SCCM Resource Library-Diagnostic Error in Pediatric Critical Care: A Multicenter Study

Video Transcription

Thank you very much for attending this session. My name is Tina Sifra. I am a pediatric intensivist and health services researcher from the University of Iowa, Harvard  College of Medicine. I'm also the lead principal investigator of DxPQ. On behalf of the DxPQ investigators, today I am presenting the results of our multicenter  effort to shed more light on diagnostic error in pediatric critical care. I have no relevant financial relationships to disclose.  This study has been funded by the Agency for Healthcare Research and Quality, the National Institutes of Health, and the University of Iowa Department of Pediatrics.  Critically ill children are uniquely vulnerable to diagnostic error. They vary in age and development and are dependent on their caregivers to articulate their history,  signs and symptoms. They also have a broad range of pathophysiology that includes congenital, genetic, and metabolic disease that makes them unique from adult patients.  The time-pressured ICU environment also presents competing demands on providers. Many times, ICU clinicians need to manage life-threatening pathophysiology while also  attempting to diagnose the patient's underlying disease. Data is abundant in the ICU, and clinicians need to understand how to separate signal from noise.  Emotions can also run high, seeing that the stakes are very high when children are critically ill. The prevalence of diagnostic error vary from 8% to 25% of ICU admissions.  This variation is highly dependent on the method employed for detecting diagnostic error. For example, we have traditionally relied on autopsy studies to determine diagnostic  error rates. However, we know that this will only include diagnostic errors found in patients who die in the hospital.  Studies do not give any information about misdiagnosis among patients who survive. Diagnostic error rates also depend on the sample population study.  For example, many studies have focused on cohorts hypothesized to be at higher risk for diagnostic error, such as readmissions in the PICU or those admitted after a rapid response.  Other studies have also been limited to specific diseases. We also know little about the factors associated with diagnostic error and the extent of harm  that misdiagnosis can cause for patients and families. Any effort to successfully prevent diagnostic error must be informed by a clear understanding  of the vulnerabilities in the diagnostic process that can be targeted by specific interventions.  The objective of this study is to determine the prevalence and characteristics of diagnostic  errors and identify factors associated with diagnostic error up to 7 days after PICU admission.  This is the first study to determine the prevalence of diagnostic error in the pediatric intensive care unit in a randomly selected population.  We performed a retrospective cohort study using standard structured medical record review.  Ultimately, the study included 882 non-elective patient admissions to four tertiary referral PICUs over one year.  We selected only non-elective patient admissions in order to include patients with acute illness who will likely require some amount of diagnostic activity.  To this effect, elective admissions were excluded, including planned post-operative patients and patients admitted for elective procedures.  The four participating PICUs were all academic tertiary referral units located in diverse geographic areas, serving both urban and rural populations.  Two PICUs are combined medical-surgical and cardiac units, while the other two are purely medical-surgical units.  We selected patients admitted over one year using stratified random sampling by season of the year to ensure a wide variety of patients with seasonal diagnosis will be included.  At each participating site, four to five PICU clinicians performed standardized record review.  The clinician reviewers were either attending pediatric intensivists or advanced practice providers who have been practicing in the PICU for at least one year.  All clinician reviewers underwent training in an effort to standardize the medical record review. 35% of records were reviewed by at least two reviewers.  Clinician reviewers did not review records of patients that they personally attended to while admitted to the PICU within the first seven days.  Patients who had the lead PI or any of the site PIs as their attending physician at any time were excluded from the study.  We collected data on patient demographics, clinical characteristics, and clinician and encounter characteristics.  The SAFER-Dx instrument was used to identify potential missed opportunities in diagnosis of the first seven days of PICU admission.  The instrument guides reviewers in identifying missed opportunities in diagnosis, which are  defined as actions that could have been performed differently to make an accurate and timely diagnosis given available information at a particular point in time.  While using the instrument, the clinician reviewer puts herself in the shoes of the diagnosing clinician with all of the available information that the clinician had at a particular  time. The reviewer then decides whether, given this information, the clinician could have arrived with the correct diagnosis at that time.  If not, the reviewer identifies any potential missed opportunity. Missed opportunities in the diagnosis could have been missed not only by the clinician  but also by the PICU team, the system, or the patient and family. The SAFER-Dx instrument is a validated tool to identify diagnostic error using structured  chart review and has been previously used in PICU settings for this purpose. The instrument is comprised of 13 items.  Reviewers rate each item on a Likert scale, ranging from strongly disagree to strongly agree.  Items identified missed opportunities in diagnosis within specific aspects of the diagnostic  process, such as data gathering, information integration, and information interpretation.  In item number 13, given the missed opportunities identified in the case, the reviewer makes a final determination as to the occurrence of diagnostic error.  All cases with an identified potential diagnostic error were reviewed by the lead PI and three site PIs at regularly scheduled consensus meetings.  At these meetings, each case with a potential error was presented to the group. Research team members were given the opportunity to ask questions about the case.  Robust discussion was encouraged and a final determination of diagnostic error was made  if at least three or four members of the consensus panel agreed diagnostic error likely occurred.  Characteristics of patients, encounters, and clinicians were compared among those patients with versus without diagnostic error using simple statistical tests, such as the Fisher's  exact test for categorical variables and the Wilcoxon rank sum test for continuous variables.  