Pediatric Diagnostic Safety: State of the Science and Future Directions

Research and improvement efforts in pediatric diagnostic safety lag behind similar efforts with adult populations. However, pediatric providers and healthcare organizations have made significant contributions to understanding the impact of MDOs in children from primary care offices to the pediatric intensive care unit (PICU). In addition, leading organizations in pediatric safety have begun efforts to educate about, study, and reduce harm from MDOs.

Primary Care

Professional organizations such as AAP strongly advocate that all children have a “medical home” in which patients, their families, and clinicians “develop a partnership of mutual responsibility and trust.”⁵⁹ Such partnerships model patient and family inclusion on the diagnostic team as described in Improving Diagnosis in Health Care.⁶⁰ Nonetheless, MDOs occur even in this setting designed to identify new diagnoses well before harm arises from them.

One-third of ambulatory pediatricians surveyed by the AAP Quality Improvement Innovations Network reported making diagnostic errors at least monthly and errors that harmed patients at least annually.⁶¹ Nearly 90 percent indicated interest in reducing diagnostic errors, especially for conditions that evolve over months to years (e.g., hypertension, depression), require subspecialist referral, or may go unnoticed due to systems issues (e.g., not addressing abnormal values).

Responding to this knowledge, primary care diagnostic safety researchers have begun to address common MDOs through a quality improvement collaborative: Reducing Diagnostic Errors in Pediatric Primary Care, or Project RedDE.¹³ Early retrospective work showed more than half of the opportunities to recognize hypertension and depression were missed. When available to the treating clinician, 11 percent of abnormal lab values requiring treatment or additional diagnostic evaluation were not addressed.

Subsequent investigation has started to describe the factors that contribute to these common MDOs.⁶² Project RedDE developed a cluster-randomized, stepped-wedge prospective investigation of a virtual quality improvement collaborative involving 31 pediatric practices.⁶³ Results of the quality improvement initiatives provide actionable steps for other pediatric practices to address these three primary care issues (hypertension, depression, lack of followup on abnormal lab values).^64-66

As with other clinical practice environments, Project RedDE focused on specific diagnoses and, in the case of missed lab results, a specific diagnostic process. However, primary care providers are responsible for identifying a wide range of conditions. While most of these conditions may be common pediatric diagnoses, rarer and atypical presentations may go unrecognized, exposing patients to more risk of harm from MDOs.⁴⁰

As more primary care practices adopt electronic health records (EHRs), they may be able to leverage these systems to detect other missed opportunities. Singh, et al., used electronic triggers based on EHR-recorded events to detect episodes of care that might include an MDO. By linking unplanned hospitalizations and ambulatory visits (office, urgent care, ED), they significantly increased identification of cases at risk of containing an MDO.⁶⁷

Another proactive and easily implemented approach leveraged the EHR to prompt diagnostic pauses for primary care clinicians seeing patients in urgent followup within 2 weeks of a prior visit; 13 percent of these encounters contained diagnostic discrepancies.⁶⁸ Clinicians reported that these pauses helped them identify opportunities to improve the diagnostic process at the followup visit.

These investigations reveal two key obstacles for pediatric primary care providers who want to reduce MDOs. First, the best evidence to date comes from one well-funded, centrally coordinated research project. This project included resources not typically available to primary care providers (i.e., statistical, technical, and logistical support from the AAP quality improvement network).

Second, these projects integrated EHRs to aid in building alerts and triggers and tracking data. The resources needed to hire clinical informaticists to support this work may not be feasible for many primary care practices. Thus, primary care practices will probably need to partner with academic medical centers, pediatric hospitals, and patient safety organizations to overcome these obstacles.

