Choosing metrics in public health assessments: attributing credit for the recent large coronary heart disease mortality decline in the US population
- 1Department of Public Health and Primary Care, University of Cambridge, Cambridge, UK
- 2The Institute of Health and Society, Newcastle University, Newcastle upon Tyne, UK
- 3The Division of Public Health, University of Liverpool, Liverpool, UK
Background
Most explanations of falls in coronary heart disease (CHD) mortality limit themselves to event based metrics (such as fewer deaths). However, time based metrics (such as life-years gained (LYG)) promise to capture more of the social value attached to the deaths averted. We have assessed the sensitivity of conclusions about the relative contributions of treatments and risk factor changes to the choice of metric.
Methods
Using a validated CHD mortality model (IMPACT), we integrated data on the number of CHD patients, treatment uptake, treatment effectiveness, risk factor trends, and median survival among US adults aged 25–84 between 1980 and 2000, in order to estimate fewer deaths and LYGs. LYG were estimated using the US general population life expectancy for CHD onsets averted in 2000 and the median survival rates from Medicare for the additional survivors after CHD onset. (All the latter gains are currently attributed to treatments). We examined how uncertainty within the model may vary according to choice of metric.
Results
Between 1980 and 2000, CHD mortality-rates halved resulting in approximately 341 745 fewer deaths in 2000; approximately 47% of the fall was attributed to treatments in patients (after clinical presentation) and 44% to population-wide risk factor reductions (independent of medication). However, this split was altered to 35%/65% when LYG was used. Taking smoking as an example, those who did not experience a smoking attributable CHD death in 2000 because they did not smoke have been given the average US life expectancy (at the age of averted death) in 2000. The life-expectancy of the hypothetical non-smoker, however, would be expected to be higher. Applying these extended years to those deaths avoided by not smoking can add more than 50 000 more life years attributable to the decline of smoking in the population. Furthermore, while the model attributes all gains from increased survival post CHD onset to improved treatments, an increasing proportion of non-smokers among CHD patients could result in additional LYG attributable to smoking reduction.
Discussion and Conclusions
Using life-years gained rather than deaths avoided strengthens the case for primary prevention by risk factor reduction, because it captures more fully the social gains that result. The additional assumptions required are all relatively minor. It seems that past versions of the model may understate the relative importance of risk factor changes when better assessed using the metric of life-years gained.
