METHODS: We compare mean scores in individuals aged 35-89 in five surveys. Preliminary data from the National Health Measurement Study (NHMS) (June 2005 to February 2006), a random digit dialed telephone survey, includes 1262 respondents. The NHMS included categorical self-rated health (CSRH), SF-36v2 scored as the SF-6D, EQ-5D, QWB-SA with VAS, HUI2/3, VAS, and HALex. The Medical Expenditures Panel Survey (2003) has 16533 respondents from a mailed questionnaire including CSRH, SF-12v2 scored as the SF-6D, and EQ-5D with VAS. The Joint Canada/United States Survey of Health (2002-2003) has 3913 respondents from a telephone survey including CSRH and HUI2/3. The US Valuation of the EuroQol EQ-5D Health States (2002) has 2590 respondents from an interviewer present, paper-and-pencil questionnaire including CRSH, EQ-5D with VAS, and HUI2/3. The National Health Interview Survey (2003) has 23227 respondents from an in-person interview including HALex and information for an imputed scale called the QWBx2. All analyses were performed in STATA 8 with weights necessary to generate nationally representative values.
RESULTS: All indices, scaled with 1=full health, 0=dead, show overall decreasing HRQOL with approximately parallel slopes as age increases in all surveys. When co-administered in the NHMS, QWB-SA has lowest scores in each age category; EQ-5D (US weights) gives highest scores. The average difference across age groups between estimates from interviewer administered surveys and self-administered surveys is 0.032; in general, estimates from interviewer-administered surveys are higher than estimates from self-administered surveys. Across surveys with the same mode of administration, nationally representative mean estimates from the same instrument are similar.
CONCLUSIONS: There are now several available datasets which provide nationally representative values for HRQoL indices. These HRQoL indices show similar patterns across age groups, although the means are offset when comparing one index to another. The mean values for the same index from different surveys are generally comparable when mode of administration is the same. When choosing among available datasets for a comparator to data collected in other studies, it appears that researchers should consider the mode of survey administration.