Abstract
Background
Permanent supportive housing (PSH) programs, which have grown over the last decade, have been associated with changes in health care utilization and spending. However, little is known about the impact of such programs on use of prescription drugs critical for managing chronic diseases prevalent among those with unstable housing.
Objective
To evaluate the effects of PSH on medication utilization and adherence among Medicaid enrollees in Pennsylvania.
Design
Difference-in-differences study comparing medication utilization and adherence between PSH participants and a matched comparison cohort from 7 to 18 months before PSH entry to 12 months post PSH entry.
Subjects
Pennsylvania Medicaid enrollees (n = 1375) who entered PSH during 2011–2016, and a propensity-matched comparison cohort of 5405 enrollees experiencing housing instability who did not receive PSH but received other housing services indicative of episodic or chronic homelessness (e.g., emergency shelter stays).
Main Measures
Proportion with prescription fill, mean proportion of days covered (PDC), and percent adherent (PDC ≥ 80%) for antidepressants, antipsychotics, anti-asthmatics, and diabetes medications.
Key Results
The PSH cohort saw a 4.77% (95% CI 2.87% to 6.67%) relative increase in the proportion filling any prescription, compared to the comparison cohort. Percent adherent among antidepressant users in the PSH cohort rose 7.41% (95% CI 0.26% to 14.57%) compared to the comparison cohort. While utilization increased in the other medication classes among the PSH cohort, differences from the comparison cohort were not statistically significant.
Conclusions
PSH participation is associated with increases in filling prescription medications overall and improved adherence to antidepressant medications. These results can inform state and federal policy to increase PSH placement among Medicaid enrollees experiencing homelessness.
Supplementary Information
The online version contains supplementary material available at 10.1007/s11606-024-08621-0.
KEY WORDS: permanent supportive housing, medicaid, adherence, prescription, difference-in-differences
Introduction
Lack of medication adherence is associated with poor health outcomes and increased hospitalizations, which contribute to excess mortality and cost.1 Severe and persistent mental health conditions are widespread among individuals who are homeless or experiencing housing instability, and the lack of stable housing can make it challenging for individuals to adhere to psychiatric and other medication regimens.2 Individuals experiencing unstable housing often suffer from multiple chronic health conditions that require prescription medications for proper management, such as cardiovascular disease, diabetes, and hypertension.3 Unfortunately, these individuals face significant barriers to medication adherence.4–6
Permanent Supportive Housing (PSH) links affordable, non-time-limited housing with voluntary support services designed to address the needs of those experiencing unstable housing or who are chronically homeless.7 PSH services vary by state, but typically include access to case management, tenant advocacy, and vocational management, in addition to medical support.8 Policymakers across multiple states have looked to PSH as a potential solution to controlling costs and improving health outcomes for this high-need and high-cost population. Over a dozen states already have some type of supportive housing services benefit in place.9 As each community has a finite number of vouchers to permanently subsidize rent for qualified individuals, not all qualified individuals receive PSH.
Existing research on PSH has focused on its effects on housing stability, overall health care utilization, and health spending.10–14 Limited evidence regarding pharmaceuticals suggests that persons placed in PSH were more likely to receive prescriptions for diabetes medications,15 and more likely to experience increases in prescription drug expenditures.11 However, no studies have evaluated utilization, as measured by proportion with prescription filled, and adherence across multiple medication classes.
This study extends prior research by utilizing Medicaid claims data between 2011 and 2016 to evaluate the impact of PSH on Medicaid beneficiary’s medication utilization and adherence in four commonly used medication classes among this population. Using a difference-in-differences (DID) approach, we compare changes to trends between our PSH cohort and a propensity score–matched cohort of adult Medicaid enrollees experiencing housing instability in Pennsylvania who did not receive PSH.
Methods
This study was reviewed by the University of Pittsburgh Institutional Review Board and deemed exempt because it was a secondary analysis of existing data.
