Impact of assisted HIV self-testing on case-finding rate and positive HIV diagnoses in Burundi: A province-level difference-in-difference analysis

Abstract

Background

HIV self-testing (HIVST) is a testing strategy to reach individuals not engaged in healthcare. We evaluated ICAP’s Reaching Impact, Saturation and Epidemic Control (RISE) program’s healthcare-worker-assisted HIVST on the number of persons testing HIV-positive (PTP) and HIV case finding rate (CFR) in “hard-to-reach” members of the general population in targeted provinces in Burundi.

Methods

Using routine data from October 2020 to September 2023, with October-December 2021 as the scale-up quarter, we used difference-in-difference analysis to assess changes in PTP and CFR after versus before scale-up-onset in “mature” versus “less mature” provinces. Provinces were chosen to approximate more and less exposed comparison units. We performed subgroup analyses by sex and age (10-24, ≥25 years).

Results

HIVST averaged 1.46% (mature) and 3.07% (less mature) of pre-scale-up testing, and 13.59% (mature) and 9.84% (less mature) of post-scale-up testing. PTP trends declined more in the mature versus less mature provinces post-scale-up (−34.25, 95% CI -84.87-16.37) although this was not significant. We found no difference in the CFR change between provinces post versus pre-scale-up (0.06%, 95% CI -0.29%-0.41%). Results were robust to subgroup and sensitivity analyses.

Conclusions

We found no changes in PTP and CFR after RISE’s HIVST scale-up. After results were reported to implementers, they implemented strategies to better meet programmatic goals.

Keywords

HIV prevention HIV self-testing program evaluation implementation science

Background

International HIV programming is guided by UNAIDS’ 2025 targets, aimed at reducing the burden of HIV on people and their communities. There are three main country-level goals (the ‘95-95-95 goals’) for completion by 2025: ensure that 95% of people living with HIV (PLWH) know their status; ensure that 95% of PLWH who know their status are on antiretroviral treatment (ART); and ensure that 95% of PLWH on ART are virally suppressed.¹ By 2023, Burundi, a densely populated country with a 0.9% HIV prevalence, had achieved 92-91-81.²

Finding the remaining status-unaware PLWH requires innovative programming and strategies geared toward reaching those who are historically unengaged in healthcare. Gaps in HIV testing are pronounced.^{3, 4} An analysis of the 2016-2017 Burundi Demographic Health Survey found that only 27.1% of women and 16.6% of men had received an HIV test in the last year.⁵ One innovative strategy for HIV testing outside of healthcare facilities is directly assisted HIV self-testing (aHIVST).⁶ Under aHIVST, sometimes referred to as supervised HIVST, rapid tests for HIV infection are taken in the presence of a healthcare worker or trained peer mentor at a location outside of health facilities, ideally one convenient to communities with testing needs. The healthcare worker or peer mentor may demonstrate the use of the test kit, help the client use the kit, and/or help interpret the result.⁶

Directly aHIVST has been found to be highly feasible, acceptable, and reliable.^7-12 In a systematic review comparing assisted and unassisted HIVST in mainly high-income settings, researchers found that both strategies were acceptable and preferred over traditional, facility-based testing, but assisted testing was more sensitive than unassisted testing.⁹ In a randomized trial comparing aHIVST to unassisted HIVST in the Democratic Republic of Congo, the rate of correctly interpreting the results was 86.9% in the unassisted arm and 93.2% in the assisted arm, and there were no differences in the case finding rate (i.e., the proportion of all administered tests that were HIV positive) between the arms.¹¹ In KwaZulu-Natal and Mozambique, studies on aHIVST found this testing strategy to be 98.7% sensitive and 100% specific⁸ and highly acceptable.¹⁰

In Burundi, aHIVST was first introduced as a testing strategy for key populations (KPs) (i.e., men who have sex with men [MSM], female sex workers [FSW]) in the country’s 2018-2022 addendum to its National HIV and STI Strategic Plan.^13,14 Through the U.S. President’s Emergency Plan for AIDS Relief (PEPFAR) “Linkages across the Continuum of HIV Services for KPs Affected by HIV” (LINKAGES) program, scale-up of aHIVST specifically for KP began in June 2018 in Bujumbura Mairie, Bujumbura Rural, Kayanza, Ngozi, Kirundo, and Gitega provinces.¹⁵ LINKAGES used peer outreach workers to provide aHIVST. In an analysis of program performance over the first 9 months, researchers found that the CFR after aHIVST was significantly higher for FSW (14.4% vs 9.26%) and MSM (13% vs 3.7%) than from traditional testing, suggesting that aHIVST was a success in these populations.¹⁵

