Impact of Missing Values in the Self-Administered Washington Group Short Set: A Cross-Sectional Study Using Secondary Data of a National Survey of Persons With Disabilities in Japan

Abstract

Introduction: The Washington Group Short Set on Functioning (WGSS), consisting of six questions on basic functioning, is recommended for interviewer-administered surveys. While self-administration is a practical alternative, its impact on response accuracy is poorly understood. This study aimed to quantify the extent of missing values in responses to the self-administered WGSS and enhanced WGSS (WGSSE), consisting of the six questions of the WGSS and four additional questions on basic functioning, and to examine the association between respondents' disability-related characteristics and missing values.

Methods: In this cross-sectional study using nationally representative survey data from individuals with disabilities in Japan (n=12,822), the individual and overall percentages of missing values for the WGSS- and WGSSE-related questions were calculated. Binomial logistic regression was used to analyze the relationship between possession of disability certificates, the independent variable, and the missingness of disability status defined by the WGSS and WGSSE, the dependent variable.

Results: Non-response rates for individual WGSS- and WGSSE-related questions were 9.71-30.69%. Missingness of disability status, as defined by the WGSS and WGSSE, which was attributed to missing values for the individual questions, was observed in 16.94% and 24.32% of the participants, respectively. After adjusting for confounders, possession of a disability certificate was significantly associated with the missingness of disability status, as defined by the WGSS and WGSSE (adjusted odds ratio: 1.176-2.931).

Conclusion: Self-administration of the WGSS resulted in a significant proportion of missing responses among individuals with disabilities. Fewer individuals with disability certificates may be represented in disability statistics if missing data are excluded, potentially underestimating disability prevalence by up to approximately 20%. Stakeholders in Japan should consider these drawbacks when interpreting and using the self-administered WGSS data. Further research is needed to explore the characteristics of self-administered WGSS data.

Keywords: disability; disability statistics; missing values; multiple imputation; washington group short set questionnaire on functioning.

Figure 1. Participant selection process

Figure 2. Percentages of participants with missing responses for individual WGSS- and WGSSE-related questions

^†. Among only those who answered the sub-questions of depression on frequency as “daily, weekly, monthly, or a few times a year” (n=10,042); ^††. Among only those who answered the sub-questions of anxiety on frequency as “daily, weekly, monthly, or a few times a year” (n=10,336); ^‡. Integrated indicator of “depression_intensity” and ” depression_frequency” following the Washington group recommendation; ^‡‡. Integrated indicator of “anxiety_intensity” and ” anxiety_frequency” following the Washington group recommendation WGSS: Washington Group Short Set on Functioning; WGSSE: WGSSE enhanced