Data
We analyzed Medicaid enrollment and claims data from the Pennsylvania Medicaid program linked to Homeless Management Information Systems (HMIS) records from 54 of 67 Pennsylvania counties between 2011 and 2016. HMIS captures Department of Housing and Urban Development-financed housing services provided to individuals and families who are homeless or at risk of homelessness.16 These data include type of housing service (e.g., PSH, overnight shelters) and service dates (e.g., PSH entry and exit) (see Appendix 1 for a more expansive definition of PSH).
HMIS and Medicaid data were matched by the Pennsylvania Department of Human Services based on social security number and date of birth (linkage described in Appendix 2).17 Our linked Medicaid-HMIS data encompassed the majority of Pennsylvania, but we were unable to obtain housing data for Philadelphia and 12 other counties that use separate HMIS systems. We identified the most frequently used medications among PSH participants in our data. We also conducted a literature search to identify common chronic conditions among the PSH population. We then identified the top four classes of medication for treatment of chronic or relapsing/remitting conditions for which a person started on pharmacotherapy treatment would not be expected to discontinue it during the study period. These four classes were antidepressant, antipsychotic, antiasthmatic, and antidiabetic medications. Sample sizes for other chronic medication classes were not sufficient to support analyses. HEDIS National Drug Code (NDC) lists were used to identify medications in the four classes in pharmacy claims (Appendix Table 6).
PSH Cohort
We identified Pennsylvania Medicaid enrollees who entered PSH between April 1, 2012, and December 31, 2015, and were age 21 or older at the time of entry. Consistent with this program’s goal of providing long-term housing, 96% stayed in PSH for at least 180 days.
To analyze changes in medication utilization and adherence before and after PSH entry, we required individuals to meet minimum cumulative Medicaid enrollment criteria in two periods: (1) for at least 6 months between 7 and 18 months before PSH entry (to establish a baseline period); and (2) at least 6 months in the 12 months immediately following PSH entry. This allowed us to measure time-varying covariates pre-PSH entry and examine changes in medication utilization and adherence before and after PSH entry while recognizing that it is common for individuals in Medicaid to experience gaps in coverage.18
Previous research in this population identified increases in health spending concentrated in behavioral health (e.g., residential treatment) in the 6 months preceding PSH placement.11 This uptick in health system engagement may reflect the need for medical documentation to qualify for PSH. By omitting the 6 months immediately preceding PSH entry in our treatment sample and the 6 months preceding the reference month in our comparison cohort, from both the propensity scoring matching and final analyses, we excluded trends likely related to, but preceding, PSH entry.11 See Appendix Table 4 for more information on study periods. A traditional approach of matching a comparison sample with similar pre-intervention trends would have underestimated differences in medication utilization.
Comparison Cohort
We constructed a comparison cohort of Medicaid enrollees who received other (non-PSH) housing services indicative of episodic or chronic homelessness with similar demographic and health characteristics as PSH recipients. First, we used propensity score matching to identify an initial comparison group of individuals who resembled PSH recipients on time-invariant characteristics including gender, race, and ethnicity, as well as chronic condition diagnoses during the baseline period as measured by the Chronic Illness & Disability Payment System and MedicaidRx. Second, we assigned a reference month (or pseudo index month) for individuals in the comparison cohort based on the PSH recipient to whom they were matched (i.e., if the PSH recipient entered PSH in June 2015, then individual matches in the comparison group were assigned June 2015 as their reference month). This controlled for secular trends in utilization and adherence associated with homelessness but unrelated to PSH, which was observed in previous research.11 We matched each PSH recipient with four comparison participants with the highest propensity scores (Appendices 3 and 4).
Outcome Measures
Our primary outcome was the differential change in the proportion with a filled prescription in the PSH and comparison cohorts between the 12-month pre- and post-PSH periods. We measured this overall (beyond the individual medication classes) and for each of the specific medication classes (antidepressants, antipsychotics, antiasthmatics, and antidiabetics). Participants were required to have ICD-9/10-CM codes related to the medication class (e.g., a diagnosis of depression was required among those with antidepressant prescriptions) in the pre-PSH period (Appendix 5).
