ASSOCIATION BETWEEN TIME HOMELESS AND PERCEIVED HEALTH STATUS AMONG THE HOMELESS IN SAN FRANCISCO

Transcription

1 Journal of Community Health, Vol. 22, No. 4, August 1997 ASSOCIATION BETWEEN TIME HOMELESS AND PERCEIVED HEALTH STATUS AMONG THE HOMELESS IN SAN FRANCISCO Mary Castle White, MPH, PhD; Jacqueline Peterson Tulsky, MD; Carol Dawson, MS; Andrew R. Zolopa, MD; and Andrew R. Moss, PhD ABSTRACT: The purpose of this study was to describe the perceived health of the homeless, and to measure the effect of time homeless on perceived health status, after controlling for sociodemographic characteristics and health conditions. The design was cross-sectional; the population was a representative sample of homeless in San Francisco, interviewed on health issues. Analysis of predictors of poor or fair health status was by logistic regression. In this sample of 2780 persons, 37.4% reported that their health status was poor or fair as compared to good or excellent. Reporting poor or fair health status was significantly associated with time homeless, after controlling for sociodemographic variables and health problems including results from screening for HIV and TB (OR= 1.49, 95%CI ). Comparisons with data from the National Health Interview Survey (NHIS) showed poorer health status among the homeless persons in this study. Standardized morbidity ratios were highest for asthma; there was twice the number of homeless persons reporting asthma, in younger as well as older adults, as would be expected using NHIS rates. There was also an excess of arthritis, high blood pressure and diabetes in those as compared to adults in the Health Interview Survey. The time spent homeless remains associated with self-reported health status, after known contributors to poor health are controlled. Persons who have been homeless for longer periods of time may be the persons to whom health care interventions should be aimed. Mary Castle White is Associate Professor, Department of Community Health Systems, School of Nursing, University of California, San Francisco, CA; Jacqueline Peterson Tulsky is Assistant Professor, Department of Medicine, University of California, San Francisco, CA; Carol Dawson is a Doctoral student, Department of Community Health Systems, School of Nursing, University of California, San Francisco, CA; Andrew R. Zolopa is Assistant Professor of Medicine, Division of Infectious Diseases and Geographic Medicine at Stanford University Medical Center, Stanford, CA; and Andrew R. Moss is Professor in Residence, Department of Epidemiology and Biostatistics, University of California, San Francisco, CA. Requests for reprints should be addressed to Mary Castle White, MPH, PhD, Department of Community Health Systems, Box 0608, School of Nursing, University of California, San Francisco, CA This study was supported by grant R01-DA from the National Institute on Drug Abuse and a grant from the State of California, University-wide AIDS Research Program. This paper was presented at the meeting of the American Public Health Association, San Diego, California, November, Human Sciences Press, Inc. 271

2 272 JOURNAL OF COMMUNITY HEALTH INTRODUCTION In recent years, there have been a number of studies undertaken to describe the homeless. Difficulties in studying this population have been in agreement on a definition of homelessness and methodologic limitations in sampling, data collection from nonrepresentative samples, and small sample sizes.1 These reports have shown both the heterogeneity of characteristics and a wide range of health conditions in this population.1-6 A recent report has documented high age-specific mortality rates among the homeless as compared to non-homeless,7 and researchers have suggested that there is increasing morbidity with time homeless.1 Measures of well-being have been shown to predict health expenditures and mortality,8 and measures of health-related quality of life have been used to evaluate the outcomes of interventions and the need for health services.9 Perceived health status, measured by overall rating of current health in general, is part of the Medical Outcomes Study Short-Form General Health Survey, and has been tested for its validity and reliability in a number of populations, including general populations and those with chronic health conditions.8,10,11 The purpose of this study was to describe the self-reported health of a large, representative sample of the homeless in San Francisco, and to answer the research question: Is time homeless associated with perceived health status? We examined likely characteristics, health conditions and behaviors that might explain perceived health status in the homeless. After controlling for such explanatory variables, we sought to measure the impact of homelessness itself, as measured by time homeless, on self-reported health status. METHODS This study is a secondary analysis of data from a large, cross-sectional study of the homeless in San Francisco, conducted in in order to examine the prevalence of tuberculosis and HIV infection in this population. A report of this analysis has appeared elsewhere.12 Sample A representative cross-sectional sample of homeless persons in San Francisco were found, interviewed and examined, including tuberculosis

