urban institute nonprofit social and economic policy research

The Role of Medicaid in Improving Access to Care for Homeless People

Read complete document: PDF


PrintPrint this page
Document date: June 01, 2002
Released online: June 01, 2002

This report was prepared under a grant from the Kaiser Commission on Medicaid and the Uninsured to the Urban Institute.

The nonpartisan Urban Institute publishes studies, reports, and books on timely topics worthy of public consideration. The views expressed are those of the authors and should not be attributed to the Urban Institute, its trustees, or it funders, or to the Kaiser Commission on Medicaid and the Uninsured.


TABLE OF CONTENTS

Highlights

Chapter 1: Introduction
Methods
   The NSHAPC Survey
   Descriptive and Independent Variables
   Dependent Variables
Overview of Rest of Report

Chapter 2: Adults in Homeless Families
   Introduction
   Characteristics of Adults in Homeless Families
   Access to and Use of Care
Results of Regression Analyses: Insurance, Use, and Access
   Insurance Receipt
      Medicaid Receipt
      Receipt of Other Health Insurance
Results from the NSHAPC Section on Physical Health Care
   Contact with a Doctor or Nurse for Health Reasons
   Ambulatory Care
   Emergency Room Care
   Hospitalization
   Inability to Obtain Care When Needed
   Summary
Treatment for Mental Health Problems
Treatment for Substance Use Problems
Implications

Chapter 2 Tables

Chapter 3: Children in Homeless Families
   Introduction
   Characteristics of Children in Homeless Families
   Access to and Use of Care
Results of Regression Analyses: Insurance, Use, and Access
   Who Gets Health Insurance?
      Medicaid Receipt
      Receipt of Other Health Insurance
Results from the NSHAPC Section on Physical Health Care
   Contact with a Doctor or Nurse for Health Reasons
   Ambulatory Care
   Emergency Room Care
   Hospitalization
   Inability to Obtain Care When Needed
   Summary

Chapter 3 Tables

Chapter 4: Single Homeless Adults
   Introduction
   Characteristics of Single Homeless Adults
   Access to and Use of Care
Results of Regression Analyses: Insurance, Use, and Access
   Predicting Insurance Receipt
      Medicaid Receipt
      Receipt of Other Health Insurance
Results from the NSHAPC Section on Physical Health Care
   Contact with a Doctor or Nurse for Health Reasons
   Ambulatory Care
   Emergency Room Care
   Hospitalization
   Inability to Obtain Care When Needed
   Summary
Treatment for Mental Health Problems
Treatment for Substance Use Problems
Implications

Chapter 4 Tables

Chapter 5: Conclusion
Chapter 5 Tables
   Does Having Insurance Help Homeless People Get Access to Care?
   Does it Matter What Type of Insurance it is?
   Does it Matter What Type of Care it is?
   Does it Matter Whether the Homeless Person is a Parent in a Homeless Family, a Child in a Homeless Family, or a Single Homeless Adult?
   How Do Chronic Physical Health Problems and Past-Year Problems with Alcohol, Drugs, and Mental Health Affect Use of Care?
   How Does Being Connected in One or More Ways to the Homeless Assistance System Affect Use of Care?
   How Does Being Connected to Major Public Benefit Programs Affect Use of Care?
Chapter 5 Tables (cont.)
   Conclusion

References


CHAPTER 1: INTRODUCTION

The Kaiser Commission on Medicaid and the Uninsured commissioned this report to learn more about the role that Medicaid plays in the ability of homeless people to get health care. This report examines the ability of homeless people to get the health care they need, and the factors that affect their receipt of care. Within the general rubric of health care we include treatment for physical health, mental health, and substance abuse problems.

Among the general population, research has long documented the relationship between having health insurance and receiving health care. Most recently, Fragale and Haley (2001) have documented this relationship for the housed low-income adult population, while Currie and Gruber (1996), Dubay and Kenney (2001), Kaestner (1999), and Newacheck et al. (1998) do the same for housed low-income children. Research on the effects of health insurance on access to care for homeless people is considerably sparser, but what little exists shows positive effects for adults (Glied et al., 1998, Kushel, Vittinghoff, and Haas, 2001, Padgett and Streuning, 1991, Padgett, Streuning, and Andrews, 1990) and children (Miller and Lin, 1988, Weinreb et al., 1998). While there is ample evidence at the national level of insurance effects for the housed low-income population, all studies of homeless populations have been at the local level except for Kushel, Vittinghoff, and Haas (2001), who used the National Survey of Homeless Assistance Providers and Clients (NSHAPC) as their source of data.