Secondary outcomes included the differential change in mean proportion of days covered (PDC) and percent adherent (PDC ≥ 80%) in the PSH and comparison cohorts between the 12-month pre- and post-PSH periods. PDC was calculated based on CDC guidance, taking the total number of days covered by refills in a measurement period and dividing by the number of days between the first fill and the end of the measurement period.19 Percent adherent represents the proportion of patients who meet a PDC threshold of 80% or greater. Both measures were calculated across individual medication classes mentioned above. To calculate secondary outcomes, participants were required to fill at least two prescriptions for the select medication class on different dates of service during the pre- and post PSH entry period, which means a minimum of four prescription fills during the study period (see Appendix 6 for additional description).
Statistical Analysis
We utilized a DID approach for our analyses.20 This allowed for the comparison of changes over time in our outcomes between the PSH and comparison cohorts, while accounting for changes in secular trends and controlling for both measured and unmeasured confounding. Unadjusted differences in outcomes were compared using a paired sample t-test. We used Student’s t tests to compare continuous variables and chi-square tests to compare categorical variables. We used logit models to estimate differential changes in proportions with a filled prescription and percent adherent (PDC ≥ 80%). Linear regression models were used to estimate the differential change in mean PDC. All models included PSH vs. comparison group indicator, pre vs. post PSH indicators, and the interaction of these two variables, along with fixed effects for year, duration of Medicaid enrollment, and months of homeless indicative services at baseline. A sensitivity analysis limited the sample to individuals continuously enrolled in Medicaid during the entire study period.
Results
Demographics
The PSH cohort included 1375 individuals and the comparison cohort consisted of 5405 individuals. Both cohorts were predominantly female, non-Hispanic White, used tobacco, suffered from at least one chronic disease, and resided in an urban area. Behavioral health issues were prevalent: 94.3% of recipients were diagnosed with a mental illness and 64.5% with a substance use disorder. Significant differences between individuals in the PSH versus the comparison cohort after matching included age category, proportion with continuous Medicaid enrollment, and proportion with utilization of homeless indicative services during the baseline period (Table 1).
Table 1.
Patient Characteristics at Baseline
| Characteristic | PSH cohort N = 1375 |
Comparison cohort N = 5405 |
P-value1 |
|---|---|---|---|
| Female, % | 786 (57.2%) | 3084 (57.1%) | 0.94 |
| Age, % in category | |||
| 55 + | 169 (12.3%) | 560 (10.4%) | < 0.01 |
| 45–54 | 362 (26.3%) | 1259 (23.3%) | |
| 35–44 | 320 (23.3%) | 1290 (23.9%) | |
| 25–34 | 405 (29.5%) | 1662 (30.7%) | |
| 21–24 | 119 (8.7%) | 634 (11.7%) | |
| Race, % in category | |||
| Non-Hispanic White | 904 (65.7%) | 3552 (65.7%) | 0.94 |
| Non-Hispanic Black | 410 (29.8%) | 1601 (29.6%) | |
| Hispanic | 38 (2.8%) | 148 (2.7%) | |
| Other | 23 (1.7%) | 104 (1.9%) | |
| Pennsylvania Medicaid managed care contracting region, % in category | |||
| Southwest | 1005 (73.1%) | 3971 (73.5%) | 0.97 |
| Southeast | 17 (1.2%) | 75 (1.4%) | |
| Northwest | 145 (10.5%) | 565 (10.5%) | |
| Northeast | 82 (6.0%) | 301 (5.6%) | |
| Lehigh | 126 (9.2%) | 493 (9.1%) | |
| Resident of Allegheny county, % | 753 (54.8%) | 2958 (54.7%) | 0.98 |
| Resident of urban areas, %2 | 947 (68.9%) | 3718 (68.8%) | 0.95 |
| Eligible for Medicaid through a disability pathway, %4 | 304 (22.1%) | 1216 (22.5%) | 0.76 |
