Appropriate prescribing and centrality measures- a social network analysis.
Sophie Wang, Nicolas Larrain, Oliver GroeneIntroduction
Physicians are naturally embedded in social networks where information exchange occurs. Research shows that where individuals are positioned and their relationship with others in the network are important predictors of the individual’s performance. Previous literature examining prescribing decision-making has largely been limited to inter-individual influences, leaving meso level factors relatively unexplored. By leveraging the wealth of administrative data collected, we construct patient sharing physician networks to better understand how relationships affect quality of care. The objective of this research is to investigate the relationship between inappropriate prescribing rates and network centrality measures.
We derived physician networks from shared patients using claims data in the year 2015 set in the Southwest of Germany. We used a threshold of 5 shared patients or more to determine a meaningful tie based on a validated method (Barnett et al 2011). We examined degree and betweenness, network analysis measures that indicate connectedness and influence, respectively. Inappropriate prescribing was captured using FORTA score, which is a patient level score for seniors (aged 65 or older) representing the number of inappropriate medications within each quarter based on a validated algorithm. We used linear regression to examine whether physician network indicators were associated with inappropriate prescribing scores after accounting for patient demographics, risk factors, and physician’s patient volume. All analyses were performed using R version 3.6.2.
Our cross-sectional network comprised of 222 physicians linked by 28,409 patients (58% female), with a network density of 0.24 and diameter of 5. Total number of inappropriate medications prescribed per patient ranged from 0 to 79.00, with a mean of 18.34. Preliminary results using linear regression show that patients under the care of physicians in an influential position within her network, as measured in betweenness score (β = 18.72, 95%CI: 14.58 to 22.86) has lower inappropriate medications prescribed within the calendar year. The average degree across all physicians seen in the year did not have a strong effect degree (β = -0.0015, 95%CI: -0.0034 to 0.0004).
We found preliminary evidence that patients whose physician occupies an influential position in her network is strongly associated with lowered inappropriate medications prescribed compared to connectedness. By occupying an influential position with increased information flow and exchange, it is conceivable that the physician may provide more coordinated care and be more up to date on recent prescribing guidelines.