirically. Based on the authors’ consensus, we categorized factors contributing to diagnostic error as either cognitive factors, defined as factors related to an individual clinician’s diagnostic process such as clinical reasoning processes or knowledge gaps, or healthcare system factors, defined as those factors relating to any of the following categories: healthcare organization characteristics, technologies and tools, physical environment, or tasks of the healthcare team. Similar to a previous review article focusing on diagnostic error [7], we excluded studies that focused on inter-rater or observer variation, the validation of instruments or tests, single case reports, assessments limited to provider satisfaction, preference, or acceptance of interventions, techniques to enhance diagnosis involving asymptomatic screening instruments, specific tests or technologies, development of risk models, or those studies that took place in an outpatient or clinic setting. We also excluded studies focused on costs and consequences of diagnostic error to be consistent with prior reviews concentrating on the epidemiology of diagnostic error and related interventions [7,19,20,24].

One author (JS) reviewed study titles and determined whether or not to include them for further abstract review. Four authors (JS, BG, DN, and RP) then reviewed the full abstracts and determined which studies to include in the full article review. Reviewers categorized abstracts as “included”, “excluded”, or “unsure”. All abstracts categorized as unsure were reviewed by all study team members, with a team consensus determining which to include.

Data Extraction

We reviewed all full articles whose abstracts were determined to meet inclusion criteria and used structured REDCap (Research Electronic Data Capture) forms [25,26] for data collection. We collected data on the focus of each article, the study design, the study setting, and the study sample size. We categorized each article’s focus as the incidence/prevalence of diagnostic error and interventions to measure diagnostic error, factors contributing to diagnostic error, interventions related to preventing or reducing diagnostic error, or other. Data relating to each study’s setting included the specific setting within a pediatric hospital the study took place (general wards, ED, PICU, CICU, NICU, or some combination) and the type of pediatric hospital (academic medical center, community medical center, or a combination of the two). We extracted additional data from studies focused on the incidence/prevalence of diagnostic error and interventions to measure diagnostic error, including how diagnostic error was measured and whether the incidence/prevalence was measured for a single diagnosis or more generally. For articles focusing on factors contributing to diagnostic error, we included whether the article focused on cognitive factors or healthcare system factors, the population under study, and the general methods used to identify contributing factors. For articles focusing on interventions, we adapted an extraction method used previously by Singh et al. [7]. As an indication of the quality of evidence supporting an intervention, we documented whether the intervention was implemented in practice or was simply suggested as a theoretical entity. We documented whether the intervention was designed to prevent or reduce diagnostic error, which population used the intervention and for how long, whether the intervention was tested against a control, and what outcomes were impacted by the intervention.

Results

An overview of the article selection process is provided in Figure 1. The initial literature search yielded 755 publications. After applying language, date, and full abstract filters, the total number of publications was limited to 538 articles. We excluded 479 articles after the titles were reviewed, leaving a total of 59 articles for abstract review. We included 28 articles for full review after applying the inclusion and exclusion criteria to the full abstracts. We determined that 6 of the included articles did not meet inclusion criteria during the full article review and thus were not included and did not undergo data extraction. In one of the included review articles, we identified an additional study that was felt to meet all inclusion criteria and thus, while not identified initially, was included in the final set of articles and underwent data extraction.

Figure 1.

Overview of Article Selection

We included a total of 23 articles in the final review, with the majority having an observational design (n=17) and taking place in academic medical centers (n=19). Nineteen of the studies focused on a single topic, with 4 focusing on both the incidence/prevalence of diagnostic error and contributing factors. A breakdown of the included studies by both hospital setting and topic is provided in Figure 2.

Figure 2.

