Abstract
Background/Aims Regular physical activity (PA) markedly improves health, physical function, independence, mood, and quality of life while reducing the risk of obesity, diabetes, cardiovascular disease, and some cancers. However, most people in the US do not meet recommended PA levels. The primary care setting is well-suited to long-term follow-up with repeated assessment and modification of treatment plans (e.g., for controlling hypertension). This approach is adaptable to exercise assessment and counseling—“exercise as medicine”. Kaiser Permanente (KP) recently introduced a PA promotion initiative that includes a new electronic medical record (EMR) feature to document current PA level (called “Exercise as a Vital Sign” or EVS). Provider counseling and referral for exercise items do not appear in structural fields in the EMR. We therefore plan to use natural language processing to assess the efficacy of EVS.
Methods We developed a method to measure the prevalence of leisure-time PA assessment, counseling, and referral in electronic medical records (EMRs) in primary care. The protocol was adapted from the “5As” of smoking cessation counseling, a motivational interviewing strategy designed to effect behavioral change, and operationalizes how clinicians document PA status and counseling for PA, including key concepts, phrases, words, numbers, orders, and procedures. Three chart reviewers used tools provided by the CER Hub to identify these measures in 400 charts each from two healthcare organizations. We assessed the internal validity of the measures through analysis of the manual chart review data.
Results Abstractors as a group (n = 3) reliably identified components of PA counseling, including non-specific advice, personalized advice, and referral for PA, in approximately 25% of encounters from a health plan that had implemented EVS. Significantly fewer instances of PA counseling were found in encounters from a second health system without EVS.
Conclusions This preliminary work supports the development of strategies to create a Mediclass program to automatically search the ambulatory component of EMR for these measures. We plan to assess PA counseling before and after the implementation of EVS, which has occurred at different times in different regions of KP, allowing a lagged baseline design for evaluation of relationships between EVS, PA counseling, and health outcomes.