| Continuous Medicaid enrollment, %5 | 1179 (85.8%) | 3117 (57.7%) | < 0.01 |
| Use of homeless indicative services, %6 | 244 (17.8%) | 655 (12.1%) | < 0.01 |
| Diagnoses, %7 | |||
| Mental health diagnosis | 1297 (94.3%) | 5081 (94.0%) | 0.65 |
| Tobacco use disorder | 1140 (82.9%) | 4468 (82.7%) | 0.83 |
| Substance use disorder | 887 (64.5%) | 3475 (64.3%) | 0.88 |
| Gastrointestinal, low | 777 (56.5%) | 3054 (56.5%) | 0.99 |
| Cardiac disease | 766 (55.7%) | 3037 (56.2%) | 0.75 |
| Cardiovascular disease, extra low | 727 (52.9%) | 2867 (53.0%) | 0.91 |
| Infectious, low | 430 (31.3%) | 1657 (30.7%) | 0.66 |
| Hepatitis C virus | 338 (24.6%) | 1301 (24.1%) | 0.69 |
| Diabetes | 223 (16.2%) | 866 (16.0%) | 0.86 |
| Human immunodeficiency virus | 36 (2.6%) | 136 (2.5%) | 0.83 |
Data is from the baseline period (i.e., 7–18 months before PSH entry) and after propensity score matching
PSH Permanent Supportive Housing
1P-value for differences in proportions between the PSH and comparison cohorts. Differences were assessed using chi-square tests for categorical variables
2Rurality was assessed at the county level according to a definition provided by The Center for Rural Pennsylvania
3If an enrollee had ≥ 1 month in the years 2011–2017 in which they were dually enrolled in Medicare and Medicaid, they were considered to have been a dual enrollee Medicare and Medicaid
4If an enrollee had ≥ 1 month in the years 2011–2017 in which they were enrolled in Medicaid through a disability pathway, they were considered to have been enrolled through a disability pathway
5Continuous Medicaid enrollment represents 12 months of coverage between 7 and 18 months before PSH entry
6For example, emergency shelters, transitional housing, day shelters, and other non-shelter homelessness services
7Categories were assessed using the Chronic Illness & Disability Payment System and Medicaid Rx (CDPS-MRX). “Low” and “extra low” refer to levels of severity. Categories listed here are the top 5 CDPS-MRX categories present in the PSH cohort
Medication Utilization
During the study period, the PSH cohort saw an 8.22% rise in the proportion of PSH recipients who filled any prescription (84.75 to 92.95%), compared to a 5.01% rise in the comparison cohort (77.26 to 82.28%), for a relative increase of 4.77% (95% CI 2.87 to 6.67%). No significant difference in proportion with a filled prescription for antidepressants, antipsychotics, antiasthmatics, and antidiabetics was observed (Table 2). Results were similar in DID models adjusting for year, duration of Medicaid enrollment, and months of homeless indicative services at baseline. Results from a sensitivity analysis among participants continuously enrolled in Medicaid were consistent with the primary analysis (Appendix 7).
Table 2.
Changes in Prescription Fills Before and After PSH
| Therapeutic class | PSH cohort (n = 1375) | Difference (95% CI)* | Comparison cohort (n = 5405) | Difference (95% CI)* | Difference in differences (95% CI)** | P-value *** | ||||
|---|---|---|---|---|---|---|---|---|---|---|
| N | Pre | Post | N | Pre | Post | |||||
| Any medication N (%) | 1375 | 1165 (84.73%) | 1278 (92.95%) | 9.64% (7.78%, 11.50%) | 5405 | 4176 (77.26%) | 4447 (82.28%) | 4.87% (3.46%, 6.27%) | 4.77% (2.87%, 6.67%) | < 0.01 |
| Antidepressants N (%) | 640 | 478 (74.69%) | 518 (80.94%) | 6.44% (1.84%, 11.03%) | 1490 | 1102 (73.96%) | 1196 (80.27%) | 6.23% (3.32%, 9.14%) | 0.21% (− 5.32%, 5.73%) | 0.94 |
| Antipsychotics N (%) | 170 | 131 (77.06%) | 133 (78.24%) | 1.26% (− 8.12%, 10.64%) | 351 | 266 (75.78%) | 278 (79.20%) | 3.30% (− 1.65%, 8.25%) | − 2.05% (− 14.31%, 10.21%) | 0.74 |
| Antiasthmatics N (%) | 195 | 136 (69.74%) | 139 (71.28%) | 1.58% (− 4.87%, 10.64%) | 485 | 318 (65.57%) | 350 (72.16%) | 6.54% (2.69%, 10.38%) | − 4.96% (− 12.42%, 2.50%) | 0.19 |
