Global Health Metrics And Evaluation

Global health metrics and evaluation is the discipline of counting health.

12 frameworks, from 1830 to today · Global Health · Medicine & Health

still a working view ended · each leads to its entry below

Overview

Global health metrics and evaluation is the discipline of counting health. It asks how many people are born, how many die, what they die of, how sick the living are, and whether the programs meant to help them actually work. Its practitioners build the statistical scaffolding that lets anyone say, with a straight face, that a disease killed 1.2 million people last year or that a vaccination campaign averted a certain number of deaths. That scaffolding is not neutral: the numbers decide where money goes, which diseases get attention, and which populations become visible to the institutions that govern global health. The history of this subfield is therefore a history of tools, but also of arguments about what should be counted and who gets to count it.

Vital Registration and the Statistical State

The oldest strand is civil registration: the routine, legal recording of births, deaths, and causes of death by a state. Its modern form took shape in nineteenth-century Europe, where compulsory registration and standardized cause-of-death lists allowed governments to see patterns in mortality for the first time. The International Statistical Institute adopted the first international classification of causes of death in 1893, building on work by Jacques Bertillon; this lineage eventually became the World Health Organization's International Classification of Diseases, whose successive revisions have anchored cause-of-death reporting since the mid-twentieth century.

The problem was, and remains, coverage. In much of the world, deaths occur without a doctor, a certificate, or a register1. Even today, a large share of deaths worldwide go unrecorded, and where they are recorded the stated cause is often unreliable. The history of global health metrics is in large part a history of working around this absence.

Household Surveys and Verbal Autopsy

From the 1970s onward, sample surveys became the workhorse for measuring health where registration failed. The World Fertility Survey, later followed by the Demographic and Health Surveys (from 1984) and UNICEF's Multiple Indicator Cluster Surveys (from 1995), gave countries repeated, comparable snapshots of child mortality, fertility, nutrition, and health-service use. These surveys were designed to be statistically representative and repeatable, which made them the backbone of global monitoring for the Millennium Development Goals and their successors.

A parallel innovation addressed the cause-of-death gap. Verbal autopsy is a method in which trained interviewers ask family members about the symptoms and circumstances preceding a death, and the responses are used to assign a probable cause. Early verbal autopsy instruments were developed in the 1980s and 1990s, largely for research settings in South Asia and sub-Saharan Africa. The method's weakness was subjectivity: different physicians could read the same interview differently. The response, from the 2000s, was to standardize questionnaires and to develop automated algorithms, including probabilistic models and, later, machine-learning classifiers, that assign causes consistently. Verbal autopsy moved from a research curiosity to a routine component of national mortality surveillance in several countries.

Measuring the Burden: Summary Measures and the Global Burden of Disease

The 1980s and 1990s saw a push to combine death and disability into single summary measures of population health. The idea was to compare the toll of conditions that kill quickly with those that disable for decades. Disability-adjusted life years (DALYs), developed in the early 1990s by Christopher Murray, Dean Jamison, and colleagues working with the World Bank, express health loss as years of life lost to premature death plus years lived with disability, weighted for severity. The 1993 World Development Report, Investing in Health, used this framework to argue for prioritizing cost-effective interventions in low-income countries.

The Global Burden of Disease (GBD) study, first published in 1996 with Murray and Alan Lopez as lead authors, applied the DALY to a comprehensive set of diseases, injuries, and risk factors across world regions. Its 1990 baseline estimates were startling: they showed that neuropsychiatric conditions and injuries accounted for far more health loss than the international agenda had assumed, and that the dominance of infectious disease in poor countries coexisted with rising noncommunicable disease burdens. The study's methods were contested—severity weights, age weighting, and discounting were all argued over—but its central move, ranking conditions by health loss rather than by mortality alone, reshaped priority-setting debates.

The GBD enterprise was institutionalized at the Institute for Health Metrics and Evaluation (IHME), founded at the University of Washington in 2007 with funding from the Bill & Melinda Gates Foundation23. Successive GBD rounds, from 2010 onward, expanded to hundreds of causes and risks, incorporated covariates and statistical models, and published annually. The enterprise also drew criticism: for its modeling choices, for the dominance of a single institution in producing the numbers everyone else uses, and for the tension between global estimates and country-owned data. The debate over who validates the numbers became part of the field's politics.

Evaluating Programs: From Quasi-Experiments to Randomized Trials

Alongside measurement of disease came evaluation of interventions. The dominant mid-century model was quasi-experimental: before-and-after comparisons, matched controls, and natural experiments. These methods were cheap and often the only option, but they struggled with confounding—if a program is placed where need is greatest, its results look worse than they are.

The 1990s and 2000s brought a shift toward randomized controlled trials in development economics and global health, associated with the Abdul Latif Jameel Poverty Action Lab (founded 2003) and researchers such as Esther Duflo and Abhijit Banerjee. Randomization—assigning programs by lottery—made it possible to attribute outcomes to the intervention rather than to selection4. Landmark trials tested deworming, insecticide-treated bednets, and conditional cash transfers, and the approach spread through global health funding. Its supporters argued that only randomization could settle causal questions; critics argued that trials were expensive, often small, sometimes ethically fraught, and poor at answering questions about scale-up and system effects. The two approaches now coexist, with quasi-experimental designs such as difference-in-differences and regression discontinuity widely used where randomization is impractical56.

