Beware of the origin of numbers: Standard scoring of the SF-12 and SF-36 summary measures distorts measurement and score interpretations

The 12-item Short Form Health Survey (SF-12) is a generic health rating scale developed to reproduce the Physical and Mental Component Summary scores (PCS and MCS, respectively) of a longer survey, the SF-36. The standard PCS/MCS scoring algorithm has been criticized because its expected dimensionality often lacks empirical support, scoring is based on the assumption that physical and mental health are uncorrelated, and because scores on physical health items influence MCS scores, and vice versa. In this paper, we review the standard PCS/MCS scoring algorithm for the SF-12 and consider alternative scoring procedures: the RAND-12 Health Status Inventory (HSI) and raw sum scores. We corroborate that the SF-12 reproduces SF-36 scores but also inherits its problems. In simulations, good physical health scores reduce mental health scores, and vice versa. This may explain results of clinical studies in which, for example, poor physical health scores result in good MCS scores despite compromised mental health. When applied to empirical data from people with Parkinson's disease (PD) and stroke, standard SF-12 scores suggest a weak correlation between physical and mental health (r_s.16), whereas RAND-12 HSI and raw sum scores show a much stronger correlation (r_s.67–.68). Furthermore, standard PCS scores yield a different statistical conclusion regarding the association between physical health and age than do RAND-12 HSI and raw sum scores. We recommend that the standard SF-12 scoring algorithm be abandoned in favor of alternatives that provide more valid representations of physical and mental health, of which raw sum scores appear the simplest.