A randomized study comparing instruments for measuring health-related quality of life in _HIV-infected patients

Abstract

To compare the feasibility, reliability, validity and sensitivity to change of the MOS-HIV and MQOL-HIV in order to determine their suitability for use in clinical research. Five hundred and fifty-eight HIV-infected patients and 80 healthy blood donors were randomly assigned to receive the MOS-HIV or MQOL-HIV. Test-retest reliability was assessed in 98 clinically stable patients, and responsiveness in 296 patients initiating or switching anti-retroviral treatment. Feasibility was assessed using mean time of administration and percentage of missing responses. Reliability was assessed using Cronbach's alpha and the intraclass correlation coefficient (ICC). Construct validity was assessed by correlating questionnaire scores with EuroQol-5D scores, number of symptoms, CD4 cell count and viral load. The area under the curve (AUC) was used for discrimination between patients and healthy donors, and HRQoL scores were compared across disease stage. Responsiveness was assessed by calculating the standardized effect size (SES). Mean administration time was 16 minutes for both questionnaires. On the MOS-HIV 18.9% patients had missing responses compared with 33.6% on the MQOL-HIV. Cronbach's alpha values were higher for MOS-HIV sub-scales (0.78-0.89) than MQOL-HIV sub-scales (0.44-0.82), and neither instrument showed good test-retest reliability (ICC values of 0.24-0.85 for MOS-HIV versus 0.48-0.82 for MQOL-HIV). AUC values for the MOS-HIV were 0.6-0.86, compared with 0.5-0.79 for the MQOL-HIV, and the MOS-HIV had higher correlations with symptoms (r = -0.28 to 0.79) and EuroQol scores (r = 0.4-0.66) than the MQOL-HIV (r = -0.15 to 0.42 and r = -0.11 to 0.59, respectively). Neither instrument discriminated well between disease stages. Eight of 11 MOS-HIV sub-scales and the Mental Health Summary Score were responsive to change (SES, 0.18-0.36), compared with six of 10 MQOL-HIV sub-scales and MQOL Index (SES, 0.16-0.27). Neither instrument demonstrated completely satisfactory psychometric properties for use in clinical research, although the MOS-HIV performed slightly better on feasibility and validity and the MQOL-HIV on test-retest reliability.