OBJECTIVES: Factors that constitute barriers to healthcare utilization (HCU) among individuals diagnosed with cancer are not always available in administrative claims data. They would be collected via survey, which can result in incomplete or missing responses. Missingness is typically characterized as ‘yes/no’ without details on the characteristics of non-responders. However, these details can inform statistical approaches to account for missing responses. Our objective was to describe missingness patterns from a survey of social and economic barriers to HCU.

METHODS: This study utilized a linkage of individual-level tumor registry and Medicare claims data from 2018-2021. Eligible individuals returned a paper questionnaire fielded in 2022-2023. The survey asked questions from three categories (number of questions): income (3), perceived discrimination (PD; 5), and psychosocial behavioral characteristics (PBC; 14). To capture emergency department visits and hospitalizations within the administrative claims, we imposed a 6-month continuous enrollment (CE) criterion to define the CE sample. We extracted the category-level count of missing responses and reported frequencies. We tested for group differences using Fisher’s exact test.

RESULTS: Sample sizes were 283 in the full sample and 163 in the CE sample. The full sample proportions for male gender, non-Hispanic White, and non-Hispanic Black were: 56%, 81%, and 16%, respectively. Missingness proportions in the full sample were 1% for income, 15% for PD, and 8% for PBC. These proportions were similar across subgroups defined by race (White vs. non-White) and gender. There were statistically significant differences by age groups with the highest missingness for PD (21%) and PBC (11%) in those aged 76+ years. In the CE sample, missingness for PBC was statistically significantly higher among individuals with HCU (13%) compared to those without (5%).

CONCLUSIONS: Among survey respondents, responses varied across question categories, age groups, and HCU groups. Additional research is needed for questions related to PD.