Search

Timely outreach to individuals in an advanced stage of illness offers opportunities to exercise decision control over health care. Predictive models built using Electronic health record (EHR) data are being explored as a way to anticipate such need with enough lead time for patient engagement. Prior studies have focused on hospitalized patients, who typically have more data available for predicting care needs. It is unclear if prediction driven outreach is feasible in the primary care setting. In this study, we apply predictive modeling to the primary care population of a large, regional health system and systematically examine the impact of technical choices, such as requiring a minimum number of health care encounters (data density requirements) and aggregating diagnosis codes using Clinical Classifications Software (CCS) groupings to reduce dimensionality, on model performance in terms of discrimination and positive predictive value. We assembled a cohort of 349,667 primary care patients between 65 and 90 years of age who sought care from Sutter Health between July 1, 2011 and June 30, 2014, of whom 2.1% died during the study period. EHR data comprising demographics, encounters, orders, and diagnoses for each patient from a 12 month observation window prior to the point when a prediction is made were extracted. L1 regularized logistic regression and gradient boosted tree models were fit to training data and tuned by cross validation. Model performance in predicting one year mortality was assessed using held-out test patients. Our experiments systematically varied three factors: model type, diagnosis coding, and data density requirements. We found substantial, consistent benefit from using gradient boosting vs logistic regression (mean AUROC over all other technical choices of 84.8% vs 80.7% respectively). There was no benefit from aggregation of ICD codes into CCS code groups (mean AUROC over all other technical choices of 82.9% vs 82.6% respectively). Likewise increasing data density requirements did not affect discrimination (mean AUROC over other technical choices ranged from 82.5% to 83%). We also examine model performance as a function of lead time, which is the interval between death and when a prediction was made. In subgroup analysis by lead time, mean AUROC over all other choices ranged from 87.9% for patients who died within 0 to 3 months to 83.6% for those who died 9 to 12 months after prediction time.

Graphical abstract, Data quality is essential to the success of the most simple and the most complex analysis. In the context of the COVID-19 pandemic, large-scale data sharing across the US and around the world has played an important role in public health responses to the pandemic and has been crucial to understanding and predicting its likely course. In California, hospitals have been required to report a large volume of daily data related to COVID-19. In order to meet this need, electronic health records (EHRs) have played an important role, but the challenges of reporting high-quality data in real-time from EHR data sources have not been explored. We describe some of the challenges of utilizing EHR data for this purpose from the perspective of a large, integrated, mixed-payer health system in northern California, US. We emphasize some of the inadequacies inherent to EHR data using several specific examples, and explore the clinical-analytic gap that forms the basis for some of these inadequacies. We highlight the need for data and analytics to be incorporated into the early stages of clinical crisis planning in order to utilize EHR data to full advantage. We further propose that lessons learned from the COVID-19 pandemic can result in the formation of collaborative teams joining clinical operations, informatics, data analytics, and research, ultimately resulting in improved data quality to support effective crisis response.