A generalized linear mixed model with a site or the PICU specified as a random effect was constructed to determine factors associated with diagnostic error.  Our study included 882 patients with a median age of 3.8 years old. Forty-four percent of patients were female, 53 percent were white, and 88 percent were non-Hispanic.  Forty-five percent of patients were admitted through the same institution's emergency department and 37 percent were admitted from another institution directly to the PICU.  The mean PRISM-3 score was 2.7 and 50 percent of patients had a complex chronic condition. Thirteen of 882 patients, or 1.5 percent of patients, had a diagnostic error up to seven  days after PICU admission. Forty-six percent of diagnostic errors were discovered during the PICU stay, but more than 24 hours after admission.  Thirty-eight percent of diagnostic errors were discovered after transfer out of the PICU.  The diagnostic error was discovered most commonly because original signs and symptoms failed to resolve or improve, or new data emerged from diagnostic testing.  Only one patient suffered temporary harm, requiring prolonged hospitalization as a result of diagnostic error.  In this table, we show the 13 misdiagnoses observed in our study with our corresponding PICU admission diagnosis.  Note that 31 percent of the misdiagnoses were additional conditions or complications of the primary reason for PICU admission.  Thus, the documented PICU admission diagnoses were correct for this group. The most common misdiagnoses included infectious, respiratory, and neurologic conditions.  This slide shows the distribution of misdiagnostic opportunities contributing to diagnostic error.  The most common misdiagnostic opportunities included failure to consider specific points in the history suggestive of an alternative diagnosis and failure to pursue additional  diagnostic evaluation given available clinical information. Other common diagnostic opportunities which were missed included failure to consider specific  points in the physical exam suggestive of an alternate diagnosis and inappropriate diagnostic reasoning given the clinical presentation.  Note that for each diagnostic error observed, there were usually more than one misdiagnostic opportunity contributing to the diagnostic error.  In this slide, we show the clinical characteristics of patients with versus without diagnostic error.  Using simple statistical comparisons, we observed that among the cohort of patients with diagnostic  error, there were significantly more atypical presentations on admission compared to those patients with no diagnostic error.  Similarly, among the cohort with diagnostic error, there were significantly more patients who had a neurologic chief complaint compared to those patients with no diagnostic error.  We did not find any significant differences among those with versus without diagnostic  error in terms of age, gender, race, ethnicity, admission source, severity of illness, presence of complex chronic conditions, and discharge diagnosis category.  Here we show the encounter in clinician characteristics of patients with versus without diagnostic error.  We noted statistically significant differences in clinician characteristics. Patients with diagnostic error were noted to have significantly more attending pediatric  intensivists who were more than or equal to 45 years of age and had attending physicians  who had significantly more weeks of PICU service in the same year as the patient was admitted compared to patients without diagnostic error.  Diagnostic uncertainty was also prominent. A cohort with diagnostic error had a significantly higher proportion of patients with uncertainty  on PICU admission compared to those without diagnostic error. Patients with diagnostic error were also observed to require more clinical resources,  with significantly more laboratory and imaging tests performed and more subspecialty services consulted in the first 24 hours of PICU admission.  There was also significantly more observed discordance in primary diagnosis between PICU admission and PICU discharge among those with diagnostic error.  Regarding patient outcomes, patients with diagnostic error had a longer median hospital stay compared to those without diagnostic error.  However, there was no significant differences in PICU length of stay and mortality in hospital. In this slide, we present the results of a generalized linear mixed model with a site  or the PICU designated as a random effect, showing two factors that are significantly associated with occurrence of diagnostic error.  These include a typical presentation on admission with an odds ratio of 4.6 and the presence of diagnostic uncertainty on admission with an odds ratio of 8.7.  Our study has several limitations. Our determinations of diagnostic error were highly dependent on the quality of documentation,  which is a common limitation in retrospective record review studies. There's also the possibility of hindsight bias in our review, although we mitigated  this by training reviewers in a standard method of reviewing. There's also the possibility that our observed diagnostic error prevalence is an underestimate  of the true prevalence for two reasons. First, we may have missed diagnostic errors since some records were reviewed by only one reviewer.  Therefore, it is possible that some of these singly reviewed records may contain diagnostic errors.  Second, we included only those diagnostic errors discovered during the hospital encounter. It remains a possibility that diagnostic errors may have been discovered after discharge.  In conclusion, our study showed that 1.5% of critically ill children had a diagnostic error seven days after PICU admission.  Diagnostic errors were associated with a typical presentation and diagnostic uncertainty on admission. These findings seem intuitive.  However, our study shows empirical evidence of this association in practice. One may surmise that patients with atypical symptoms and diagnostic uncertainty would  encourage more diagnostic activity and motivate clinicians above and beyond their usual practice to identify an accurate diagnosis.  However, our study affirms an opposing view, that patients with these characteristics in PICU admission remain at higher risk for diagnostic error.  Future research directions should be aimed at identifying targeted interventions to prevent diagnostic error and diagnostic error-related harm.  To develop effective interventions, investigators need to consider atypical presentations and  diagnostic uncertainty as potential markers for patients at higher risk for misdiagnostic opportunities. Thank you very much for your attention.  I would like to thank all the members of my mentoring team and collaborators for their unwavering support and encouragement.  Thank you to all of the members of the DxPICU research team in Iowa City and beyond.  Finally, thank you to the University of Iowa, especially the Division of Critical Care, for being the heart and home of this work. I would love to answer any questions you may have.  Thank you.

