Download presentation
Presentation is loading. Please wait.
Published byMelissa Phillips Modified over 8 years ago
1
1 | Copyright © 2015 Intellectual Ventures Management, LLC (IV). All rights reserved. Measuring the Benefit of Effective HIV/AIDS Programs in Sub-Saharan Africa: Healthy Life Gained or Death and Sickness Averted? Anna Bershteyn 1, Daniel J. Klein 1, Gesine Meyer-Rath 2, Philip A. Eckhoff 1, and Mead Over 3 1 Institute for Disease Modeling, Bellevue, United States 2 Boston University School of Public Health, Global Health, Johannesburg, South Africa 3 Center for Global Development, Washington, United States Presented at ICASA 2015, Harare, Zimbabwe. Session WEAE0701: Track E/7, Stretching the dollar. Thanks to David Bernotas 1 for his support in preparing this talk.
2
2 | Copyright © 2015 Intellectual Ventures Management, LLC (IV). All rights reserved. Measuring HIV Burden HIV causes immense burden: 1.2 million deaths & 37 million infected in 2014 1 Goal: maximize burden reduction using available tools, given budget constraint – Approach: Use mathematical models to guide policy Question: How do we achieve the most impact per dollar? – Answer: depends on which metric we choose to optimize! We translated various health economic concepts to estimate HIV burden in a detailed individual-based model of HIV 1. UNAIDS Fact Sheet 2015
3
3 | Copyright © 2015 Intellectual Ventures Management, LLC (IV). All rights reserved. Policy Metrics for Burden and Impact
4
4 | Copyright © 2015 Intellectual Ventures Management, LLC (IV). All rights reserved. Simple Measures of Burden & Impact Simple burden (impact) measures – Prevalence (reduction): Percent of population that is infected – New infections (averted): Number of newly infected individuals – HIV-cause deaths (averted): Number of deaths due to HIV-cause Compute impact from model-based comparison of two scenarios 1.Baseline: serves as the “reference” trajectory 2.Intervention: contains the alternate scenario to be measured Time discounting typically 3% from start of intervention
5
5 | Copyright © 2015 Intellectual Ventures Management, LLC (IV). All rights reserved. Health-Adjusted Measures Disability-based (DALYs)Health-based (HALYs) Measures “Sickness” Less is better Sum YLD and YLL Measures “Happiness” More is better Sum of health-adjusted life-years InfectionDeathInfectionDeathL.E. Birth
6
6 | Copyright © 2015 Intellectual Ventures Management, LLC (IV). All rights reserved. The impact of impact metrics MetricOptimal PolicyComments Infections averted Target ART to individuals at risk of transmitting, e.g. young/risky. Scale up prevention: medical male circumcision, PrEP, condoms, etc. Eventually reduces deaths, but only after a significant delay. Required to achieve long-term control of HIV. Deaths averted Target ART to individuals at risk of death, e.g. low CD4, older, higher VL. Minimize treatment interruptions. Less prevention (depending on time horizon / discounting). Tends to prioritize those at high risk of death over those at high risk of transmitting or acquiring HIV. DALYs / HALYs Avoid young deaths, e.g. in those vertically infected. Treatment for “burdened” individuals. Similar to deaths averted for HIV with higher priority for younger individuals, who happen to also have higher transmission potential on average.
