Missed acute myocardial infarction in the emergency department-standardizing measurement of misdiagnosis-related harms using the SPADE method

Computable phenotype for diagnostic error: developing the data schema for application of symptom-disease pair analysis of diagnostic error (SPADE)

OBJECTIVES

Diagnostic errors are the leading cause of preventable harm in clinical practice. Implementable tools to quantify and target this problem are needed. To address this gap, we aimed to generalize the Symptom-Disease Pair Analysis of Diagnostic Error (SPADE) framework by developing its computable phenotype and then demonstrated how that schema could be applied in multiple clinical contexts.

METHODS

We created an information model for the SPADE processes, then mapped data fields from electronic health records (EHR) and claims data in use to that model to create the SPADE information model (intention) and the SPADE computable phenotype (extension). Later we validated the computable phenotype and tested it in four case studies in three different health systems to demonstrate its utility.

RESULTS

We mapped and tested the SPADE computable phenotype in three different sites using four different case studies. We showed that data fields to compute an SPADE base measure are fully available in the EHR Data Warehouse for extraction and can operationalize the SPADE framework from provider and/or insurer perspective, and they could be implemented on numerous health systems for future work in monitor misdiagnosis-related harms.

CONCLUSIONS

Data for the SPADE base measure is readily available in EHR and administrative claims. The method of data extraction is potentially universally applicable, and the data extracted is conveniently available within a network system. Further study is needed to validate the computable phenotype across different settings with different data infrastructures.

Has the COVID-19 pandemic changed existing patterns of non-COVID-19 health care utilization? A retrospective analysis of six regions in Europe

BACKGROUND

Resilience of national health systems in Europe remains a major concern in times of multiple crises and as more evidence is emerging relating to the indirect effects of the COVID-19 pandemic on health care utilization (HCU), resulting from de-prioritization of regular, non-pandemic healthcare services. Most extant studies focus on regional, disease specific or early pandemic HCU creating difficulties in comparing across multiple countries. We provide a comparatively broad definition of HCU across multiple countries, with potential to expand across regions and timeframes.

METHODS

Using a cross-country federated research infrastructure (FRI), we examined HCU for acute cardiovascular events, elective surgeries and serious trauma. Aggregated data were used in forecast modelling to identify changes from predicted European age-standardized counts via fitted regressions (2017-19), compared against post-pandemic data.

RESULTS

We found that elective surgeries were most affected, universally falling below predicted levels in 2020. For cardiovascular HCU, we found lower-than-expected cases in every region for heart attacks and displayed large sex differences. Serious trauma was the least impacted by the COVID-19 pandemic.

CONCLUSION

The strength of this study comes from the use of the European Population Health Information Research Infrastructure's (PHIRI) FRI, allowing for rapid analysis of regional differences to assess indirect impacts of events such as pandemics. There are marked differences in the capacity of services to return to normal in terms of elective surgery; additionally, we found considerable differences between men and women which requires further research on potential sex or gender patterns of HCU during crises.

Longitudinal Changes in Diagnostic Accuracy of a Differential Diagnosis List Developed by an AI-Based Symptom Checker: Retrospective Observational Study

BACKGROUND

Artificial intelligence (AI) symptom checker models should be trained using real-world patient data to improve their diagnostic accuracy. Given that AI-based symptom checkers are currently used in clinical practice, their performance should improve over time. However, longitudinal evaluations of the diagnostic accuracy of these symptom checkers are limited.

OBJECTIVE

This study aimed to assess the longitudinal changes in the accuracy of differential diagnosis lists created by an AI-based symptom checker used in the real world.

METHODS

This was a single-center, retrospective, observational study. Patients who visited an outpatient clinic without an appointment between May 1, 2019, and April 30, 2022, and who were admitted to a community hospital in Japan within 30 days of their index visit were considered eligible. We only included patients who underwent an AI-based symptom checkup at the index visit, and the diagnosis was finally confirmed during follow-up. Final diagnoses were categorized as common or uncommon, and all cases were categorized as typical or atypical. The primary outcome measure was the accuracy of the differential diagnosis list created by the AI-based symptom checker, defined as the final diagnosis in a list of 10 differential diagnoses created by the symptom checker. To assess the change in the symptom checker's diagnostic accuracy over 3 years, we used a chi-square test to compare the primary outcome over 3 periods: from May 1, 2019, to April 30, 2020 (first year); from May 1, 2020, to April 30, 2021 (second year); and from May 1, 2021, to April 30, 2022 (third year).

