Big data to analyze patterns of care and improve outcomes for children with cerebral palsy

Abstract

In their paper, Kurowski et al. chose a particularly relevant example of complex clinical practice, as children with CP need variable, targeted, and adaptive care from multiple caregivers and in diverse settings. Based on a retrospective assessment of data made available through electronic health records, 6369 children (mean age 8y 2mo; range: 0–21y) were selected. Machine learning hierarchical clustering was used to determine clusters of care and the ratio of in-person to care-coordination visits calculated for each specialty. Seven clusters of care were identified, including musculoskeletal and function, neurological, high frequency and urgent care, procedural, comorbid diagnoses, developmental and behavioral, and primary care. Network analysis showed shared membership in several clusters. Inperson to care-coordination visit ratio was 1:5 overall for health care encounters, implying that most interactions with care teams occur outside of in-person visits. These results illustrate how health care data, when entered into an electronic health record, are not only collected, but also structured and analyzed. Such an analysis is not only beneficial for the patient through assessment of how care is provided for this particular individual, but also provides substance for a more global assessment of care models and how, overall, health care is provided in practice. The described elucidation of the ‘breadth and depth’ of the interaction of specific patient populations with the health care system can therefore lay the basis for streamlining care. Although less common than other chronic diseases, CP is particularly relevant for its complexity and its need for multidisciplinary care. Less complex but more common chronic diseases, such as congestive heart failure, diabetes mellitus, chronic obstructive pulmonary disease, or dementia, would be accessible to the same kind of analyses, with potential impacts with a magnitude that might at least equal the one reported for patients with CP. Ideally, then, the broader and systematic changes that could result from this approach would lead to benefits both for the individual patient, through more structured interventions, and for the system, through higher efficiency. Furthermore, the authors are correct to point out the potential benefits beyond measurable efficacy, for example in terms of potential decrease in stressors like uncoordinated and urgent visits to emergency departments. Although the population of children with CP may represent the psychological distress associated with illness particularly well, this relevant point, although not specifically addressed by the study, is indeed crucial in the care we provide for any patient. With the cost of providing health care increasing at more than the rate of gross domestic product in every industrialized country, smart systems are likely to play a significant part in the future of health care. By more accurately predicting the demand for services, waste can be reduced and insurance can become more efficient. However, adequate management of big data implies a significant evolvement of the hospitals’ information systems, with a strong interaction between medical practitioners and engineers to avoid irrelevant clinical data being analyzed and providing irrelevant results.