One can approach the basic research question of this project in (at least) two ways. First, one can seek to establish a relationship between having insurance and getting care, holding constant as many other potentially relevant factors as possible. This is the approach followed by Kushel, Vittinghoff, and Haas (2001). These researchers used the National Survey of Homeless Assistance Providers and Clients (NSHAPC), the only available national database with individual-level client information, to look at all homeless adults (that is, they did not differentiate by their family status). They also did not differentiate by type of insurance, grouping Medicaid, veterans health care, private insurance, and other insurance together in one dummy variable). They found that having health insurance significantly increased the likelihood of using ambulatory care, being hospitalized, being able to get medical care when needed, and being able to comply with medications regimes. However, having health insurance did not affect the likelihood of using emergency rooms. These results were obtained while holding constant many factors that might themselves be expected to affect need for and use of health care, including age, gender, race/ethnicity, veteran status, housing status, locale, number of comorbid chronic physical health conditions, and past-year alcohol, drug, and mental health problems.

The results obtained by Kushel, Vittinghoff, and Haas (2001) are important, showing as they do that insurance aids homeless people to get health care at least as much as it helps housed people. However, the analytic approach masks rather than reveals potentially significant interaction effects among the variables Kushel, Vittinghoff, and Haas held constant or did not include. These interactions may alter the association of having health insurance and getting health care. Equally important, they may explain how the association was created, and thus be relevant for policies seeking to increase health care access. In this report we explore the different routes to health care access that may depend on one's family status, connection to case managers or other aspects of program and service systems, and eligibility for cash benefits that entail eligibility for Medicaid. As part of this exploration we conduct separate analyses for homeless families and single individuals. Within families we look separately at insurance and access for parents and children. Among single individuals we look separately at the linkage between insurance and health care access for those with SSI and those without SSI. While almost 9 in 10 single homeless people with SSI have Medicaid, twice as many single homeless adults have Medicaid without having SSI. It is quite likely that their route into health care as well as their needs and patterns of service use differ. In all analyses, we treat having Medicaid, the primary health insurance for the homeless population, and having any other type of health insurance as two separate dummy variables, to see whether type of insurance makes a difference.

METHODS

The NSHAPC Survey

Our analyses are based on data from NSHAPC, a survey conducted by the U.S. Bureau of the Census in 1996 at the behest of the federal Interagency Council on the Homeless. The NSHAPC methodology has been described in detail elsewhere (Burt et al., 1999, Chapter 2 and Appendixes A - D). Briefly, the survey was nationally representative of people using homeless assistance programs in fall 1996. It was conducted in 76 primary sampling units (PSUs). The first 28 were the 28 largest metropolitan statistical areas (MSAs) in the United States. The next 24 were MSAs randomly selected from 12 strata defined by geographical region and MSA size. The final 24 were randomly selected groups of counties outside of MSAs (i.e., rural areas), defined by the catchment areas of community action agencies (which completely cover the non- MSA parts of the United States). All homeless assistance programs in these MSAs falling into one of 16 types were catalogued and assessed for size. A sample of programs was selected from this list proportional to program size, and to assure representation of each PSU and of all types of programs within the PSU. Clients were recruited for interviewing at selected programs, using a fixed skip interval after a random start. Five to eight clients completed interviews during each of about 700 program visits, with a response rate of 96 percent. Clients were paid $10.00 once they completed the interview. The final analytic sample included 4,133 respondents, of whom 2,938 were currently homeless, 677 were formerly homeless, and 518 were other service users who reported never having been homeless. In this report we focus on parents and children in currently homeless families (N = 465), and currently homeless single people (N = 2,473).

Descriptive and Independent Variables

Gender was recorded by observation. Respondents reported both their age in years and their date of birth, which was used to calculate age if age was not reported. They reported their race and Hispanic origin on two standard Census Bureau questions (race and Hispanic origin were asked separately). We combined responses to create the categories of non-Hispanic white, non-Hispanic black, Hispanic, American Indian/Native American, Asian/Pacific Islander, and other. For the present analysis we collapsed the last four categories into "other" because sample sizes did not make it feasible to use all categories. Location was determined by the site from which the respondent was recruited for NSHAPC. Veteran status was determined by a question asking whether the respondent had ever been on active-duty military service in the U.S. Armed Forces, Reserves, or National Guard.

Family Status. Respondents were classified as head of a homeless family if they reported that one or more of their own minor children were homeless with them. Otherwise respondents were classified as single, even if they had children residing elsewhere. Respondents were asked the number and ages of minor children who were homeless with them. Number of children was coded as 1, 2, or 3 or more children homeless with the respondent; Age of youngest child was classified as under 2, 2 to 5, and 6 or older.

Defining Alcohol, Drug, and Mental Health (ADM) Status. Individuals are classified as having an ADM problem if they have had at least one alcohol use, drug use, or mental health problem during the past year. Presence of each problem is defined completely on the basis of self-report, without verification by any official records.

