Search

Abstract Background Delays in diagnostic and therapeutic processes are a potentially preventable cause of morbidity and mortality. Process improvement depends on accurate knowledge about as-is processes, historically collected from front-line workers and summarized in flowcharts. Such flowcharts can now be generated by process discovery algorithms supplied with chronological records from real-world cases. However, these algorithms may generate incomprehensible flowcharts when applied to complex unstructured processes, which are common in healthcare. The aim of this study is to evaluate methods for analysing data from real-world cases to determine causes of delay in complex healthcare processes. Methods Physician insurance claims and hospital discharge data were obtained for patients undergoing carotid endarterectomy at a single tertiary hospital between 2008 and 2014. All patients were recently symptomatic with vision loss. A chronological record of physician visits and diagnostic tests (activities) was generated for each patient using claims data. Algorithmic process discovery was attempted using the Heuristic Miner. The effect of activity selection on treatment delay was investigated from two perspectives: activity-specific effects were measured using linear regression, and patterns of activity co-occurrence were identified using K means clustering. Results Ninety patients were included, with a median symptom-to-surgery treatment time of 34 days. Every patient had a unique sequence of activities. The flowchart generated by the Heuristic Miner algorithm was uninterpretable. Linear regression models of waiting time revealed beneficial effects of emergency and neurology visits, and detrimental effects of carotid ultrasound and post-imaging follow-up visits to family physicians and ophthalmologists. K-means clustering identified two co-occurrence patterns: emergency visits, neurology visits and CT angiography were more common in a cluster of rapidly treated patients (median symptom to surgery time of 18 days), whereas family physician visits, carotid ultrasound imaging and post-imaging follow-up visits to eye specialists were more common in a cluster of patients with treatment delay (median time of 57 days). Conclusions Routinely collected data provided a comprehensive account of events in the symptom-to-surgery process for carotid endarterectomy. Linear regression and K-means clustering can be used to analyze real-world data to understand causes of delay in complex healthcare processes.

Abstract Background and purpose Studies of carotid endarterectomy (CEA) require stratification by symptomatic vs asymptomatic status because of marked differences in benefits and harms. In administrative datasets, this classification has been done using hospital discharge diagnosis codes of uncertain accuracy. This study aims to develop and evaluate algorithms for classifying symptomatic status using hospital discharge and physician claims data. Methods A single center’s administrative database was used to assemble a retrospective cohort of participants with CEA. Symptomatic status was ascertained by chart review prior to linkage with physician claims and hospital discharge data. Accuracy of rule-based classification by discharge diagnosis codes was measured by sensitivity and specificity. Elastic net logistic regression and random forest models combining physician claims and discharge data were generated from the training set and assessed in a test set of final year participants. Models were compared to rule-based classification using sensitivity at fixed specificity. Results We identified 971 participants undergoing CEA at the Vancouver General Hospital (Vancouver, Canada) between January 1, 2008 and December 31, 2016. Of these, 729 met inclusion/exclusion criteria (n = 615 training, n = 114 test). Classification of symptomatic status using hospital discharge diagnosis codes was 32.8% (95% CI 29–37%) sensitive and 98.6% specific (96–100%). At matched 98.6% specificity, models that incorporated physician claims data were significantly more sensitive: elastic net 69.4% (59–82%) and random forest 78.8% (69–88%). Conclusion Discharge diagnoses were specific but insensitive for the classification of CEA symptomatic status. Elastic net and random forest machine learning algorithms that included physician claims data were sensitive and specific, and are likely an improvement over current state of classification by discharge diagnosis alone.