NIH disease funding levels and burden of disease

Abstract

Background: An analysis of NIH funding in 1996 found that the strongest predictor of funding, disability-adjusted life-years (DALYs), explained only 39% of the variance in funding. In 1998, Congress requested that the Institute of Medicine (IOM) evaluate priority-setting criteria for NIH funding; the IOM recommended greater consideration of disease burden. We examined whether the association between current burden and funding has changed since that time.

Methods: We analyzed public data on 2006 NIH funding for 29 common conditions. Measures of US disease burden in 2004 were obtained from the World Health Organization's Global Burden of Disease study and national databases. We assessed the relationship between disease burden and NIH funding dollars in univariate and multivariable log-linear models that evaluated all measures of disease burden. Sensitivity analyses examined associations with future US burden, current and future measures of world disease burden, and a newly standardized NIH accounting method.

Results: In univariate and multivariable analyses, disease-specific NIH funding levels increased with burden of disease measured in DALYs (p = 0.001), which accounted for 33% of funding level variation. No other factor predicted funding in multivariable models. Conditions receiving the most funding greater than expected based on disease burden were AIDS ($2474 M), diabetes mellitus ($390 M), and perinatal conditions ($297 M). Depression ($719 M), injuries ($691 M), and chronic obstructive pulmonary disease ($613 M) were the most underfunded. Results were similar using estimates of future US burden, current and future world disease burden, and alternate NIH accounting methods.

Conclusions: Current levels of NIH disease-specific research funding correlate modestly with US disease burden, and correlation has not improved in the last decade.

Figure 1. Ten-year Comparison of Differences Between Actual and Expected Disease-Specific NIH Funding Relative to US Burden of Disease in DALYs.

A comparison of differences between actual and expected funding values as predicted by DALYs burden alone in 1996 (light blue) and 2006 (navy). Negative values reflect actual funding dollars less than expected and positive values represent actual funding dollars more than expected.

Figure 2. NIH Funding in 2006 and US Disease Burden in DALYs in 2004 for 29 Common Medical Conditions.

The solid line represents the results of a traditional multivariable analysis, showing the relationship between US disease-specific DALYs burden and actual 2006 NIH funding dollars. The dashed line projects NIH funding levels in a similar multivariable model that requires that a disease with no burden receives no funding (constrained model). Though the models produce similar results, several diseases that would be considered overfunded in one model are considered underfunded in the other. For example, cervical cancer appears to be overfunded relative to the dashed line, while it is underfunded relative to the solid one.

Figure 3. Differences Between Actual and Expected Disease-Specific Funding in 2006.

Determinations of actual funding relative to expected funding were generally similar among separate analytic models predicting funding levels from disease burden measures. Univariate results are based on DALYs alone (navy), the only variable retained in a stepwise forward multivariable model. A traditional multivariable model including public interest variables (grey-blue) retained only DALYs and total charity revenue in the model. A constrained multivariable (light blue) model required an intercept of zero-zero to impose a requirement that conditions with no burden received no funding and retained DALYs, total number of US hospital discharges, and mean charge per hospitalization in 2004.