Partner UCSF Medical Center
Project Welcome Home Pay for Success Housing Initiative
Notes 1 Santa Clara County Pay For Success Project Welcome Home 3-year report November 1, 2018 Maria Raven, MD, MPH; Associate Professor, Dept. of Emergency Medicine, UCSF Margot Kushel, MD; Professor, Div. of General Internal Medicine, UCSF Matthew Niedzwiecki, PhD; Assistant Professor, Dept. of Emergency Medicine, UCSF; Health Researcher, Mathematica Policy Research Sarah Sabbagh, MPH; Health Policy Research Associate, Dept. of Emergency Medicine, UCSF 2 I. INTRODUCTION Santa Clara County is home to the sixth largest population of people experiencing homelessness in the country,i with nearly 7,400 homeless individuals reported in 2017. ii Systemic factors that contribute to high rates of homelessness in Santa Clara County include a high cost of living and a real estate market that renders construction of affordable housing challenging. Homeless individuals, particularly those who are chronically homeless, have multiple health challenges, including chronic diseases, substance use problems, and mental health conditions. This results in frequent use of multiple acute, emergent, and behavioral health services, such as emergency department (ED) visits, hospitalizations, mental health and substance use treatments, and jails. Providing low-barrier permanent housing, or subsidized housing that does not require commitments to sobriety or engagement in care, has been shown to house chronically homeless people effectively. Low-barrier subsidized housing with on-site or closely linked voluntary supportive services (such as case management, medical, mental health, and substance use treatment) known as permanent supportive housing,iii – has become a primary strategy for housing chronically homeless individuals. A growing literature highlights the ability of permanent supportive housing to create cost offsets by housing chronically homeless individuals whose needs result in high costs to the health and social care delivery systems. For example, one study found that supportive housing resulted in reductions in shelter use, hospitalizations, length of hospital stay, and time incarcerated among homeless people with severe mental illness.iv Another study found that the use of permanent supportive housing led to decreases in emergency and inpatient hospital services.v However, this literature has been marred by low quality research, which compares use before and after housing, and doesn’t take into account regressions to the mean and selection bias. Further, access to permanent supportive housing remains limited due to the associated costs and scarcity of available housing units, and more rigorous evaluation of its impact is needed. We are now in the fourth year of Santa Clara County’s Pay for Success project entitled Project Welcome Home (PWH). PWH aims to evaluate the provision of permanent supportive housing for chronically homeless individuals who are frequent users of health and mental health care systems and jails in Santa Clara County. The success outcome for PWH is stable tenancy for individuals placed in permanent supportive housing. We have also undertaken a randomized controlled trial (RCT) that is examining differences in utilization of health services and the criminal justice system, and monitoring the changes in use of longitudinal care (e.g. regular attendance at primary care). Our RCT compares the outcomes for chronically homeless individuals with high utilization of multiple County services (acute medical, mental health, and jail) who receive permanent supportive housing (intervention) to those who are randomized to receive usual care (control). This report documents the analysis of PWH data through the first three years of the project and presents findings on months of continuous stable tenancy, primary care utilization, ED visits, inpatient medical hospital care, utilization of mental health and substance use services, criminal justice system use, and ambulance use. We hypothesize that individuals receiving the PWH 3 intervention of immediate referral to PSH administered by Abode Services will experience a decrease in the utilization of acute, emergent, and criminal justice services over time (e.g. ED visits, inpatient hospitalizations, use of ambulance services, inpatient psychiatric stays, and jail stays), along with an increase in other more sustaining health services associated with improved health and social outcomes (e.g. primary care chronic disease management, outpatient primary care, and outpatient mental health and substance use services). II. METHODS RCT framework: Our RCT employs an intention to treat (ITT) framework, which means that all patients randomly assigned to one of the treatments are analyzed together, regardless of whether or not they completed or received that treatment. In this case, individuals who are randomized to the permanent supportive housing intervention will be retained in the intervention group for our analysis even if they are not able to be located, if they are located but never enter housing, or are located, engaged, and enter housing, but are not retained in housing long-term. We follow individuals randomized to the usual care group administratively using data provided by the County in the Palantir platform in order to track study outcomes in comparison to those in the intervention group. This is the most rigorous method to assess results, because it maintains the original randomization scheme. The findings are not subject to the bias introduced if we “break” randomization. This method tends to offer conservative estimates of effect. Recruitment: Individuals are screened for eligibility by County staff use a screening tool developed for this study that uses administrative data on medical and behavioral health conditions, length of homelessness, and health and social services utilization history. The tool identifies residents of Santa Clara County who are chronically homeless frequent users of acute County services and are predicted to remain frequent users in the future. County staff at County facility/referral