Putting the Patient in Patient Reported Outcomes: A Robust Methodology for Health Outcomes Assessment

When analyzing many health-related quality-of-life (HRQoL) outcomes, statistical inference is often based on the summary score formed by combining the individual domains of the HRQoL profile into a single measure. Through a series of Monte Carlo simulations, this paper illustrates that reliance solely on the summary score may lead to biased estimates of incremental effects, and I propose a novel two-stage approach that allows for unbiased estimation of incremental effects. The proposed methodology essentially reverses the order of the analysis, from one of 'aggregate, then estimate' to one of 'estimate, then aggregate'. Compared to relying solely on the summary score, the approach also offers a more patient-centered interpretation of results by estimating regression coefficients and incremental effects in each of the HRQoL domains, while still providing estimated effects in terms of the overall summary score. I provide an application to the estimation of incremental effects of demographic and clinical variables on HRQoL following surgical treatment for adult scoliosis and spinal deformity.