Paul Grootendorst1, Emmanuelle Piérard, Minsup Shim. 1. Faculty of Pharmacy, University of Toronto and Department of Economics, McMaster University, Hamilton ON, 144 College Street, Toronto, ON, M5S 3M2, Canada. paul.grootendorst@utoronto.ca
Abstract
OBJECTIVES: Several studies suggest that, on the basis of life-expectancy regressions, pharmaceutical drugs are responsible for much of the marked gains in life expectancy observed over the last 50 years. In this article, we critically appraise these studies. METHODS: We point out several modeling issues: identification of the contribution of new drugs from advances in disease management, changes in the distribution of healthcare and other confounding factors. RESULTS: We suggest that some models produce estimates of pharmaceutical productivity that are implausibly high. Other models have very large forecast errors. Finally, the models that we replicated were found to be sensitive to seemingly innocuous changes in specification. CONCLUSION: It is difficult to estimate the biomedical determinants of life expectancy using aggregate data. Analyses using individual level data or perhaps disease-specific data will probably produce more compelling results.
OBJECTIVES: Several studies suggest that, on the basis of life-expectancy regressions, pharmaceutical drugs are responsible for much of the marked gains in life expectancy observed over the last 50 years. In this article, we critically appraise these studies. METHODS: We point out several modeling issues: identification of the contribution of new drugs from advances in disease management, changes in the distribution of healthcare and other confounding factors. RESULTS: We suggest that some models produce estimates of pharmaceutical productivity that are implausibly high. Other models have very large forecast errors. Finally, the models that we replicated were found to be sensitive to seemingly innocuous changes in specification. CONCLUSION: It is difficult to estimate the biomedical determinants of life expectancy using aggregate data. Analyses using individual level data or perhaps disease-specific data will probably produce more compelling results.