Joint Modeling of DALYs and QALYs – Center for Health Decision Science

Disability-adjusted life years (DALYs) and quality-adjusted life years (QALYs) are typically treated as alternative outcome measures. John A. Graves presented three approaches to incorporating these multidimensional health outcome measures in common decision analytic modeling frameworks in a CHDS seminar. Graves is Professor of Health Policy and Medicine at Vanderbilt University School of Medicine and Vanderbilt University Medical Center, and Professor of Management at Vanderbilt Owen Graduate School of Management. The work he presented was co-authored with Jinyi Zhu, a CHDS alumni.

DALYs are the primary choice of health outcome for economic evaluations in global health settings, but there is limited guidance on proper estimation of DALYs in decision analytic models. Standard economics education on modeling tends to focus on QALYs. DALYs, the sum of years lost to disability (YLD) and years lost to disease (YLL), have important methodological differences compared to QALYs. One of the differences relevant to modeling lies in the need to separately track the timing and number of deaths due to disease for YLL calculation. To address this gap, Graves and his collaborators showed three approaches tailored to beginner, intermediate, and advanced modelers. To illustrate these approaches, Graves presented an example Markov cohort model with generic health states. Moreover, Graves noted that the methods extend to other modeling frameworks such as microsimulations and discrete event simulations.

The beginner “Separate Death State” approach adds a cause-specific death state into the model structure and uses the Markov Trace to track state occupancy. The intermediate “Non-Markovian Trackers” approach augments the transition matrix with a non-Markovian tracking component without adding a separate state. The advanced “Markov Chain with Rewards” approach also adds an absorbing state and defines a block matrix Markov chain with rewards for occupancy and disease related death transitions, which is more efficient and also allows for other moments of interest. Graves compared the results of these three approaches with two “shortcut” approaches used in the field and showed that the shortcut approaches incorrectly estimated YLLs.

Learn more: Read the publication, Modeling Disability-Adjusted Life Years for Policy and Decision Analysis
Learn more: Resource Pack: Models for Health Decision Science