Automating assessment of obesity care quality

点击翻译按钮获取中文摘要

DESCRIPTION (provided by applicant): Current trends suggest that obesity prevalence will continue to rise and that costs of treating obesity-related disease will dramatically increase as the population ages. Despite NHLBI guidelines for preventing, diagnosing, and treating obesity among adults, most health care systems have been slow to respond to this looming public health problem. This slow response is partly due to the inability to assess adherence to obesity diagnosis and treatment guidelines. In particular, the lack of appropriate Health IT for integrating diverse clinical data on obesity, even within state-of-the-art electronic medical record systems (EMRs), makes it difficult to evaluate the quality of care, measure the effectiveness of new intervention programs, and make rational decisions at system, organization, and individual patient levels. EMRs offer the potential to efficiently assess large populations, however much of the data necessary for obesity care assessment are unavailable to automated methods because they reside in the text clinical notes of the EMR. Previous studies have shown that although some data of interest are recorded in easily retrievable fields (e.g., body weights recorded as a vital sign; standard diagnosis codes), much of the treatment information is found only in free-text clinical notes. This research aims to develop, validate, apply, and evaluate a scalable method for routine and comprehensive measurement of outpatient obesity care quality. To accomplish this, we will extend MediClass (a "Medical Classifier"), which is a proven technology for extracting care quality data from both coded data and free-text clinical notes in the EMR. This research will perform retrospective analysis of adult primary care from the EMR data of two distinct health systems: a mid-sized HMO (Kaiser Permanente Northwest, KPNW) and a consortium of public health clinics (OCHIN) including a diverse sample of patients, providers, and health care practices of the West Coast states (primarily Oregon, but also Washington and California). We propose to use Health IT to integrate diverse data and knowledge that advance quality improvement for both insured and the indigent, uninsured, and underinsured populations of this region. We will first develop obesity care quality (OCQ) measures using up to-date NHLBI guidelines for diagnosis and treatment of obesity. Next, we will develop and validate an automated method for applying these measures to comprehensive EMR data. At each study site, the Medi Class system will extract coded data and use natural language processing (NLP) on free-text clinical notes to identify OCQ-relevant clinical events in the EMR. Then we will apply the OCQ measures to assess current levels of obesity care quality in the two health systems. Finally, we will evaluate the associations between OCQ measures of recommended obesity care and provider characteristics as well as clinical outcomes for patients, including change in weight.