Introduction

Persons experiencing homelessness are faced with medical, social, and environmental challenges to their physical and mental health. Moreover, persons experiencing homelessness disproportionately suffer from medical illness, problematic substance use, and psychiatric disorders [1–2]. High rates of psychiatric disorders within homeless populations can make health management access and utilization more complex [3–4]. For instance, persons with mental illness access primary care less frequently than patients without mental illness, yielding worse chronic disease management outcomes, and disproportionately high premature mortality rates [5].

For homeless individuals, these challenges are compounded by health service underutilization with significant unmet needs and problems accessing primary and specialty care, [6–8] in turn increasing use of hospital and emergency departments [9–11]. Even when payment is assured, as in Canada [8, 12] or the Veterans Health Administration, [13] this remains the case.

Efforts to remediate these challenges include initiatives to extend the patient-centered medical home (PCMH) concept to homeless individuals. As summarized by Strange et al., the PCMH embodies core principles of primary care (access, comprehensiveness, integration, and relationship), new ways of organizing practice and its payment, and attention to a wide range of performance indicators in which patients play a crucial role in reporting whether key attributes of primary care are in fact being attained [14].

In recent years, both the US Department of Veterans Affairs and the US Department of Health and Human Services have recently advanced PCMH models for persons experiencing, or emerging from, homelessness. The Department of Veterans Affairs funded 37 such programs starting in 2012, [15] and their use appears to facilitate reductions in emergency department services [16]. The Health Care for the Homeless Program, operating with the US Department of Health and Human Services (HHS), has long included strong examples of this nature, [17] and HHS has strongly encouraged the adoption of the PCMH model by federally qualified health centers, and advanced a special initiative multiply diagnosed HIV positive homeless individuals [18].

By definition, patients’ assessments of their experience in care serve as crucial indicators of success in fostering a PCMH, [14] and their positive assessments correlate with retention, adherence and improvements in some conditions [19–21]. For homeless patients with mental health conditions in particular, there is ample qualitative evidence that seeking care can be an adverse experience colored by stigma, and lack of coordination or disrespect [22]. Research efforts to systematically understand what patient or system characteristics help to produce a favorable care experience for homeless patients with mental health conditions are scarce. The present study utilized a new patient-reported 33-item instrument developed specifically to capture homeless persons’ experience of primary care, the Primary Care Quality-Homeless survey [23]. Drawing on data from a sample of patients receiving care at 5 sites across the country, we sought to identify determinants of primary care experiences for homeless persons with mental health conditions across various delivery models.

Methods

Measures

Conceptual model

Sofaer and Firminger’s patient perceptions model guided choice of predictors related to primary care experience [29]. As shown in Fig. 1, nodes within the model include the following measured variables: characteristics of patient (demographics, health, depression/anxiety, severe psychiatric symptoms, drug and alcohol severity, and housing status), social support, perceived extent of choice among providers, primary care service tailoring for homeless persons, and patient experience seeking and using services. Within this model, the nodes highlight the predictive relationship between the nodes and outcome variable, patient perceptions of the primary care experience. In modifying Sofaer and Firminger’s model, we added two particular nodes of interest: social support and primary care service tailoring for homeless persons (based on receiving care at either of two tailored primary care programs, compared to three mainstream programs). Recent studies have highlighted the importance of both variables in similarly vulnerable populations [30–31]. Our adaptation of the model is intended to be conceptually broad and generalizable to other vulnerable populations. However dimensions within the adapted model highlight important characteristics in populations with mental health conditions.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 1. Adapted conceptual framework for predictors of patient care experiences among individual experiencing homelessness.

This framework was adapted from Shosanna Sofaer and Kirsten Firminger model of patient perceptions. For the purpose of our analysis, we retained some original variables as well as the general relationship expressed in this model, although the variables are specific to our sample population.

https://doi.org/10.1371/journal.pone.0117395.g001

Results

The majority of the sample was male (83%). With regard to race, 33% were White, 56% African-American, 2% American Indian/Alaskan Native, <1% Asian/Pacific Islander, and 9% other race. Mean age was 52 years old (± 16), years of education were 12.9 (SD = 2), 82% of the sample had an annual income of less than $15,000 (Table 1). The mean PCQ-H score for all persons included in the analysis was 3.13 (SD = 0.38) on a 4-point Likert Scale.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 1. PCQ-H sample population demographics across five clinic sites.

