Homeless with Schizophrenia Presumptive Disability Pilot Evaluation

Introduction

Having a disability is a factor that increases the risk of becoming homeless. In 2009, almost 38 percent of the homeless population had a disability, compared with about 16 percent of the total U.S. population (Department of Housing and Urban Development 2010). Individuals with a serious mental illness are particularly vulnerable to homelessness. Additionally, the nature of mental illness prevents many from applying for assistance. Two serious and chronic mental illnesses—“schizophrenia” and “schizoaffective disorder”—together affect about 1 out of every 100 people (National Alliance on Mental Illness 2012, 2013). Individuals with those disorders face formidable challenges to gaining much needed support, such as adequate housing and treatment, and accessing public benefits.

Supplemental Security Income (SSI)—a potential source of income for this population—is a means-tested program that makes monthly payments to individuals who have limited income and resources and who are aged 65 or older, blind, or disabled. Section 223 of the Social Security Act defines disability as, “the inability to engage in any substantial gainful activity by reason of a medically determinable physical or mental impairment, which can be expected to result in death or which has lasted or can be expected to last for a continuous period of not less than 12 months.” For this study—the Homeless with Schizophrenia Presumptive Disability (HSPD) pilot—we focused on homeless SSI applicants in specific geographic locations who alleged schizophrenia or schizoaffective disorder.

Many homeless individuals with a serious mental illness are potentially eligible for SSI payments, but the nature of their impairment poses obstacles to completing the SSI application process. For an applicant to meet the Social Security Administration's (SSA's) definition of a disability, the evidence presented must be thorough. However, the treatment history of applicants who are homeless and have a serious mental illness may be intermittent, inaccurate, or incomplete, and involve multiple locations and doctors. Additionally, the lack of stable housing makes it difficult for homeless individuals to maintain or safeguard required documentation, such as identification and medical records, and to provide accurate contact information. These complications in turn affect the individual's access to many social services and his or her ability to schedule and keep appointments, such as the consultative examination (CE), that SSA may require to make a disability determination. In most cases, an SSI award for an adult depends on the degree of functional limitation, not solely on a medical diagnosis (Wixon and Strand 2013). The evaluation of a disability on the basis of schizophrenia or schizoaffective disorder requires documentation of the medically determinable impairment(s), consideration of the degree of limitation such impairment(s) may impose on the individual's ability to work, and consideration of whether those limitations have lasted or are expected to last for a continuous period of at least 12 months.

In this article, we assess whether helping individuals in our target population with completing the SSI application process, coupled with providing presumptive disability (PD) payments, helps to improve several programmatic outcomes. Specifically, we compare the processing times and payment outcomes for individuals receiving application assistance and PD payments prior to SSA's final disability determination with individuals not receiving those services in nearby locations, in a prior period, or in the same location and time period. Our results are not causal; however, we find that the intervention is associated with a shorter application process and an increase in SSI payments over a defined follow-up period. We also discuss the implications of our findings for national policy.

The HSPD Pilot

SSA designed the HSPD pilot to address the factors that prevent homeless adults with schizophrenia or schizoaffective disorder from receiving SSI payments. In 2012, SSA's San Francisco Regional Office partnered with community health agencies in San Francisco and Santa Cruz, California, to implement the project. In 2013, SSA expanded the project by bringing onboard an additional community health agency in Los Angeles. These partners—the San Francisco Department of Public Health, the Human Services Agency of San Francisco, the County of Santa Cruz Health Services Agency, and the Los Angeles County Department of Health Services—all had experience with providing services both to individuals with mental illnesses and to those experiencing homelessness. Additionally, all partners were established institutions providing comprehensive and multidisciplinary programs and services to address public-health issues in their communities. They also employed staff experienced in working with individuals in specialized programs who could implement the HSPD interventions by connecting persons in the target population to an array of services to help address their medical, psychological, advocacy, and housing needs. A crucial step for developing the study populations for the project evaluation was identifying individuals who were potentially homeless during the period of interest. For details on the measures used in determining homelessness, see Appendix A.