RESULTS

A total of 381 patients were included. Common diseases comprised 257 (67.5%) cases, and typical presentations were observed in 298 (78.2%) cases. Overall, the accuracy of the differential diagnosis list created by the AI-based symptom checker was 172 (45.1%), which did not differ across the 3 years (first year: 97/219, 44.3%; second year: 32/72, 44.4%; and third year: 43/90, 47.7%; P=.85). The accuracy of the differential diagnosis list created by the symptom checker was low in those with uncommon diseases (30/124, 24.2%) and atypical presentations (12/83, 14.5%). In the multivariate logistic regression model, common disease (P<.001; odds ratio 4.13, 95% CI 2.50-6.98) and typical presentation (P<.001; odds ratio 6.92, 95% CI 3.62-14.2) were significantly associated with the accuracy of the differential diagnosis list created by the symptom checker.

CONCLUSIONS

A 3-year longitudinal survey of the diagnostic accuracy of differential diagnosis lists developed by an AI-based symptom checker, which has been implemented in real-world clinical practice settings, showed no improvement over time. Uncommon diseases and atypical presentations were independently associated with a lower diagnostic accuracy. In the future, symptom checkers should be trained to recognize uncommon conditions.

Diagnostic error among vulnerable populations presenting to the emergency department with cardiovascular and cerebrovascular or neurological symptoms: a systematic review

BACKGROUND

Diagnostic error (DE) is a common problem in clinical practice, particularly in the emergency department (ED) setting. Among ED patients presenting with cardiovascular or cerebrovascular/neurological symptoms, a delay in diagnosis or failure to hospitalise may be most impactful in terms of adverse outcomes. Minorities and other vulnerable populations may be at higher risk of DE. We aimed to systematically review studies reporting the frequency and causes of DE in under-resourced patients presenting to the ED with cardiovascular or cerebrovascular/neurological symptoms.

METHODS

We searched EBM Reviews, Embase, Medline, Scopus and Web of Science from 2000 through 14 August 2022. Data were abstracted by two independent reviewers using a standardised form. The risk of bias (ROB) was assessed using the Newcastle-Ottawa Scale, and the certainty of evidence was evaluated using the Grading of Recommendations Assessment, Development, and Evaluation approach.

RESULTS

Of the 7342 studies screened, we included 20 studies evaluating 7436,737 patients. Most studies were conducted in the USA, and one study was multicountry. 11 studies evaluated DE in patients with cerebrovascular/neurological symptoms, 8 studies with cardiovascular symptoms and 1 study examined both types of symptoms. 13 studies investigated missed diagnoses and 7 studies explored delayed diagnoses. There was significant clinical and methodological variability, including heterogeneity of DE definitions and predictor variable definitions as well as methods of DE assessment, study design and reporting.Among the studies evaluating cardiovascular symptoms, black race was significantly associated with higher odds of DE in 4/6 studies evaluating missed acute myocardial infarction (AMI)/acute coronary syndrome (ACS) diagnosis compared with white race (OR from 1.18 (1.12-1.24) to 4.5 (1.8-11.8)). The association between other analysed factors (ethnicity, insurance and limited English proficiency) and DE in this domain varied from study to study and was inconclusive.Among the studies evaluating DE in patients with cerebrovascular/neurological symptoms, no consistent association was found indicating higher or lower odds of DE. Although some studies showed significant differences, these were not consistently in the same direction.The overall ROB was low for most included studies; however, the certainty of evidence was very low, mostly due to serious inconsistency in definitions and measurement approaches across studies.

CONCLUSIONS

This systematic review demonstrated consistent increased odds of missed AMI/ACS diagnosis among black patients presenting to the ED compared with white patients in most studies. No consistent associations between demographic groups and DE related to cerebrovascular/neurological diagnoses were identified. More standardised approaches to study design, measurement of DE and outcomes assessment are needed to understand this problem among vulnerable populations.

TRIAL REGISTRATION NUMBER

The study protocol was registered in the International Prospective Register of Systematic Reviews PROSPERO 2020 CRD42020178885 and is available from: https://www.crd.york.ac.uk/prospero/display_record.php?ID=CRD42020178885.