| Antidiabetics N (%) | 138 | 76 (55.07%) | 85 (61.59%) | 6.64% (2.68%, 10.61%) | 463 | 245 (52.92%) | 276 (59.61%) | 6.64% (2.42%, 10.93%) | − 0.03% (− 4.97%, 4.91%) | 0.99 |
PSH Permanent Supportive Housing, N number
*Pre-period is the reference
**The comparison cohort is the reference
***P-value < 0.05 is considered significant; the model incorporated fixed effects for year, duration of Medicaid enrollment, and months of homeless indicative services at baseline
Medication Adherence
During the study period, the percent adherent (PDC ≥ 80%) among individuals with antidepressant use in the PSH cohort rose by 13.93% (18.66 to 32.59%,) compared to a 4.96% increase in the comparison cohort (19.71 to 24.66%), for a relative increase of 7.41% (95% CI 0.26 to 14.57%), although the mean PDC DID estimate of 3.75% (95% CI − 0.10 to 7.59%) was not significantly different. Measurements of mean PDC and percent adherent (PDC ≥ 80%) among other therapeutic medication classes in this population (i.e., antipsychotics, antiasthmatics, antidiabetics) showed no significant difference between the PSH and comparison cohorts (Table 3).
Table 3.
Changes in Percent of Days Covered and Adherence Before and After PSH
| Therapeutic class | PSH cohort N = 1375 |
Difference (95% CI)* | Comparison cohort N = 5405 |
Difference (95% CI)* | Difference in differences (95% CI)** | P-value*** | ||
|---|---|---|---|---|---|---|---|---|
| Pre | Post | Pre | Post | |||||
| Antidepressants | ||||||||
| Sample size (N) | 359 | 359 | 746 | 746 | – | – | ||
| Mean PDC (SD) | 51.57% (26.24%) | 62.47% (25.63%) | 10.91% (7.23%, 14.59%) | 48.89% (27.44%) | 56.05% (25.43%) | 7.16% (4.95%, 9.37%) | 3.75% (− 0.10%, 7.59%) | 0.06 |
| Percent adherent (PDC ≥ 80%) | 18.66% | 32.59% | 12.64% (6.13%, 19.15%) | 19.71% | 24.66% | 5.22% (1.77%, 8.68%) | 7.41% (0.26%, 14.57%) | 0.04 |
| Antipsychotics | ||||||||
| Sample size (N) | 97 | 97 | 186 | 186 | ||||
| Mean PDC (SD) | 62.6% (27.91%) | 69.64% (27.53%) | 7.04% (1.06%, 13.01%) | 55.66% (28.41%) | 61.06% (26.11%) | 5.39% (1.27%, 9.51%) | 1.64% (− 4.27%, 7.56%) | 0.57 |
| Percent adherent (PDC ≥ 80%) | 41.24% | 49.48% | 7.56% (− 0.36%, 15.48%) | 28.49% | 32.26% | 3.95% (− 2.81%, 10.08%) | 3.61% (− 7.44%, 14.66%) | 0.52 |
| Antiasthmatics | ||||||||
| Sample size (N) | 73 | 73 | 167 | 167 | ||||
| Mean PDC (SD) | 39.36% (24.44%) | 48.96% (28.86%) | 9.60% (2.93%, 16.28%) | 36.90% (23.09%) | 46.91% (26.10%) | 10.01% (6.13%, 13.88%) | − 0.40% (− 8.45%, 7.64%) | 0.92 |
| Percent adherent (PDC ≥ 80%) | 5.48% | 20.55% | 14.40% (6.49%, 22.31%) | 6.59% | 14.97% | 8.62% (4.88%, 12.36%) | 5.78% (− 3.18%, 14.74%) | 0.21 |
| Antidiabetics | ||||||||
| Sample size (N) | 59 | 59 | 181 | 181 | ||||
| Mean PDC (SD) | 59.21% (22.77%) | 59.19% (25.08%) | − 0.02% (− 5.53%, 5.50%) | 50.25% (27.64%) | 59.27% (25.84%) | 9.02% (2.73%, 15.30%) | − 9.03% (− 18.41%, 0.34%) | 0.06 |
| Percent adherent (PDC ≥ 80%) | 22.03% | 23.73% | 2.02% (− 12.40%, 16.44%) | 19.89% | 29.83% | 9.35% (− 0.37%, 19.08%) | − 7.33% (− 26.87%, 12.20%) | 0.46 |
PSH Permanent Supportive Housing, PDC percent of days covered, N number, SD standard deviation
*Pre-period is the reference
**The comparison cohort is the reference
***P-value < 0.05 is considered significant; the model incorporated fixed effects for year, duration of Medicaid enrollment, and months of homeless indicative services at baseline
The results of a sensitivity analysis limited to participants continuously enrolled in Medicaid was largely consistent with the primary analysis. However, there was a statistically significant relative increase in mean PDC among individuals with antidepressant use (8.49%, 95% CI 4.87 to 12.11%) in the PSH compared to the comparison cohort (Appendix 7). All analyses adjusted for year, duration of Medicaid enrollment, and months of homeless indicative services at baseline.