3 M.C. White, J.P. Tulsky, C. Dawson, A.R. Zolopa, and A.R. Moss 273 and HIV testing, between 1990 and Persons were located by quota sampling from all major facilities (shelters and meal lines) serving the homeless. The sampling methodology of Burnam and Koegel was adapted for use in this study, and this method was described in an earlier paper.13 Prior to study initiation, the sampling method was tested to determine its validity in providing a sample that was broadly representative of the homeless in San Francisco.12 Interview The 300-item English or Spanish interview, conducted by trained interviewers following informed consent for participation, included a number of demographic, health condition, and health behavior questions. Demographic information asked and used in this analysis included date of birth, gender, years of education completed, race or ethnic group, military service, whether they had been jailed, and length of homelessness. General health status was reported by self-ranking on a scale (poor, fair, good, excellent). Health conditions were reported as present or absent by the questions "Which of the following health problems has a doctor or other health care professional told you that you now have or are under treatment for now?", for the following list: arthritis, heart disease, heart attack, high blood pressure, diabetes, cancer, emphysema/chronic bronchitis, asthma, tuberculosis, cirrhosis, fatty liver, hepatitis, kidney disease, stroke, and seizure disorder. Mental health problems were assessed by answers to the question "Have you ever stayed overnight or longer in a hospital or treatment program for emotional or mental problems?" Current alcohol abuse was defined by the CAGE criteria14 and a self-report of having had a drink in the last 30 days. Current crack cocaine use and injection drug use were each defined as use in the last 30 days. Smoking status was determined by the question "Do you smoke cigarettes?" Time homeless was defined as total time ever homeless. Data Analysis Sociodemographic data were tabulated to describe the characteristics of the sample overall and with respect to self-reported health status. Frequencies of individual health conditions were tabulated and examined, and then a single variable for the presence of any physical condition was made from the list of conditions asked. Variables were examined, by Student's t test or by Chi Square test, to determine their relationship to selfreported health status (poor or fair as compared to good or excellent).

4 274 JOURNAL OF COMMUNITY HEALTH Multiple logistic regression, including variables significant in the bivariate analyses, was done to indicate the variables independently associated with a self-rating of poor or fair health. Alpha was set at 0.10 for identification of variables to include in the multiple logistic regression, and then set at 0.05 to assess statistical significance of logistic regression results. SAS software was used to perform the analyses. Some of the conditions reported by the homeless in this sample could be compared to normative US population data from the National Health Interview Survey (NHIS), a survey of noninstitutionalized persons in the US population.15 Age-specific Standardized Morbidity Ratios (SMRs) were computed for questions that were the same for both surveys, with the following exception: for bronchitis, the homeless survey included a single question about chronic bronchitis or emphysema, whereas in the NHIS these diseases were separated. In the NHIS the rates for bronchitis were higher than those for emphysema, and therefore bronchitis rates were used. Standard methodology was used to calculate SMRs16 and 95% confidence intervals." Characteristics of the Sample RESULTS The study sample included 2780 individuals, 2135 (76.8%) of whom were men and 645 (23.2%) women. By ethnicity, 50.4% were Black, 34.8% were White, 7.6% were Hispanic, and the remaining 7.2% were categorized as "other", including additional ethnic categories or mixed ethnicity. The average age was 38 years, with men averaging one year older than women. The mean number of years of education reported by the subjects was 12.2, and this did not differ by gender. Over half of the subjects (62.8%) had spent time in jail, and of these, the mean time spent in jail for men was 18.7 months, and for women 7.3 months. Men reported 29 months of total time ever homeless, on average, and women reported a mean of 21 months; in both groups, however, the distribution of time homeless was was skewed left with a median of 12 months. Health Conditions, Behaviors and Health Status Of the total sample, 37.4% reported that their health status was poor or fair, while 62.6% reported that their health was good or excellent. Sociodemographic and health variables were examined separately for their associations with reported health status. Sociodemographic variables in addition to the variable of interest, time homeless, are shown in Table 1.