points (i.e., Valley Medical Center (VMC), Emergency Psychiatric Services (EPS), Valley Homeless HealthCare Program (VHHP), Custody Health / Re-Entry Center) can identify potential participants in real-time and refer them to trained study staff for further evaluation of eligibility to enroll. After assessing individuals’ eligibility based on the screening tool, trained study staff arranges to meet potential participants in person and verify eligibility. To be eligible for consent and enrollment, an individual must be chronically homeless as determined by the County, live in Santa Clara County, and meet a pre-determined threshold of a combination of emergency department visits, hospital admissions, psychiatric emergency department visits, psychiatric hospitalizations, and county jail days. Individuals are ineligible for consent and enrollment if they are: incarcerated; connected to and choose not to leave another Specialty Mental Health or other intensive case management program (because Abode is considered an intensive case management program); hospitalized and the treating physician(s) plans to discharge them to a skilled nursing facility or inpatient hospice; or unable to give informed consent as determined by inability to complete teach-back (see below). Eligible individuals consented using the teach-back method to assure understanding of the study.vi They are then randomized to either the intervention or control group using a 4 computerized random number generator that sends an automatic email to study staff indicating assignment. Individuals randomized to the intervention are handed-off to Abode for outreach and engagement in the permanent supportive housing intervention program. After the initial consent and randomization, individuals who consent to be in the study are not contacted for the purposes of this evaluation. Individuals randomized to the control group (usual care) may be offered housing via other County-based channels, per usual care. Data sources: Our evaluation team uses the Palantir data platform to extract and analyze data from multiple county data sources. Palantir developed a platform that pulls and matches data from HealthLink (all physical health services within the Santa Clara County system, including ambulatory care, Emergency Department and inpatient hospitalizations), CJIC (County jail systems), HMIS (Santa Clara County homeless information), and both Unicare and DADS, the County mental health and substance use treatment data systems. These data are pulled hourly and accessible through encrypted, password-protected sites accessible only to study staff. Our evaluation team is the only group who has access to CJIC data (for all clients) and data for individuals randomized to the control group. We are working with the County to gather data on housing placements for individuals in the control group as well as individuals randomized to the intervention group who were never placed in housing. At three years, the study sample is large enough sample to examine results at 24 months post enrollment. We do not present results for the period of 12 months post enrollment because this shorter follow-up period will be dominated by the time that it takes to house individuals after they are enrolled. We would not expect to find effects within this short follow-up period. In addition, because behavior change regarding the use of health services can be a long process of engagement and trust-building, we examine results for the period starting 12 months after enrollment for all participants enrolled for at least 24 months (13-24 months post-enrollment). Analysis To evaluate the randomization, we compared the mean characteristics (e.g., demographics and pre-enrollment utilization) of the intervention and control groups using chi-square and t-tests. To evaluate the causal effect assignment to the treatment group using an intent-to-treat framework, we compared the mean outcomes of all individuals assigned to the treatment group to those individuals assigned to the control group. Outcomes included health services use (e.g., emergency department, other hospital services, outpatient behavioral health care), other social services use (shelter and single room occupancy (SRO) housing), and criminal justice encounters (arrests and time spent in county jail). In additional analyses, we used negative binomial regression models to incorporate additional control variables. We chose negative binomial models to account for the fact that the outcome variables were counts of service utilization over a certain time period and were not normally distributed. We compare outcomes at α = 0.95 with two-tailed tests. We analyze two time periods: (1) a 24-month period after enrollment and (2) a 12-month period that began one year after enrollment (enrolled months 13 through 24). By excluding the first year of enrollment, we aimed to isolate medium-term outcomes that may emerge after individuals were housed for several months. 5 We only include individuals who have evidence of remaining in the County or who are alive at the start of our analysis period and who are “present” in the data (no evidence of death or moving out of county) for at least half of the analytic period. We define attrition as not having any utilization in any sector for a 6+ month period that extends through the end of the study period but having no record of death. We assume that people with no contact in County-based services for six months have moved out of county. For those who have died and who have evidence of attrition, we censor their data after their last receipt of service. Specifically, we prorate their utilization for the study period, based on their utilization when they were present in the data. For example, in the 0-24 month analysis, we include only individuals with at least 12 months of the data. For an individual who died or have evidence of leaving the county moved out of county in month 16, we prorate their utilization in the first 15 months and apply it to the final months of the analytic period. Suspected deaths are confirmed with data from Abode