Development of the Primary Care Quality-Homeless (PCQ-H) instrument: a practical survey of homeless patients' experiences in primary care

Abstract

Background: Homeless patients face unique challenges in obtaining primary care responsive to their needs and context. Patient experience questionnaires could permit assessment of patient-centered medical homes for this population, but standard instruments may not reflect homeless patients' priorities and concerns.

Objectives: This report describes (a) the content and psychometric properties of a new primary care questionnaire for homeless patients; and (b) the methods utilized in its development.

Methods: Starting with quality-related constructs from the Institute of Medicine, we identified relevant themes by interviewing homeless patients and experts in their care. A multidisciplinary team drafted a preliminary set of 78 items. This was administered to homeless-experienced clients (n=563) across 3 VA facilities and 1 non-VA Health Care for the Homeless Program. Using Item Response Theory, we examined Test Information Function (TIF) curves to eliminate less informative items and devise plausibly distinct subscales.

Results: The resulting 33-item instrument (Primary Care Quality-Homeless) has 4 subscales: Patient-Clinician Relationship (15 items), Cooperation among Clinicians (3 items), Access/Coordination (11 items), and Homeless-specific Needs (4 items). Evidence for divergent and convergent validity is provided. TIF graphs showed adequate informational value to permit inferences about groups for 3 subscales (Relationship, Cooperation, and Access/Coordination). The 3-item Cooperation subscale had lower informational value (TIF<5) but had good internal consistency (α=0.75) and patients frequently reported problems in this aspect of care.

Conclusions: Systematic application of qualitative and quantitative methods supported the development of a brief patient-reported questionnaire focused on the primary care of homeless patients and offers guidance for future population-specific instrument development.

Figure 1

Depiction of Test Information Function curves, by subscale, for the initial set of items field-tested (solid lines) and for the items selected after reduction (dotted lines). The “latent trait” θ refers to the underlying strength of satisfaction for the relevant trait, with θ for most persons falling between -1.5 and +1.5, with small numbers falling outside these bounds. For the Patient-Clinician Relationship subscale, 5.2% of respondents had latent trait value θ≤ -1.5, and 8.2% had θ≥ +1.5.For the (perceived) Cooperation subscale, 7.9% had latent trait value θ≤ -1.5, and 17% had θ≥ +1.0. For the Access/Coordination subscale, 2.5% had latent trait value θ≤ -1.5, and 8.7% had θ≥ +1.5.For the Homeless-specific needs subscale, 4.5% had latent trait value θ≤ -1.5 and 16% had θ≥ +1.0.