|Type||Thesis or Dissertation - Doctor of Philosophy|
|Title||Feasibility, effectiveness, budget impact and surveillance of partner services for HIV in Kenya|
HIV assisted partner services (aPS), is widely practiced in the United States and Europe but
less so in Africa largely due to limited data on their effectiveness and feasibility in routine health care
settings. Yet aPS could increase HIV testing rates, reduce STI/HIV exposure, and assure prompt linkage
to antiretroviral therapy (ART) initiation. We report the effect of immediate aPS in improving 1) the rate of
HIV testing, 2) case-finding of HIV-infected individuals, and 3) linkages to HIV care for their partners. The
trial determined the number needed to interview to identify one new case of HIV and explored,
geographical differences in case finding rates. Additionally we assess the budget impact of scaling up
aPS in Kisumu County, the region with the third highest number of HIV-infected persons in Kenya. We
also present results of the pilot of a surveillance system for aPS.
Methods: A cluster-randomized design was used to recruit eligible HIV-infected index cases from 18
clusters allocated to two study arms, immediate and delayed. The intervention was elicitation of sexual
history from index cases and enumeration of sexual partners of HIV infected index cases in the preceding
three years, notification, testing and referral to care if HIV-infected, of the named sexual partners.
Participants in the delayed arm received a similar service only that this was delivered six weeks later. We
used generalized estimating equations to evaluate the effect of the intervention on rates of HIV testing,
identification of new HIV tests, HIV infections and enrollment to HIV care. The number of index cases
needed to interview and the case finding rates were also evaluated using a similar approach. The study
was registered in ClinicalTrials.gov as number NCT01616420.
To estimate the budget impact, we constructed an Excel-based costing tool to simulate the budget impact
analysis of HIV partner services on an annual basis over a 5 year time horizon. HIV Testing and
Counseling (HTC) and aPS unit and total costs were estimated and allocated using ingredient-based
approaches. Time motion was used to determine full-time equivalent of tracing sexual partners of index
patients. Weighted costs of ART, clinic visits and hospitalizations that accrued due to aPS were
generated through decision tree modeling. We estimated a range, where the lower-bound cost assumed
that all sexual partners tested were HIV-negative and the upper-bound cost assumed that all sexual
partners were HIV-positive. All costs were undiscounted and reported separately for the task-shifting
scenarios. Appendix I outlines the assumptions regarding the patient workload and the number of
providers available to do aPS in Kisumu County. Appendices II & III are the assumptions for calculating
HIV testing costs and budget impact respectively. Appendix IV is the decision trees for determining
expected costs for antiretroviral therapy, clinic visits and hospitalization visits.
For the pilot of the surveillance system, we revised the national HTC data collection tool to include
specific questions on whether clients testing for HIV were doing so due to an exposure from an HIVinfected
Results: The study enrolled 1119 index cases from 18 different clusters (550 in the intervention arm and
569 in the control arm) who mentioned 1872 sexual partners. Of the sexual partners, 1292 (69%), [620 in
the intervention arm and 672 in the control arm] were enrolled. Enrollment and follow-up data were
available for 579 (63%) of sexual partners mentioned in the immediate arm and enrollment data available
for 672 (70%) in the delayed arm. 388 in the immediate arm during enrolment and within the study after
enrollment, and 118 in the delayed arm in the preceding two months before enrollment (Incidence Rate
Ratio (IRR) 3.78, 95% CI: 3.08-4.65). The incidence rate ratio comparing rates of new testing for HIV
between the immediate and delayed groups was 11.50 (95% CI: 5.56-23.78). Immediate aPS also
increased the number testing positive and those enrolled in HIV care, IRR 3.22 (95% CI: 2.26-4.61) and
3.95 (95% CI: 2.48-6.28) respectively. The number of index cases needed to interview (NNTI) to identify
an HIV infection in the partners was 4.08, and that to identify a newly testing partner was 3.34. No studyrelated
intimate partner violence was reported.
The average annual aPS costs are US$ 1,092,161 and US$ 753,547 for Kisumu County using nurses
and CHWs, respectively. The weighted average cost of scaling up aPS over a five period using nurses
was 45% higher compared to CHWs (US$ 5,460,837 and US$ 3,767,738 respectively). Overall, the
differences between the upper and lower bound costs were 8.7% for nurse-based aPS and 2.5% for the
CHW-based approach. Over the time horizon, the total budget impact of nurse-model was US$
1,726,832, 69.2% and 29.5% of which were accounted for by aPS costs and ART costs respectively. The
CHW model incurred an incremental cost of US$ 1,184,640, 68.6% lower than the nurse-based model.
Proportional distribution of impact across budget categories was similar in the two models, although
CHWs model had lower aPS related impact
The weighted unit costs of HIV testing across the three levels of facilities for HIV-infected index clients
using nurses were US$ 25.36 and US$ 17.86 using CHWs. Costs for testing sexual partners of infected
index clients were higher overall, with an HIV test costing US$ 19.18 per person if all tests were negative
and US$31.07 per person if all tests were positive for nurses and US$ 11.74 per person and US$ 14.14
per person for CHWS respectively.
Median time for data capture using the HTC form was 4 minutes (IQR: 3-15), with a longer duration for
HIV-infected participants, and there was no reported data loss.
Interpretation: aPS is safe, effective and feasible at the population level and should be implemented as
part of HIV Testing and Counseling (HTC delivery). In addition to early ART initiation, aPS may have
considerable effect on HIV transmission at the population level. Furthermore, aPS is affordable although
not cost-saving and routine Health Information Systems (HIS) may be used to monitor aPS outcomes.
|»||Kenya - AIDS Indicator Survey 2012-2013|