Abstract
Objectives
As young people enter adulthood, the interchangeable use of child and adult outcome measures may inaccurately capture changes over time. This study aimed to use item response theory (IRT) to model a continuous score for functional ability that can be used no matter which questionnaire is completed.
Methods
Adolescents (11‐17 yrs) in the UK Childhood Arthritis Prospective Study self‐completed an adolescent‐Childhood Health Assessment Questionnaire (A‐CHAQ) and a Health Assessment Questionnaire (HAQ). Their parents completed the proxy‐CHAQ (P‐CHAQ). Those with at least two simultaneously completed questionnaires at initial presentation or one year were included.
Psychometric properties of item responses within each questionnaire were tested using Mokken analyses to assess the applicability of IRT modelling. A previously developed IRT model from the Pharmachild‐NL registry was validated in CAPS participants. Agreement and correlations between IRT‐scaled functional ability scores were tested using intra‐class correlations and Wilcoxon signed‐ranked tests.
Results
In 303 adolescents, median age at diagnosis was 13 years and 61% were female. CHAQ scores consistently exceeded HAQ scores.
Mokken analyses demonstrated high scalability, monotonicity and that each questionnaire yielded reliable scores. There was little difference in item response characteristics between adolescents enrolled to CAPS and Pharmachild (maximum item residual 0.08). Significant differences were no longer evident between IRT‐scaled HAQ and CHAQ scores.
Conclusion
IRT modelling allows the direct comparison of function scores regardless of different questionnaires being completed by different people over time. This facilitates ongoing assessment of function as adolescents transfer from paediatric clinics to adult services.
As young people enter adulthood, the interchangeable use of child and adult outcome measures may inaccurately capture changes over time. This study aimed to use item response theory (IRT) to model a continuous score for functional ability that can be used no matter which questionnaire is completed.
Methods
Adolescents (11‐17 yrs) in the UK Childhood Arthritis Prospective Study self‐completed an adolescent‐Childhood Health Assessment Questionnaire (A‐CHAQ) and a Health Assessment Questionnaire (HAQ). Their parents completed the proxy‐CHAQ (P‐CHAQ). Those with at least two simultaneously completed questionnaires at initial presentation or one year were included.
Psychometric properties of item responses within each questionnaire were tested using Mokken analyses to assess the applicability of IRT modelling. A previously developed IRT model from the Pharmachild‐NL registry was validated in CAPS participants. Agreement and correlations between IRT‐scaled functional ability scores were tested using intra‐class correlations and Wilcoxon signed‐ranked tests.
Results
In 303 adolescents, median age at diagnosis was 13 years and 61% were female. CHAQ scores consistently exceeded HAQ scores.
Mokken analyses demonstrated high scalability, monotonicity and that each questionnaire yielded reliable scores. There was little difference in item response characteristics between adolescents enrolled to CAPS and Pharmachild (maximum item residual 0.08). Significant differences were no longer evident between IRT‐scaled HAQ and CHAQ scores.
Conclusion
IRT modelling allows the direct comparison of function scores regardless of different questionnaires being completed by different people over time. This facilitates ongoing assessment of function as adolescents transfer from paediatric clinics to adult services.
Original language | English |
---|---|
Pages (from-to) | 947-954 |
Journal | Arthritis Care & Research |
Volume | 73 |
Issue number | 7 |
Early online date | 14 Apr 2020 |
DOIs | |
Publication status | Published - 1 Jul 2021 |