Optimizing measurement of misdiagnosis-related harms using symptom-disease pair analysis of diagnostic error (SPADE): comparison groups to maximize SPADE validity

Diagnostic errors in medicine represent a significant public health problem but continue to be challenging to measure accurately, reliably, and efficiently. The recently developed Symptom-Disease Pair Analysis of Diagnostic Error (SPADE) approach measures misdiagnosis related harms using electronic health records or administrative claims data. The approach is clinically valid, methodologically sound, statistically robust, and operationally viable without the requirement for manual chart review. This paper clarifies aspects of the SPADE analysis to assure that researchers apply this method to yield valid results with a particular emphasis on defining appropriate comparator groups and analytical strategies for balancing differences between these groups. We discuss four distinct types of comparators (intra-group and inter-group for both look-back and look-forward analyses), detailing the rationale for choosing one over the other and inferences that can be drawn from these comparative analyses. Our aim is that these additional analytical practices will improve the validity of SPADE and related approaches to quantify diagnostic error in medicine.

A comprehensive framework to estimate the frequency, duration, and risk factors for diagnostic delays using bootstrapping-based simulation methods

BACKGROUND

The incidence of diagnostic delays is unknown for many diseases and specific healthcare settings. Many existing methods to identify diagnostic delays are resource intensive or difficult to apply to different diseases or settings. Administrative and other real-world data sources may offer the ability to better identify and study diagnostic delays for a range of diseases.

METHODS

We propose a comprehensive framework to estimate the frequency of missed diagnostic opportunities for a given disease using real-world longitudinal data sources. We provide a conceptual model of the disease-diagnostic, data-generating process. We then propose a bootstrapping method to estimate measures of the frequency of missed diagnostic opportunities and duration of delays. This approach identifies diagnostic opportunities based on signs and symptoms occurring prior to an initial diagnosis, while accounting for expected patterns of healthcare that may appear as coincidental symptoms. Three different bootstrapping algorithms are described along with estimation procedures to implement the resampling. Finally, we apply our approach to the diseases of tuberculosis, acute myocardial infarction, and stroke to estimate the frequency and duration of diagnostic delays for these diseases.

RESULTS

Using the IBM MarketScan Research databases from 2001 to 2017, we identified 2,073 cases of tuberculosis, 359,625 cases of AMI, and 367,768 cases of stroke. Depending on the simulation approach that was used, we estimated that 6.9-8.3% of patients with stroke, 16.0-21.3% of patients with AMI and 63.9-82.3% of patients with tuberculosis experienced a missed diagnostic opportunity. Similarly, we estimated that, on average, diagnostic delays lasted 6.7-7.6 days for stroke, 6.7-8.2 days for AMI, and 34.3-44.5 days for tuberculosis. Estimates for each of these measures was consistent with prior literature; however, specific estimates varied across the different simulation algorithms considered.

CONCLUSIONS

Our approach can be easily applied to study diagnostic delays using longitudinal administrative data sources. Moreover, this general approach can be customized to fit a range of diseases to account for specific clinical characteristics of a given disease. We summarize how the choice of simulation algorithm may impact the resulting estimates and provide guidance on the statistical considerations for applying our approach to future studies.

Death rates following emergency department chest pain assessment

OBJECTIVES

We sought to define the rate of unexpected death from acute coronary syndrome or arrhythmia in chest pain patients directly discharged from the ED.

METHODS

Retrospective audit of all chest pain patients at a tertiary ED for 7 years. Medical and post-mortem records of the deceased were reviewed with independent cardiologist adjudication to determine outcomes. Primary outcome measure was 28-day unexpected death secondary to acute coronary syndrome or arrhythmia.

RESULTS

During the study period, 25 924 patients presented with chest pain, 292 (1.1%, 95% confidence interval [CI] 0.99-1.01%) died within 28 days. Of these, 16 680(64%, 95% CI 63.88-64.12%) were discharged by ED, two (0.01%, 95% CI 0-0.011%) of this group died from the primary outcome.

CONCLUSION

Unexpected death is very uncommon after ED discharge of chest pain patients.

A clinically-guided unsupervised clustering approach to recommend symptoms of disease associated with diagnostic opportunities

OBJECTIVES

A first step in studying diagnostic delays is to select the signs, symptoms and alternative diseases that represent missed diagnostic opportunities. Because this step is labor intensive requiring exhaustive literature reviews, we developed machine learning approaches to mine administrative data sources and recommend conditions for consideration. We propose a methodological approach to find diagnostic codes that exhibit known patterns of diagnostic delays and apply this to the diseases of tuberculosis and appendicitis.