The HSPD project included two intervention components: SSI application assistance and a PD recommendation. First, the community partners used their established outreach processes to identify homeless individuals who had schizophrenia or schizoaffective disorder, then they helped those individuals with the SSI application process. Throughout the process, community-partner staff helped individuals with a confirmed diagnosis by scheduling and coordinating necessary appointments, gathering medical evidence, and ensuring that the local participating SSA field office received the application.

Second, community-partner staff recommended PD payments for SSI recipients who were homeless and had a confirmed diagnosis of schizophrenia or schizoaffective disorder. The PD policy allows an individual applying for SSI based on a disability to receive payments for up to 6 months prior to SSA's initial disability determination; the existence of certain disabilities “presume” approval for SSI. Generally, the field office may approve PD payments for persons with conditions that fall under a limited number of specific categories, such as an amputated leg or an allegation of total deafness, for which the evidence strongly reflects that the impairment would meet SSA's definition of a disability. Repayment of any monies received is not required, as long as SSA does not deny the application for nonmedical reasons (SSA 2014b).

For the HSPD pilot, three SSA field offices—San Francisco Downtown, Santa Cruz, and Los Angeles Downtown—authorized PD payments based on a confirmed diagnosis of schizophrenia or schizoaffective disorder (SSA 2014a). Community partners used the PD recommendation form created for this project—the Schizophrenia Presumptive Disability Recommendation Form (SSA-121)—on which licensed physicians or psychologists were required to attest whether the individual's condition met criteria consistent with SSA's medical listings for schizophrenia or schizoaffective disorder. SSA's standard PD process does not require such a recommendation form.

The HSPD Process

To learn more about how the HSPD process was actually implemented, all local partners responded (via e-mail or telephone) to a standard set of questions about their processes. From their responses, we learned that the length of the application process varied with each individual case, but usually took from several days to a few months to complete. During the outreach process, partners identified individuals who potentially met the HSPD pilot criteria and referred them to staff and clinicians for individual case management and professional assessments. The case manager reviewed existing medical records, obtained additional information from treatment providers and family, and scheduled an appointment with a physician or psychologist to further document the nature of the disability. If the assessment indicated that the disability was schizophrenia or schizoaffective disorder, then the physician or psychologist completed the PD recommendation form, certifying that the individual showed certain symptoms and correlated functional limitations and that the applicant's condition was not caused by substance abuse (alcohol or drugs). The case manager submitted the completed SSI application, PD form, and supporting evidence to the participating SSA field office and the Disability Determination Services (DDS) then expedited HSPD cases through the determination process. Additionally, community-partner staff provided assistance throughout the full adjudication process, when needed.

At each site, the intervention process involved intensive case-management and follow-up services. Staff members conducted face-to-face meetings several times with individuals in the pilot and monitored their cases closely. They reminded individuals of and accompanied them to various appointments, coordinating activities with other members of the team, when needed. They also assisted individuals with finding other support services that could help them with their housing, transportation, and other basic needs.

The pilot operated for 24 months, from April 2012 to April 2014. During that time, SSA and its partners assisted 260 homeless individuals in California with their SSI applications and PD recommendations: 78 in San Francisco, 24 in Santa Cruz, and 158 in Los Angeles.

Data and Methodology

At the outset, we decided that it was not feasible to use a randomized design because of the vulnerability of the homeless population and the obligations of SSA's partners and service providers. Instead, we chose a quasi-experimental design aimed at identifying the effects of application assistance and PD payments on the outcomes of interest. The primary outcomes from the research questions we focused on in this article address the extent to which the pilot had the following effects:

1. Increased SSI allowance rates at the initial adjudicative level (and increased SSI payment receipt after 6 and 12 months)
2. Reduced the need for CEs
3. Reduced the time required to adjudicate the claim (including specific segments of the application process)
4. Reduced appeals
5. Increased total payments
6. Reduced deaths

Specifically, we wanted to compare the outcomes of individuals who received SSI application assistance and PD payments—the treatment group—with the outcomes of individuals in the three comparison groups—main group (C1), second group (C2), and third group (C3). Table 1 summarizes the four groups observed during the pilot evaluation. All individuals included in our analyses met the selection criteria in Appendixes A and B.