Number of Articles Based on Setting and Topic Focus

Abbreviations: CICU, cardiac intensive care unit; ED, emergency department; NICU, neonatal intensive care unit; PICU, pediatric intensive care unit

^aMultiple settings refers to studies that took place in at least two of the following locations: general wards, PICU, NICU, CICU, or ED

Incidence/Prevalence and Interventions to Measure Diagnostic Error

Autopsy Studies

The majority of autopsy studies used the Goldman criteria to classify discrepancies between histopathologic diagnoses and clinical diagnoses. Per the Goldman criteria, Class I errors involve missed diagnoses directly related to a patient’s death that had a potentially adverse impact on survival and would have changed medical management. Class II errors also involve missed diagnoses directly related to a patient’s death but that had no potential impact on survival and would not have changed therapy [27]. Rates of diagnostic error in studies comparing clinical diagnoses to histopathologic diagnoses ranged from 3.8% for patients with a single diagnosis of appendicitis in the ED [28] to 53.7% of patients in the PICU who died while receiving Extracorporeal Membrane Oxygenator (ECMO) support [29].

Retrospective Cohort Studies

We identified 7 studies investigating the incidence/prevalence of diagnostic error in pediatric hospital settings using retrospective electronic health record review. The diagnostic error rate was estimated to be between 3.8% and 6.4% in patients re-presenting to medical care after ED discharge [30,31], 3.1% to 12.1% in PICU and ED patients determined to be at a high-risk of diagnostic error [32,33], 5% in general wards patients [34] and 8% in general PICU patients [35]. Sundberg et al. defined high risk patients based on certain ICD diagnostic codes [33] while Davalos et al. defined high risk PICU patients as those who either were autopsied, were seen as outpatients within 2 weeks before PICU admission, or were transferred to the PICU from an acute care floor after a rapid response event [32].

Five of the 7 studies investigating the incidence/prevalence of diagnostic error through retrospective chart review defined diagnostic error as missed or delayed diagnoses and used discrepancies between admitting and discharge diagnostic codes or personal judgment as measurement tools. Only two studies used validated frameworks to determine if a diagnostic error occurred [32,35].

Other Studies

While the majority of included studies used either retrospective chart review or autopsy to estimate the rate of diagnostic error, one study conducted by Bhat et al. [18] utilized interviews with providers and nurses working in a pediatric CICU. The study demonstrated that 92% of providers and 68% of nurses perceived that diagnostic errors directly resulting in patient harm occurred at least 5 times per year [18].

Tables 1 and 2 summarize the selected articles focusing on the incidence or prevalence of diagnostic error in the pediatric hospital.

Table 1:

Summary of Articles Focusing on Incidence/ Prevalence of Diagnostic Error and Interventions to Measure Diagnostic Error

References	Year	Hospital System	Study Design	Sample Size	Study Setting	Summary of Findings
Newton et al. [36]	2004	Academic and community medical centers	Observational	61 autopsies (39 from a tertiary children’s hospital, 22 from referring community hospitals)	Multiple inpatient settingsa	21% prevalence of diagnostic error (determined as discrepancy between pre and postmortem diagnosis)
Rodrigues et. al. [37]	2021	Academic medical center	Observational	31 autopsies	PICU	35.4% disagreement between final major clinical diagnosis and anatomopathological diagnosis (Class Ib)
Blanco et al. [29]	2014	Academic medical center	Observational	54 autopsies	PICU	Major discrepancy (Class Ib) in 53.7% of autopsies
Ostrow et al. [30]	2020	Academic medical center	Observational	1546 patients admitted to the hospital with 72 hours of an ED visit	ED	Misdiagnosis determined to be the root cause of admission within 72 hours after initial ED visit in 6.4%
Connors et al. [38]	2001	Academic and community medical centers	Observational	58 patients from pediatric trauma registry	ED	Delays in diagnosis identified in 1% of blunt trauma patients
Galai et al. [28]	2016	Academic medical center	Observational	400 patients	ED	Misdiagnosis in 3.8% of all histo-pathologically confirmed cases of appendicitis over a 4-year period
Salamati et al. [39]	2008	Academic medical center	Observational	84 autopsies	Multiple inpatient settingsa	12% of autopsies changed the clinical diagnosis
Depiero et al. [31]	2001	Academic medical center	Observational	261 patients returning to the ED within 72 hours of a previous visit	ED	3.8% of patients classified as having a missed diagnosis at initial ED presentation
Sundberg et al. [33]	2018	Academic medical center	Observational	55233 ED encounters for hospitalized children, narrowed to 2161 records of patients with one of 10 high risk diagnoses	ED	Computerized tool was developed and implemented which identified a 20% discordant rate between ED and discharge diagnoses, compared to a 3.1% rate determined via manual chart review
Cifra et al. [35]	2020	Academic medical center	Observational	50 patients consecutively admitted to the PICU	PICU	8% prevalence of diagnostic error within the first 12 hours of PICU admission using the Safer Dx tool
Custer et al. [20]	2014	Academic medical center	Systematic review	13 studies (7 PICU, 6 NICU); 1,063 deaths 498 autopsies in PICU studies, 2,124 deaths and 1259 autopsies in NICU studies	Multiple inpatient settings	Major diagnostic errors by Goldman criteria in 19.6% of autopsied deaths
Davalos et al. [32]	2017	Academic medical center	Observational	214 “high- risk” PICU patient records	PICU	Using the Safer Dx instrument, diagnostic errors were identified in 12.1% of high-risk patient records
Marshall et al. [40]	2021	Academic medical center	Experimental	Not specified, intervention delivered to entire PHM division	General wards	Increased attending physician reporting of diagnostic learning opportunities after implementation of an electronic reporting form with reminders, scheduled reflection time, and monthly progress reports