Assisted HIVST was further scaled up under PEPFAR by ICAP-Columbia University through its Reaching Impact, Saturation and Epidemic Control (RISE) program focusing on “hard-to-reach” members of the general population – adolescents, young women and men, and more – that may not be engaged in healthcare. Starting in October 2020 in Kirundo, Ngozi, Kayanza, Karuzi, Muyinga, Gitega, Ruyigi, and Cankuzo provinces (Supplemental Figure 1), the goal of this program was to optimize efficiency and case finding using targeted HIV testing services, including index testing and aHIVST, to reduce missed opportunities for care, focus on hotspots, and reach people where they are – at home, in their communities, and where they work.¹⁶ RISE used community relay workers (CRWs) to provide aHIVST. Unassisted HIVST was not available.

This study’s objective was to evaluate the impact of ICAP’s aHIVST efforts with regard to the quarterly CFR and the number of persons testing positive for HIV (PTP) in a segment of the general population. Trends in CFR and the number of PTP were declining in RISE provinces before ICAP’s aHIVST implementation. It was hypothesized that aHIVST would reach PLWH in the general population who were unaware of their status and not engaged in care thus improving or reducing the downward trend observed over time for both outcomes.

Methods

Data source

We used aggregate PEPFAR monitoring, evaluating, and reporting (MER) indicators that are routinely collected each quarter.¹⁷ From ICAP-supported HIV testing, care, and treatment facilities, aHIVST data was captured on paper HIVST registers by CRWs who distributed aHIVST kits and supervised use in the community. CRWs were trained peers. If a person’s aHIVST test result was reactive, the CRW accompanied them to a health facility to report their result, receive a confirmatory HIV test using the national HIV testing algorithm, and, if positive, link to ART (Supplemental Figure 2). The positive test result was then captured in the MER indicator HTS_TST_POS. This data was reported monthly using the national reporting form containing country-level and PEPFAR-required reporting indicators. HTS_TST_POS results from aHIVST were not disaggregated from HTS_TST_POS results from other HIV testing modalities. The data was checked for coherence and entered in an ICAP-managed database for analysis and reporting. Rigorous data quality checks occurred quarterly, and data quality assurance exercises were performed annually. For this analysis, we aggregated aHIVST data at the province level from October-December 2020 through July-September 2023. Use of data for program evaluation and research was approved by the Columbia University IRB as not human subjects research; no personally identifying information was collected or used.

Measures

This analysis assessed changes in HIV testing volume, case identification, and CFR after the province-level scale-up of aHIVST kit distribution. Kit distribution increased in each province substantially starting in the October-December 2021 quarter (Figure 1) – the start of a new fiscal year. This aligned with guidance ICAP received to increase aHIVST efforts.

Figure 1.

aHIVST kits distributed in RISE-Supported provinces, October-December 2020 to July-September 2023.

The primary outcome of interest was the change in the number of PTP reported quarterly (MER indicator HTS_TST_POS). The secondary outcome of interest was the change in the quarterly CFR. We estimated the number of HIV tests performed in each quarter (the denominator for CFR) using two operationalizations: (1) summing the number of aHIVST kits distributed (MER indicator HTS_SELF) plus the number of HIV tests conducted at facility-based testing locations (MER indicator HTS_TST), and (2) only including the number of HIV tests conducted at facility-based testing locations. For both operationalizations, the number of PTP (the numerator for CFR) was estimated using the MER indicator HTS_TST_POS.

Two definitions of CFR were used because one underestimates, and the other overestimates, our measures of CFR. The first definition (including the number of aHIVST kits distributed in the denominator) may underestimate the CFR because not all aHIVST distributed are used, while the second definition (only tests occurring at facility-based locations in the denominator) would overestimate this proportion because the denominator includes no non-reactive results from aHIVST.

We conducted secondary subgroup analyses disaggregating our results by sex (male/female) and age (10-24 vs ≥25 years of age). As RISE is focused on the general population, KP were not disaggregated in this analysis. Sex and age were operationalized as a categorical variable with four levels – younger female, younger male, older female, and older male.