Services and County death records. Because the rate of death for participants enrolled in the evaluation was higher than expected in both groups, we also compared pre-enrollment data between individuals who died and those who remain alive to explore whether demographic characteristics or patterns of health and social services utilization differed prior to study enrollment and randomization. III. RESULTS Study Enrollment: Cumulatively by the end of Q12, 499 of the 763 individuals who screened as potentially eligible based on our UCSF triage tool had undergone assignment to study staff and outreach. A total of 372/499 (74.5%) of these individuals were found to be eligible based on further assessment by study staff during the reporting period. By the end of the reporting period, 372 individuals were randomized and enrolled in the study (199 to usual care and 173 to the PWH intervention). The other 264 did not meet enrollment criteria. The primary reasons that led to ineligibility include the client already being housed; preferring to stay with their current case manager rather than enroll in PWH; being too cognitively impaired to give consent; not meeting the chronically homeless definition; requiring a higher level of care; no longer living in the County; passing away; or refusing services. Program Exits/Non-Conforming Referrals: Sixty-five (65) individuals who were randomized to the treatment group and referred to Abode were categorized as non-conforming referrals and were exited from the program or died. These individuals remain in the analysis as members of the intervention group, even though Abode did not offer them services. Reasons for being declared non-conforming include: death between the time of randomization and housing (23); unable to contact after randomization despite significant efforts by Abode (17); declined Abode’s services despite having consented to study enrollment (3); required hospice or nursing facility placement (3); requested to disenroll from the program (3); left county (3); violence towards staff and asked to leave program (2); 6 other reasons, such as long term incarceration (13). Two individuals were eventually eligible and enrolled in the control group. As expected, many of the PWH enrollees did not remain in their first housing placement, or in some cases in subsequent housing placements. Among the clients enrolled in the program the end of the quarter, 66% had been re-housed once; 24% had been re-housed two or three times; and 10% had been rehoused 5-9 times. The program staff estimates that, at any point in time, there are approximately 10% of enrollees who are unhoused and in housing search after initial placement. Deaths and Attrition (See Appendix 1) In the control group, we confirmed that 18 individuals who were not present in any of our datasets for 6 or more months had died based on records from Santa Clara County. An additional 20 had no record of death and we therefore assume they have moved out of the County. In the treatment group, we found that 23 individuals had died based on reporting from Abode and Santa Clara County, whereas only 7 individuals that were not present in the data had no death record, and were presumed to have moved out of the county. We present data examining differences in individuals who died and remain alive (regardless of what group they are assigned to for the evaluation) on page 7 and in Appendix 4. RCT FINDINGS (see Appendix 2) We present data from the randomized control trial, along with narrative explanations. Our main results consist of analyses that censor individuals who died or experienced attrition (moved out of County) (see Appendix 1, Table 1), and include all other participants, as well as regression results that control for small observable differences among the treatment and control groups after randomization. This intention to treat analysis that includes all participants randomized to either treatment or control groups, regardless of whether individuals randomized to the treatment were ever able to be contacted by Abode and/or housed. Baseline data prior to enrollment: Demographics and pre-enrollment utilization is balanced across treatment and control groups, as would be expected in a randomized controlled trial. As expected, with many variables, they differed in one variable. Individuals randomized to the treatment group were 14.9 percentage points (p<0.01) less likely to have reported a usual source of care (doctor or nurse practitioner) in the two years prior to enrollment. (Appendix 2, Table 1) Twenty-four months post enrollment with censoring for death and attrition (n=122 usual care, n=108 treatment) Table 2 in Appendix 2 displays data for the 24-month period post enrollment for all individuals enrolled in the evaluation, censoring those who attrited or died. We see statistically significant 7 decreases in the number of psychiatric ED visits (4.9 control vs. 2.9 treatment, p=0.03) and the number of ED visits that occur via police transport (1.4 control vs. 0.7 treatment, p=0.01). We continue to see non-statistically significant reductions in ED visits (15.7 control vs. 13.1 treatment), inpatient hospital visits (3.9 control vs. 3.1 treatment) and shelter days (32.4 control vs. 18.2 treatment) comparing control and treatment groups. We also see statistically significant differences in outpatient mental health services use, with a greater number of visits in the treatment group across multiple categories (43.6 control vs. 83.8 treatment, p<0.01 for any mental health visits). Differences in outpatient substance use treatment are not statistically different between the two groups. 