7
7 | Copyright © 2015 Intellectual Ventures Management, LLC (IV). All rights reserved. Application to Individual-Based Modeling
8
8 | Copyright © 2015 Intellectual Ventures Management, LLC (IV). All rights reserved. Application to Individual-Based Modeling Individual-Based Modeling Model each individual – Vital dynamics – Network of relationships – HIV infections: stage, CD4, … Individual factors – Age, gender, risk, … Know everything about everyone – Exact age & cause of death – Future time of natural death The EMOD-HIV Model Individual-based network model of generalized HIV epidemics Model, source code, documentation, and examples are publically available online: idmod.org idmod.org Learn more at our satellite session: – Thu@7am in Committee Room 5a & b
9
9 | Copyright © 2015 Intellectual Ventures Management, LLC (IV). All rights reserved. Quantifying Burden
10
10 | Copyright © 2015 Intellectual Ventures Management, LLC (IV). All rights reserved. Simple Burden Metrics New InfectionsHIV-Cause Deaths
11
11 | Copyright © 2015 Intellectual Ventures Management, LLC (IV). All rights reserved. Quantifying Burden: DALYs Lost to HIV Quantify post-2016 HIV burden by comparing to hypothetical cure in 2016 – An artificial point of reference – Results in zero burden Burden front-loaded at death – E.g., if a person dies 10 years prematurely, all 10 YLL applied at time of death – Discounting depends on time of death, not number of YLLs Delay end of epidemic until 2026: gap can close
12
12 | Copyright © 2015 Intellectual Ventures Management, LLC (IV). All rights reserved. Quantifying Burden: DALYs Lost to HIV Quantify post-2016 HIV burden by comparing to hypothetical cure in 2016 – An artificial point of reference – Results in zero burden Burden front-loaded at death – E.g., if a person dies 10 years prematurely, all 10 YLL applied at time of death – Discounting depends on time of death, not number of YLLs Delay end of epidemic until 2026: gap can close
13
13 | Copyright © 2015 Intellectual Ventures Management, LLC (IV). All rights reserved. Quantifying Burden: HALYs Lost to HIV HALY burden requires counterfactual: 2016 cure Burden spread in time – E.g., if a person dies 10 years prematurely, apply 1 YLL per year for next 10 years – Discounting depends on YLLs: more discounting of very premature deaths HALYs lost – Grow over time – Can never close gap
14
14 | Copyright © 2015 Intellectual Ventures Management, LLC (IV). All rights reserved. Quantifying Burden: HALYs Lost to HIV HALY burden requires counterfactual: 2016 cure Burden spread in time – E.g., if a person dies 10 years prematurely, apply 1 YLL per year for next 10 years – Discounting depends on YLLs: more discounting of very premature deaths HALYs lost – Grow over time – Can never close gap
15
15 | Copyright © 2015 Intellectual Ventures Management, LLC (IV). All rights reserved. Quantifying Burden: HALYs Lost to HIV HALY burden requires counterfactual: 2016 cure Burden spread in time – E.g., if a person dies 10 years prematurely, apply 1 YLL per year for next 10 years – Discounting depends on YLLs: more discounting of very premature deaths HALYs lost – Grow over time – Can never close gap
16
16 | Copyright © 2015 Intellectual Ventures Management, LLC (IV). All rights reserved. Quantifying Burden: HALYs Lost to HIV HALY burden requires counterfactual: 2016 cure Burden spread in time – E.g., if a person dies 10 years prematurely, apply 1 YLL per year for next 10 years – Discounting depends on YLLs: more discounting of very premature deaths HALYs lost – Grow over time – Can never close gap
17
17 | Copyright © 2015 Intellectual Ventures Management, LLC (IV). All rights reserved. Quantifying Program Impact
18
18 | Copyright © 2015 Intellectual Ventures Management, LLC (IV). All rights reserved. Performance of HIV Program: Burden Reduction
19
19 | Copyright © 2015 Intellectual Ventures Management, LLC (IV). All rights reserved. Hypothetical Burden Reduction Infections AvertedHIV-Cause Deaths Averted
20
20 | Copyright © 2015 Intellectual Ventures Management, LLC (IV). All rights reserved. Hypothetical Burden Reduction DALYs AvertedHALYs Gained
21
21 | Copyright © 2015 Intellectual Ventures Management, LLC (IV). All rights reserved. Conclusion and Recommendations HIV burden can be estimated using mathematical models Optimal programming is sensitive to burden metric and discounting – There is no “right” answer for the metric – Burden estimation using HALYs requires cure counterfactual – HALYs account for lost life due to unborn population Program Impact: – DALY response is more immediate than HALYs – Due to DALY YLL concentrated at time of death – Result: $/DALY Averted < $/HALY Gained, especially with high discounting – Be patient for delayed-impact interventions like male circumcision Recommendations – Modelers: Be exceedingly clear about how burden/impact is calculated – Policy Makers: Think carefully about impact metrics
