The test-retest reliability of the modified Patient Generated Index

OBJECTIVES: The purpose of this study was to determine the test-retest reliability of the recently modified Patient Generated Index (PGI) when used as a self-completed postal questionnaire by older people with arthritis. METHODS: Two sets of postal questionnaires were used, sent two weeks apart, followed by interviews with a theoretically chosen subsample of respondents to both. Reliability for respondents whose health had not changed in the intervening two weeks was assessed using intraclass correlation coefficients. Data from interviews were used to develop theories to select subgroups for analysis. RESULTS: One thousand one hundred and twenty-nine people over 65 years old with arthritis were surveyed and 53 were selected for interview. The response rates to the two surveys were 78% and 83%, respectively. Complete data for the test-retest analysis were available from 236 (29.2%) respondents. The intraclass correlation coefficient (r(i)) was 0.55 95% confidence interval (CI) 0.43 to 0.64, P < 0.0001]. The most optimistic range for the limits of agreement was 64.66 around the difference between two total PGI scores. The intraclass correlation coefficient decreased considerably if there had been more than six changes to the areas chosen for inclusion on the second occasion (r(i) = 0.37, 95% CI 0.01 to 0.65, P < 0.05). Interview data showed that some respondents incorrectly interpreted the scoring instructions, often inconsistently. When their data were excluded, the intraclass correlation coefficient increased to 0.67 (95% CI 0.56 to 0.76, P < 0.0001). CONCLUSIONS: The results indicate that the modified PGI may be sufficiently reliable for within-group comparison of elderly people with arthritis, but not for assessing individual patients. However, it also elicits individual patient concerns and priorities about disease that other questionnaire-based instruments may not identify.