Treatment Group

Our treatment group consisted of 238 homeless individuals who met the criteria for the pilot, established a claim during the pilot period (April 20, 2012–April 18, 2014), and received assistance. We excluded 22 other applicants from the group for various reasons, including establishing a claim outside the pilot period, applying in a nonparticipating field office, and not having schizophrenia or schizoaffective disorder indicated on the application. We also excluded applicants who applied for Disability Insurance rather than SSI and those who did not meet the criteria for homelessness. For SSA, the date a claim is established is the date on which the agency officially enters the applicant's claim into its records. This is typically later than the date that the applicant filed the claim.

Comparison Groups

In total, our comparison groups consisted of 2,571 individuals. The largest comparison group C1 (with 1,038 members) included individuals who had applied for benefits in the prior 2-year period (April 20, 2010–April 18, 2012). Comparison group C2 (with 676 members) and comparison group C3 (with 857 members) consisted of individuals who had established their claims during the pilot period.

In our main comparison group (C1), SSI applicants alleged either schizophrenia or schizoaffective disorder according to their applications, met the criteria for homelessness, and had applied for SSI payments in one of the three pilot field offices in the 2 years before the pilot. Individuals included in C1 did not receive PD payments based on an allegation of schizophrenia or schizoaffective disorder, but may have received some assistance from SSA's community partners. Comparing individuals in the treatment group with those in comparison group C1 provided us with an estimate of the impact of the PD payments, without the confounding influence of location differences. However, some bias in the results may remain, as there may be year-specific differences between the two groups. Additionally, we note that this does not necessarily separate the effects of the PD payments from the application assistance provided to the target population as part of our designed intervention.

SSI applicants in our second comparison group (C2) alleged schizophrenia or schizoaffective disorder, met the criteria for homelessness, and applied for SSI payments in the surrounding area field offices during the study period. Individuals in C2 might have been eligible for PD payments had they received assistance, but they were not in a participating location served by SSA's community partners. Thus, comparing the differences between the treatment group and comparison group C2 should avoid any year-specific distinction and identify the effects of the PD payments along with the assistance given by the providers. However, some selection into those two groups based on location may bias our estimated effect.

Our third comparison group (C3) included SSI applicants who allegedly had schizophrenia or a schizoaffective disorder, met the criteria for homelessness, and had applied for SSI benefits in one of the three pilot field offices during the pilot period. Individuals in C3 did not receive the schizophrenia or schizoaffective disorder PD payments or the same application support received by the treatment group. However, they may have been eligible for application assistance and PD payments had the community partners identified them and provided assistance. Alternatively, they may have received some assistance from the partners, but were not considered eligible for PD payments. As with the treatment-to-C2 comparison, the difference between the treatment group and comparison group C3 avoids any year-specific factors.

Estimation Methods

We estimated the unadjusted means and proportions for the outcomes for each study group—treatment, C1, C2, and C3—calculated the difference between the groups, and applied the appropriate statistical tests to determine if the differences were significant. For continuous and binary outcomes, such as benefit amounts or elapsed days, we used a standard two-sample t-test (or proportion test) on the equality of means or proportions. Because of the exploratory nature of the study, we tested at the 10 percent significance level. (We did not use regressions or other means to adjust our estimates for observed characteristics. In future research, we may explore regression-adjusted, difference-in-differences, and propensity-score-based estimates.)

Data Sources

We combined administrative data from three SSA sources—the Structured Data Repository, the Supplemental Security Record, and the Numerical Identification System (Numident)—to answer our research questions. The Structured Data Repository, which includes demographic and programmatic information on SSI applications, was the primary data source for the study. Specifically, this source provided field office codes; alleged diagnosis descriptions; primary diagnosis codes selected by the DDS examiners or medical consultants; CE requests; application dates; appeals data, including decisions at each level of adjudication with corresponding dates; and field office and DDS case processing dates. The Supplemental Security Record provided us with information on current-pay statuses and total SSI payments in the first year after application. The Numident gave us information on deaths that occurred within the first 12 months after applicants had established their claims.

Characteristics of the Treatment and Comparison Groups

Table 2 presents selected demographic, geographic, and disability-related characteristics of the treatment group and three comparison groups. For our analyses, we combined the San Francisco Downtown and Santa Cruz field offices and surrounding area field offices into one Northern California location because of the smaller number of participants in those geographic areas.

