(121) Identifying Clustering of Clinic-level Prescribing Patterns for Patients with Differing Health Coverage in Primary Care in Ireland: a Cohort Study

Background: Ireland has a mixed system of healthcare entitlement; based on age/income some patients have coverage for physician visits and prescriptions. The remainder pay out of pocket. In primary care, general practitioners (GPs) provide care to both patient groups, giving an opportunity to explore how prescribing patterns may differ based on health coverage.

Objectives: To determine the number and types of drugs making up the majority of prescribing in primary care in Ireland by health coverage, and explore clustering of prescribing patterns across clinics.

Methods: This cohort study was set in 43 GP clinics in Ireland. It used anonymous patient record data extracted from these clinics for all patients aged ≥65 years for 2011-2018. Prescribing data was coded using the WHO Anatomical Therapeutic Chemical (ATC) classification system, and health coverage of each patient was recorded (classed as ‘public’, for full public coverage of physician visits and drugs, and ‘private’ for all other patients). The Drug Utilisation 90% indicator (DU90%) was applied, which identifies the number and type of drugs that make up 90% of prescribing at the clinic/prescriber level. This was applied overall and at clinic level, and separately by health coverage. We compared drugs included in the DU90% segment for public patients with that for private patients. K-means clustering (k=3) was used to group clinics based on inclusion of drugs in the clinic’s DU90% segment, and these clusters were compared based on prescribing and other characteristics.

Results: For all clinics, the DU90% contained 162 drugs (160 and 166 for public and private prescribing respectively). At the clinic level, the median number of drugs in the DU90% was 136.5 (range 105-163) for public prescribing, and 145 (range 100-170) for private prescribing. Thirty-nine drugs were included in a clinic’s public DU90% and not the private, most often cardiovascular (CV) drugs (35.9%, n=14). For the 369 drugs in a clinic’s private but not public DU90%, central nervous system (CNS) and CV were most common (24.1% and 23.3% respectively). Clustering analysis identified three clusters: a high number of drugs in the clinic DU90% cluster (n=7 practices), and intermediate (n=13) and low (n=23) cluster. Differences in mean number of drugs from therapeutic subgroups included in the DU90% across clusters were noted for antibacterials, lipid lowering agents, and several CNS subgroups. The high cluster had the highest proportion of prescriptions for GMS patients (80.1%),
and the low cluster had the lowest (76.4%).

Conclusions: A higher number of distinct drugs are typically used for private prescribing, due to a greater variety of mainly CV and CNS drugs. This may be driven by private patient expectations and has potential implications for prescribing quality.