Разработанный алгоритм на базе электронных медицинских карт с высокой точностью выявляет детей с высокой медицинской потребностью

A standardized approach that employs electronic health record (EHR) data was able to identify children and youth with medical complexity (CMC). The algorithm was used to recruit adolescent patients to a program for transitioning their care to adult services. These study results were published in BMJ Health & Care Informatics. A team of researchers developed a novel algorithm to identify CMC among pediatric patients at the Hospital for Sick Children (SickKids) in Toronto, Canada. Their goals were to identify pediatric patients aged 17 to 18 years for potential referral to a program for transitioning CMC to adult care, and in a second phase, applied the algorithm to a separate cohort of children aged 1 to 17 years to assess its generalizability across the broader pediatric population. The data set used to identify records for retrospective abstraction and chart review was the Sick-Kids Enterprise-wide Data in Azure Repository (SEDAR), an EHR-sourced data set at SickKids. Using the criteria for CMC set out by the referral-based program Complex Care for Kids Ontario (CCKO), the research team translated that criteria into Structured Query Language (SQL) code and created the algorithm CMC-ID, which was then refined through an iterative process. The unique charts identified were reviewed by clinical research staff to make sure they truly met the criteria. Ultimately, 6 versions of the algorithm were reviewed across 5174 eligible patients (280 unique charts across all versions of the algorithm). Version 1 generated the highest number of true positives on chart review, or 67 of 235 patients who were identified as transitioning CMC (positive predictive value [PPV], 28.5%; 95% CI, 22.8-34.7). However, version 6 performed best, identifying 47 true positives out of 55 identified patients (PPV, 85.5%; 95% CI, 73.3-93.5). Of those 47, 19 (40%) were female, and the mean age was 17.5 years. In the second phase of the study, version 6 identified 947 true positives in the younger cohort (PPV, 89.1%; 95% CI, 86.9-91.0). Inter-rater reliability for this cohort was high, with a corresponding Cohen’s kappa of 0.73. In this cohort, 362 (43%) were female, and the mean age was 8.0 years. “The application of algorithms to EHR-sourced data sets represents a promising approach to identifying CMC for a range of purposes, including recruitment to targeted interventions or clinical programmes, quality improvement and population health monitoring,” the researchers wrote. The researchers explained that the CMC-ID was developed, optimized for PPV, to reliably identify CMC in EHRs within a large pediatric hospital in Canada. Its best use may be as a high-alert index to support recruitment into complex care programs. They noted that their EHR-based approach may be “particularly well suited for identifying more severely affected, high-needs CMC” more accurately and efficiently. Although the algorithm should not replace clinician referrals, it may substantially reduce manual screening burden and improve access to targeted programs and services among the CMC population. Disclosures: The study authors have no funding or conflicts of interest to disclose.