5 M.C. White, J.P. Tulsky, C. Dawson, A.R. Zolopa, and A.R. Moss 275 TABLE 1 Sociodemographic Characteristics of Homeless Persons in San Francisco by Self-Reported Health Status (Poor or Fair as Compared to Good or Excellent) San Francisco, by Bivariate Analyses Characteristic Gender female male Age < >50 Race/ethnicity White Black Hispanic Other Education > 12 years < 12 years Time homeless < 1 year > 1 year Ever been jailed no yes Total Persons Reporting Poor or Fair Health # (%) OR 95 %C.I. P 276 (42.8) 763 (35.7) 205 (32.3) 694 (37.1) 136 (51.3) 392 (40.5) 497 (35.5) 79 (37.6) 71 (35.2) 684 (33.8) 353 (47.2) 386 (33.1) 578 (43.6) (33.8) 690 (39.5) Ever been in military service (males only) no (35.9) yes (35.5) Time homeless, examined as a continuous variable, was significantly associated with reported poor or fair health status; however, in quartile analysis of the variable18 the observations suggested a nonlinear relationship; for this reason, subsequent analyses were done with time homeless grouped

6 276 JOURNAL OF COMMUNITY HEALTH into a dichotomous variable at the median, <1 year and >l year. Persons reporting their ethnicity as other than White were less likely to report poor or fair health status as compared to Whites, as were those who had been jailed. The proportion of men reporting poor or fair health status was virtually the same regardless of prior military service. Health conditions and behaviors were also examined for their relationship to self-reported health status (Table 2). Over a third (36,3%) reported at least one of the health conditions asked in the survey. Arthritis was the most frequent single condition, reported by 12.4% (10.6% of men and 18.3% of women). For health related behaviors, the majority of subjects (61.2%) reported either crack cocaine, injection drugs, or alcohol abuse, and of these individually, current alcohol abuse was the most fre- TABLE 2 Health Conditions and Behaviors Among Homeless Persons by Self- Reported Health Status (Poor or Fair as Compared to Good or Excellent), San Francisco, by Logistic Regression Analysis Characteristic Physical condition* Mental health problem Positive HIV status Injection drug use Alcohol abuse Cigarette smoking Crack use Persons Reporting Characteristic (% of Total Answering the question) 1008 (36.3) 624 (22.6) 245 (8.8) 378 (13.9) 1108 (41.1) 2194 (79.2) 980 (35.4) Persons Reporting Poor or Fair Health (% of Those with char.) 561 (55.7) 306 (49.0) 117 (47.8) 183 (48.4) 477 (43.1) 848 (38.7) 397 (40.5) OR %C.I P *includes any of the following: arthritis, heart disease, heart attack, high blood pressure, diabetes, cancer, emphysema/chronic bronchitis, asthma, tuberculosis, cirrhosis, fatty liver, hepatitis, kidney disease, stroke, or seizure disorder

