Profiles of HIV Care Disruptions Among Adult Patients Lost to Follow-up in Zambia: A Latent Class Analysis

Abstract

Background: Patients report varied barriers to HIV care across multiple domains, but specific barrier patterns may be driven by underlying, but unobserved, behavioral profiles.

Methods: We traced a probability sample of patients lost to follow-up (>90 days late) as of July 31, 2015 from 64 clinics in Zambia. Among those found alive, we ascertained patient-reported reasons for care disruptions. We performed latent class analysis to identify patient subgroups with similar patterns of reasons reported and assessed the association between class membership and care status (ie, disengaged versus silently transferred to a new site).

Results: Among 547 patients, we identified 5 profiles of care disruptions: (1) "Livelihood and Mobility" (30.6% of the population) reported work/school obligations and mobility/travel as reasons for care disruptions; (2) "Clinic Accessibility" (28.9%) reported challenges with attending clinic; (3) "Mobility and Family" (21.9%) reported family obligations, mobility/travel, and transport-related reasons; (4) "Doubting Need for HIV care" (10.2%) reported uncertainty around HIV status or need for clinical care, and (5) "Multidimensional Barriers to Care" (8.3%) reported numerous (mean 5.6) reasons across multiple domains. Patient profiles were significantly associated with care status. The "Doubting Need for HIV Care" class were mostly disengaged (97.9%), followed by the "Multidimensional Barriers to Care" (62.8%), "Clinic Accessibility" (62.4%), "Livelihood and Mobility" (43.6%), and "Mobility and Family" (23.5%) classes.

Conclusion: There are distinct HIV care disruption profiles that are strongly associated with patients' current engagement status. Interventions targeting these unique profiles may enable more effective and tailored strategies for improving HIV treatment outcomes.

FIGURE 1.

Patient flowchart. As of July 1, 2015, 28,117 patients who had ever initiated ART were considered LTFU across 64 sites and 2898 were randomly selected for active tracing from 32 sites. Among patients selected for tracing, 1007 (34.8%) were found alive without a care disruption, 412 (14.2%) had died, and we were unable to trace 932 (32.2%). We ascertained patient-reported reasons for care disruptions among the 547 (18.9%) patients who we found alive with a confirmed care disruption.

FIGURE 2.

Profiles of care disruptions (n = 547). Patient profiles of care disruption are based on latent class models based on the patient-reported reasons for care disruptions and the number of reasons a patient reported. The estimated proportion of patients in each latent class are in parentheses at the top, and bars correspond to the probability of reporting a particular reason for care disruption within each class. Models used population-representative sampling weights after tracing a random sample of patients who were considered lost to follow-up as of July 31, 2015.

FIGURE 3.

Estimated prevalence of disengagement by latent class and sex (n = 547). Estimated prevalence of disengagement based on marginal estimates from an adjusted Poisson regression that included an interaction term latent class and sex. Regression model incorporated population-representative sampling weights after tracing a random sample of patients who were considered lost to follow-up as of July 31, 2015. The P-value for the interaction term was 0.246.