Discussion
Among a cohort of adults enrolled in Pennsylvania Medicaid, we found that PSH was associated with a statistically significant increase of 4.77% in the proportion PSH participants filling any prescription medication relative to a comparison cohort matched on observable characteristics. We also found improvements in one of two measures of adherence to antidepressant therapy. Having at least 80% of days covered with antidepressants increased by 7.41% among those receiving PSH relative to a comparison group. These findings were robust to sensitivity analyses.
The PSH cohort also experienced increases in the percent with filled prescriptions for antidepressants; mean PDC for antidepressants, antipsychotics, and antiasthmatics; and percent adherent for antidepressants and antiasthmatics. However, these increases were similar in magnitude to those observed in the comparison cohort, and thus were not significantly different in the DID analyses.
This study fills important gaps regarding the potential impact of PSH on medication therapy among individuals experiencing unstable housing. No previous study on PSH has evaluated the proportion of individuals with a filled prescription, mean change in PDC, or percent adherent across individual classes of medications although previous research reported that PSH was associated with increased prescription drug spending overall.10,11 In New Jersey, pharmacy spending associated with PSH was sensitive to model specifications and results ranged from no difference to a $175.61 (P = 0.002) relative increase in pharmacy spending per person per quarter.10 In Pennsylvania, PSH was associated with a $32.34 (95% CI $1.03 to $64.71) increase in pharmacy spending per person-month relative to a comparison population.11 Our findings extend this prior work to examine patterns of use within specific medication classes commonly used among people receiving PSH. Our findings on increased utilization of prescription drugs are consistent with previous studies reporting increases in pharmacy spending following PSH entry.
Adherence to antidepressants was only observed to be improved in the PSH cohort. However, our results were not consistent in direction and magnitude across all drug classes. While overall we observed a statistically significant relative increase in percentage of participants with a prescription fill, analyses of individual medication classes found no significant differences. This may be due to decreased power due to smaller sample sizes at the class-level. Percent of days covered generally increased across all classes of medication across groups, although we were unable to detect significant differences between the PSH and comparison cohorts. Future studies drawing on data from multiple state Medicaid programs and including other important medication classes (e.g., HCV, HIV) should seek to replicate and extend these findings.
The implications of increased medication utilization and adherence in some classes among PSH participants can be considered a positive outcome from a clinical perspective because increased medication adherence has been associated with improved health outcomes for chronic conditions.21 A common adage is that the best medication is the one your patient will take. Increased housing stability provided through PSH may improve a patient’s chances of filling their prescriptions. Potential mechanisms through which PSH might affect medication adherence include having a safe place to store your medications, a refrigerator for medications requiring cold storage, and access to food for medication that requires a full stomach. Other mechanisms might include the supportive services available to PSH enrollees. Some housing programs offer employment services, legal services, behavioral health specialists, or case managers to help clients navigate access to community services. Utilization of these resources may, in conjunction with stable housing, improve medication adherence.22 The finding that PSH produces a modest improvement in prescription utilization overall, but inconsistent improvements in specific adherence measures, suggests that housing may be a necessary but insufficient condition for improving chronic disease care.