METHODS

We used the IBM MarketScan Research Databases, and consider the initial symptoms of cough before tuberculosis and abdominal pain before appendicitis. We analyze diagnosis codes during healthcare visits before the index diagnosis, and use k-means clustering to recommend conditions that exhibit similar trends to the initial symptoms provided. We evaluate the clinical plausibility of the recommended conditions and the corresponding number of possible diagnostic delays based on these diseases.

RESULTS

For both diseases of interest, the clustering approach suggested a large number of clinically-plausible conditions to consider (e.g., fever, hemoptysis, and pneumonia before tuberculosis). The recommended conditions had a high degree of precision in terms of clinical plausibility: >70% for tuberculosis and >90% for appendicitis. Including these additional clinically-plausible conditions resulted in more than twice the number of possible diagnostic delays identified.

CONCLUSIONS

Our approach can mine administrative datasets to detect patterns of diagnostic delay and help investigators avoid under-identifying potential missed diagnostic opportunities. In addition, the methods we describe can be used to discover less-common presentations of diseases that are frequently misdiagnosed.

Development and validation of deep learning ECG-based prediction of myocardial infarction in emergency department patients

Myocardial infarction diagnosis is a common challenge in the emergency department. In managed settings, deep learning-based models and especially convolutional deep models have shown promise in electrocardiogram (ECG) classification, but there is a lack of high-performing models for the diagnosis of myocardial infarction in real-world scenarios. We aimed to train and validate a deep learning model using ECGs to predict myocardial infarction in real-world emergency department patients. We studied emergency department patients in the Stockholm region between 2007 and 2016 that had an ECG obtained because of their presenting complaint. We developed a deep neural network based on convolutional layers similar to a residual network. Inputs to the model were ECG tracing, age, and sex; and outputs were the probabilities of three mutually exclusive classes: non-ST-elevation myocardial infarction (NSTEMI), ST-elevation myocardial infarction (STEMI), and control status, as registered in the SWEDEHEART and other registries. We used an ensemble of five models. Among 492,226 ECGs in 214,250 patients, 5,416 were recorded with an NSTEMI, 1,818 a STEMI, and 485,207 without a myocardial infarction. In a random test set, our model could discriminate STEMIs/NSTEMIs from controls with a C-statistic of 0.991/0.832 and had a Brier score of 0.001/0.008. The model obtained a similar performance in a temporally separated test set of the study sample, and achieved a C-statistic of 0.985 and a Brier score of 0.002 in discriminating STEMIs from controls in an external test set. We developed and validated a deep learning model with excellent performance in discriminating between control, STEMI, and NSTEMI on the presenting ECG of a real-world sample of the important population of all-comers to the emergency department. Hence, deep learning models for ECG decision support could be valuable in the emergency department.

Are Mental Health and Substance Use Disorders Risk Factors for Missed Acute Myocardial Infarction Diagnoses Among Chest Pain or Dyspnea Encounters in the Emergency Department?

STUDY OBJECTIVE

To assess if having a mental health and/or substance use disorder is associated with a missed acute myocardial infarction diagnosis in the emergency department (ED).

METHODS

This was a retrospective cohort analysis (2009 to 2017) of adult ED encounters at Kaiser Permanente Southern California. We used the validated symptom-disease pair analysis of diagnostic error methodological approach to "look back" and "look forward" and identify missed acute myocardial infarctions within 30 days of a treat-and-release ED visit. We use adjusted logistic regression to report the odds of missed acute myocardial infarction among patients with a history of mental health and/or substance use disorders.

RESULTS

The look-back analysis identified 44,473 acute myocardial infarction hospital encounters; 574 (1.3%) diagnoses were missed. The odds of missed diagnoses were higher in patients with mental health disorders (odds ratio [OR] 1.48, 95% confidence interval [CI] 1.23 to 1.77) but not in those with substance abuse disorders (OR 1.22, 95% CI 0.91 to 1.62). The highest risk was observed in those with co-occurring disorders (OR 1.90, 95% CI 1.30 to 2.76). The look-forward analysis identified 325,088 chest pain/dyspnea ED encounters; 508 (0.2%) were missed acute myocardial infarctions. No significant associations of missed acute myocardial infarction were revealed in either group (mental health disorder: OR 0.92, 95% CI 0.71 to 1.18; substance use disorder: OR 1.22, 95% CI 0.80 to 1.85).