Clients are classified as having a past year alcohol use problem if any of the following conditions were met: (1) they scored 0.17 or higher on a modified Addiction Severity Index1 (ASI) measure, (2) they reported drinking to get drunk three or more times a week within the past year, (3) they reported being treated for alcohol abuse within the past year, or (4) they reported ever having been treated for alcohol abuse and drinking three or more times a week within the past year.

Clients are classified as having a past year drug use problem if any of the following conditions were met: (1) they scored 0.10 or higher on a modified ASI measure, (2) they reported being treated for drug abuse within the past year, (3) they reported using drugs intravenously (shooting up), or (4) they reported using any of a variety of specific drugs three or more times a week within the past year.

Clients are classified as having a past year mental health problem if any of the following conditions were met: (1) they scored 0.25 or higher on a modified Addiction Severity Index (ASI) measure, (2) they reported receiving treatment or counseling or being hospitalized for emotional or mental problems within the past year, (3) they reported on the ASI taking prescribed medications for psychological or emotional problems within the past year, (4) they reported that a mental health condition was the single most important thing keeping them from getting out of homelessness, or (5) they reported receiving treatment or counseling or being hospitalized for emotional or mental problems at some point in their lives and having one or more of the ASI's seven emotional or psychological conditions within the past year.2

System Involvement. Respondents were asked for all sources of cash income, and also whether they were receiving a variety of public benefits at the time of the interview. Responses indicating receipt of AFDC, GA, Supplemental Security Income (SSI), food stamps, or a housing subsidy form the basis of variables indicating receipt of specific benefits or combinations of benefits. All variables are based on self-report, without verification from program records. A series of questions asked whether the respondent had received help getting a variety of things within the past 30 days. For the variable "received case manager (CM) help in the past month," we selected items most indicative of serious involvement with a case manager or other persistent staff position, including: assistance with getting financial or other public benefits, help finding a job, assistance with finding affordable housing, and assistance with rent, mortgage, or utilities for securing permanent housing. The variable was scored 1 if the respondent reported any of these types of assistance, and 0 if none were reported.

Health Insurance. Respondents were asked whether they had health insurance and, if they had children with them, whether their children had health insurance. If the answer was yes, they were asked what kind, with the options being Medicaid, VA medical care, private insurance, and "other." "Other" responses were examined and recoded into one of the primary categories if it was clear where they belonged. Respondents could identify more than one type of insurance, and some did. For this report we created dummy variables for Medicaid, any other insurance, and no insurance.

Length of Current Homeless Spell. Respondents' answers, which might be given in years, months, weeks, or days, were converted to consistent 30-day months or portions of months.

Where Slept in Last Seven Days Plus Day of Interview. Respondents were asked in detail where they had slept or rested during the seven days preceding their interview, and also where they were staying "today." Responses for "today" were checked against the type of program from which they were recruited (their sampling frame). Responses plus frame information was used to create variables indicating whether they had slept or rested in emergency shelters during the eight days in question, in homeless assistance programs offering transitional housing, and/or in places not meant for human habitation. Respondents could have slept in one, two, or all three of these venues during the time period, and often did. Thus each location is treated as its own dummy variable.

Current Health Conditions. A number of health conditions were included in the NSHAPC questionnaire. These were selected because other research indicated their frequency of occurrence among homeless populations. Thus conditions such as skin disease, infection, sores, and ulcers, and lice and scabies were included and some other conditions were not. Acute infectious conditions on the questionnaire included chest infection, cold, cough, bronchitis (URI), pneumonia, tuberculosis, and sexually transmitted diseases other than AIDS. Acute noninfectious conditions included the skin conditions and infestations just mentioned. Chronic conditions included diabetes, anemia, high blood pressure, heart disease or stroke, problems with one's liver, arthritis/rheumatism/joint problems, cancer, problems walking/lost limb/other handicap, and AIDS. In this report acute infectious conditions are represented by a dummy variable indicating the presence of any such condition. Acute noninfectious conditions are also represented by a dummy variable signifying presence or absence. Chronic conditions are represented by an ordinal variable with the values of none, one, two, and three or more conditions.

Dependent Variables

Physical Health Care. The respondent reported the last time she or he was treated or examined by a doctor or nurse for health problems, including routine checkups. If the respondent had children with her/him, the question was repeated with respect to the children. If a health care visit was reported within the past 12 months, the variable "received physical health care" was coded 1; otherwise it was coded 0.

Physical Health Care Setting. Respondents were asked where they (their children) received care, both during the most recent visit and for any care received within the past 12 months. Every setting in which they received care was recorded. Ambulatory care settings included a hospital outpatient clinic; a Veteran Affairs (VA) outpatient clinic; a community health clinic; a Health Care for the Homeless clinic; a doctor or nurse in a shelter, soup kitchen, or other program; a private doctor's office, or a migrant health care facility. Inpatient settings included a VA hospital as an inpatient or any other hospital as an inpatient. The emergency room setting was coded if the respondent said the care was received at a hospital emergency room.