Comparison units

Because aHIVST is a programmatic intervention organized at the provincial level and recommended by Burundi’s government for universal adoption, there were no “unexposed” provinces for comparison in this analysis. Visual inspection of trends in aHIVST between provinces identified clear differences in the scale of implementation (Figure 1). To approximate more and less exposed units, we calculated the ratio of the average number of aHIVST kits distributed in January to June 2022 (post scale-up) to the average number of aHIVST kits distributed in October 2020 to December 2021 (pre-scale-up) for each province. We then identified the two provinces with the highest ratios and combined them (Kayanza and Kirundo, referred to collectively as the “mature” unit) and compared them to the two provinces with the lowest ratios, combined (Karuzi and Ruyigi, the “less mature” unit). These provinces were similar on other baseline characteristics (Supplemental Table 1).^{18, 19}

Analysis

We performed descriptive analyses to examine trends in the number of aHIVST kits distributed, the number of HIV positive tests, and the CFR operationalizations across quarters and provinces. We also examined the proportion of all tests that were aHIVST, operationalized as the proportion of aHIVST kits distributed in each quarter divided by the proportion of aHIVST kits that were distributed plus the total number of HIV tests completed in each quarter (i.e., HTS_SELF/(HTS_SELF + HTS_TST)). As with the first operationalization of CFR, this calculation likely underestimates the proportion of HIV tests that are aHIVST, as non-reactive aHIVST test results are not reported to the facility.

We further analyzed the data using difference-in-difference (DiD) analysis with linear regression. DiD is a quasi-experimental study design used to estimate the average effect of an intervention by comparing changes in the outcome over space and time. Key assumptions of the DiD design are common pre-intervention trends in the outcome between comparison groups and any time-varying drivers of outcome trends are group invariant. These are known collectively as the ‘parallel trends’ assumption. Central to the DiD interpretation is the assumption that, if the intervention had not occurred, the parallel trends observed in the comparison groups in the pre-intervention period would have been maintained after the intervention.²⁰ We conducted a visual inspection of trends in the outcomes in the pre-scale-up period to determine if the data met the parallel trends assumptions.

We performed four DiD analyses with linear regression. For the primary analysis, we examined the number of PTP and the first CFR operationalization, comparing the mature to the less mature unit, post versus pre-aHIVST scale-up. The equation for the primary model was as follows, with the outcome being either the number of PTP or the CFR, where the coefficient for the interaction term represents the treatment effect:

Y = β0+ β1(Intervention) + β2(Post Period) + β3(Intervention ∗ Post Period) + ε

In the secondary subgroup analysis, we examined each of these same outcomes disaggregated by age and sex at the province level using the described age-sex categorical variable. Not every facility reporting HTS_TST_POS and HTS_TST reported age and sex disaggregations, resulting in smaller quarterly total HTS_TST_POS and HTS_TST results and impacting quarterly CFRs as well. These facilities were excluded from the age-sex analysis. The equation for the secondary model was as follows, where the coefficient for the interaction term, now including the age-sex variable, represents the treatment effect:

Y = β0 + β1(Intervention) + β2(Post Period) + β3(Age_Sex) + β4(Intervention ∗ Post Period ∗ Age_Sex) + ε

Furthermore, we performed two sensitivity analyses. The first, the “denominator sensitivity analysis,” assessed the robustness of our operationalization of CFR to uncertainty using the second operationalization of the CFR. We performed DiD analyses for this alternative CFR outcome at the province level and at the province level disaggregated by our age-sex categorical variable. The second, the “parallel trends sensitivity analysis,” accounted for a violation of the parallel trends assumption (see Results). The statistical significance threshold was set at p<0.05. All analyses were performed in R 4.2.2.²¹

Results

Figure 2 demonstrates the trend in the number of persons testing for HIV during the study period. The number of persons testing for HIV in the mature and less mature units averaged 20,722 and 8933 a quarter, respectively, with trends remaining stable over time. Figure 3 demonstrates the trend in the proportion of all HIV tests that were attributable to aHIVST over time. For both units, the proportion increased starting in the October-December 2021 quarter. This increase was greater in the mature unit until the July-September 2022 quarter, at which point there was a substantial decline in the mature unit and the trend lines became close and overlapped until the end of the study period. For the mature unit, the average proportion of tests that were aHIVST was 1.46% in the pre-scale-up period and 13.59% in the post-scale-up period. For the less mature unit, the average proportion of tests that were aHIVST was 3.07% in the pre-scale-up period and 9.84% in the post-scale-up period.

Figure 2.

Number of persons testing for HIV in the mature and less mature units, October-December 2020 to July-September 2023. Note: the vertical dotted line denotes the scale-up quarter.

Figure 3.

