Impact and cost of an HIV/AIDS improvement intervention in Cote d’Ivoire
The USAID Health Care Improvement (HCI) Project supported the Ministry of Health and implementing partners in Cote d’Ivoire from 2010-2013 to implement improvement interventions at selected HIV treatment facilities. This evaluation sought to assess the impact and cost-effectiveness of the improvement interventions supported by HCI in terms of program-attributable changes in patient utilization and morbidity.
The analysis used a retrospective cohort design based on patient records at 26 primary- and secondary-level facilities in Abidjan (public and private) that provide HIV care and treatment. Half were sites where HCI interventions had been implemented, and the other half did not participate in the intervention. Data were abstracted from patient and facility records by trained research assistants. Each facility contributed approximately 45 records from patients who had initiated antiretroviral therapy (ART) on or before March 2013, plus a subset of 10 records for an assessment of data completeness, and facility costs (medicines, laboratory supplies, human resources, plus HCI program expenses).
The study was led by a researcher from Harvard University who collaborated with HCI staff in Cote d’Ivoire and headquarters to design and implement the study.
Unadjusted analyses showed that patients in HCI site received better care in the first six months of ART initiation: 46% of patients at HCI sites received follow-up care within six months of initiating ART, versus 40% of patients at non-HCI sites (p=0.03). But this difference did not persist for more than the six months.
However, the model adjusting for confounders found no such associations in shorter or longer time periods. Patients at HCI sites had about the same odds of care within six months of initiating ART (odds ratio = 1.3-1.4 in models with different sets of covariates, p>0.05). On average, patients had a CD4 increase of 158 cells/mm3 and an average weight gain of 6.5 kilograms, with no difference between HCI and non-HCI sites (p>0.05).
This lack of a significant program-related effect may be due to the high proportion of missing information. Very few patient records included care after baseline, and although patient records were better for certain items at HCI-supported sites, this was not consistent and the overall prevalence of missing data was still high. For instance, only one-third of records from HCI sites and two-thirds of records from non-HCI sites included baseline CD4 values; and 5% of records from HCI sites and 24% from non-HCI sites reported 6-month exam data. Cost-effectiveness was ultimately not analyzed because a program effect was not found and because of missing data on relevant costs.