https://doi.org/10.1371/journal.pone.0117395.t001

The likelihood ratio test determined that demographic variables did not increase predictive validity and were therefore excluded from the model χ² = -5.5, df = 7, p < 0.05. The linear combination of the remaining predictive factors was significantly related to the primary care experience score, F (21, 354) = 8.95, p ≤0.0001 and accounted for approximately 36% of the variation in the sample. In Table 2, we present the individual predictors. The relationships between the predictive measures and the primary care experience score were in the expected direction; site of care with tailored sites and Mainstream A (the one with some tailoring) obtaining more favorable scores (F = 2.80, p = 0.03), with perceived extent of choice (F = 23.29, p<0.0001), and housing status (F = 2.91, p = 0.03), proving significant. There was a significant interaction of severe psychiatric symptoms by site [F(df = 4x1) = 3.61, p = 0.01)], further treated below. Non-significant covariates included depression/anxiety (F = 1.79, p = 0.18), severe psychiatric symptoms (F = 2.39, p<0.12), drug severity (F = 0.81, p<0.37), alcohol severity (F = 0.01, p<0.93, and social support (F = 0.77, p<0.38). General health status approached but did not attain significance (F = 3.38, p≤0.07).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 2. The effect of patient and site characteristics on predictors of care experiences among PCQ-H sample population

¹.

https://doi.org/10.1371/journal.pone.0117395.t002

Discussion

Homeless persons with mental health conditions represent a particularly vulnerable population in which factors conducive to a positive primary care experience remain understudied and yet potentially important to ongoing efforts to foster patient-centered care for vulnerable populations. The present analysis sought to identify factors that predict experience across primary care settings that varied in the degree to which they tailored service delivery for homeless persons. Tailoring ranged from standard mainstream clinics to settings that utilized dedicated, specialized staff, special hours and locations of operation, and unique services such as facilitated access to psychiatric care, food, or shelter. Our initial conceptual model, guided by an existing model of patient perceptions of care, [29] is partially supported by our results. We evaluated whether patient characteristics, social support, extent of choice in providers, and primary care tailoring by site, determined experiences in primary care for homeless persons with mental health conditions. The following factors predicted positive primary care experiences: homeless-specific tailoring, perceived extent of choice in providers, and housing status. We also found a unique interaction effect between severe psychiatric symptoms and site suggesting that the relationship between severe psychiatric symptoms and primary care experiences differed depending on site of care. Some patient characteristics (specifically health, depression/anxiety, severe psychiatric symptoms, drug and alcohol severity), and social support were not significantly associated with the primary care experiences in this population.

Tailored services predicted more positive experiences of primary care among homeless persons with severe psychiatric symptoms. Less favorable experiences of primary care were associated with two mainstream primary care settings. The stronger performance of one mainstream site that had some identifiable aspects of homeless-specific tailoring may be due to that tailoring itself, or to other organizational characteristics that we were unable to measure.

Our findings do suggest an association between primary care experience and degree of service tailoring when the sites are considered along a continuum of service tailoring. Since care service deficits for homeless persons with mental health conditions have been widely documented [1, 41], our findings tend to support what recent demonstration projects have also suggested [17, 24–25]; positive primary care experiences appear to be influenced by the service design characteristics of the agency delivering primary care. While studies have reported on specific sites or programs, to our knowledge, no prior study has incorporated data from mainstream and tailored primary care settings in their analysis of patient care experience. Furthermore, findings for site are patient-centered, being based on person-oriented data, rather than utilization or mortality [42–43].

Patients’ perceived ability to switch providers on demand is the strongest predictor of primary care experience in this analysis. Our finding is particularly interesting due to the nature of this variable; a perceived choice in care calls attention to the importance assigned to flexible and patient-centered care. Moreover, the belief that a patient can exercise such control over their care may be an extension of their experience of their care overall. This underscores the importance of the provider-patient relationship [44], as well as the importance of control where uncertainty to predict or anticipate the availability of resources is commonplace [22].

Our findings for housing status, which was categorically broken into domiciled, recently homeless, chronically homeless, and chronically and recently homeless, illustrates the importance of having stable housing to one’s experience of primary care. This is not surprising given the known association between socioeconomic status and health [45–46]. From our findings we might infer that attainment of a stable domicile is associated with a more favorable primary care experience, compared to the least stable (chronically and recently homeless) arrangement. Unfavorable primary care experiences could occur for various reasons, including transinstitutionalization, and other health complications due to unsheltered lifestyle [47].