Proportion of all HIV Tests that are aHIVST (Distributed Kits) in the Mature and Less Mature Units, October-December 2020 to July-September 2023. Note: the vertical dotted line denotes the scale-up quarter.

In Figure 4, we exhibit the trend in the outcomes – the number of PTP and the CFR (operationalization 1) – for the study period for both the mature and less mature units. In the pre-scale-up period, the number of PTP and the CFR for both units declined over time. A visual inspection demonstrated that, except for the first quarter of data, the trends for both outcomes were parallel between the mature and less mature units, meeting the parallel trends assumption. To ensure the robustness of our analysis, we performed a sensitivity analysis truncating the data to eliminate the first quarter. In terms of trends in the outcomes in the post-scale-up period, it appears that the trends in PTP and the CFR remained constant in the mature unit (Figure 4). For the less mature unit, the trends in both outcomes seemed to decline more steeply in the January-March 2023 quarter and then levelled out. Trends in the outcomes for the subgroup analysis by sex and age can be found in Supplemental Figure 3.

Figure 4.

Number of persons testing positive for HIV and CFR in the mature and less mature units, October-December 2020 to July-September 2023. Note: the vertical dotted line denotes the scale-up quarter.

We conducted DiD analysis to determine if there was a statistical difference in the number of PTP and the CFR in the post-scale-up period, compared to the pre-scale-up period, between the mature and less mature units (Table 1). We found that, on average, there were 34 fewer PTP post versus pre-scale-up in the mature unit compared to the less mature unit (95% CI -84.87-16.37, p = 0.17). We also found a non-statistically significant difference of 0.06% in CFR between the units in the post-scale-up period compared to the pre-scale-up period (95% CI -0.29%-0.41%, p = 0.72).

Table 1.

Differences-in-differences main analysis and sensitivity analyses results.

Outcome	Change in outcome	95% Confidence interval	P-value
Primary analysis
Persons testing positive	−34.25	−84.87, 16.37	0.17
Parallel Trends Sensitivity Analysis	−9.83	−58.43, 38.77	0.68
CFR (%)	0.06%	−0.29%, 0.41%	0.72
Denominator Sensitivity Analysis	0.08%	−0.27%, 0.44%	0.63
Parallel Trends Sensitivity Analysis	0.19%	−0.19, 0.57%	0.30
Secondary subgroup analysis
Persons testing positive, younger males (ref. younger females)	8.13	−12.60, 28.85	0.44
Parallel Trends Sensitivity Analysis	1.71	−17.40, 20.82	0.86
Persons testing positive, older females (ref. younger females)	−7.13	−27.86, 13.60	0.50
Parallel Trends Sensitivity Analysis	−0.96	−20.07, 18.15	0.92
Persons testing positive, older males (ref. younger females)	4.63	−16.10, 23.35	0.66
Parallel Trends Sensitivity Analysis	1.36	−17.73, 20.48	0.87
CFR (%), younger males (ref. younger females)	−0.15%	−1.12%, 1.11%	0.81
Denominator Sensitivity Analysis	−0.01%	−1.37%, 1.36%	0.99
Parallel Trends Sensitivity Analysis	−0.23%	−1.68%, 1.23%	0.76
CFR (%), older females (ref. younger females)	0.06%	−1.21%, 1.33%	0.93
Denominator Sensitivity Analysis	0.06%	−1.31%, 1.42%	0.94
Parallel Trends Sensitivity Analysis	0.06%	−1.40%, 1.51%	0.94
CFR (%), older males (ref. younger females)	0.13%	−1.14%, 1.40%	0.84
Denominator Sensitivity Analysis	0.31%	−1.05%, 1.68%	0.65
Parallel Trends Sensitivity Analysis	0.03%	−1.43%, 1.48%	0.97

In subgroup analysis and sensitivity analyses, we also found no statistically significant differences in the number of PTP and the CFR between the mature and less mature units in the post-compared to the pre-scale-up period with the inclusion of an additional term accounting for sex and age (Table 1). However, we did observe non-statistically significant differences, particularly for the number of PTP. For example, while the overall difference in the number of PTP between the mature and less mature units post-scale-up was −34.25, this difference, when examined by sex and age was 8.13 for younger males, −7.13 for older males, and 4.63 for older females compared to younger females. The extent of the difference in the number of PTP between mature and less mature units in the post-scale-up period varied by sex and age. In the denominator sensitivity analysis, with the alternative operationalization of CFR, changes in CFR, comparing the mature and less mature units pre- and post-scale-up, disaggregated by age-sex, remained non-significant. In the parallel trends sensitivity analysis, without the first quarter of data, the results were also similar, with the same conclusions.