22
22 | Copyright © 2015 Intellectual Ventures Management, LLC (IV). All rights reserved. Thank you
23
23 | Copyright © 2015 Intellectual Ventures Management, LLC (IV). All rights reserved. How to Calculate Health-Adjusted HIV Burden? HIV-Cause DALYs ⁺0.053 * [Infected CD4 > 350] ⁺0.221 * [Infected CD4 200-350] ⁺0.547 * [Infected CD4 <200] ⁺0.053 * [On ART] ⁺0.00 * [Uninfected Population] ⁺Years of Life Lost, if HIV Cause – Based on fixed reference age – Based on life expectancy at death Total HALYs ⁺0.947 * [Infected CD4 > 350] ⁺0.779 * [Infected CD4 200-350] ⁺0.453 * [Infected CD4 <200] ⁺0.947 * [On ART] ⁺1.00 * [Uninfected Population] HALYs lost due to HIV is computed relative to a HIV-cure counterfactual Eaton, J. W., Menzies, N. A., Stover, J., Cambiano, V., Chindelevitch, L., Cori, A.,... & Easterbrook, P. J. (2014). Health benefits, costs, and cost-effectiveness of earlier eligibility for adult antiretroviral therapy and expanded treatment coverage: a combined analysis of 12 mathematical models. The lancet global health, 2(1), e23-e34.
24
24 | Copyright © 2015 Intellectual Ventures Management, LLC (IV). All rights reserved. Hypothetical Intervention Simulations
25
25 | Copyright © 2015 Intellectual Ventures Management, LLC (IV). All rights reserved.
26
26 | Copyright © 2015 Intellectual Ventures Management, LLC (IV). All rights reserved.
27
27 | Copyright © 2015 Intellectual Ventures Management, LLC (IV). All rights reserved.
28
28 | Copyright © 2015 Intellectual Ventures Management, LLC (IV). All rights reserved.
29
29 | Copyright © 2015 Intellectual Ventures Management, LLC (IV). All rights reserved. Abstract An effective HIV program prolongs lives and prevents HIV infections. Over the long term, these two effects produce significant changes in the structure and rate of growth of the beneficiary population. Two commonly used approaches to calculating the health benefits of HIV/AIDS programs differ in their treatment of the longer term demographic effects. One approach values incremental healthy life-years (HLYs) gained by implementing a program, including those lived by people who would otherwise not have been born. The other approach values only the years of death and disability, or disability-adjusted life-years (DALYs), averted by the program. By embedding HIV/AIDS epidemiology and interventions within a demographic model, we calculated the size of the divergence between the two measures and suggest policy questions that each can address. Methods: The long-term burden of HIV/AIDS in South Africa was estimated under different health care expansion scenarios using EMOD-HIV, an epidemiological model that provides detailed age-stratified outputs including the HIV-free lifespan for all individuals. The full EMOD-HIV model is available online. For the same model runs, we compared the HLYs and DALYs associated with expanded HIV services. Results: The divergence between HLYs gained and DALYs averted manifests within one decade and widens over time when children are born in the intervention scenario who otherwise would not have been born. Applying an annual discount rate diminishes the cumulative difference between the two benefit measured. With a 3% annual discount rate, in the most optimistic intervention scenario (annual testing for 80% of the population with linkage to ART for all those testing positive, in addition to baseline antenatal, couples, and symptom-driven testing), both benefit measures improved, but the gap between them continued to grow until 2050, when it stabilized but never declined. Conclusions and Recommendations: Policy analysts interested in both the health and the demographic impact of HIV/AIDS programs may prefer to study life-years gained rather than valuing averted death and disability in order to fully capture the program's impact on the entire population. However, since dynamic population models do not yet exist for many other health risks, analysts wishing to compare interventions across diseases will prefer the static approach which values an intervention by the death and disability it averts.
Similar presentations
© 2025 SlidePlayer.com. Inc.
All rights reserved.