Tests of scaling assumputions, construct validity and reliability of the Chinese child health questionnaire, partent form (CHQ-PF50) and child form (CHQ-CF87)

Abstract

Validated instruments for assessment of quality of life in Chinese children are as yet unavailable. We determined the construct validity and reliability of the translated Chinese versions of the Child Health Questionnaires (Landgraf, Abetz & Ware, 1999) designed for completion by parents (CHQ-PF50) and children (CHQ-CF87). The Chinese versions were developed through iterative forward and backward translation processes by independent parties. The feasibility, as rated by degree of difficulty using a 4-point scale, and time for completion were evaluated for the Chinese CHQ-PF50 and CHQ-CF87 in 15 and 11 subjects, respectively. To assess the construct validity and reliability, 1143 parents of healthy children and 823 school children were invited to complete the Chinese CHQ-PF50 and CHQ-CF87, respectively. The results showed that both the Chinese CHQ-PF50 (mean rating 1.66) and CHQ-CF87 (mean rating 1.33) were easy to complete, with completion times of 14.23±5.23 minutes and 13.82±3.52 minutes, respectively. Psychometric analysis on item convergent validity and discriminant validity showed perfect or near perfect (>99%) rates of success for all ten scales in the CF87 and >94% for all but one scale in the PF50. The exception was the general health scale (86%). Minimal floor effects were observed for both questionnaires. However, substantial ceiling effects were observed for the five scales in both questionnaires (physical functioning, role-emotional, behavioral, role-physical, bodily pain and family activities). The median alpha coefficient of reliability for CF87 was 0.90 (range 0.85 to 0.94). The median alpha coefficient for PF50 was 0.80 (range 0.44 to 0.88), with the mental health scale falling just below the minimum criterion for group level analysis (0.68) and the general health scale being the lowest (0.44). These findings suggest that the Chinese translations of CHQ-PF50 and CHQ-CF87 are robust and sufficient. Additional work with regard to ceiling effects is required to assess the performance of the measures in condition groups.