Scoring and Population Norms.

Scoring:

The scoring of PCS and MCS is based on weights derived from the VR-36© administered to 1.4 million veteran enrollees with 877,775 respondents in the 1999 Large Health Survey of Veteran Enrollees (Veterans Health Study), the largest federal survey ever conducted in the VA. The weights were obtained from the large sample to create PCS and MCS scores. That is, dummy indicators were defined for response choices for each of the 12 items in the VR-12© and these were then entered into multivariate regression models to predict PCS and MCS scores based on the VR-36©. The resulting weights and the constant term can be used to compute PCS and MCS scores from the VR-12©. The two summary component scales derived from the VR-12© explained over 90% of the variance in PCS and MCS scales of the VR-36© used in the 1999 Large Health Survey of Veterans Enrollees.

Population norms:

PCS and MCS summary scores are standardized using a t-score transformation and normed to a U.S. population (based on a 1990 norm) of a score of 50 and a standard deviation of 10. Standard-based scoring makes it possible to interpret scores by comparing them to those of a reference population. Contemporary norms have been established and validated based on the Medical Expenditure Panel Survey (MEPS). The updated standard is widely available to serve as a contemporary standard for future applications for health-related quality of life (HRQoL) assessments.

Missing values:

Advancements made in the scoring algorithms of the VR-12© include validated methods for “imputation of missing values” that builds upon a foundation previously developed for the VR-36©. These algorithms allow for additional cases that are not scored due to one or more missing items to be imputed. The method called the Modified Regression Estimate (MRE) uses regression models for the item responses and regresses PCS and MCS on the indicator variables for available responses for items. The method is also corrected for regression towards the mean, given that the regression prediction will trend toward the average in the sample from which the prediction is created. The MRE method uses complete cases to estimate a regression equation where only those items that are present are used. This algorithm is able to compute PCS and MCS for as few as three items available. Use of the MRE approach for national VA and CMS surveys has been able to score as many as 98% of the cases. The algorithm can be scored using a conventional desktop computer.