Methods: Focus groups were conducted with homeless women recruited from service sites in the Boston, MA area to capture the diversity of the population, including young mothers, older single women, and substance users. Boston’s state-mandated universal health insurance and a strong homeless health care network enabled focus on decisional components in the absence of financial access barriers. Groups were conducted in English and Spanish using a structured discussion guide.
Results: Forty-three women participated in 8 groups between November 2012 and April 2013. The mean age of the participants was 40 years; 47% were white and 28% Hispanic/Latina; 54% had a maximum educational achievement of high school or less; nearly all (98%) had had at least one lifetime cervical cancer screen. Decision-relevant choice attributes and levels that addressed intervention-feasible factors were identified as follows: provider characteristics (patient’s familiarity with provider vs. unknown, respectful and accepting attitude vs. neutral); test timing/scheduling (offered at walk-in clinic vs. scheduled appointment; completed during provision of other care vs. independently); results provision (provided in-person by nurse with counseling vs. by phone/mail without counseling); sample collection (clinician-collected vs. self-collected for HPV screen); testing setting (homeless health clinic vs. general clinic vs. public hospital). Choice attributes addressing societal or environmental factors that were outside the realm of feasible intervention strategies included women’s fear of life-threatening diagnoses, disregard for health care during addiction episodes, and competing priorities for food and shelter.
Conclusions: Homeless women’s experience of cervical cancer screening is unique and empiric determination of their choice elements is critical to discrete choice experimental design. These qualitative data reveal both the elements of choice and their variability across subpopulations of the homeless population, by age, comorbid conditions, and housing status/history. The inclusion of empirically-identified attributes will enable the DCE to inform development of targeted screening interventions to increase uptake among this population.