Evolent Health partners with health systems across the United States to improve quality of care. Polypharmacy is a focus area due to its prevalence in the elderly and chronically ill patients, and its association with increased cost, adverse drug effects, medication non-adherence and other suboptimal outcomes. Definitions of polypharmacy range from measuring total number of medications with various cut-offs to the use of high-risk or unnecessary medications.
Before SNOMED CT encoded data were available from clinical data systems, Evolent had tested and implemented a traditional polypharmacy rule with a cut-off of nine medications based on pharmacy claims data. Although this is a standard industry approach, it has several limitations. It does not account for over-the-counter medications, it can identify a large number of patients without the ability to prioritize them for medication review services, and it can miss patients who are just under the threshold but are taking high-risk or unnecessary medications.
Recently Evolent began to integrate inpatient clinical data from a large integrated health system on the east coast of US. We examined the prevalence of “polypharmacy” in a SNOMED CT encoded problem list of the health system’s Electronic Health Record (EHR). This information provides an opportunity to enrich the claims-based rule with additional clinical judgement. An analysis was carried out to determine the utility of this SNOMED CT clinical finding and the best implementation strategy.
How SNOMED CT will be used
Firstly, chronic medication data were extracted from EHR and pharmacy claims. Data were merged by semantic mapping between drug classification systems and then aggregated on ingredient level.
Given the ambiguous definition of “polypharmacy”, the next step was to quantify how this finding agrees with other data. A challenge is the absence of start date and abatement date, which were inferred from admission history. The number of chronic medications were counted in each 3-month window between the inferred date ranges.
Among the approximate 1200 patients with this clinical finding, 93% had 4 or more medications at least once, and 55% had 9 or more. This is a moderately good agreement and, importantly, suggests that other clinical knowledge may have contributed to this finding. However, only 11% of the findings appeared to be still active in 2017. A cross walk to procedural data found no record pertaining to relevant SNOMED CT procedures such as “medication review”. Therefore, the setting where this finding was made is not entirely clear.
Why SNOMED CT will be used
National mandation through the Meaningful Use programme, and the hierarchical structure of SNOMED CT.