Men made up the largest percentage of the treatment and comparison groups (71–76 percent) with no statistically significant differences between those groups. The distribution of ages at the time of application ranged primarily from age 18 to 59. Persons aged 30 to 49 accounted for more than half of all groups combined. At the time of their application, individuals who had applied at the same three treatment field offices in both the prior period (C1) and pilot period (C3) were slightly older, compared with those in the treatment group, with differences significant at the 5 percent and 10 percent levels.

Most of the cases in the treatment group originated in the Los Angeles Downtown field office (about 65 percent), followed by San Francisco Downtown (26 percent), and Santa Cruz (9 percent). These three field offices also managed the cases in comparison groups C1 and C3: Los Angeles Downtown (79–80 percent), San Francisco Downtown (16–18 percent), and Santa Cruz (3–4 percent). We found the differences between the treatment group and comparison groups C1 and C3 statistically significant at the 1 percent level for these field offices. For the surrounding areas in comparison group C2, the percentage of cases from Northern California (42 percent) was slightly less than those from Los Angeles (58 percent), which reflects the manner in which we chose these field offices for the study.

As would be expected, nearly all cases in the treatment group indicated schizophrenia or schizoaffective disorder in the allegation text field (97 percent), and the majority of those cases had a primary diagnosis code of 2950 (95 percent), indicating schizophrenic, paranoid, and other psychotic disorders. We found some variations of schizoaffective (54 percent) in the allegation text field more often than we found variations of schizophrenia (42 percent). A small share of the treatment group did not receive a PD payment based on having a primary diagnosis code of 2950 (5 percent). Those individuals had alternative diagnosis codes for disabilities, such as affective disorders, anxiety-related disorders, and substance addiction disorders (alcohol or drugs), although not all received an allowance under those categories (not shown).

A notably smaller percentage of comparison-group cases had an allegation of schizophrenia or schizoaffective disorder (around 77 percent) or a primary diagnosis code of 2950 (ranging from 57 to 67 percent) on which SSA made a disability determination. Differences between the treatment and comparison groups (each significant at the 1 percent level) suggested that under the normal process, an allegation of schizophrenia did not consistently result in a determination based on a diagnosis of schizophrenia. However, we did not examine secondary diagnoses and because schizophrenia may be difficult to document, we may simply have observed that the medical determinations relied on thorough diagnoses of schizophrenia or schizoaffective disorder for the treatment group, but comorbidities for the comparison groups. We also emphasize that this was not a randomized control trial so our comparison groups were subjected to selection bias.

Results for Research Questions

The HSPD evaluation design report identified 10 research questions. However, with the data available, we could not answer two questions concerning cost savings and reductions in homelessness. In this section, we provide the results for the remaining research questions, some of which are combined, and related findings.

The Intervention Led to a Significantly Higher Allowance Rate at the Initial Disability Adjudication Level

The allowance rate for the entire treatment group was 94 percent, ranging from 87 percent in Northern California to 97 percent in Los Angeles (Table 3). Overall, the treatment group saw a higher allowance rate at the initial-decision level than the three comparison groups, with differences of 28 percentage points (C1), 36 percentage points (C3), and 53 percentage points (C2). According to SSA's records, at the national level, 7 percent of PD findings in fiscal years 2012, 2013, and 2014 did not result in an eventual SSI payments allowance.¹ With a 95 percent ultimate allowance rate for the treatment group, the reversal-of-PD-finding rate for the pilot is in line with the national rate during the same period.

When we reviewed SSI awards at any level, the percentage of individuals in the HSPD pilot with an allowance increased slightly for the treatment group (less than 1 percentage point), rising a little more in comparison groups C3 (3 percentage points) and C2 (5 percentage points). Comparison group C1, which had a longer time for processing appeals, experienced the highest allowance-rate increase (8 percentage points). The difference in allowance rates between the treatment and C1 groups was 28 percentage points at the initial level, falling to 21 percentage points using the allowance rate at any level. We expect the difference to shrink over time, as all appeals are fully processed given the pattern we have observed during the 2-year follow-up period used for comparison group C1; however, we do not expect it to decline too much.

