A Community HIE-based Hospital Readmission Risk Prediction & Notification System

点击翻译按钮获取中文摘要

v Project Summary Hospitals account for nearly $1 trillion (~30%) of US health care spending. Preventing avoidable hospital readmissions is considered a key opportunity for reducing waste in health care. Around 18% of patients are readmitted to hospitals within 30 days. It is estimated that almost 14% of Medicare readmissions - costing $12 billion annually - are preventable. Despite hospital-centric interventions 'avoidable' readmissions have increased steadily over the last decade. Ambulatory care is integral to reducing hospital readmissions. Patients lacking timely Primary care physician (PCP) follow ups are 10 times more likely to be readmitted. Around 75% of discharge-summaries are never received by PCPs and 50% of readmitted Medicare patients never had a follow up office visit. An integrated health IT system (e.g., a health information exchange) can increase PCP engagement in reducing avoidable readmissions. HIEs can efficiently coordinate care between hospitals and PCPs thus improving the promptness and reliability of follow-up care. As of 2012, Maryland's HIE (known as 'CRISP') collects real-time data of ~4 million unique patients, including 250k patients that have been explicitly registered by more than 400 PCPs for special cross-practice monitoring using CRISP's Encounter Notification System (ENS). ENS notifies participating providers of readmission and other sentinel events. The reporting system lacks any actionable tool to 'predictively' identify the patient's risk of readmission. The activities of this research will be significantly catalyzed by Maryland hospital and PCP community's advanced stage of focus and programmatic readiness to avoid readmission. Most hospitals in Maryland are not paid for all readmissions due to state's all-payer commission. All hospital discharge teams are especially trained and virtually all PCPs in the state are now part of an advanced patient centered care medical home (PCMH) model. The project will be based within The Johns Hopkins ACGs (adjusted clinical groups, formerly ambulatory care groups) that is one of the most advanced in the nation in developing and testing predictive models. ACG has been used for two decades across the nation and in 15 other countries, and applied to 60+ million patients to help predict various healthcare events using claims and admin data. While readmission risk predictive models (RRPM) are very common post-facto using health plan claims and hospital administrative databases, they have rarely utilized HIE data for derivation or validation. In this exploratory research, Aim 1 is to develop and evaluate an RRPM based on the population covered by CRISP to calculate a risk score for each patient discharged from Maryland hospitals in real-time. Aim 2 will be to test the RRPM's accuracy through an iterative enhancing process that will integrate non-HIE data sources to explore the most valuable future data sources (soon to be, but not yet available to CRISP); and, to test the generalizability of CRISP's RRPM at CCBC (Crescent City Beacon Community). Finally, in Aim 3 the RRPM notification system will be prototyped and integrated into CRISP's ENS. A qualitative study will be conducted to evaluate the potential effectiveness and usability of the HIE-derived-and-delivered RRPM among participating PCPs.