13-24 months post enrollment with censoring for death and attrition (n=116 usual care, n=100 treatment) Table 3 in Appendix 2 displays data for the 12-month period starting one year after a participant was enrolled, censoring for those who attrited or died. We look at these data because based on existing literature and our discussions with Abode’s direct services staff, it is clear that it can take several months for individuals to become stably housed, and then begin to shift health-seeking behaviors. The time frame one year after enrollment allows us to examine for differences after this first year has occurred. Here, we see statistically significant differences in psychiatric ED visits (2.3 control vs. 1.1 treatment, p=0.01) and shelter days (13.6 control vs. 3.5 treatment, p=0.01), and statistically significant differences in police transport to the ED (0.7 control vs. 0.3 treatment, p=0.01). We continue to see non-statistically significant trends towards fewer ED visits and inpatient hospital admissions. We continue to see more use of mental health rehabilitation services and (to a lesser degree), and fewer substance use services visits in the treatment group but these results are not statistically significant. NEGATIVE BINOMIAL REGRESSION MODELS (see Appendix 3) Appendix 3, Table 1: 24 months post-enrollment Table 1 shows results of our model at 24 months, for individuals enrolled in the evaluation for at least 24 months’ time. This model includes all individuals enrolled in the evaluation regardless of their receipt of housing. Treatment group findings (Table 1, Column 1) In our regression model, we find that overall, for those who had been enrolled in the evaluation for at least 2 years, at 24 months those in the treatment group had statistically significantly lower rates of shelter use (IRR 0.63; 95% CI [0.39, 1.00] for shelter stays; IRR 0.55; 95% CI [0.35, 0.87] for shelter days – p<0.05 for both) and statistically significantly higher rates of outpatient mental health services use (IRR 2.60; 95% CI [1.77, 3.80] – p<0.01). In addition, those in the treatment group were more likely--at the 0.10 level of significance-- to have lower rates of nonpsychiatric hospital admissions (IRR 0.70; 95% CI [0.46,1.06]), lower rates of combined medical and psychiatric hospital admissions (IRR 0.67; 95% CI [0.43,1.04]), and lower rates of ED visits that result in hospital admissions (IRR 0.68; 95% CI [0.44,1.06]). 8 Other findings (Table 1, additional columns) The remainder of Table 1 examines the rate of outcomes for specific participant characteristics, regardless of their assignment to the treatment or control group. These findings may be helpful to guide staff and programs that interface with similar populations by pointing to specific demographic groups/characteristics associated health and social services use. As an example, men enrolled in the evaluation (regardless of assignment to the treatment or control group) had statistically significantly lower rates of psychiatric hospital admissions (IRR 0.11; 95% CI [0.02,0.52], p<0.01), higher rates of non-psychiatric hospital admissions (IRR 1.58; 95% CI [1.05,2.37], p<0.05), and higher rates of arrests (IRR 1.76; 95% CI [1.07,2.89], p<0.05). Those enrolled in the evaluation who identify as Latinx have statistically significantly lower rates of total hospital admissions (IRR 0.57; 95% CI [0.33,0.98], p<0.05) and ED visits (IRR 0.70; 95% CI [0.50,0.96], p<0.05). Appendix 3, Table 2: 13-24 months post-enrollment Table 2 shows results of our model from 13-24 months post-enrollment, for individuals enrolled in the evaluation for at least 24 months. As mentioned previously, we are exploring the period 1-year after enrollment because behavior change regarding the use of health services can be a long process of engagement and trust-building: being placed in a stable housing environment and accompanying behavior change may not occur until several months to a year after enrollment. Treatment group findings (Table 1, Column 1) Examining the 12-month time period starting one year after enrollment, we find that overall, those in the treatment group had statistically significantly lower rates of police transport to the ED (IRR 0.43; 95% CI [0.24, 0.77], p<0.01), shelter use (IRR 0.11; 95% CI [0.05, 0.24] for shelter stays; IRR 0.10; 95% CI [0.05, 0.23] for shelter days, p<0.01 for both), and substance use treatment visits (IRR 0.38; 95% CI [0.18, 0.84], p<0.05), and higher rates of outpatient mental health services use (IRR 1.91; 95% CI [1.20,3.04], p<0.01). Again, while not statistically significant at the 0.05 level, we see, at the 0.10 level of significance, lower rates of combined inpatient and psychiatric hospital admissions (IRR 0.64; 95% CI [0.38,1.08]). DEATH DATA (see Appendix 4) We compared the pre-enrollment demographics and utilization data between participants who died and participants who are still alive regardless of the group to which they were randomized. (Appendix 4, Table 1). Individuals who died were more likely to be male (82.9% died vs. 71.1% alive, p=0.11), older (54.5 years died vs. 50.2 years alive, p=0.02), have inpatient admissions (4.6 died vs. 2.4 alive, p<0.01), have ED visits delivered by ambulance (8.2 died vs. 5.6 alive, p=0.08), and DADS (substance use services) crisis visits (0.3 died vs. 0.1 alive, p=0.03). Those who died were less likely to have police transports to the ED (0.4 died vs. 1.5 alive, p=0.05). We plan to explore these findings in more depth (see Future Analyses, below). 9 IV. DISCUSSION The Santa Clara County Pay for Success Project Welcome Home evaluation has entered its fourth year. Our evaluation is unique in that it focuses on a subset of chronically homeless individuals who are the most frequent users of the county’s services. It also presents an opportunity to advance the research about the effectiveness of permanent supportive housing (PSH), and by doing so, inform future initiatives to address chronic homelessness in Santa Clara County and elsewhere. The study will also fit in a growing body of literature that uses randomized controlled trials to examine outcomes of the provision of permanent supportive housing services.vii,viii,ix,x The findings from this study could be used to inform policy and programmatic decisions for all governmental and non-governmental entities that provide services for the homeless. This 3-year report, which summarizes our results to date, is encouraging. Ours is the first RCT we are aware of to focus on the impact of permanent supportive housing for frequent users of acute