The Intervention Reduced Requests for CEs at the Initial Level of Application

The DDSs requested fewer CEs for cases in the treatment group (4 percent) than for any comparison group. The differential impact was largest when comparing the treatment group to comparison group C2 (31 percentage points). The treatment/comparison group differences remained strong across both regions and were statistically significant at the 1 percent level for nearly all comparisons. Additionally, we observed large differences in CE requests for the Northern California and Los Angeles regions, which were likely attributable to differences in either the intake processes (at the partner, field office, or DDS levels) or in the population characteristics.

The Intervention Reduced the Time Required to Adjudicate the Claim

For the treatment and comparison groups, we compared the processing time for three individual time segments: 1) the earliest filing date to the date the claim was established; 2) the date the claim was established to the date the field office released the case to the DDS; and 3) the date the field office released the case to the DDS to the initial decision. The pilot appeared to have a modest impact on the time between the earliest filing date to the date the claim was established. Compared with the C1 group, the pilot reduced that time by 7 days (from 27 to 20 days); however, that difference was not significant when comparing the treatment group with the other comparison groups (Table 4).

The average number of days for the second segment—the date the claim was established to the date the field office released the case to the DDS—was only reduced by the pilot in Los Angeles. This was not completely unexpected as there should be little reason for the claim to remain at the field office once it had been established.

For the third segment—from the date the field office released the case to the DDS to the initial decision—the HSPD intervention reduced the processing time by 66 to 77 percent, to 30 days, on average. The processing time averaged 58 days in Northern California and 15 days in Los Angeles for the treatment group. All differences were statistically significant at the 1 percent level. It is important to note that the California DDS had a significant backlog of claims from 2010 to present, resulting in longer processing times for many cases. Treatment cases were not subject to the backlog, which may have led to larger differences in processing times between treatment and comparison group cases. By contrast, for the comparison groups, it took 86 days, on average, in the same field offices during the prior period (C1), compared with 107 days in the same field offices during the pilot period (C3) and 131 days in the surrounding field offices during the same period (C2).

In addition to these specific segments of the application process, we also looked at two combined time segments or overall time periods. We saw large, statistically significant reductions in the time between the date a claim was established and the individual's first SSI payment. For the treatment group, this averaged just 10 days. For the comparison groups, the average number of days for this measure was significantly higher: C1 (91 days), C2 (144 days), and C3 (106 days). Thus, the pilot reduced the time between the date the claim was established and the applicant's first SSI payment by 3 to 5 months. As would be expected, we find similar results when we look at the time between the date a claim was established and the initial disability decision, although these were somewhat smaller differences, as treatment members tended to receive their first SSI payment before their initial decision.

The Intervention Did Not Have a Significant Impact on the Rate of Appeals

We also wanted to examine whether the intervention would have an effect on the rate of appeals. The percentage of all initially denied cases appealed to the reconsideration level or higher in the comparison groups was between 45 and 50 percent, while the appeal rate in the treatment group was 64 percent; these differences are not statistically significant (Table 5). We caution that the appeal rate for the treatment group was based on only 14 denials at the initial level, whereas each comparison group had more than 300 denials at the initial level. Secondly, we might expect to see a higher appeal rate for the treatment group because the intervention was designed to select cases with a high likelihood of approval, and treatment group members were already connected to representative and advocate resources.

The Intervention Led to an Increased Likelihood of Being in Current-pay Status

To examine the impact of the intervention over time, we analyzed cases at two intervals—6 months and 12 months after the claims were established—to learn whether individuals were in current-pay status.² Individuals in current-pay status were due a payment contingent upon meeting the reporting requirements during the month.

At the 6-month mark, a larger share of the treatment group received an SSI payment (81 percent), compared with those in the comparison groups: C1 (44 percent); C3 (35 percent); and C2 (22 percent). (Table 6). These findings were statistically significant at the 1 percent level, with differences between the treatment and comparison groups ranging from 37 to 59 percentage points. We continued to find statistically significant (albeit somewhat smaller) differences at the 1-year mark. About 74 percent of the treatment group received an SSI payment at 12 months, with differences between the treatment and comparison groups ranging from 23 to 39 percentage points. The share of the treatment group who received payments declined in the interval between the 6- and 12-month marks. For the comparison groups, the shares receiving payments rose during that interval.