Paying for Results

A distinct evaluation tradition grew out of financing reform. Results-based financing ties payment to measured performance: providers or governments receive funds when they hit targets such as vaccination coverage or facility deliveries7. The approach drew on performance-based contracting experiments in the 1990s and 2000s, including Rwanda's performance-based financing scheme for health facilities and later the World Bank's Health Results Innovation Trust Fund. Evaluating these schemes required measuring both the intended effect and the unintended ones—gaming of indicators, neglect of unmeasured services—and the evidence has been mixed, with effects varying by context and design. The evaluation question shifted from "does the program work?" to "does paying for the indicator change behavior, and in which direction?"

Excess Mortality and the Data-Science Turn

The COVID-19 pandemic exposed the fragility of routine reporting and pushed excess mortality estimation to the center of the field. Excess mortality compares observed deaths with an expected baseline, capturing both direct and indirect pandemic deaths and correcting for underreporting. The Economist's model and IHME's estimates, alongside academic work by researchers such as Karlinsky and Kobak, produced widely cited global figures suggesting that official COVID-19 death counts captured only a fraction of true mortality. The methods—time-series baselines, regression models, and the reconciliation of incomplete registration—drew on the same statistical toolkit as the GBD.

More broadly, the 2010s and 2020s brought computational and data-science methods into the field: machine learning for verbal autopsy classification, Bayesian hierarchical models for small-area estimation, satellite and mobile-phone data for population denominators, and open-source platforms for reproducible analysis. These methods made it possible to estimate health indicators for districts and years where no survey exists, but they also raised questions about transparency and the authority of modeled numbers.

The Present

Today the field runs on a layered infrastructure: civil registration where it functions, surveys and verbal autopsy where it does not, and models that fill the remaining gaps. The GBD remains the dominant synthesis, but its estimates are increasingly contested by country-led analyses and by critics who argue for stronger national ownership of data. Randomized trials and quasi-experiments continue side by side, with growing attention to external validity and to evaluating health systems rather than single interventions. Excess mortality methods, honed during COVID-19, are now applied to other crises. The open questions are familiar but unresolved: how to close the registration gap, how to make modeled estimates accountable to the people they describe, and how to measure what matters without letting the measurable crowd out the rest.

References

  1. Verbal autopsy, Wikipedia
  2. Institute for Health Metrics and Evaluation (introduction, History), Wikipedia
  3. Global Burden of Disease study, Wikipedia
  4. Randomized controlled trial, Wikipedia
  5. Difference in differences, Wikipedia
  6. Regression discontinuity design, Wikipedia
  7. Pay for performance (healthcare), Wikipedia

The frameworks of Global Health Metrics And Evaluation

Each framework is a way of working the field has used: when it held, and the question it answers.

  1. 1830-Present
  2. 1950-Present

    Quasi-experiment

    Did the program cause the change?

    Fly to its tower in Nousfarer

  3. 1970-Present
  4. 1970-Present

    Household Survey Programs

    How can health be measured comparably across countries?

    Fly to its tower in Nousfarer

  5. 1985-Present

    Verbal Autopsy

    What did this person die of, without a doctor?

    Fly to its tower in Nousfarer

  6. 1990-Present

    Summary Measures of Population Health

    How do death and disability compare in one number?

    Fly to its tower in Nousfarer

  7. 1995-Present

    Randomized controlled trial

    Does the intervention work when assigned by chance?

    Fly to its tower in Nousfarer

  8. 1996-Present

    Global Burden of Disease study

    Which diseases and risks cause the most health loss?

    Fly to its tower in Nousfarer

  9. 2000-Present

    Pay for performance

    Does paying for results change provider behavior?

    Fly to its tower in Nousfarer

  10. 2005-Present

    Computational and Data-Science Health Metrics

    Can models estimate health where data are missing?

    Fly to its tower in Nousfarer

  11. 2010-Present
  12. 2020-Present

    Excess mortality

    How many more died than would have been expected?

    Fly to its tower in Nousfarer

How the ideas connect

Each arrow runs from an earlier idea to the later one it shaped. The list below says what passed between them.

Civil Registration and Vital Statisti…Quasi-experimentHousehold Survey and Verbal Autopsy M…Household Survey ProgramsVerbal AutopsySummary Measures of Population HealthRandomized controlled trialGlobal Burden of Disease studyPay for performanceComputational and Data-Science Health…Computational and Bayesian Small-Area…Excess mortality

Beyond the field

Where Global Health Metrics And Evaluation's ideas came from in other fields, and what they led to there: the strongest 12 of 63 links.

A city in Nousfarer: a red-carpeted stair climbing terraces to a crowned tower on an acropolis, more towers round it, forest and snowy mountains beyond.
A city in Nousfarer, the game played in the browser.

Global Health Metrics And Evaluation is a planet in Nousfarer

Its city has a tower for each of these 12 frameworks. Land, walk in, and climb one concept at a time.

Land on Global Health Metrics And Evaluation

Related fields

Fields its ideas are linked to

More in Global Health

See where Global Health Metrics And Evaluation sits on the Map of Human Knowledge