Conclusions

In the age of 95-95-95, the goal of aHIVST as a testing strategy is to reach individuals for testing who are not otherwise engaged by health facilities, and to thus find PLWH who are unaware of their status. In this evaluation of ICAP’s aHIVST scale-up efforts, we found no change in the number of PTP nor the CFR, comparing mature and less mature units, post- versus pre-scale-up. The declining trend for both these outcomes remained constant. The results were robust to disaggregations by age and sex, different CFR operationalizations, and truncating the data to ensure parallel trends. In fact, for the outcome of PTP, the number decreased more in the mature unit than the less mature unit in the post-scale-up period compared to the pre-scale-up period. In part, this is surprising given the success of Burundi’s aHIVST scale-up for KP.¹⁵ But ICAP’s aHIVST scale-up as part of the RISE program specifically targeted aHIVST for non-KPs – adolescents, young women and men, and other members of the general population. We did find that PTP increased for younger and older men compared to younger women, but the results were not significant. It may be that the rest of the general population is adequately healthcare-engaged and that aHIVST in Burundi makes no difference. Alternatively, it is possible that the program was not adequately targeting more “hard-to-reach” members of the general population.

These results have programmatic implications for future implementation. This evaluation was conducted as part of routine monitoring and evaluation efforts; consequently, results were promptly communicated to implementers as soon as they were available. That feedback was then used to alter the program in terms of target populations and aHIVST delivery strategies. Guidance was given to the CRWs to use the national risk screening tool to distribute kits to persons at increased risk for HIV. The kits were also reserved for community testing campaigns targeting men at risk. This points to the importance of rapid, timely, routine monitoring and evaluation using advanced epidemiologic methods. It also emphasizes the value of collaboration between implementers, evaluators, and researchers. These were strengths of this analysis. The use of DiD analysis was also a strength of this study, as this method controls for time-varying confounding by units and unit-varying confounding by time. It is also suited to province-level analysis, and it is a quick, feasible, and easy method for use in the field. The choice to use comparison units based on ratios of post- and pre-scale-up trends is also a strength. This method for choosing more and less exposed comparison units is well-suited to the reality implementers face on the ground, where implementation sites may not be contiguous or similar on many baseline characteristics, and implementation may not be staggered.

However, a limitation of this analysis was that these units may not approximate counterfactuals. Future analyses could consider other ways to match provinces, use different units of analysis, like the district or health facility levels, or compare aHIVST to unassisted HIVST in the same population. Another limitation was that mature and less mature units did not fully meet the parallel trends assumption because of the data in the October-December 2020 quarter. We performed a sensitivity analysis to account for this.

In conclusion, this DiD analysis did not find a change in the number of PTP or the CFR after the scale-up of aHIVST kit distribution for adolescents and young women and men, but it did result in shifts in implementations strategies to better meet programmatic goals.

Supplemental Material

Supplemental Material - Impact of assisted HIV self-testing on case-finding rate and positive HIV diagnoses in Burundi: Aa province-level difference-in-difference analysis

Supplemental Material for Impact of assisted HIV self-testing on case-finding rate and positive HIV diagnoses in Burundi: A province-level difference-in-difference analysis by Sara Wallach, Suzue Saito, Greet Vandebrie, Eugenie Poirot, Ruby Fayorsey, Kouassi Jean-Jacques M’Bea, Million Hailu Tesema, Bonaparte Nijirazana, Aime Ndayizeye, Christelle Kaze, Matthew R. Lamb in International Journal of STD & AIDS

Footnotes

ORCID iD

Sara Wallach

Ethical considerations

This paper involves only aggregate, province-level data. Ethical approval for this study was not required.

Funding

The authors disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: Research reported in this publication was supported by the National Institute of Allergy and Infectious Diseases of the National Institutes of Health under Award Number T32AI114398. The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health.

Declaration of conflicting interests

The authors declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Data Availability Statement

The data that support the findings of this study are available from the corresponding author upon reasonable request.*

Supplemental Material

Supplemental material for this article is available online.

References

2025 AIDS targets - UNAIDS. https://aidstargets2025.unaids.org/#section-about (accessed 12 August 2024).

Burundi 2023 Factsheet. https://www.unaids.org/en/regionscountries/countries/burundi (accessed 12 August 2024).