For all groups, the percentage of individuals in current-pay status was higher than the percentage who actually received a payment. For this measure, the impact of the intervention follows the same trend with the treatment group having a much higher percentage in current-pay status. The differences were not quite as large, but were still statistically significant.³

It was also important to examine why a person who had received the PD payments and subsequent SSI payments would have moved into nonpay status at those same two intervals. The most common reasons were the recipient's income exceeded the allowable threshold; the recipient was placed in a “failed to cooperate” or “unable to locate” category; the recipient had become an inmate of a penal institution during that time; or his or her payment-status code was missing (Table 7).

When we looked at the most common reasons why some individuals fell into nonpay status at the 6-month mark, we noted several differences between the groups. About 67 percent of persons in the treatment group had income that exceeded federal and state SSI thresholds, compared with 21 to 43 percent in the comparison groups. By contrast, no one in the treatment group “failed to cooperate” or was “unable to locate,” compared with 12 to 29 percent of the comparison groups. The percentage of individuals who were inmates of a penal institution was higher in the treatment group than in the comparison groups in the Los Angeles subset, but not in the Northern California subset. However, the differences for the last reason—payment-status missing—were less consistent and not statistically significant across most group comparisons. The impacts for the most common reasons for nonpay status at the 12-month mark were similar, but generally smaller and less significant.

The Intervention Resulted in Larger Cumulative Payments for the Treatment Group

We found large, statistically significant (at the 1 percent level) differences between the cumulative payments received by the treatment and comparison groups at the 6- and 12-month marks (Table 6). After 6 months, average SSI payments for the treatment group totaled about $3,700, which was $2,000 to $3,000 more than the $700 to $1,700 received by the comparison groups. Average cumulative payments for the treatment group were similar between the two regions (Northern California and Los Angeles); however, we saw a large difference between the two regions in average cumulative payments for comparison group C3—individuals who had applied in the same field offices during the same period. Cumulative payments for the C3 group averaged $704 for recipients in the San Francisco Downtown and the Santa Cruz locations and $1,539 for those in the Los Angeles Downtown location, leading to a smaller relative impact for the Los Angeles region. We expect differences between the treatment and comparison groups to lessen as more comparison group cases receive allowances during the appeals process and eventually receive back payments.

The Intervention Did Not Appear to Have a Strong Effect on Mortality

There were no deaths in the treatment group and a very small percentage of individuals in the comparison groups died within 12 months of establishing their claims with SSA. The average mortality rate during that period was less than 1 percent for comparison groups C1 and C2 and just over 1 percent for comparison group C3 (Table 6).

HSPD Case Reviews

SSA's Office of Quality Review (OQR) reviewed almost all (223 of the first 225) treatment cases in the pilot. Of the 215 cases allowed (96 percent), OQR cited deficiencies in 48 percent of the cases, determining that three cases had been incorrectly allowed. Of the eight denied cases, OQR cited four as deficient, with two incorrect denials. The most common deficiency cited was that the cases relied on one piece of medical evidence for establishing disability, which OQR suggested was insufficient for a determination. However, this suggestion does not imply that the DDS made an incorrect determination on these cases.

In response to OQR's review, SSA's San Francisco Center for Disability (SFCD) reviewed 54 of the 108 cases that OQR cited with deficiencies. SFCD concurred with OQR for 33 cases (61 percent), acknowledging the potential for quality issues in the adjudication of those cases and noting the variation in deficiency rate by community partner. SFCD suggested that one piece of medical evidence may be sufficient to adjudicate a claim and that OQR may not have fairly weighed the evidence from third parties, such as case managers, which can be important for the population in this study.

Discussion

Overall, the HSPD pilot appears to have been successful. The group that received the PD payments was more likely to have received an initial allowance and less likely to have required a CE than were the comparison groups. They also received their decisions and first SSI payments sooner than did the comparison groups, along with higher cumulative payment amounts in the 12 months after establishing a claim. We were not able to observe other important outcomes, such as decreased homelessness, that the pilot was intended to address.