county health and social services. We are encouraged by our findings to date. We selected individuals for participation in Project Welcome Home because they were the most complicated, with multiple indications of mental health, substance use and physical health problems, as well as prolonged homelessness. Their success at housing and our consistent finding of statistically significant reductions in use of shelter stays contradicts the narrative that individuals prefer homelessness to housing. Our study has a relatively small sample size, which limits our power to detect subtle differences in our outcomes between participants who are randomized to the treatment and control groups. However, we are seeing multiple statistically significant differences between these groups in areas of importance for policy makers and providers including decreased use of shelter and decreases in police transport to the ED. We also see multiple other encouraging findings that are significant at the 0.1 level. The low point estimates (rate reductions at or above 30%) for hospital admissions and ED visits, suggest a potentially high level of clinical significance. While not all our findings are not statistically significant at the 0.05 level, their significance at the 0.1 level gives us makes it likely that these lower rates of hospital admissions are true: there is a 90% chance we can reject the null hypothesis. As more years of data are collected, our estimates should become more precise and also speak to longer-term effects of housing on the utilization of important health care and social services. It should be noted that the higher rates of outpatient mental health services use in the treatment group is likely due to the high degree of interaction with Abode, as Abode bills for their services using outpatient mental health codes. We recognize that our estimates are conservative estimates of the true effects of providing PSH on health service utilization. Because this is an intention-to-treat framework, we know that many individuals assigned to the treatment group did not receive the treatment (permanent supportive housing), while many individuals assigned to the control group have been placed in housing through programs offered by the county and other mechanisms (see future analyses section below). This means that there is some “misclassification” – some individuals assigned 10 (and analyzed) to the PSH group were not housed while others assigned (and analyzed) to “usual care” received PSH. Misclassification biases results towards the null, meaning that it makes it more difficult to see differences between groups. Due to this, the results should be considered to be underestimates of the true effect of housing. Over the past three years, there has been a large increase in the number of permanent supportive housing units in Santa Clara County. According to the Santa Clara County Office of Supportive Housing the number of units either available or under construction increased from 1900 in 2015 to 3577 as of August 2018. As a result, we anticipate that more individuals assigned to the control group will receive housing. This may decrease our ability to detect differences as the study continues, and it may be that we are in or near the peak time in our evaluation to detect differences. We will continue to examine the impact of this intervention in the coming months and years and have plans to add on to our analysis (outlined below) that will bring additional breadth and depth to our knowledge of the impact of providing permanent supportive housing to this vulnerable population. We were able to confirm more deaths in the treatment group, whereas more individuals in the control group are presumed to have moved out of the county. It may be that more individuals in the control group chose to leave the county due to an inability to find housing compared to those who were randomized to the treatment group and direct referral to Abode Services. It is also possible that there is a lag in death reporting to the County, and that we will find in the coming months that some of the individuals we assume moved out of county died. V. FUTURE ANALYSES Impact of housing on those who received it: We are working with the County on two additional data elements. First, we are working to obtain complete data from the County regarding housing placements among individuals randomized to the control group and for individuals randomized to the intervention group who were never housed by Abode. These data will need to be reviewed for completeness and quality. Assuming the data are valid, we will estimate the average treatment effect on the treated using instrumental variables regression to better account for the impact of the Abode PSH “treatment” on those who actually received it. The average treatment effect on the treated (ATET) estimate takes into account the fact that adherence with treatment assignment was not perfect, i.e. some assigned to housing were not housed and some assigned to the control group were housed. We will also report descriptive statistics on housing entry and exit dates, and the number of moves for each group as a measure of housing stability that can be examined for both treatment and control groups. Participant deaths: Individuals enrolled in Project Welcome Home had higher rates of death than we anticipated. We are exploring this further by obtaining death certificates from the County for all individuals enrolled in the study who have died. We will review these data, and examine their utilization after enrollment and up to the time of death to uncover possible predictors for death and to better understand why the death rate in this population is higher than expected. We will also determine whether we must undertake chart review (which would require an IRB amendment) to better understand this phenomenon. 11 References Cited: i The U.S. Department of Housing and Urban Development. The 2017 Annual Homeless Assessment Report (AHAR) to Congress: Point-in-Time Estimates of Homelessness. ii Applied Survey Research. 