Cluster analysis to identify patient profiles and substance use patterns among pregnant persons with opioid use disorder

Abstract

The study objective was to identify distinct profiles of pregnant persons with opioid use disorder (PP-OUD) using cluster analysis and examine difference in substance use patterns between profiles. We examined data from 104 PP-OUD ≤ 32 weeks of gestation who were recruited into a behavioral health clinical trial at two academic medical centers. We used Partitioning Around Medoids analysis to identify clusters and explored patterns of substance use and substance use treatment between clusters using bivariate statistical tests and regression methods. We identified two distinct clusters of participants, including ‘Group A’ (n = 68; 65.4 %) and ‘Group B’ (n = 36; 34.6 %). Group A had fewer members who were not employed (38 % vs 58 %) and incarcerated (3 % vs 8 %) compared to Group B. Group A compared with Group B included more members with: a history of overdose (72 % vs 50 %); anxiety (85 % vs 25 %); ≥moderate pain (76 % vs 22 %); ≥moderate depression (75 % vs 36 %); ≥moderate drug use severity (94 % vs 78 %); and, more days of cannabis (mean: 6.2 vs 2.3 days), stimulant (mean: 4.5 vs 1.3 days), and injection heroin (mean: 1.3 vs 0 days) use in the past 30 days (P < 0.05 for all comparisons). Clusters of PP-OUD differed with respect to sociodemographic characteristics, mental health conditions, and substance use patterns. More research is needed to confirm identified profiles and assess treatment outcomes associated with cluster membership.