Although the pilot was generally successful, its scalability to the national level is unclear. The community partners who developed the cases had experience working with individuals who were homeless or had mental impairments, largely because of the high volume of similar cases in the target areas and prior involvement with SSA outreach efforts. Although many other locales have similar public-health agencies performing similar functions, it is uncertain how the services provided in the pilot will transfer to other settings.

Our community partners, particularly in Northern California, were somewhat conservative in their diagnoses, signing off on the PD form shown in Appendix C only after careful review to ensure that there was sufficient medical evidence (and the absence of drug abuse or alcoholism), consistent with SSA's medical listings. However, even with experienced partners making careful diagnoses, SSA's OQR reported issues with insufficient medical evidence for many cases.

The requirements for the PD finding followed SSA's medical listings for schizophrenia or schizoaffective disorder. As such, it required the applicant to have medically documented evidence of certain persistent symptoms resulting in increased restrictions or difficulty with specific functions or a history of a chronic schizophrenic, paranoid, or other psychotic disorder. It is unclear whether gathering such information for homeless individuals suffering from schizophrenia or schizoaffective disorder is generally feasible or cost effective, regardless of the legal requirement. One alternative to PD payments may be to require a shorter longitudinal medical history for homeless individuals alleging schizophrenia or schizoaffective disorder and to have SSA conduct a continuing disability review after 2 years that waives the medical improvement review standard. SSA could potentially combine such a policy with two existing fast-track programs—Quick Disability Determination and Compassionate Allowance—to expedite homeless cases. Changing the medical improvement review standard and required longitudinal history would likely require a statutory change. Because this study focused only on a small population of homeless individuals alleging schizophrenia or schizoaffective disorder, the appropriateness of such a policy change for the SSI program as a whole is unclear. However, regardless of the policy implemented, it may be helpful for all disability adjudicators to receive additional training on the evidentiary requirements for claims with no longitudinal treatment history of a mental impairment or diagnosis, as suggested by SSA's San Francisco Regional Office.

We note that other locales have tested similar interventions. For example, in 1993, SSA initiated the Maryland SSI Outreach Project in the city of Baltimore, which also successfully awarded PD payments to homeless individuals meeting certain impairment criteria. Some of the recommendations from that project continue to be appropriate. For example, as the HSPD pilot demonstrated, replicating similar outreach projects would require SSA to work closely with organizations that are capable of both diagnosing and supporting homeless individuals with mental impairments (National Alliance to End Homelessness 2015).

The SSI/Social Security Disability Insurance (SSDI) Outreach, Access and Recovery (SOAR) project funded by the Substance Abuse and Mental Health Services Administration (SAMHSA) in 2001 aimed to increase access to federal disability benefits for adults who are homeless or at risk of being homeless and have a mental illness, medical impairment, and/or a co-occurring substance abuse disorder. SOAR programs have helped increase the award rate and reduce the time from application to decision for this vulnerable population. SAMHSA continues to fund the SOAR Technical Assistance Center, which facilitates state- and local-based SOAR programs (SAMHSA, n.d.). Based on experiences from the SOAR projects, the authors of a National Academy of Social Insurance (NASI) report suggested three policy changes: 1) expanding the list of acceptable medical sources for DDS examiners, 2) allowing individuals who have been homeless for at least 6 months and who have schizophrenia to qualify for PD payments, and 3) modifying SSA's processes to address the needs of homeless adults (Perret, Dennis, and Lassiter 2008). The authors also recommended that SSA improve its tracking of residential statuses and assign homeless cases to field office and DDS staff who have received additional training in working with this population.

As we noted earlier, detailed and longitudinal medical evidence often does not exist for the homeless population, whose records are sporadic or difficult to obtain. The authors of the NASI report suggest that professionals, such as licensed social workers, certified nurse practitioners, or certified physician assistants, should be able to provide evidence that is weighed as heavily as other evidence provided by physicians and psychologists. These individuals are often more likely to provide treatment for this population, making them better at providing the necessary information.