Emergency Medicine

The ED, perhaps more than any other clinical location, challenges clinicians with multiple threats to diagnostic performance. Medford-Davis, et al., detail the domains in which these challenges lie:

• Brief, ad hoc patient encounters lacking established rapport.
• Limited clinical information due to alterations in mental status, extremes of age, delirium, or poor information exchange between health systems.
• Time pressures for the recognition and treatment of life-threatening illness as well as those related to patient volume and rapid disposition.
• Frequent interruptions.
• A nearly infinite list of diagnostic possibilities.
• Resource limitations, especially for pediatric patients.⁶⁹

Unsurprisingly, ED studies consistently indicate that adverse events frequently involve MDOs.^70-72 Although specialized for the care of children, pediatric EDs are not immune to MDOs, with diagnostic issues being second only to management concerns as the source of adverse events (19.3% and 52.4%, respectively).⁷³

ED encounters provide limited opportunity to accurately diagnose many of the conditions that patients first bring to medical attention there. Thus, return visits have received considerable attention as possible MDOs,^74,75 although return visits resulting in admission are more likely to involve MDOs.^71,76

Fewer data exist for children but their experiences parallel data in adult populations. Nearly 10 percent of unplanned ED 48-hour return visits in children involved a change in diagnosis; however, the presence of MDOs was not assessed.⁷⁷ Another single-center study evaluated unplanned admission within 14 days of an ED or urgent care visit to identify discrepancies between ED index encounter and subsequent hospital discharge summary diagnoses.⁹ Twenty percent of cases reviewed involved MDOs. This method allowed researchers to aggregate similar MDOs that were not identified through traditional incident reporting structures.

Aggregating cases with similar diagnostic process failures may allow patient safety leaders to implement or modify clinical decision support resources (e.g., clinical care pathways) to mitigate the risks to future patients. Evaluating ED return visits with admission may augment hospital systems’ ability to identify diagnostic improvement opportunities.

A less labor-intensive way to identify MDOs leverages administrative data to identify patterns of recurrent MDOs in pediatric ED encounters. For example, administrative data collected in four states showed that 8.1 percent of children admitted for sepsis within 7 days of an ED encounter experienced a probable MDO.⁷⁸ This approach also holds promise for nonpediatric hospitals interested in reducing MDOs.⁷⁹

Other data indicate that methods for detecting MDOs using administrative data will likely require different time horizons depending on the specific condition of interest.⁸⁰ Further, attempts to quantify MDOs in the ED must account for return visits to a different ED or risk missing approximately one in six return visits.⁸¹ Overall, the existing literature suggests that identifying, learning from, and reducing harm associated with MDOs in the ED will require multiple measurement strategies.

Hospital Medicine

Few studies have examined the epidemiology of MDOs among hospitalized children. One preliminary investigation that used a structured chart review to describe the prevalence of MDOs in pediatric patients admitted to a community hospital over a 90-day period showed that they affect approximately 5 percent of hospitalized children.¹² Similarly, 6.3 percent of children readmitted within 15 days at a single freestanding children’s hospital within 15 days of discharge experienced an MDO.⁸²

More recently, a quality improvement initiative successfully increased physician reporting of “diagnostic learning opportunities.” The project coupled a simple, dedicated reporting mechanism for suspected MDOs while patients were still hospitalized with a systematic review process to confirm the presence of a diagnostic learning opportunity.⁸ This approach consistently identified MDOs; 66 percent of reported events were determined to be MDOs. Further, the reports also generated insights about where in the diagnostic process errors often occur, which highlighted that most identified MDOs were multifactorial.

Nationwide Children’s Hospital has leveraged a combination of data streams, including physician event reporting, an abdominal pain-related electronic chart review trigger, morbidity and mortality conferences, and autopsy results, to build a diagnostic error index.¹¹ This approach allowed a more thorough and nuanced view of the incidence of MDOs and was used to identify targets for quality improvement initiatives. It also enabled consistent tracking of progress made toward improved diagnostic safety.

Relatively few published studies in pediatric hospital medicine (PHM) describe interventions or approaches to reduce MDOs. Recognition and open discussion of diagnostic uncertainty has been proposed as a strategy to prevent MDOs due to inappropriately applied heuristics (see box below), such as premature closure.

A group of eight PHM providers developed and piloted diagnostic timeout over a 12-month period.⁸³ In general, the diagnostic pause was well received. Notably, in half of cases studied, the timeout did not confirm the initial diagnosis, often led to new actions being taken, and was rarely seen as a time burden. It also engaged learners in diagnostic reasoning.

