Estimated mortality on HIV treatment among active patients and patients lost to follow-up in 4 provinces of Zambia: Findings from a multistage sampling-based survey

Abstract

Background

Survival represents the single most important indicator of successful HIV treatment. Routine monitoring fails to capture most deaths. As a result, both regional assessments of the impact of HIV services and identification of hotspots for improvement efforts are limited. We sought to assess true mortality on treatment, characterize the extent under-reporting of mortality in routine health information systems in Zambia, and identify drivers of mortality across sites and over time using a multistage, regionally representative sampling approach.

Methods and findings

We enumerated all HIV infected adults on antiretroviral therapy (ART) who visited any one of 64 facilities across 4 provinces in Zambia during the 24-month period from 1 August 2013 to 31 July 2015. We identified a probability sample of patients who were lost to follow-up through selecting facilities probability proportional to size and then a simple random sample of lost patients. Outcomes among patients lost to follow-up were incorporated into survival analysis and multivariate regression through probability weights. Of 165,464 individuals (64% female, median age 39 years (IQR 33–46), median CD4 201 cells/mm ³ (IQR 111–312), the 2-year cumulative incidence of mortality increased from 1.9% (95% CI 1.7%–2.0%) to a corrected rate of 7.0% (95% CI 5.7%–8.4%) (all ART users) and from 2.1% (95% CI 1.8%–2.4%) to 8.3% (95% CI 6.1%–10.7%) (new ART users). Revised provincial mortality rates ranged from 3–9 times higher than naïve rates for new ART users and were lowest in Lusaka Province (4.6 per 100 person-years) and highest in Western Province (8.7 per 100 person-years) after correction. Corrected mortality rates varied markedly by clinic, with an IQR of 3.5 to 7.5 deaths per 100 person-years and a high of 13.4 deaths per 100 person-years among new ART users, even after adjustment for clinical (e.g., pretherapy CD4) and contextual (e.g., province and clinic size) factors. Mortality rates (all ART users) were highest year 1 after treatment at 4.6/100 person-years (95% CI 3.9–5.5), 2.9/100 person-years (95% CI 2.1–3.9) in year 2, and approximately 1.6% per year through 8 years on treatment. In multivariate analysis, patient-level factors including male sex and pretherapy CD4 levels and WHO stage were associated with higher mortality among new ART users, while male sex and HIV disclosure were associated with mortality among all ART users. In both cases, being late (>14 days late for appointment) or lost (>90 days late for an appointment) was associated with deaths. We were unable to ascertain the vital status of about one-quarter of those lost and selected for tracing and did not adjudicate causes of death.

Conclusions

HIV treatment in Zambia is not optimally effective. The high and sustained mortality rates and marked under-reporting of mortality at the provincial-level and unexplained heterogeneity between regions and sites suggest opportunities for the use of corrected mortality rates for quality improvement. A regionally representative sampling-based approach can bring gaps and opportunities for programs into clear epidemiological focus for local and global decision makers.

Abstract

To improve estimates of mortality for patients who initiate HIV treatment, Charles Holmes and colleagues use a multistage sampling approach to find and trace health outcomes for patients lost to follow-up after starting antiretroviral therapy in Zambia.

Author summary

Why was this study done?

• Previous studies from cohorts in South Africa and parts of East Africa have suggested that site-level reporting of mortality is incomplete.

• We wanted to understand the degree to which this phenomenon was impacting HIV outcomes at a broader scale, in this case at the provincial level in Zambia, a country with one of the highest burdens of HIV.

• We also wanted to gain an in-depth understanding of differences between the outcomes of clinical care sites in order to assess the role of mortality as a potential quality improvement target.

What did the researchers do and find?

• From a source population of patients in 4 provinces (Lusaka, Southern, Eastern, and Western) who visited government-operated HIV treatment sites in these provinces, we conducted a multistage sampling approach of a stratified selection of sites and a random sample of patients lost to follow-up.

• Lost patients were traced and their vital status was ascertained, which was used to enable a corrected regionally representative estimate of survival after starting antiretroviral therapy (ART) as well as corrected site-specific mortality estimates.

• Of 165,464 individuals, the 2-year cumulative incidence of mortality increased from 1.9% to 7.0% for all ART users and from 2.1% to 8.3% for new ART users, and provincial-level mortality rates rose 3- to 8-fold once corrected for true outcomes.

• Being late (>14 days late for appointment) or lost (>90 days late for an appointment) was associated with death.

What do these findings mean?

• Deaths are under-reported within the Zambian HIV program, and mortality rates are highly variable across sites and provinces.

• Our findings enable national- and global-level policy makers to correct existing underestimates of mortality, link these data to quality improvement efforts, and reprioritize interventions to target regional and site-level reductions in mortality as a goal of HIV programs.

• We have also established that this methodology is feasible for use as a representative surveillance tool for accurate monitoring of provincial and potentially national levels of mortality, even as vital status registries and data systems are further developed and strengthened.