CONCLUSION

The look-back analysis identified patients with mental illness at increased risk of missed acute myocardial infarction diagnosis, with the highest risk observed in those with a history of comorbid substance abuse. Having substance use disorders alone did not increase this risk in either cohort. The look-forward analysis revealed challenges in prospectively identifying high-risk patients to target for improvement.

Rate of sepsis hospitalizations after misdiagnosis in adult emergency department patients: a look-forward analysis with administrative claims data using Symptom-Disease Pair Analysis of Diagnostic Error (SPADE) methodology in an integrated health system

OBJECTIVES

Delays in sepsis diagnosis can increase morbidity and mortality. Previously, we performed a Symptom-Disease Pair Analysis of Diagnostic Error (SPADE) "look-back" analysis to identify symptoms at risk for delayed sepsis diagnosis. We found treat-and-release emergency department (ED) encounters for fluid and electrolyte disorders (FED) and altered mental status (AMS) were associated with downstream sepsis hospitalizations. In this "look-forward" analysis, we measure the potential misdiagnosis-related harm rate for sepsis among patients with these symptoms.

METHODS

Retrospective cohort study using electronic health record and claims data from Kaiser Permanente Mid-Atlantic States (2013-2018). Patients ≥18 years with ≥1 treat-and-release ED encounter for FED or AMS were included. Observed greater than expected sepsis hospitalizations within 30 days of ED treat-and-release encounters were considered potential misdiagnosis-related harms. Temporal analyses were employed to differentiate case and comparison (superficial injury/contusion ED encounters) cohorts.

RESULTS

There were 4,549 treat-and-release ED encounters for FED or AMS, 26 associated with a sepsis hospitalization in the next 30 days. The observed (0.57%) minus expected (0.13%) harm rate was 0.44% (absolute) and 4.5-fold increased over expected (relative). There was a spike in sepsis hospitalizations in the week following FED/AMS ED visits. There were fewer sepsis hospitalizations and no spike in admissions in the week following superficial injury/contusion ED visits. Potentially misdiagnosed patients were older and more medically complex.

CONCLUSIONS

Potential misdiagnosis-related harms from sepsis are infrequent but measurable using SPADE. This look-forward analysis validated our previous look-back study, demonstrating the SPADE approach can be used to study infectious disease syndromes.

Antecedent treat-and-release diagnoses prior to sepsis hospitalization among adult emergency department patients: a look-back analysis employing insurance claims data using Symptom-Disease Pair Analysis of Diagnostic Error (SPADE) methodology

OBJECTIVES

The aim of this study was to identify delays in early pre-sepsis diagnosis in emergency departments (ED) using the Symptom-Disease Pair Analysis of Diagnostic Error (SPADE) approach.

METHODS

SPADE methodology was employed using electronic health record and claims data from Kaiser Permanente Mid-Atlantic States (KPMAS). Study cohort included KPMAS members ≥18 years with ≥1 sepsis hospitalization 1/1/2013-12/31/2018. A look-back analysis identified treat-and-release ED visits in the month prior to sepsis hospitalizations. Top 20 diagnoses associated with these ED visits were identified; two diagnosis categories were distinguished as being linked to downstream sepsis hospitalizations. Observed-to-expected (O:E) and temporal analyses were performed to validate the symptom selection; results were contrasted to a comparison group. Demographics of patients that did and did not experience sepsis misdiagnosis were compared.

RESULTS

There were 3,468 sepsis hospitalizations during the study period and 766 treat-and-release ED visits in the month prior to hospitalization. Patients discharged from the ED with fluid and electrolyte disorders (FED) and altered mental status (AMS) were most likely to have downstream sepsis hospitalizations (O:E ratios of 2.66 and 2.82, respectively). Temporal analyses revealed that these symptoms were overrepresented and temporally clustered close to the hospitalization date. Approximately 2% of sepsis hospitalizations were associated with prior FED or AMS ED visits.

CONCLUSIONS

Treat-and-release ED encounters for FED and AMS may represent harbingers for downstream sepsis hospitalizations. The SPADE approach can be used to develop performance measures that identify pre-sepsis.