Abbreviations: CICU, cardiac intensive care unit; ED, emergency department; NICU, neonatal intensive care unit; PICU, pediatric intensive care unit

^aMultiple inpatient settings refers to studies that took place in at least two of the following locations: general wards, PICU, NICU, CICU, or ED

^bGoldman classification, missed major diagnoses are categorized as Class I and II diagnostic errors. Class I errors had a potential adverse impact on survival and would have changed medical management, Class II errors had no potential impact on survival and would not have changed therapy [29].

Table 2:

Summary of Articles Focusing on Both Incidence of and Factors Contributing to Diagnostic Error

Cifra et al. [41]	2015	Academic medical center	Observational	96 morbidity and mortality cases	PICU	21% of reviewed cases had identified misdiagnoses. 35% were discovered at autopsy, 55% reported at morbidity and mortality conference; system related factors were solely responsible in 40%, cognitive factors were solely responsible in 20%, 35% had a combination of both
Cifra et al. [19]	2021	---	Systematic review	17 studies	PICU	Autopsy studies identified 10–23% rate of missed major diagnoses, retrospective record review reported diagnostic error rates of 8% to 25%; system factors identified in 40–67% of diagnostic error cases, cognitive factors in 20–33%, and a combination in 40%
Warrick et al. [34]	2014	Community hospital	Observational	378 patients admitted to the hospital with an acute illness	General wards	Retrospective chart review identified 5% prevalence of misdiagnosis with cognitive factors identified as the main contributor; clinicians identified system factors as contributing more commonly
Bhat et al. [18]	2018	Academic medical center	Observational	200 ICU providers and nurses	CICU	92% of providers and 68% of nurses perceived that diagnostic errors result in patient harm 5 or more times per year; respondents perceived diagnostic errors as being associated more with system factors than cognitive factors

Abbreviations: CICU, cardiac intensive care unit; ED, emergency department; NICU, neonatal intensive care unit; PICU, pediatric intensive care unit

Interventions to Measure Diagnostic Error

Two studies investigated interventions to measure diagnostic error (Table 1). One was observational in nature and focused on developing a computerized tool to measure the rate of discordant ED and discharge diagnoses [33]. The other was experimental in design and involved the utilization of a quality improvement framework to increase physician reporting of diagnostic learning opportunities [36].

Factors Contributing to Diagnostic Error

Summaries of articles focused on factors contributing to diagnostic error and articles focused on a combination of the incidence of diagnostic error and contributing factors are displayed in Tables 2 and 3. The majority of studies took place in either an ICU or general wards setting, with no studies performed in the ED (Figure 2). In the PICU setting, system-related factors were identified as being solely responsible for diagnostic errors in 40–67% of cases, cognitive factors as solely responsible in 20–33% of cases, and a combination of factors in 35–40% of cases [35,41]. Cognitive factors were also found to play a prominent role in the NICU setting. A small case series of NICU patients identified cognitive factors as the sole cause of error in 6 of 10 cases [42]. One study focused on factors contributing to diagnostic errors on the general wards [34]. In reviewing 378 medical records of patients admitted to the hospital with an acute illness, the authors identified that cognitive factors were the main contributor to diagnostic errors. The finding contrasts with the results elicited from interviews with physicians, who identified system-related factors contributing more frequently.

