Consider a paradox that haunts global health: we have more sophisticated tools than ever to treat disease, yet in many parts of the world, we cannot answer the most basic question—what are people actually dying from? This isn't a minor administrative inconvenience. It represents a fundamental blind spot that undermines every intervention, every policy decision, and every dollar spent on improving population health.

The consequences ripple outward in ways that are difficult to overstate. Ministries of health allocate resources based on incomplete pictures. International donors fund programs targeting diseases they assume are prevalent, without robust evidence. Epidemics gain footholds because surveillance systems fail to detect early signals. We are, in essence, practicing global health with one eye closed.

The uncomfortable truth is that data quality is not merely a technical problem awaiting a technological fix. It reflects deeper structural issues—underinvestment in health systems, misaligned incentives, colonial legacies that privileged extraction over institution-building, and a global health architecture that often treats information as an afterthought rather than a prerequisite. Until we confront these realities, even our most ambitious health initiatives will rest on foundations of sand.

Roughly two-thirds of deaths worldwide occur without registration of their cause. Let that sink in. In an era of precision medicine and genomic sequencing, the majority of humans who die each year simply disappear from the epidemiological record. They leave behind grieving families but no data point, no contribution to our understanding of population health.

This isn't merely about bureaucratic completeness. Vital registration systems—the infrastructure that records births, deaths, and causes of death—constitute the bedrock upon which rational health policy must be built. Without knowing what kills people, how can we prioritize? Should we invest in cancer treatment or maternal health services? Cardiovascular disease prevention or infectious disease control? These questions demand answers grounded in evidence, not intuition.

The geography of this gap maps distressingly onto historical patterns of underinvestment. Sub-Saharan Africa registers fewer than 10% of deaths with medically certified causes in many countries. South Asia fares only marginally better. Meanwhile, high-income countries have maintained comprehensive vital registration for generations, giving them a perpetual advantage in health planning.

The consequences extend beyond national borders. Global burden of disease estimates—those influential rankings that shape international funding priorities—must rely heavily on statistical modeling to fill the gaps. These models are sophisticated, but they remain educated guesses. When donors decide that malaria deserves more funding than pneumonia, or vice versa, they're often acting on extrapolations from limited data.

Creating functional vital registration systems requires sustained commitment over decades. It demands trained personnel, legal frameworks, community engagement to overcome cultural barriers to registration, and political will to maintain systems that provide no immediate electoral benefit. There are no shortcuts, and therein lies the challenge.

Takeaway

You cannot manage what you cannot measure, and for most of humanity, we cannot measure the most fundamental health outcome—what causes death.

Even where health facilities exist and collect data, the information they generate is often systematically unreliable. This isn't primarily about incompetent record-keeping—it's about incentive structures that actively corrupt data quality.

Consider the facility manager whose budget depends on reported patient volumes. The temptation to inflate numbers is structural, not personal. Or consider the vertical disease program that ties funding to case detection. Staff quickly learn which diagnoses unlock resources. Tuberculosis reported? Funding flows. Pneumonia reported? Nothing special happens. The data begins to reflect the funding landscape rather than the disease landscape.

Under-reporting proves equally pernicious. Health workers in understaffed facilities may lack time for proper documentation. Maternal deaths might go unreported when they occur outside facility walls, or when staff fear accountability systems that punish rather than support. Vaccine coverage rates get inflated when performance targets create pressure to demonstrate success.

The digital revolution has not solved these problems—in some cases, it has amplified them. Electronic health information systems can generate impressive dashboards and visualizations, but they remain only as good as the data entered at the point of care. Garbage in, garbage out, now with better graphics.

Quality assurance mechanisms exist, but they require resources and institutional capacity that remain scarce. Data validation through supervisory visits, cross-checking between data sources, and triangulation with external information all demand investments that compete with direct service delivery. When budgets are tight, data quality rarely wins the competition.

Takeaway

Health data from facilities reflects not just disease patterns but the incentive structures, resource constraints, and accountability mechanisms within which health workers operate.

In response to vital registration failures and facility data limitations, the global health community has developed an elaborate toolkit of alternative data sources. Household surveys, sentinel surveillance systems, verbal autopsy methods—each attempts to fill specific gaps, but none provides a complete substitute for robust routine systems.

The Demographic and Health Surveys (DHS) program exemplifies both the power and limitations of this approach. These nationally representative surveys have generated invaluable data on maternal and child health across low- and middle-income countries for four decades. They allow cross-country comparisons and trend analysis that would otherwise be impossible. Yet they occur only every five years in most countries, capture self-reported rather than measured health outcomes for many indicators, and cannot detect rare events or rapid changes.

Sentinel surveillance—placing intensive monitoring at selected sites—offers another partial solution. By choosing facilities or populations for deep data collection, researchers can generate high-quality information about specific conditions. HIV surveillance through antenatal clinics pioneered this approach. But sentinel sites are, by definition, not representative. Extrapolating from them to national estimates requires assumptions that may or may not hold.

Verbal autopsy methods attempt to determine causes of death through structured interviews with family members. When trained interviewers ask systematic questions about symptoms preceding death, algorithms can assign probable causes. The approach has improved dramatically with standardized instruments and machine learning, but it remains a poor substitute for medical certification, particularly for distinguishing between conditions with similar presentations.

The fundamental challenge is that each method captures different populations, uses different definitions, and operates on different timescales. Reconciling these sources requires sophisticated statistical methods and substantial judgment. The synthesis often reveals more about the methods than about the underlying health reality.

Takeaway

Alternative data sources are essential stopgaps, but they function best as complements to—not replacements for—investments in strengthening routine health information systems.

The path forward requires a fundamental reorientation in how we value health information. Data quality cannot remain an afterthought, addressed only after service delivery priorities are funded. It must be understood as the foundation upon which effective service delivery depends.

This means sustained investment in civil registration and vital statistics systems, even when the payoff horizon extends beyond electoral cycles. It means redesigning incentive structures so that accurate reporting is rewarded rather than punished. It means building domestic capacity for data analysis rather than perpetuating dependence on external technical assistance.

International funders bear particular responsibility. When vertical programs create parallel reporting systems that serve donor accountability rather than national planning needs, they actively undermine the information infrastructure that countries need. The global health community must move beyond extractive data practices toward genuine partnership in building sustainable information systems. Without this shift, we will continue treating symptoms while remaining blind to the disease.