7 M.C. White, J.P. Tulsky, C. Dawson, A.R. Zolopa, and A.R. Moss 277 quently reported, by 41.1% (45.1% for men and 27.9% for women). Over three quarters (79.2%) of the sample were current smokers. In a logistic regression to examine health-related conditions and behaviors associated with self-reported health status, the presence of a physical condition, as expected, was strongest (Table 2). Controlling for the presence of physical conditions, mental health problems and HIV, those who reported current injection drug use and those who were current alcohol abusers were more likely to report poor or fair health status (OR 1.45 and 1.40, respectively), while current crack use or cigarette smoking were not significantly associated with self-reported health status. Sociodemographic and health-related variables (conditions and behaviors) that were significantly associated with self-reported health status were entered into the final logistic regression model with the variable of interest, time homeless (Table 3). Physical and mental health conditions, positive HIV status and older age were most strongly associated with reporting poor or fair health status (Table 3), as might be expected. Other variables that might explain poor self-reported health status remained important in the multiple regression model, including alcohol abuse and injection drug use and less than a high school education. Controlling for these variables, time homeless (one year or longer) remained indepen- TABLE 3 Characteristics Significantly Associated with Self-Reported Health Status (Poor or Fair as Compared to Good or Excellent) Among Homeless Persons, San Francisco, by Logistic Regression Analysis Characteristic Physical condition* Positive HIV status Age (> 50 vs. < 30) Mental health problem Time homeless (> 1 yr vs. < 1 yr) Education (< 12 yrs vs. > 12 yrs) Injection drug use Alcohol abuse Gender (female vs. male) OR , % C.I P *includes any of the following: arthritis, heart disease, heart attack, high blood pressure, diabetes, cancer, emphysema/chronic bronchitis, asthma, tuberculosis, cirrhosis, fatty liver, hepatitis, kidney disease, stroke, or seizure disorder

8 278 JOURNAL OF COMMUNITY HEALTH dently associated with poorer self-reported health (OR 1.49, 95% CI , p = ). Comparisons with the US Population Comparisons with data from the NHIS showed that the homeless persons in this study differed considerably from US norms (Table 4). In particular, there was twice the number of persons reporting asthma, in younger as well as older adults. There was also an excess of arthritis, high blood pressure and diabetes in those. TABLE 4 Rate of Self-Reported Health Conditions from the National Health Interview Survey, 1990 and Excess Morbidity Among Homeless Persons, San Francisco, Condition Asthma Arthritis High blood pressure Diabetes Bronchitis/emphysema Heart disease Kidney disease NHIS Rate/ SMR* % CI *SMR (standardized morbidity ratio) is calculated as the observed number among homeless persons divided by the number expected in the homeless using NHIS rates.

9 M.C. White, J.P. Tulsky, C. Dawson, A.R. Zolopa, and A.R. Moss 279 DISCUSSION Homelessness is associated with higher mortality rates', and perceived health status has been linked to morbidity and mortality.1-6 Our study asked the question of the association between time homeless and perceived health status in a representative sample of homeless persons in San Francisco. We examined likely contributors to perceived poor health, such as sociodemographic variables and health conditions and behaviors. As might be expected, those homeless who were older or who indicated a physical or mental health condition were more likely to report only poor or fair health status. In addition, female gender and less education were associated with poor or fair health status. While controlling for the effects of these likely contributors to poor health status, however, time homeless remained significantly associated with poor or fair perceived health status. Explanations for this finding may be that longer time homeless causes or exacerbates health problems; that healthy persons do not remain homeless for long, therefore those remaining homeless tend to be in poorer health; or that the finding is spurious and other variables account for poor or fair health status. These data are cross-sectional, and therefore associations can not be causally linked. Limitations of this study include the potential biases in self-report of health status, health conditions, and behaviors. Data from homeless persons, however, have been shown to be as reliable as that from subjects in other settings for comparisons of perceived health status.19,20 Another important issue, especially in characterizing the homeless, is the sampling methodology and the representativeness of the sample selected for study. The methods used in this study are thought to be successful in capturing a representative sample of the homeless in San Francisco, and the sociodemographic characteristics did not differ substantially from those found by others in studies of this population.1,3,5,21 Health Status Over a third of the homeless in our cross-sectional study reported poor or fair health status. By contrast, Gelberg and her colleagues found that 21% of persons in a medical clinic serving the homeless reported their health as fair or poor, compared to 13% among those who were seen in the clinic but who were not homeless.22 Differences in their findings from ours may be because their population did not include those homeless persons who do not have access to or do not seek health care. The work of Ropers and Boyer23 and of Robertson and Cousineau3 demonstrated that health status was rated as fair or poor in over one third of homeless surveyed, and that over half had no regular source of care.3