Burundi PLACE Report . Priorities for local AIDS control efforts — MEASURE evaluation. https://www.measureevaluation.org/resources/publications/tr-14-99.html (accessed 20 August 2024).

Coulaud

Mujimbere

Nitunga

, et al. An assessment of health interventions required to prevent the transmission of HIV infection among men having sex with men in bujumbura, Burundi. J Community Health 2016; 41: 1033–1043.

Nshimirimana

Vuylsteke

Smekens

, et al. HIV testing uptake and determinants among adolescents and young people in Burundi: a cross-sectional analysis of the Demographic and Health Survey 2016–2017. BMJ Open 2022; 12: e064052.

World Health Organization . Guidelines on HIV self-testing and partner notification: supplement to consolidated guidelines on HIV testing services. : World Health Organization, 2016. https://iris.who.int/handle/10665/251655 (accessed 14 August 2024).

Krause

Subklew-Sehume

Kenyon

, et al. Acceptability of HIV self-testing: a systematic literature review. BMC Public Health 2013; 13: 735.

Martínez

Steele

Govender

, et al. Supervised oral HIV self‐testing is accurate in rural K wa Z ulu‐ N atal, S outh A frica. Tropical Med Int Health 2016; 21: 759–767.

Pant Pai

Sharma

Shivkumar

, et al. Supervised and unsupervised self-testing for HIV in High- and low-risk populations: a systematic review. PLoS Med 2013; 10: e1001414.

10.

Hector

Davies

M-A

Dekker-Boersema

, et al. Acceptability and performance of a directly assisted oral HIV self-testing intervention in adolescents in rural Mozambique. PLoS One 2018; 13: e0195391.

11.

Tonen-Wolyec

Kayembe Tshilumba

Batina-Agasa

, et al. Comparison of practicability and effectiveness between unassisted HIV self-testing and directly assisted HIV self-testing in the Democratic Republic of the Congo: a randomized feasibility trial. BMC Infect Dis 2020; 20: 830.

12.

Figueroa

Johnson

Ford

, et al. Reliability of HIV rapid diagnostic tests for self-testing compared with testing by health-care workers: a systematic review and meta-analysis. Lancet HIV 2018; 5: e277–e290.

13.

Ministere De La Sante Publiqueet De La Lutte Contre Le Sida . Plan strategique national de lutte contre le SIDA 2014-201. Republique Du Burundi. https://minisante.gov.bi/wp-content/uploads/pnls/PSN-VIH_2014-2017.pdf (July 2014).

14.

Ministere De La Sante Publique Et De La, Lutte Contre Le Sida . Addendum Au Plan Strategique National Vih/Sida 2018-2022. Republique Du Burundi. https://files.aho.afro.who.int/afahobckpcontainer/production/files/Addendum_au_plan_strat%C3%A9gique_national_VIHSida_2018-2022.pdf (March 2020).

15.

Lillie

Boyee

Kamariza

, et al. Increasing testing options for key populations in Burundi through peer-assisted HIV self-testing: descriptive analysis of routine programmatic data. JMIR Public Health Surveill 2021; 7: e24272.

16.

Reaching impact, saturation and epidemic control (RISE). https://icap.columbia.edu/cpt_projects/reaching-impact-saturation-and-epidemic-control-rise/ (accessed 20 August 2024).

17.

Fiscal

PEPFAR

Year 2023 Monitoring, Evaluation, and Reporting (MER) Indicators. United States Department of State. Accessed February 20, 2024. https://www.state.gov/pepfar-fy-2023-mer-indicators/

18.

U.S. President’s emergency plan for AIDS relief. Burundi Country Operational Plan (COP) 2020 Strategic Direction Summary . https://www.state.gov/wp-content/uploads/2020/07/COP-2020-Burundi-SDS-FINAL.pdf (19 May 2020).

19.

U.S. President’s emergency plan for AIDS relief. Burundi Country Operational Plan (COP) 2022 Strategic Direction Summary . https://www.state.gov/wp-content/uploads/2022/09/Burundi-COP22-SDS-.pdf (6 May 2022).

20.

Wing

Coady

Simon

Kosali

Bello-Gomez

Ricardo A.

Designing Difference in Difference Studies: Best Practices for Public Health Policy Research. Annual Review of Public Health. 2018; 39(1): 453–469. doi:10.1146/annurev-publhealth-040617-01350729328877.

21.

Team RC . R: a language and environment for statistical computing.

Supplementary Material

Please find the following supplemental material available below.

For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.

For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.

0.00 MB

1.12 MB