This study has several limitations. First, although this is one of the largest studies on PSH, our sample was small, limited to select counties in a single state, and limited to data through 2016. This potentially reduces generalizability and although there have been no significant changes to Pennsylvania PSH programs since 2016, external factors (e.g., COVID-19) may have influenced program success in recent years. Second, our study may be underpowered to address its secondary objective of assessing refill adherence in specific classes. While we used a validated measure of adherence, the measure’s requirement for multiple medication fills may bias our results away from the null. Third, our comparison group consisted of individuals using non-PSH housing services. We observed differential increases in health care use in the PSH group (largely for behavioral health) in the 6 months immediately preceding PSH placement that were not observed in the comparison group. While we sought to address this issue by removing that 6-month period, we cannot rule out the possibility that some of the relative changes we observe between the PSH and comparison cohorts are explained by factors other than receipt of PSH. Also, we may underestimate PSH effects if individuals in the comparison group later achieved stable housing. Fourth, we lacked data on eligibility and opportunity for PSH among the comparison group. Fifth, medication acquisition from the pharmacy may not capture true rates of treatment because it does not directly measure whether patients are taking the medication or if they are filling them through another insurer. Sixth, Medicaid enrollment gaps may contribute to lower refill adherence compared to continuously enrolled populations. However, based on our sensitivity analysis, enrollment gaps are unlikely to explain differences between the PSH and comparison cohort. Seventh, our study was limited to individuals receiving homelessness services as captured by HMIS and is not representative of individuals who are unstably housed but not receiving services.
As PSH programs continue to grow, in both expenditures and populations served, its importance to policymakers has increased. PSH was associated with higher utilization overall and increased adherence to antidepressant medications—indicating that promoting PSH placements among Medicaid enrollees could enhance medication utilization and adherence.
Supplementary Information
Below is the link to the electronic supplementary material.
Funding
Donald Bourne received funding from the National Institute of General Medical Sciences of the National Institutes of Health (T32GM008208); Clinical and Translational Science Institute, University of Pittsburgh (1 TL1 TR001858); and Evan Cole received funding from the National Institute on Minority Health and Health Disparities grant (R01MD015261). Mara Hollander received funding from the National Institute of Mental Health (T32MH109436-05). Drs. Cole, Donohue, and Xue received funding for related work from an intergovernmental agreement with the Pennsylvania Department of Human Services.
Declarations:
Conflict of Interest:
None.
Footnotes
Publisher's Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
References
- 1.Walsh CA, Cahir C, Tecklenborg S, Byrne C, Culbertson MA, Bennett KE. The association between medication non-adherence and adverse health outcomes in ageing populations: A systematic review and meta-analysis. British journal of clinical pharmacology. 2019;85(11):2464–2478. doi: 10.1111/bcp.14075. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 2.Ayano G, Tesfaw G, Shumet S. The prevalence of schizophrenia and other psychotic disorders among homeless people: a systematic review and meta-analysis. BMC psychiatry. 2019;19(1):370. doi: 10.1186/s12888-019-2361-7. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 3.Arnold EM, Strenth CR, Hedrick LP, et al. Medical comorbidities and medication use among homeless adults seeking mental health treatment. Community mental health journal. 2020:1–9. [DOI] [PubMed]
- 4.Hunter CE, Palepu A, Farrell S, Gogosis E, O’Brien K, Hwang SW. Barriers to prescription medication adherence among homeless and vulnerably housed adults in three Canadian cities. Journal of primary care & community health. 2015;6(3):154–161. doi: 10.1177/2150131914560610. [DOI] [PubMed] [Google Scholar]
- 5.Wilder ME, Kulie P, Jensen C, et al. The impact of social determinants of health on medication adherence: a systematic review and meta-analysis. Journal of general internal medicine. 2021;36:1359–1370. doi: 10.1007/s11606-020-06447-0. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 6.Wilder ME, Zheng Z, Zeger SL, et al. Relationship between social determinants of health and antihypertensive medication adherence in a medicaid cohort. Circulation: Cardiovascular Quality and Outcomes. 2022;15(2):e008150. [DOI] [PMC free article] [PubMed]
- 7.National Academies of Sciences, Engineering, Medicine. Permanent supportive housing: evaluating the evidence for improving health outcomes among people experiencing chronic homelessness. National Academies Press; 2018. [PubMed]
- 8.Dickson-Gomez J, Quinn K, Bendixen A, et al. Identifying variability in permanent supportive housing: A comparative effectiveness approach to measuring health outcomes. Am J Orthopsychiatry. 2017;87(4):414–424. doi: 10.1037/ort0000232. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 9.The Corporation for Supportive Housing. Summary of state Action: Medicaid & Housing Services. 2022.