2017 Santa Clara County Homeless Census & Survey Comprehensive Report. iii National Health Care for the Homeless Council. Permanent Supportive Housing. Available at: https://www.nhchc.org/policy-advocacy/issue/permanent-supportive-housing/. Accessed August 9, 2018. iv Culhane D, Metrauex S, Hadley T. Public Service Reductions Associated with Placement of Homeless Persons with Severe Mental Illness in Supportive Housing. Housing Policy Debate. 2002;13(1):107-163. v Sadowski LS, Kee RA, VanderWeele TJ, Buchanan D. Effect of a Housing and Case Management Program on Emergency Department Visits and Hospitalizations among Chronically Ill Homeless Adults: A Randomized Trial. JAMA. 2009;301(17):1771-1778. vi Sudore RL, Landefeld CS, Williams BA, Barnes BE, Lindquist K, Schillinger D. Use of a Modified Informed Consent Process Among Vulnerable Patients: A Descriptive Study. J Gen Intern Med. 2006;21(8):867-73. vii Rog DJ, Marshall T, Dougherty RH, George P, Daniels AS, Shoma Ghose S, Delphin-Rittmon ME. Permanent Supportive Housing: Assessing the Evidence. Psychiatr Serv. 2014;65(3):287-294. viii Stergiopoulos V, Hwang SW, Gozdzik A, Nisenbaum R, Latimer E, Rabouin D, et al. Effect of Scattered-Site Housing Using Rent Supplements and Intensive Case Management on Housing Stability Among Homeless Adults With Mental Illness: A Randomized Trial. JAMA. 2015;313(9):905-915. ix Sadowski LS, Kee RA, Vander Weele TJ, Buchanan D. Effect of a Housing and Case Management Program on Emergency Department Visits and Hospitalizations Among Chronically Ill Homeless Adults: A Randomized Trial. JAMA. 2009;301(17):1771-1778. x Sanbonmatsu L, Ludwig J, Katz LF, Gennetian LA, Duncan GJ, Kessler RC, et al. Moving to Opportunity for Fair Housing Demonstration Program -- Final Impacts Evaluation. U.S. Department of Housing and Urban Development, Office of Policy Development and Research. 2011. Available at: https://www.huduser.gov/publications/pdf/mtofhd_fullreport_v2.pdf. Accessed August 7, 2018. 1 Appendix 1: Deaths and Attrition Deaths: Based on the data contained within our information systems (including Abode and County health systems) we have identified 41 deaths (23 in the treatment and 18 in the control group) prior to the end of the reporting period on 06/30/18. The 23 deaths in the treatment group were reported by Abode, and the control group deaths were verified with the help of the County. To determine county deaths, we submitted names of individuals who had not had any type of encounter in any of our data sources for at least six months and these individuals were checked against the county death records. We censored individuals from analyses at the date of confirmed death. The higher rate of death in the treatment group may be due to timelier death records reported by Abode. Given the rate of other attrition in the control group (see below) we expect that over time, these numbers will become more balanced. Other attrition: Secondly, we examine usage patterns to assess for evidence that an individual was living in Santa Clara County and alive. If an individual had no evidence of use of services in any of our data systems for >180 days and is not known to have died, we assume that the individual has moved or may have died, but that death may have not yet been documented. Table 1: Cumulative deaths and other attrition in treatment and control groups Evaluation Group Control Treatment Death* Attrition# TOTAL Death* Attrition# TOTAL 18 20 38 23 7 30 * Based on record of a participant’s death based on County death records or reporting from Abode # No encounter in any segment of UCSF dataset in the last 6 months of the analysis period 2 Appendix 2: Demographic and Utilization Data Table 1: Demographic and Utilization Data for the Two Years Pre-Enrollment: “Balance Check” Control Mean (n=199) Treatment Mean (n=173) Difference p-value Characteristics Male 72.6% 72.3% 0.3% 0.94 Hispanic 26.3% 25.2% 1.2% 0.80 Black 15.6% 12.7% 2.9% 0.43 Other Race 51.3% 54.9% -3.7% 0.48 Age 50.3 51.1 -0.9 0.45 Smoker 66.8% 65.3% 1.5% 0.76 Medi-Cal Coverage 79.4% 80.3% -0.9% 0.82 Medicare Coverage 17.1% 16.2% 0.9% 0.82 Primary Care Regular Source of Care 81.4% 66.5% 14.9% 0.00 Primary Care Office Visits 8.8 7.4 1.4 0.25 Hospital Use Inpatient Total Inpatient Visits 5.2 5.3 -0.1 0.89 Inpatient Psychiatric Admissions 0.3 0.2 0.1 0.18 Inpatient Admissions 2.6 2.7 -0.03 0.94 Number Of Inpatient Days 16.0 15.6 0.5 0.88 Emergency Department (ED) Total ED Visits 21.4 19.2 2.2 0.27 ED Treat-and-Release Visits 19.2 16.8 2.4 0.22 Admitted ED Visits 2.2 2.4 -0.2 0.62 Psychiatric ED Visits 5.8 5.2 0.6 0.52 ED Visits with Ambulance Transport 6.3 5.3 1.1 0.26 ED Visits with Police Transport 1.3 1.5 -0.1 0.67 Criminal Justice Data Arrests 2.8 3.8 -1.0 0.07 Sentence Days 52.1 55.9 -3.8 0.71 Shelter Data Shelter Stays 41.2 35.9 5.3 0.47 Shelter Days 46.2 43.5 2.8 0.75 Unicare 3 Control Mean (n=199) Treatment Mean (n=173) Difference p-value Any Substance Use Service (DADS) Visits 5.8 7.4 -1.6 0.28 DADS, Intake and Individual Treatment Visits 3.0 3.0 0.0 0.99 DADS Treatment Planning Visits 2.7 2.8 -0.1 0.88 DADS Crisis Intervention 0.2 0.1 0.1 0.15 DADS Group Visits 1.6 3.2 -1.6 0.01 DADS Medication Visits 0.4 0.2 0.2 0.40 Any Mental Health (MH) Visits 30.1 28.5 1.6 0.81 MH Management Visits 16.3 15.0 1.2 0.73 MH Assessment Visits 1.6 1.4 0.2 0.49 MH Testing Visits 0.2 0.2 0.0 0.85 MH Individual Treatment 0.5 0.4 0.1 0.75 MH Group Treatment 1.8 0.4 1.4 0.30 MH Rehab Visits 3.5 3.8 -0.3 0.82 MH Medication (MD) 2.5 2.4 0.1 0.88 MH Medication (Non-MD) 4.0 5.1 -1.1 0.56 4 Table 2: Main Analysis: Demographic and County Services Utilization 24 Months Post-Enrollment Control Mean (n=122) Treatment Mean (n=108) Difference p-value Sample Characteristics Days Enrolled 943.5 948.9 -5.3 0.71 Months Enrolled 31.0 31.2 -0.2 0.68 Male 71.9% 72.2% -0.3% 0.96 Hispanic 33.1% 26.2% 6.8% 0.27 Black 13.1% 13.0% 0.2% 0.97 Other Race 43.4% 48.1% -4.7% 0.48 Age 49.0 50.8 -1.8 0.21 Smoker 68.0% 65.7% 2.3% 0.71 Medi-Cal Coverage 82.0% 80.6% 1.4% 0.79 Medicare Coverage 13.9% 16.7% -2.7% 0.57 Primary Care Regular Source of Care 80.3% 76.9% 3.5% 0.52 Primary Care Office