The interaction between severe psychiatric symptoms and site of care was not expected. At Tailored VA and Mainstream VA-A sites, presence of severe psychiatric symptoms were associated with a more positive rating of the primary care experience. However at Mainstream VA-B and Mainstream VA-C sites, presence of severe psychiatric symptoms were associated with less favorable rating of the primary care experience. There may be something specific to these populations or site of care with regard to how patients with psychiatric symptoms are received or cared for. Mainstream VA-C served as a regional referral site for inpatient psychiatric care and likely had a particularly severe psychiatric case-mix. Based on discussions with Mainstream VA-C clinicians, patients with past inpatient psychiatric hospitalizations were more likely to stay in the vicinity of Mainstream VA-C. The uniquely low primary care experience scores may reflect a particularly extreme population in terms of psychiatric severity.

Meanwhile, the Tailored Non-VA site offers an obviously strong experience for all clients. Its experience for patients with severe psychiatric symptoms was not low, but it did not attain parity of experience for patients with severe psychiatric symptoms. The similarity of scores across both groups (presence and absence of severe psychiatric symptoms) at the Tailored VA site may reflect the reality that this particular tailored VA program was housed within mental health space where the services for mental care were very tightly integrated. In fact, both Mainstream VA-A and Tailored VA sites are full service health systems where outpatient care is provided alongside the full range of psychiatric services, including inpatient care. This does not apply to the Mainstream VA-B and VA-C sites. Mainstream VA-B lacks inpatient psychiatric care and ordinarily sends patients requiring such care to Mainstream VA-C, roughly 80 miles distant. This pattern of service distribution may have implications for the types of formerly homeless patients who reside in close proximity to each of these two facilities. However, we do acknowledge that these findings, exploratory as they are, primarily invite future research directed at understanding how the presence of severe psychiatric symptoms might influence the experience of primary care relationships and services.

Health status approached but did not attain statistical significance as a predictor of primary care experiences. This aligns with prior studies that have had mixed results among similarly homeless vulnerable populations. Some prior studies have found that better quality of care leads to greater satisfaction, and presumably better health [48]. However, other studies have not observed a significant association between health status and satisfaction with care [49–50].

Although it was important to account for depression and anxiety in this population, our findings hint at a complicated relationship with other aspects of care. Since depression and anxiety are risk factors for non-compliance of medical treatment and diminished health status, it is difficult to account for the myriad of direct and indirect reasons why patient experience is not significant. Perhaps assessing changes in depression or anxiety symptoms rather than a static score of depression/anxiety would predict care experience in a more meaningful way.

Interestingly, social support was not a significant predictor of primary care experience among homeless persons with mental illness. While it was assumed that greater social support could be viewed as resource capital with potential to improve healthcare access and outcomes, [51–52] our measure of social support (companionship ratings per Strong Ties scale) may not capture forms of valued social capital in this sample population (i.e., community residential networks related to severe psychiatric symptoms or veteran status, substance disorder support groups, ACCESS, etc.). For example, a patient’s likelihood of forming a strong relationship in primary medical care appears to have little to do with the quality of companionship they obtain outside of care. Other types of social capital not measured here, including group membership, merit future study.

Limitations to this study include the correlational nature of the analysis, requiring caveats regarding causal reasoning. Additionally, our scoring of severe psychiatric symptoms reflect only a part of the symptom constellations in the diagnosis of mental illness, substance misusing, and PTSD. To avoid misleading diagnostic interpretations, scores are translated into “presence of severe psychiatric symptoms,” or “absence of severe psychiatric symptoms” for our sample population. Moreover, data available to this analysis could not cover all variables anticipated by Sofaer and Firminger's conceptual model (e.g. “social/cultural norms,” “reputation of provider”). A qualitative approach could help explore the subtleties of primary care processes and experiences that may otherwise be lost with limited response criteria. Lastly, four out of five sites were VA, and therefore those lacking financial coverage for care was minimal. Exploration of financial barriers may be an important determinant of experience in other settings.

Given ongoing healthcare reform, there is increasing attention to the care of traditionally underserved populations, to patient-centered care, and to patient perceptions of care. We studied organizational and patient factors that affect perception of care among homeless persons with mental health conditions. Strengths of this study include a strong response rate and use of in-person assessment in an underserved, marginalized population that can be very difficult to interview. Results support the importance of tailored care, perceived choice among providers, and housing status. Although there is no formal definition of tailored primary care, the following service design characteristics are believed to be important: co-location, same-day services, and integration of primary care with mental health and substance disorder services. Continued implementation research that offers insight into what service characteristics are valued, as well as how those service processes are best delivered is needed. To improve performance on patient experience surveys, organizations may want to deliver services that offer greater flexibility and choice to patients experiencing homelessness.

Acknowledgments

Disclaimer: The opinions expressed are those of the authors and do not represent positions of the Department of Veterans Affairs or other affiliated institutions.