Table 3:

Summary of Articles Focusing on Factors that Contribute to Diagnostic Error

References	Year	Hospital System	Study Design	Sample Size	Study Setting	Summary of Findings
Shafer and Suresh [42]	2018	Academic medical center	Observational	10 cases of nonlethal diagnostic error	NICU	Cognitive factors identified as solely responsible for diagnostic error in 60%, systems factors identified in 10%, and a combination identified in 20%, one case identified as a “no fault” error
Smink et al. [43]	2004	Academic and community medical centers	Observational	37109 nonincidental appendectomies with a misdiagnosis rate of 8.4%	Multiple inpatient settingsa	Lowest and low volume hospitals had a significantly increased likelihood of misdiagnosis, OR 1.5 (95% CI 1.0–2.2) for lowest volume hospitals and OR 1.6 (95% CI 1.1–2.3) for low volume hospitals
Grubenhoff et al. [16]	2019	Academic medical center	Experimental	70 providers (physicians and physician assistants)	Multiple inpatient settingsa	Providers demonstrated limited familiarity with common heuristics leading to diagnostic error in case vignettes and identified negative peer and personal perceptions of diagnostic performance as barriers to discussing errors publicly

Abbreviations: CICU, cardiac intensive care unit; ED, emergency department; NICU, neonatal intensive care unit; PICU, pediatric intensive care unit

^aMultiple inpatient settings refers to studies that took place in at least two of the following locations: general wards, PICU, NICU, CICU, or ED

Interventions to Reduce, Prevent, or Mitigate Diagnostic Error

We identified 3 articles that described interventions to reduce, prevent, or mitigate diagnostic error which are summarized in Table 4. Two studies took place in the ED setting and described interventions that could theoretically reduce diagnostic error but were not tested empirically. The one experimental study that was conducted in the clinical setting described a reduction in the prevalence of delayed diagnosis or injury following implementation of a trauma response team designed to evaluate patients in the ED [46].

Table 4:

Summary of Articles Focusing on Interventions to Reduce, Prevent or Mitigate Diagnostic Error

References	Year	Hospital System	Study Design	Sample Size	Study Setting	Summary of Findings
Searns et al. [44]	2020	Academic medical center	Observational	162 patients	Multiple inpatient settingsa	A novel antimicrobial stewardship model, known as handshake stewardship, identifies and intervenes on diagnostic error by offering a second look by a stewardship team that directly communicates with the primary team
Alvarez et al. [45]	2006	Academic and community medical centers	Experimental	118 records of infants undergoing ultrasound to rule out pyloric stenosis (88 to train a Bayesian decision network, 28 to test the network)	ED	Physicians using the Bayesian decision network better predicted the probability of pyloric stenosis; would have ordered 22% fewer ultrasounds and missed no cases of pyloric stenosis
Perno et al. [46]	2005	Academic medical center	Observational	3265 hospitalized trauma patients	ED	Implementation of a designated trauma response team, evaluating trauma patients in the ED, was associated with a decrease in the prevalence of delayed diagnosis of injury

Abbreviations: CICU, cardiac intensive care unit; ED, emergency department; NICU, neonatal intensive care unit; PHM, pediatric hospital medicine; PICU, pediatric intensive care unit

^aMultiple inpatient settings refers to studies that took place in at least two of the following locations: general wards, PICU, NICU, CICU, or ED

Discussion

Our literature review demonstrated a significant gap in knowledge of diagnostic error in pediatric hospital settings, with limited data describing factors contributing to diagnostic error and interventions to reduce diagnostic error. There were a significant number of studies that used autopsy or retrospective review to describe the incidence and prevalence of diagnostic error. Retrospective study findings were limited as few validated frameworks exist to guide the identification and classification of diagnostic errors. We also identified that a significant portion of studies took place in either an ED or intensive care setting, with fewer studies focusing on the general inpatient wards. Nevertheless, the review confirms that diagnostic error is a prevalent type of error within the pediatric hospital, with various contributing factors.