Visits 11.8 9.3 2.5 0.23 Hospital Use Inpatient Total Inpatient Visits 3.9 3.1 0.8 0.37 Inpatient Psychiatric Admissions 0.07 0.05 0.02 0.59 Inpatient Admissions 2.1 1.7 0.4 0.36 Number of Inpatient Days 10.5 11.3 -0.9 0.77 Emergency Department (ED) Total ED Visits 15.7 13.1 2.6 0.23 ED Treat-and-Release Visits 13.9 11.6 2.3 0.27 Admitted ED Visits 1.8 1.4 0.3 0.42 Psychiatric ED Visits 4.9 2.9 2.0 0.03 ED Visits with Ambulance Transport 5.7 5.5 0.2 0.86 ED Visits with Police Transport 1.4 0.7 0.7 0.01 Criminal Justice Data Arrests 2.6 2.6 0.0 0.96 Sentence Days 64.6 63.8 0.8 0.96 Shelter Data Shelter Stays 29.9 17.9 11.9 0.10 Shelter Days 32.4 18.2 14.2 0.06 Unicare Any Substance Use Service(DADS) Visits 5.4 3.5 1.8 0.19 5 Control Mean (n=122) Treatment Mean (n=108) Difference p-value DADS, Intake and Individual Treatment Visits 2.1 1.6 0.5 0.47 DADS Treatment Planning Visits 1.7 1.3 0.4 0.56 DADS Crisis Intervention 0.04 0.01 0.03 0.20 DADS Group Visits 2.4 1.5 0.9 0.26 DADS Medication Visits 0.2 0.1 0.1 0.34 Any Mental Health (MH) Visits 43.6 83.8 -40.2 0.00 MH Management Visits 23.1 15.7 7.4 0.10 MH Assessment Visits 1.8 2.6 -0.8 0.03 MH Testing Visits 0.7 1.5 -0.8 0.00 MH Individual Treatment 1.8 1.3 0.5 0.48 MH Group Treatment 0.1 0.0 0.1 0.22 MH Rehab Visits 7.6 58.1 -50.5 0.00 MH Medication (MD) 3.4 3.3 0.1 0.84 MH Medication (Non-MD) 5.0 1.3 3.7 0.01 6 Table 3: Main Analysis: Demographic and County Services Utilization 13-24 Months Post-Enrollment Control Mean (n=116) Treatment Mean (n=100) Difference p-value Sample Characteristics Days Enrolled 944.7 953.2 -8.5 0.56 Months Enrolled 31.0 31.3 -0.3 0.54 Male 71.6% 71.0% 0.6% 0.93 Hispanic 33.0% 27.4% 5.7% 0.38 Black 12.9% 13.0% -0.1% 0.99 Other Race 44.8% 46.0% -1.2% 0.86 Age 49.2 51.0 -1.8 0.22 Smoker 67.2% 65.0% 2.2% 0.73 Medi-Cal Coverage 81.0% 81.0% 0.0% 0.99 Medicare Coverage 14.7% 16.0% -1.3% 0.79 Primary Care Regular Source of Care 67.2% 68.0% -0.8% 0.91 Primary Care Office Visits 5.0 4.7 0.3 0.79 Hospital Use Inpatient Total Inpatient Visits 1.8 1.2 0.6 0.14 Inpatient Psychiatric Admissions 0.03 0.02 0.01 0.83 Inpatient Admissions 0.9 0.6 0.3 0.15 Number of Inpatient Days 4.8 4.4 0.4 0.80 Emergency Department (ED) Total ED Visits 7.1 5.7 1.4 0.27 ED Treat-and-Release Visits 6.3 5.2 1.1 0.36 Admitted ED Visits 0.8 0.5 0.3 0.15 Psychiatric ED Visits 2.3 1.1 1.2 0.01 ED Visits with Ambulance Transport 2.7 2.5 0.2 0.82 ED Visits with Police Transport 0.7 0.3 0.4 0.01 Criminal Justice Data Arrests 1.3 1.0 0.3 0.35 Sentence Days 32.3 32.0 0.3 0.97 Shelter Data Shelter Stays 13.1 3.5 9.5 0.02 Shelter Days 13.6 3.5 10.1 0.01 Unicare Any Substance Use Service(DADS) Visits 2.5 1.7 0.9 0.33 DADS, Intake and Individual Treatment Visits 0.9 0.8 0.1 0.86 7 Control Mean (n=116) Treatment Mean (n=100) Difference p-value DADS Treatment Planning Visits 0.7 0.7 0.0 0.95 DADS Crisis Intervention 0.01 0.00 0.01 0.35 DADS Group Visits 1.2 0.7 0.5 0.34 DADS Medication Visits 0.2 0.04 0.16 0.21 Any Mental Health (MH) Visits 22.8 31.2 -8.4 0.12 MH Management Visits 11.9 3.4 8.5 0.00 MH Assessment Visits 0.9 0.7 0.3 0.25 MH Testing Visits 0.3 0.5 -0.2 0.09 MH Individual Treatment 1.0 0.6 0.4 0.47 MH Group Treatment 0.0 0.0 0.0 MH Rehab Visits 4.1 24.1 -20.0 0.00 MH Medication (MD) 1.8 1.3 0.4 0.28 MH Medication (Non-MD) 2.7 0.6 2.1 0.02 8 Appendix 3: Negative Binomial Regression Models Table 1: Negative binomial regression model: 24 months post-enrollment Treatment Group Male Hispanic Black Other Race Age Age squared Smoker Medi-Cal Medicare N Inpt psych adm 0.9 [0.31,2.62] 0.11** [0.02,0.52] 1.61 [0.44,5.90] 0.46 [0.06,3.33] 1.38 [0.35,5.43] 0.89 [0.68,1.18] 1 [1.00,1.00] 0.83 [0.22,3.05] 0.64 [0.08,5.01] 5.09 [0.55,47.41] 245 Inpt med adm 0.70+ [0.46,1.06] 1.58* [1.05,2.37] 0.61+ [0.36,1.04] 1.96* [1.02,3.73] 1.46 [0.92,2.33] 1.06 [0.95,1.18] 1 [1.00,1.00] 0.67 [0.41,1.09] 1.64 [0.53,5.15] 1.19 [0.34,4.11] 245 Total adm (med+psych) 0.67+ [0.43,1.04] 1.47+ [0.97,2.24] 0.57* [0.33,0.98] 1.89+ [0.98,3.65] 1.50+ [0.94,2.39] 1.06 [0.94,1.18] 1 [1.00,1.00] 0.65+ [0.39,1.08] 1.47 [0.49,4.43] 1.07 [0.32,3.56] 245 Inpatient LOS 0.71 [0.42,1.19] 0.98 [0.55,1.76] 0.65 [0.35,1.23] 1.81 [0.74,4.43] 1.21 [0.71,2.06] 1.05 [0.90,1.21] 1 [1.00,1.00] 0.69 [0.37,1.28] 4.17* [1.29,13.44] 2.77 [0.77,9.91] 245 Total ED 0.9 [0.65,1.24] 1.18 [0.86,1.63] 0.70* [0.50,0.96] 1.34 [0.86,2.06] 1.71** [1.21,2.43] 0.99 [0.92,1.06] 1 [1.00,1.00] 0.9 [0.60,1.37] 0.79 [0.30,2.04] 0.95 [0.35,2.54] 245 ED T&R 0.93 [0.65,1.32] 1.15 [0.81,1.63] 0.70* [0.50,0.99] 1.3 [0.79,2.12] 1.74** [1.19,2.57] 0.98 [0.91,1.06] 1 [1.00,1.00] 0.92 [0.59,1.43] 0.73 [0.28,1.91] 0.92 [0.34,2.49] 245 ED admit 0.68+ [0.44,1.06] 1.65* [1.06,2.56] 0.53* [0.31,0.93] 1.82+ [0.92,3.57] 1.54+ [0.94,2.51] 1.07 [0.95,1.21] 1 [1.00,1.00] 0.66 [0.39,1.12] 1.34 [0.43,4.22] 0.98 [0.28,3.40] 245 Psych ED 0.96 [0.60,1.52] 1.01 [0.59,1.74] 0.95 [0.53,1.66] 0.8 [0.37,1.74] 1.39 [0.82,2.38] 1 [0.89,1.12] 1 [1.00,1.00] 1.69+ [0.94,3.03] 0.59 [0.21,1.61] 1.27 [0.43,3.77] 245 Ambulance transport 0.85 [0.54,1.32] 0.91 [0.56,1.47] 0.61+ [0.36,1.04] 1.35 [0.75,2.44] 2.01** [1.21,3.34] 1 [0.88,1.13] 1 [1.00,1.00] 0.99 [0.60,1.64] 0.68 [0.20,2.38] 0.78 [0.21,2.94] 245 Police transport 0.66 [0.39,1.14] 1 [0.60,1.66] 0.85 [0.51,1.42] 1.44 [0.67,3.11] 1.26 [0.73,2.15] 0.87 [0.74,1.03] 1 [1.00,1.00] 1.61 [0.85,3.07] 0.93 [0.38,2.32] 2.25 [0.75,6.72] 245 Arrests 1.2 [0.80,1.76] 1.76* [1.07,2.89] 1 [0.61,1.65] 1.08 [0.64,1.80] 1.17 [0.75,1.84] 0.94 [0.85,1.04] 1 [1.00,1.00] 2.09** [1.31,3.33] 0.87 [0.46,1.65] 1.06 [0.48,2.32] 245 Sentence days 1.12 [0.64,1.96] 2.50* [1.15,5.44] 1.22 [0.57,2.62] 0.55 [0.24,1.27] 0.52* [0.28,1.00] 1.05 [0.94,1.17] 1.00+ [1.00,1.00] 1.7 [0.82,3.51] 0.66 [0.21,2.04] 0.89 [0.22,3.54] 245 Shelter stays 0.63* [0.39,1.00] 1.3 [0.76,2.20] 1.29 [0.71,2.32] 0.66 [0.30,1.46] 1.05 [0.62,1.77] 1.14 [0.97,1.33] 1 [1.00,1.00] 1.22 [0.70,2.12] 3.69** [1.79,7.61] 2.24+ [0.93,5.40] 245 Shelter days 0.55* [0.35,0.87] 1.39 [0.83,2.32] 1.19 [0.67,2.12] 0.67 [0.31,1.43] 1.19 [0.71,1.99] 1.11 [0.94,1.30] 1 [1.00,1.00] 1.22 [0.71,2.11] 3.63** [1.74,7.59] 2.06 [0.85,5.03] 245 Substance use treatment visits 0.68 [0.35,1.32] 3.20** [1.52,6.73] 0.57 [0.28,1.16] 0.30* [0.11,0.85] 0.49+ [0.23,1.01] 1.40** [1.18,1.65] 1.00** [0.99,1.00] 1.61 [0.83,3.11] 5.20* [1.20,22.63] 2.14 [0.35,12.94] 245 Mental health treatment visits 2.59** [1.77,3.79] 0.66* [0.44,1.00] 0.91 [0.59,1.37] 0.94 [0.55,1.61] 1.03 [0.69,1.55] 1.07 [0.94,1.21] 1 [1.00,1.00] 1.21 [0.83,1.77] 2.19 [0.81,5.90] 2.91+ [0.97,8.79] 245 Abbreviations: outpt = outpatient; inpt = inpatient; psych = psychiatric; med=medical; adm = admission; LOS = length of stay; T&R = treat and release; admit = admitted; ED = emergency department P-values: +p<0.10; *p<0.05; p<**0.01 9 Table 2: Negative binomial regression model: 13-24 months post-enrollment Treatment Group Male Hispanic Black Other Race Age Age squared Smoker Medi-Cal Medicare N Inpt psych adm 5.5 [0.11,277.58] 0.00** [0.00,0.00] 10.57 [0.33,336.44] 8.82 [0.62,125.05] 7.22 [0.49,105.70] 2.4 [0.52,11.13] 0.99 [0.97,1.01] 0.98 [0.03,37.14] 0.03 [0.00,9.30] 0.4 [0.01,28.60] 220 Inpt med adm 0.67 [0.40,1.10] 1.27 [0.77,2.09] 0.84 [0.48,1.45] 2.74* [1.18,6.35] 1.19 [0.70,2.03] 1.15+ [1.00,1.33] 1 [1.00,1.00] 0.79 [0.46,1.35] 1.52 [0.36,6.33] 1.16 [0.26,5.23] 220 Total adm (med+psych) 0.64+ [0.38,1.08] 1.19 [0.69,2.03] 0.74 [0.41,1.31] 2.87* [1.18,6.96] 1.32 [0.75,2.33] 1.15+ [1.00,1.33] 1 [1.00,1.00] 0.7 [0.39,1.25] 1.01 [0.22,4.52] 0.69 [0.14,3.40] 220 Inpatient LOS 0.78 [0.41,1.47] 0.65 [0.30,1.40] 1.08 [0.48,2.40] 3.84* [1.10,13.47] 1.44 [0.72,2.87] 1.19+ [0.98,1.45] 1 [1.00,1.00] 0.86 [0.40,1.87] 0.91 [0.09,9.21] 0.46 [0.04,4.97] 220 Total ED 0.82 [0.56,1.19] 1.15 [0.80,1.65] 0.71+ [0.48,1.05] 1.86* [1.04,3.35] 1.76** [1.18,2.62] 1.01 [0.91,1.12] 1 [1.00,1.00] 1.12 [0.70,1.78] 0.89 [0.35,2.29] 1.03 [0.39,2.76] 220 ED T&R 0.86 [0.57,1.29] 1.1 [0.75,1.61] 0.69+ [0.46,1.03] 1.7 [0.90,3.21] 1.84** [1.20,2.83] 1.01 [0.91,1.11] 1 [1.00,1.00] 1.16 [0.70,1.93] 0.78 [0.29,2.06] 0.97 [0.35,2.65] 220 ED admit 0.65 [0.38,1.12] 1.43 [0.83,2.47] 0.68 [0.37,1.26] 2.80* [1.14,6.84] 1.29 [0.73,2.29] 1.18+ [0.99,1.40] 1 [1.00,1.00] 0.74 [0.40,1.36] 1.2 [0.28,5.03] 0.88 [0.19,3.99] 220 Psych ED 0.70 [0.42,1.19] 1.16 [0.65,2.06] 1.06 [0.53,2.13] 1.01 [0.39,2.58] 0.98 [0.53,1.82] 1.03 [0.90,1.18] 1 [1.00,1.00] 2.10+ [0.99,4.46] 0.87 [0.26,2.90] 1.44 [0.43,4.82] 220 Ambulance transport 0.9 [0.54,1.50] 0.92 [0.54,1.56] 0.55+ [0.30,1.03] 1.74 [0.86,3.53] 2.53** [1.41,4.55] 1.06 [0.91,1.23] 1 [1.00,1.00] 0.9 [0.44,1.83] 0.9 [0.21,3.87] 1.01 [0.22,4.63] 220 Police transport 0.43** [0.24,0.77] 1.51 [0.81,2.80] 1.59 [0.87,2.91] 1.74 [0.63,4.84] 1.05 [0.56,1.96] 0.96 [0.84,1.11] 1 [1.00,1.00] 2.25* [1.05,4.79] 1.06 [0.23,4.98] 1.38 [0.27,7.02] 220 Arrests 1.24 [0.72,2.15] 1.71* [1.03,2.85] 0.79 [0.43,1.45] 1.27 [0.70,2.32] 1.55 [0.89,2.70] 1.05 [0.91,1.22] 1 [1.00,1.00] 1.13 [0.56,2.28] 1.21 [0.57,2.58] 0.96 [0.41,2.25] 220 Sentence days 1.23 [0.62,2.44] 2.85* [1.25,6.51] 0.88 [0.37,2.08] 0.40+ [0.15,1.06] 0.73 [0.33,1.61] 1.08 [0.87,1.34] 1 [1.00,1.00] 2.96* [1.20,7.33] 1 [0.29,3.51] 0.71 [0.15,3.46] 220 Shelter stays 0.11** [0.05,0.24] 1.48 [0.59,3.72] 2.70* [1.02,7.11] 0.43 [0.10,1.84] 0.25** [0.10,0.64] 1.23 [0.93,1.63] 1 [1.00,1.00] 4.61* [1.39,15.32] 2.63 [0.55,12.62] 1.02 [0.17,6.07] 220 Shelter days 0.10** [0.05,0.23] 1.51 [0.60,3.77] 2.63* [1.01,6.89] 0.41 [0.10,1.74] 0.25** [0.10,0.63] 1.24 [0.93,1.64] 1 [1.00,1.00] 4.87** [1.52,15.62] 2.58 [0.53,12.51] 0.99 [0.17,5.90] 220 Substance use treatment visits 0.38* [0.18,0.84] 11.07** [4.62,26.52] 0.99 [0.39,2.50] 1.15 [0.31,4.30] 1.05 [0.45,2.44] 1.16 [0.95,1.41] 1 [1.00,1.00] 1.25 [0.55,2.87] 2.67 [0.53,13.55] 0.76 [0.09,6.45] 220 Mental health treatment visits 1.91** [1.20,3.04] 0.65* [0.43,0.99] 0.75 [0.45,1.26] 0.93 [0.50,1.76] 0.88 [0.57,1.38] 1.08 [0.93,1.26] 1 [1.00,1.00] 1.22 [0.76,1.94] 1.09 [0.37,3.23] 1.41 [0.40,4.97] 220 Abbreviations: outpt = outpatient; inpt = inpatient; psych = psychiatric; med=medical; adm = admission; LOS = length of stay; T&R = treat and release; admit = admitted; ED = emergency department P-values: +p<0.10; *p<0.05; p<**0.01 10 Appendix 4: Death Data Table 1: Demographic and Utilization for the 24 Months Pre-Enrollment Comparing Participants who Died to Those Who Remained Alive Alive Mean (n=331) Died Mean (n=41) Difference p-value Sample Characteristics Days Enrolled 766.5 860.7 -94.1 0.05 Months Enrolled 25.1 28.1 -3.0 0.06 Male 71.1% 82.9% -11.8% 0.11 Hispanic 25.9% 25.0% 0.9% 0.91 Black 14.8% 9.8% 5.0% 0.38 Other Race 52.9% 53.7% -0.8% 0.92 Age 50.2 54.5 -4.3 0.02 Smoker 66.8% 61.0% 5.8% 0.46 Medi-Cal Coverage 78.9% 87.8% -9.0% 0.18 Medicare Coverage 17.2% 12.2% 5.0% 0.42 Primary Care Regular Source of Care 73.7% 80.5% -6.8% 0.35 Primary Care Office Visits 8.1 8.9 -0.8 0.68 Hospital Use Inpatient Total Inpatient Visits 4.8 8.8 -4.1 0.00 Inpatient Psychiatric Admissions 0.3 0.0 0.2 0.07 Inpatient Admissions 2.4 4.6 -2.2 0.00 Number of Inpatient Days 14.2 28.8 -14.6 0.00 Emergency Department (ED) Total ED Visits 19.9 24.1 -4.2 0.18 ED Treat-and-Release Visits 17.8 20.0 -2.1 0.49 Admitted ED Visits 2.1 4.1 -2.1 0.00 Psychiatric ED Visits 5.8 3.2 2.6 0.09 ED Visits with Ambulance Transport 5.6 8.2 -2.6 0.08 ED Visits with Police Transport 1.5 0.4 1.1 0.05 Criminal Justice Data Arrests 3.4 2.5 0.9 0.33 Sentence Days 56.3 34.1 22.2 0.17 Shelter Data Shelter Stays 40.5 24.3 16.2 0.16 Shelter Days 46.6 31.3 15.4 0.26 11 Alive Mean (n=331) Died Mean (n=41) Difference p-value Unicare Any Substance Use Service(DADS) Visits 6.4 7.7 -1.3 0.59 DADS, Intake and Individual Treatment Visits 2.9 3.7 -0.8 0.51 DADS Treatment Planning Visits 2.6 3.9 -1.3 0.31 DADS Crisis Intervention 0.1 0.3 -0.2 0.03 DADS Group Visits 2.4 2.2 0.2 0.84 DADS Medication Visits 0.3 0.5 -0.3 0.37 Any Mental Health (MH) Visits 30.7 18.7 12.0 0.25 MH Management Visits 16.4 9.6 6.9 0.23 MH Assessment Visits 1.5 0.9 0.6 0.18 MH Testing Visits 0.2 0.1 0.1 0.38 MH Individual Treatment 0.5 0.1 0.4 0.34 MH Group Treatment 1.2 0.0 1.2 0.56 MH Rehab Visits 3.7 3.3 0.4 0.84 MH Medication (MD) 2.5 1.7 0.8 0.37 MH Medication (Non-MD) 4.7 3.0 1.6 0.60
Updated over 4 years ago