Although we did not conduct a formal cost-benefit analysis, the PD recommendation reduced the time spent by SSA to develop a case, and fewer CEs also clearly reduced costs for the agency. One study suggests that a CE for mental health impairments costs over $235 (Wittenburg and others 2012). However, the exams and tests required for a CE can vary and the costs in California in particular may differ from this average. The HSPD pilot demonstrated that the number of CEs requested for the treatment group was 14 percentage points less than the number requested for comparison group C1. A back-of-the-envelope calculation suggests that in the absence of the pilot, SSA would have requested an additional 33 CEs for the treatment group, which translates to a potential savings of about $7,755. This, combined with the higher initial allowance rate and reduced number of appeals for the treatment group, indicates that other administrative savings were likely as well. SSA incurred few administrative costs for the PD payments other than the fixed cost of setting up the process. We did not consider one-time cost items, such as staff training. SSI payments to individuals did not provide a cost in this setting because it is SSA's mission to administer such payments.

The results presented in this study are from a quasi-experimental design and are not causal in nature. The demographic characteristics presented in Table 2 suggest the treatment group is somewhat different from the comparison groups. In future work, policy analysts could use more rigorous statistical techniques that would control for these differences and provide estimates that are more robust than those presented here.

Finally, we note that many individuals in the treatment group also filed an application in the prior period, and those applications were initially included in the comparison groups. To avoid double counting these treatment group members, we removed their prior applications from the comparison groups. As noted in the report on the Maryland SSI Outreach Project, helping qualified individuals to receive SSI payments the first time they apply is likely more cost effective than granting an award after the second or third application.

Appendix A: Identification of Homeless Individuals

A crucial step to developing comparison groups for the HSPD pilot evaluation was identifying individuals who were potentially homeless during the period of interest. The community partners identified the individuals in the treatment groups as homeless (a criterion for participating in the pilot). For uniformity, we used the same selection process for both the treatment and comparison groups in our analyses. This inevitably meant removing some treatment group members from the analysis who did not have a clear indication of homelessness in SSA's administrative data. The five selection criteria for identifying homelessness follow:

1. Homeless flag on the SSI application.
2. A residence type of “transient” listed as the most recent residence type, with a start date on or before the date that the SSI claim was established in SSA's records. The five transient data files from 2010 to 2014 came from SSA's Office of Systems.
3. Residential address field contained a word or phrase from Keyword Set A1 or Keyword Set A2.
4. Residential address field contained the name of an emergency shelter from the Department of Housing and Urban Development's (HUD's) list of emergency shelters in California. Organization names and program names were pulled from HUD's list, a few abbreviations were removed (ERT and STAR), and shelter names were shortened, for example, MSC-South Shelter was shortened to MSC.
5. Remarks section in the Field Office Disability Report (SSA Form 3367) or Disability Report—Adult (SSA Form 3368) contained a word from Keyword Set A1.

Appendix B: Identification of Schizophrenia and Schizoaffective Disorder

The evaluation of the HSPD pilot required identifying SSI applicants who allegedly had, or had been diagnosed with, schizophrenia or schizoaffective disorder. We apply this same identification process to the treatment group for consistency. To be included in the evaluation, each case must have met at least one of the following criteria:

1. Allegation description of schizophrenia or schizoaffective disorder

   We mined the allegation text field for root words and various misspellings of “schizo” found in Keyword Set B1 (below). From this list, we searched again for root words more specific to “schizophrenia” and “schizoaffective” found in Keyword Set B2 and Keyword Set B3 to differentiate these two categories. The second search picked up one invalid observation, which we removed. Lastly, we used a “sounds like” function to search the text field for “schizophrenia” and “schizoaffective,” to catch any additional common misspellings. This last procedure did not find any additional observations.

2. Primary diagnosis code: 2950

   We flagged any observations with a “2950” primary diagnosis code as a potential indicator of schizophrenia or schizoaffective disorder. The “2950” impairment code covers the Mental Disorder listing 112.03: Schizophrenic, Paranoid, and Other Psychotic Disorders. We included any individuals who received a denial based on this code, in addition to those who were approved, to capture as many individuals as possible who may have schizophrenia or schizoaffective disorder.

Appendix C