Estimates of the incidence and prevalence of diagnostic error varied widely and depended on the study setting and measurement technique. For example, autopsy studies and studies set in intensive care units tended to have higher estimates of diagnostic error prevalence. Retrospective chart reviews offered more generalizable estimates of diagnostic error compared to autopsy studies, but were limited by inconsistent use of a structured, validated measurement or identification tool. Despite the heterogeneity of these results, overall estimates were similar to prior studies conducted in adult patients [9–11] and pediatric studies conducted in the outpatient setting [14,15].

Our review did identify that a quality improvement methodology can increase physician reporting of diagnostic learning opportunities, defined as potential opportunities to make a better or more timely diagnosis [40]. A logical next step may be to couple this increased reporting with either the Safer Dx tool or Diagnostic Error Evaluation and Research (DEER) taxonomy [47,48] to confirm that the identified learning opportunities meet a validated definition of diagnostic error. In general, future work related to the epidemiology of diagnostic error in pediatric hospital settings should use validated measurement tools and focus on prospectively identifying this type of error in areas with a more limited evidence base.

Similar to those studies focused on incidence or prevalence, a minority of studies focusing on the factors contributing to diagnostic error utilized a structured framework and categorization scheme, a limitation that likely led to significant heterogeneity. Despite this heterogeneity, individual cognitive factors were found to play a prominent role, being implicated in anywhere from 20 to 60% of diagnostic error cases. While our review identified the prevalence of cognitive and system factors in diagnostic error cases, it offered less clarity on the origin of these factors and how to intervene to ensure they have less influence over clinicians’ decision-making processes. Given the complexity of the diagnostic process and the prominent role the external work environment plays in diagnosis [49,50], future work should be geared toward evaluating the process in real-time in the clinical environment to better assess all contributing factors. Qualitative methods, such as focused ethnography, which utilizes third party observation of individuals’ behaviors to achieve a better understanding of underlying processes [51], has been used to better define the diagnostic process in adult medical centers [52] and is a technique that could offer valuable insight in pediatric hospital settings.

We identified only 3 studies that focused on reducing diagnostic error in the pediatric hospital setting, with only 1 published in the last 5 years [44] and none prospectively evaluating an intervention’s impact on clinically significant diagnostic error. Thus, there is a significant need to develop and evaluate diagnostic error-related interventions in pediatric hospital settings. Interventions that embed tools such as diagnostic checklists or timeouts at specific times in the patient care process, while not extensively studied, have shown promise [24,53,54] and may be a starting point for further research.

Our review has several important limitations. We utilized a restrictive search strategy to identify literature, and while this helped ensure only the most relevant and up-to-date literature was included, we likely missed several key papers. We also excluded many papers through a simple review of study titles and, therefore, could have missed additional key articles found in the initial search. The topic of diagnostic error is large, and we focused our review only on three key aspects: prevalence or incidence, contributing factors, and interventions. As such, we cannot comment on the cost or impact of diagnostic errors and their direct effects on patients cared for in pediatric hospitals. Finally, while we included multiple settings within the pediatric hospital, we likely missed studies taking place in other areas, such as the operating room or imaging suite, and therefore further limited the scope of the review.

Conclusion

In conclusion, our review summarizes the current evidence base of the prevalence, interventions related to, and factors contributing to diagnostic error in pediatric hospitals. We identified several knowledge gaps related to the measurement of diagnostic error and the contributing factors and noted a paucity of studies focused on evaluating potential interventions. We believe future studies should utilize a standard definition of diagnostic error and employ validated measurement tools that can be used in real-time and in various settings. Research focused on mitigating diagnostic error and better understanding the contributing factors requires more robust study designs and precise outcomes. By closing these knowledge gaps and moving forward in these future research directions, diagnostic error will be better understood and mitigated, thereby improving overall patient safety within pediatric hospitals.