(112) Exploring Anticholinergic Co-Medication Networks in HIV and Frailty: An Exploratory Random Graph Modelling Approach

Background: Medications with anticholinergic properties can exacerbate frailty. Although network analysis has been used to describe anticholinergic co-prescription patterns in HIV-related frailty, there's a gap in inferential network analysis. Exponential Random Graph Models (ERGM) enable us to make inferences about whether specific network signatures will appear more often than we would by chance alone.

Objectives: To estimate the effect of anticholinergic properties of medications on the probability of connections within a medication co-prescription network among individuals at different stages of frailty within a population aging with HIV.

Methods: Using baseline data from 824 individuals living with HIV, aged 35+, from the Canadian Positive Brain Health Now cohort, we examined the impact of anticholinergic drug use on frailty. Anticholinergic exposure was measured using the Anticholinergic Cognitive Burden scale, and frailty was assessed with Modified frailty phenotype criteria. ERGM estimated the probability of co-prescription ties in the network based on anticholinergic properties and burden score, using Markov Chain Monte Carlo Maximum Likelihood for parameter estimation. Network coefficient differences across frailty states were analyzed with a z-test. Statistical analysis was conducted using ergm package in R.

Results: An increase in anticholinergic burden reduced the odds of similar burden medications being co-prescribed by 9% in frail, 20% in prefrail, and 22% in robust individuals. Medications with higher anticholinergic burden are less likely to be part of multiple prescriptions. When considering frailty states, anticholinergic medications are generally less likely to be co-prescribed, with the odds being lowest in robust individuals (OR=0.63 [0.57, 0.69]), slightly higher in prefrail (OR=0.66 [0.61, 0.71]), and highest in frail individuals (OR=0.84 [0.77, 0.92]). Anticholinergics were less “isolated” in frail networks than in robust (z = -4.33, p< 0.001) and prefrail networks (z = -3.98, p< 0.001). There was a significant difference in co-prescription patterns of anticholinergics in frail patients as opposed to robust or prefrail ones, with frail patients being less likely to have these medications prescribed in isolation.

Conclusions: The distinct patterns of anticholinergic use and burden in individuals with HIV who are frail compared to robust and prefrail individuals indicate tailored prescription practices based on health needs. This study underscores the importance of medication review and potential deprescribing to mitigate adverse effects from high anticholinergic loads.