Outcomes of a Program to Support Outpatient Antibiotic Adherence Among People Experiencing Housing Instability With Serious Injection Drug-Related Infections in Boston, MA: A Propensity Score Matched Study
Nicole C McCann, Elizabeth Glaser, Jake R Morgan, Dominic Hodgkin, Ghulam Karim Khan, Simeon D KimmelAbstract
Background
Antibiotics are effective for treating serious injection drug-related infections (SIRIs). However, completion rates are often low among people who inject drugs. We evaluated Boston Healthcare for the Homeless Program Infectious Disease Outreach Liaison Team (BOLT), an outpatient liaison program for people experiencing housing instability hospitalized with SIRI, intended to increase intravenous or oral antibiotic completion.
Method
We included BOLT enrollees (1 June 2023 to 31 August 2024) and historical controls (1 June 2020 to 28 February 2023). Primary outcomes were the number of re-hospitalizations and emergency department (ED) visits in the 90-day period following the SIRI index hospitalization. The secondary outcome was probability of patient-directed discharge (PDD) from the index hospitalization. BOLT enrollees and historical controls were matched on demographic and clinical characteristics. In regression analyses, we further adjusted for observed characteristics in a doubly robust approach.
Results
Prior to matching, BOLT enrollees and historical controls were significantly different on most characteristics, with BOLT enrollees having higher prevalence of osteomyelitis and epidural abscess SIRIs, and opioid and stimulant use disorders. The matched analytic sample included 74 BOLT enrollees and 121 historical controls. The matching approach achieved balance on most characteristics except osteomyelitis prevalence and methadone receipt. In regression analyses, 90-day re-hospitalization was similar between BOLT enrollees and controls (P = .982). BOLT enrollees had an adjusted mean of 0.35 fewer ED visits per enrollee versus controls over the 90-day postindex hospitalization period (P = .288). Probability of PDD was similar between groups (P = .946).
Conclusions
BOLT was not significantly associated with changes in acute health care utilization. This may reflect unobserved differences between BOLT enrollees and historical controls and suggests a need for additional research to understand program implementation and more proximal health outcomes.