Missing data in spinal cord injury survey research: racial differences
1Chen Y, 2Lin H, 2Tseng T, 1Wen H, 1DeVivo M
1University of Alabama at Birmingham, Birmingham, Alabama, USA; 2Louisiana State University Health Sciences Center, New Orleans, Louisiana, USA
Objective: Among people with spinal cord injury (SCI), minorities experience a disproportionately higher burden of diseases. Knowledge of data quality by race/ethnicity will help better design racial health disparity research and understand potential errors/biases. This study was conducted to document the extent of incompleteness in response to questions of sensitive topics and examine the role of race/ethnicity on data quality among individuals enrolled in a longitudinal SCI database.
Design/Method: This study included 7507 participants (5483 non-Hispanic whites, 1414 non-Hispanic blacks, and 610 Hispanics) enrolled in the National SCI Database who returned for follow-up between March 2001 and September 2006 when various psychosocial outcomes were included in the database. For those with more than one follow-up during the study period, status at the latest contact was included in the analyses. Missing data were defined as any missing, unknown or refusal response to the interview items. Socio-demographic variables were updated at each follow-up. Gender and neurologic deficits were assessed during initial hospital care. The mode of data collection was also documented. Multiple logistic regression was conducted to assess the effect of race/ethnicity on missing data for each outcome variable after accounting for potential confounding factors.
Results: The overall missing rate was 29.7%, 9.5%, 9.7%, 10.7%, 12.0%, and 9.8% for the Craig Handicap Assessment and Reporting Technique-Short Form (CHART) economic self-sufficiency subscale, CAGE questionnaire, drug use, Diener’s Satisfaction with Life Scale, Patient Health Questionnaire, and pain severity, respectively. The missing rate for the CHART measure was significantly higher among non-Hispanic blacks and Hispanics than among non-Hispanic whites, after controlling for demographics, injury factors, mode of data collection, and study sites. The missing data in the other outcome measures examined were also significantly higher among non-Hispanic blacks than among non-Hispanic whites but were not significantly different between Hispanics and non-Hispanic whites. Later post-injury years and a higher education level were associated with a lower missing data rate across all outcome variables. The study sites that had a higher loss to follow-up rate tended to have a lower missing data rate among respondents for all outcome measures except the CHART economic self-sufficiency scale. For most of the outcome variables, a higher missing rate was more common in males and in people having SCI at older age and more severe injury. Compared with other modes of data collection, the self-administered data collection method associated with a higher missing data rate for the depression variable (P = 0.004), but a lower missing rate for the drug use variable (P = 0.056).
Conclusion: Our study highlights the importance of research methodology designed to improve non-response or response incompleteness, particularly in non-Hispanic blacks. It also provides insight into the potential bias in SCI health disparity research that results from racial variation in data quality.
Support: National Institutes on Disability, Independent Living, and Rehabilitation Research (NIDILRR 90DP0083).
You Must Be Logged In
This Reminders tab will display for your scheduled lectures 15 minutes before start time and for new messages waiting.