- 10.DeLia D, Nova J, Chakravarty S, Tiderington E, Kelly T, Cantor JC. Effects of Permanent Supportive Housing on Health Care Utilization and Spending Among New Jersey Medicaid Enrollees Experiencing Homelessness. Medical care. 2021;59:S199–S205. doi: 10.1097/MLR.0000000000001443. [DOI] [PubMed] [Google Scholar]
- 11.Hollander MA, Cole ES, Donohue JM, Roberts ET. Changes in Medicaid Utilization and Spending Associated with Homeless Adults’ Entry into Permanent Supportive Housing. Journal of General Internal Medicine. 2021:1–8. [DOI] [PMC free article] [PubMed]
- 12.Aubry T, Bloch G, Brcic V, et al. Effectiveness of permanent supportive housing and income assistance interventions for homeless individuals in high-income countries: a systematic review. The Lancet Public Health. 2020;5(6):e342–e360. doi: 10.1016/S2468-2667(20)30055-4. [DOI] [PubMed] [Google Scholar]
- 13.Bohnhoff JC, Xue L, Hollander MAG, et al. Healthcare Utilization Among Children Receiving Permanent Supportive Housing. Pediatrics. 2023;151(4). [DOI] [PMC free article] [PubMed]
- 14.Williams JL, Keaton K, Phillips RW, Crossley AR, Glenn JM, Gleason VL. Changes in Health Care Utilization and Associated Costs After Supportive Housing Placement by an Urban Community Mental Health Center. Community Ment Health J. 2023:1–10. [DOI] [PMC free article] [PubMed]
- 15.Lim S, Miller-Archie SA, Singh TP, Wu WY, Walters SC, Gould LH. Supportive housing and its relationship with diabetes diagnosis and management among homeless persons in New York City. American journal of epidemiology. 2019;188(6):1120–1129. doi: 10.1093/aje/kwz057. [DOI] [PubMed] [Google Scholar]
- 16.PA Homeless Management Information System. https://pennsylvaniacoc.org/homeless-management-information-system. Accessed.
- 17.Czajka JL, Verghese S. Social Security Numbers in Medicaid Records: Reporting and Validity, 2009. Mathematica Policy Research;2013.
- 18.Sommers BD, Rosenbaum S. Issues in health reform: how changes in eligibility may move millions back and forth between Medicaid and insurance exchanges. Health affairs. 2011;30(2):228–236. doi: 10.1377/hlthaff.2010.1000. [DOI] [PubMed] [Google Scholar]
- 19.Centers for Disease Control & Prevention. Calculating proportion of days covered (PDC) for antihypertensive and antidiabetic medications: an evaluation guide for grantees. In:2015.
- 20.Wing C, Simon K, Bello-Gomez RA. Designing difference in difference studies: best practices for public health policy research. Annual review of public health. 2018;39. [DOI] [PubMed]
- 21.Marcum ZA, Hanlon JT, Murray MD. Improving medication adherence and health outcomes in older adults: an evidence-based review of randomized controlled trials. Drugs & aging. 2017;34(3):191–201. doi: 10.1007/s40266-016-0433-7. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 22.MacKinnon L, Kerman N, Socías ME, Brar R, Bardwell G. Primary care embedded within permanent supportive housing for people who use substances: a qualitative study examining healthcare access in Vancouver, Canada. Health & Social Care in the Community. 2022;30(6):e5062–e5073. doi: 10.1111/hsc.13921. [DOI] [PMC free article] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.