Diagnostic uncertainty, especially when inadequately communicated within the healthcare team and to patients and families, may contribute to MDOs.^91,92 A quality improvement initiative to increase shared situational awareness among the healthcare team about diagnostic uncertainty for children admitted to the PHM service resulted in a novel “uncertain diagnosis” or UD label added to the EHR.⁹³ The UD label was built on existing situational awareness infrastructure at this institution to identify risk for clinical deterioration. It included communication strategies to facilitate contingency planning for patients with uncertain diagnoses.⁹⁴

Building on this work regarding prospectively identifying diagnostic uncertainty in hospitalized children, additional studies have sought to better identify patients with uncertain diagnoses using EHR data such as diagnosis codes and clinical documentation.^95,96 Despite promising results for improving identification and situational awareness about diagnostic uncertainty, reductions in patient harm have yet to be clearly linked to these intermediate outcomes.

Feedback on diagnostic accuracy is often inconsistent or lacking in clinical practice. Without consistent feedback on diagnostic outcomes, physicians cannot calibrate their clinical reasoning for future, similar cases.⁹⁷ To address this gap, a prospective educational study sought to provide structured feedback to residents on subsequent diagnostic changes for patients admitted overnight. The goal was to facilitate learning and diagnostic calibration.⁹⁸ Notably, they found that 12.7 percent of all cases had major diagnostic changes. In addition, residents reported that this approach increased their comfort with giving feedback and improved self-efficacy in identifying and mitigating cognitive biases.

Critical Care

Focused attention on diagnostic safety has resulted in increased knowledge on the epidemiology of MDOs in pediatric critical care over the past few years.⁹⁹ Autopsy studies have historically served as the primary source of information on the prevalence of pediatric intensive care unit (PICU) MDOs¹⁰⁰ despite pediatric autopsy rates being below 50 percent and autopsies providing a limited and biased sample.⁹⁹

Recognizing these limitations, more recent investigations have attempted to determine MDO frequency, causes, and impact across the larger PICU population beyond the small percentage of critically ill children who die. These epidemiologic studies have used a variety of methods and sources to characterize MDOs, including clinician surveys,¹⁰¹ morbidity and mortality conference reviews,^102,103 incident reports,¹⁰⁴ and structured medical record review.^7,105,106

In contrast, research to understand the pediatric critical care diagnostic process remains underdeveloped. Early efforts using qualitative, quasi-experimental, and mixed methods studies have provided interesting insights into vulnerabilities of the PICU diagnostic process, including investigations of:

• Team cognition for diagnosis in daily PICU work.¹⁰⁷
• Impact of PICU transfer communication on the diagnostic process.¹⁰⁸
• Effect of hierarchy and professionalism on diagnosis during operating room-to-PICU handoffs.¹⁰⁹
• Quality and meaning of pediatric intensivists’ diagnosis narratives in clinical notes.¹¹⁰ 
• Impact of subspecialty consulting services on PICU diagnosis.¹¹¹

Much work is still needed to translate these observations into effective and feasible interventions that can prevent misdiagnosis-related harm in critically ill children.

Since most research on diagnostic performance in the PICU has been disease oriented, almost all of the current interventions promoting diagnostic safety in daily PICU practice are disease specific, such as electronic sepsis alerts and clinical pathways.^112-114 However, momentum is building for more disease-agnostic systems interventions, with early investigations of:

• Clinical decision support tools to detect patients at high risk for clinical deterioration.¹¹⁵
• PICU feedback systems to referring clinicians.¹¹⁶
• Communication of diagnostic uncertainty.^93,117 
• Standardization of referral communication for PICU transfer.^118,119

Despite these early achievements in primary and acute care, the achievements have been relatively isolated to specific practice locations. Integrated study of the impact of care transitions from the office to the ED to the hospital wards and ICUs on the diagnostic process has received little attention. In addition, outpatient subspecialty clinics have pursued little research on MDOs in children.