This dissertation uses three types of routinely collected data from HIV care and treatment facilities in sub-Saharan Africa to investigate the association between the availability of adherence support and active outreach services on patient non-retention, loss to follow-up, and measured death after ART initiation. Following a literature review summarizing the state of knowledge concerning the influence of programmatic services on patient retention in care and survival, these relationships are first examined in an aggregate analysis of over 232,000 patients at 349 HIV care and treatment facilities initiating ART between January 2004 and December 2008. Key findings are that several adherence support and outreach services are associated with reduced rates of non-retention, loss to follow-up, and death. Specifically, facilities offering three or more adherence support services, written educational materials promoting ART adherence, one-on-one or group adherence counseling sessions, reminder tools, and food rations to promote ART adherence were associated with reduced non-retention and loss to follow-up, while facilities offering on-site support groups for HIV+ patients, peer educators, provision of reminder tools, and food rations to promote ART adherence were associated with reduced death rates. In sub-analyses investigating six- and 12-month retention after ART initiation, facilities offering three or more separate adherence support services, routine review of medication pickup and/or dedicated ART pharmacists, and active patient outreach to trace patients missing visits had lower non-retention. Taken together, this analysis provides evidence that program-level services found efficacious in experimental settings are also effective in operational settings.
Next, a sub-analysis is conducted among facilities also providing electronic patient-level data to investigate similarities and differences in the association between adherence support and outreach services and patient non-retention, loss to follow-up, and measured death using aggregate vs. patient-level estimates of these outcomes, and to assess whether adjustment for patient-level differences between facilities change these measures of association. In multivariate analyses, clinics offering active patient outreach had lower rates of non-retention in both the ART cohort analysis and the patient-level analysis, and clinics offering food rations to promote ART adherence were associated with a lower risk of ascertained death in both the facility-level and patient-level analyses, but this association was diminished after adjustment for patient-level covariates. In contrast, various adherence counseling or support services were associated with lower non-retention in the ART cohort analyses but not in the patient-level data analyses. When compared with the results in the first paper, fewer associations were observed, suggesting either that the countries with patient-level databases are not representative of the entire range of HIV care and treatment facilities assessed in the first paper, and/or the specific facilities with electronic databases are more similar to each other than they are to facilities without electronic databases.
Finally, the dissertation concludes with an investigation into the relationship between loss to follow-up and measured death. For this analysis, estimates of the death probability among patients lost to follow-up are created under varying assumptions (either assuming that the death probability among those lost to follow-up is equivalent to the death probability within various strata of covariates, or assuming that the probability of death is greater among patients lost to follow-up). Key findings from this analysis are that ratio comparisons of death rates between facilities offering different services are robust to changes in the death probability if patients lost to follow-up are assumed to have a similar probability of death, conditioned on covariates, as those not lost to follow-up, but that associations between facility services and death rates are masked under the scenario where the facility service is associated with loss to follow-up and the death probability is assumed to be higher, conditioned on covariates, then the death probability among patients not lost to follow-up.
| Identifer | oai:union.ndltd.org:columbia.edu/oai:academiccommons.columbia.edu:10.7916/D82R4ZJH |
| Date | January 2011 |
| Creators | Lamb, Matthew Raymond |
| Source Sets | Columbia University |
| Language | English |
| Detected Language | English |
| Type | Theses |
Page generated in 0.0021 seconds