Video Summary

This presentation summarizes the results of a study on diagnostic error in pediatric critical care. The study aimed to determine the prevalence and characteristics of diagnostic errors and identify factors associated with these errors. The study involved a retrospective cohort analysis of 882 patient admissions to four tertiary referral pediatric intensive care units over one year. The results showed that 1.5% of patients had a diagnostic error within seven days of admission. Factors associated with diagnostic errors included atypical presentations and diagnostic uncertainty on admission. The study suggests the need for targeted interventions to prevent diagnostic errors in this population.

Asset Subtitle

Quality and Patient Safety, Worldwide Data, Pediatrics, 2022

Asset Caption

INTRODUCTION: Diagnostic errors (DEs) harm critically ill children. We determined factors associated with DE in the 1st 7 days after pediatric ICU (PICU) admission. METHODS: We performed a retrospective study of 854 randomly-selected non-elective admissions over 1yr to 4 tertiary-referral PICUs. 4-5 PICU clinicians/site reviewed charts using the Safer Dx instrument to identify potential missed opportunities in diagnosis (MODs) during the 1st 7 days after admission; 35% of charts were reviewed by >2 reviewers. Cases with MODs at all sites were reviewed by 4 site PIs as a group who made final DE determinations. We used the Wilcoxon rank-sum test and Fisher’s exact test to compare characteristics of patients with vs. without DE. RESULTS: 13 of 854 (1.5%) patients had a DE within the 1st 7 days after admission. Reviews for DE were 98.6% concordant among multi-reviewer charts, however Κ=0.28 (95%CI: 0.02, 0.84) likely due to low DE prevalence. Most common missed diagnoses were respiratory (31%) and infectious (31%) conditions. Most DEs were discovered during the PICU stay but >24 hrs after admission (46%). One DE caused harm with prolonged hospital stay. Common MODs included failure to consider the diagnosis despite a suggestive history (46%) and failure to broaden diagnostic testing given clinical information (46%). Demographics, presence of complex chronic conditions, illness severity, and admission source were not associated with DE. Patients with DE had more: 1) atypical presentations (23%vs.4%, p=0.012), 2) neurologic chief complaints (46%vs.19%, p=0.025), 3) admitting intensivists >45 yrs old (92%vs.65%, p=0.041), 4) admitting intensivists with more service wks/yr (median 12.8 vs. 10.3 wks, p=0.023), 5) diagnostic uncertainty on admission (77%vs.25%, p < 0.001), and 6) lab tests performed on 1st day (median 20 vs. 8 tests, p=0.013). Patients with DE had more diagnostic discordance between admission and discharge (68%vs.5%, p < 0.001) and had a longer hospital stay (median 14 vs. 4 days, p=0.002). CONCLUSIONS: 1.5% of critically ill children had DEs within the 1st 7 days of PICU admission. DEs were associated with atypical presentations, neurologic complaints, and diagnostic uncertainty, which can help clinicians identify patients on admission who are at high risk for failures in diagnostic reasoning and MODs/DE.

Meta Tag

Content Type Presentation

Knowledge Area Quality and Patient Safety

Knowledge Area Pediatrics

Knowledge Area Worldwide Data

Knowledge Level Advanced

Membership Level Select

Tag Evidence Based Medicine

Tag Epidemiology Outcomes

Tag Pediatrics

Year 2022

Keywords

diagnostic error

pediatric critical care

prevalence

characteristics

factors associated

	Society of Critical Care Medicine 500 Midway Drive Mount Prospect, IL 60056 USA Phone: +1 847 827-6888 Fax: +1 847 439-7226 Email: support@sccm.org	Contact Us About SCCM Newsroom Advertising & Sponsorship DONATE	MySCCM LearnICU Patients & Families Surviving Sepsis Campaign Critical Care Societies Collaborative	GET OUR NEWSLETTER
© Society of Critical Care Medicine. All rights reserved. \| Privacy Statement \| Terms & Conditions The Society of Critical Care Medicine, SCCM, and Critical Care Congress are registered trademarks of the Society of Critical Care Medicine.