Unmet Need. Responses to the question "Have you needed to see a doctor/nurse in the last year but were not able?" were coded as a dummy variable for which a yes response coded 1 and a no response was coded 0. A similarly worded question asked about unmet need for the respondent's children, and was likewise coded as 1 if the children had needed a doctor/nurse but been unable to see one, and as 0 if this had not happened within the last year. Respondents with no children were not asked the question, and are not included in the analyses of children's access to health care.

Mental Health Care. The respondent reported whether he or she had "ever received outpatient treatment or counseling for emotional or mental problems." If the answer was yes, the most recent treatment episode was recorded. A dummy variable was created for which an episode within the past 12 months was coded 1and anything else was coded 0. For inpatient treatment the question read "have you ever been HOSPITALIZED for emotional or mental problems?" and was followed by questions about frequency and recency if the answer was yes. A dummy variable was created for which a hospitalization within the past 12 months was coded 1 and anything else was coded 0. A summary variable for any mental health treatment was created to reflect outpatient and/or inpatient treatment within the past year. Anyone who had not been classified as having a mental health problem was coded as "missing" for these three dummy variables. No questions were asked about children's receipt of mental health care.

Substance Abuse Treatment. The respondent reported whether he or she had "ever received OUTPATIENT treatment for problems with alcohol (drugs)." If the answer was yes, the most recent treatment episode was recorded. A dummy variable was created for which an outpatient episode within the past 12 months was coded 1and anything else was coded 0. For inpatient treatment the question read "have you ever received INPATIENT treatment (including detox) for problems with alcohol (drugs)?" and was followed by questions about frequency and recency if the answer was yes. A dummy variable was created for which inpatient treatment within the past 12 months was coded 1 and anything else was coded 0. A summary variable for any substance abuse treatment was created to reflect outpatient and/or inpatient treatment within the past year. Anyone who had not been classified as having an alcohol or drug problem was coded as "missing" for these three dummy variables. No questions were asked about children's receipt of care related to drug or alcohol use.

OVERVIEW OF REST OF REPORT

The next three chapters of this report present results related to access to and use of treatment for physical and mental health and substance use problems. They follow a general pattern. They first describe the subsample of interest on personal and other factors that may affect access to and receipt of care, both as a whole and in subgroups reflecting benefit receipt (AFDC or GA for families, and SSI for singles). They then present descriptive information about receipt of physical health care and mental health and substance abuse treatment. They conclude with regression analyses examining factors affecting receipt of care. Chapter 2 focuses on the responding parent in homeless families. Bivariate and regression results are reported separately for households that did and did not receive AFDC or GA (62 percent received one or the other, among whom 93 percent reported receiving Medicaid). Chapter 3 focuses on children homeless with their parents, also splitting the sample by the parent's AFDC/GA status. Chapter 4 reports findings for single homeless adults. Bivariate and regression results are reported separately for people who did and did not report receiving SSI (11 percent did get SSI). Nine out of ten of these individuals also reported receiving Medicaid. The report ends with a chapter summarizing results and discussing their implications. It will describe the clear associations that have emerged, the factors that seem to be important in assuring access to health care, and what might be done to improve access.

This report is available in its entirety in the Portable Document Format (PDF), which many find convenient when printing.


1 The Addiction Severity Index is an instrument developed by the National Institute on Drug Abuse (Fureman, Parikh, Bragg, and McLellan, 1990). It contains subscales to measure a client's level of problems with alcohol, with drugs, and with mental or emotional problems. Cutoff levels used in this report are slight modifications of the means reported in Zanis, McLellan, Cnaan, and Randall (1994).

2 The eighth ASI item, "taking prescribed medications for psychological or emotional problems," is a criterion in its own right for classifying a client as having a mental health problem.


Topics/Tags: | Health/Healthcare | Housing


Usage and reprints: Most publications may be downloaded free of charge from the web site and may be used and copies made for research, academic, policy or other non-commercial purposes. Proper attribution is required. Posting UI research papers on other websites is permitted subject to prior approval from the Urban Institute—contact [email protected].

If you are unable to access or print the PDF document please contact us or call the Publications Office at (202) 261-5687.

Disclaimer: The nonpartisan Urban Institute publishes studies, reports, and books on timely topics worthy of public consideration. The views expressed are those of the authors and should not be attributed to the Urban Institute, its trustees, or its funders. Copyright of the written materials contained within the Urban Institute website is owned or controlled by the Urban Institute.

Email this Page