10 280 JOURNAL OF COMMUNITY HEALTH Health Conditions and Behaviors The homeless in San Francisco have a number of chronic conditions which may account for poor self-reported health status. At least one physical condition was reported by 30% of the sample; comparisons with other research are difficult because of different conditions surveyed and different methodologies used, such as interview versus physical exam. Alcohol abuse was diagnosed in 68.1% and 31.6% of men and women, respectively, in Baltimore homeless, 5 which is higher than our findings of 45.1% for men and 27.9% for women. This may have been due to a difference in definitions or in the sampling that occurred in the jails in the Baltimore study, 5 which may have selected for those with alcohol problems. Our definition of alcohol abuse was having had a drink in the last 30 days, along with positive criteria as defined by the CAGE measures of alcohol use. 14 When we examined those who fulfilled the CAGE criteria alone, the proportions were higher and resembled those of Breakey. 5 Comparisons with the US Population Comparisons with national data indicate that the homeless persons in this study considered their health to be poorer than did persons in the general population. Data from the Behavioral Risk Factor Surveillance System (BRFSS), a continuous, random-digit-dialed telephone survey of noninstitutionalized persons over age 18, indicated that 13.4% of all persons and 13.0 % of Californians rated their health as poor or fair in 1993, 24 as compared to 37.4% of the persons in our sample. In data from the National Health Interview Survey in 1990, between 4.1% in those years old and 27.6% in those age 65 and older characterized their health as fair or poor. 15 In the poorest family income stratum, under $10,000, the range of proportions was closer to that seen in this study, from 5.9% in those to 38.7% in those age 65 and older reporting poor or fair health status. Excess self-reported morbidity compared to the US general population was seen for asthma, for which hospitalization rates have been inversely correlated with income and education. 25 Excess morbidity, in particular among younger adults, was seen for arthritis, which may be a reflection of living situations and exposure to the elements, and that seen for bronchitis in older adults may be a result of high smoking rates in this population. Lower self-reported morbidity from heart disease in the homeless as compared to the general population may reflect lower diagnostic rates because of poorer access to care rather than lower rates of heart disease, or because of earlier deaths selecting for healthier individuals. These compar-

11 M.C. White, J.P. Tulsky, C. Dawson, A.R. Zolopa, and A.R. Moss 281 isons should be viewed with caution, however, because the data were only adjusted for age; differences in the gender and income distributions in the homeless as compared to a national sample may have resulted in biased estimates of SMRs. Further work on specific diseases found among the homeless is needed. CONCLUSIONS This study reports on the characteristics and health status of the homeless, using a large sample representative of the homeless in San Francisco. Homeless adults in this study characterized their health as poorer and reported more morbidity than did other populations surveyed in the US. The time spent homeless remained associated with self-reported health status, after known contributors to poor health were controlled. Implications of this study are in targeting interventions for this diverse group. Documenting time homeless may be an effective screening question to identify subgroups of homeless who are at high risk of morbidity and mortality. REFERENCES 1. Winkelby MA. Comparison of risk factors for ill health in a sample of homeless and nonhomeless poor. Public Health Rep. 1990; 105: Roth D, Bean J. New perspectives on homelessness: Findings from a statewide epidemiological study. Hosp Commun Psychiatry. 1986; 37: Robertson MJ, Cousineau MR. Health status and access to health services among the urban homeless. Amer J Public Health. 1986; 76: Linn LS, Gelberg L. Priority of basic needs among homeless adults. Psych Epidemiol 1989; 32: Breakey WR, Fischer PJ, Kramer M, et al. Health and mental problems of homeless men and women in Baltimore. J Amer Med Assoc 1989; 262: Stephens D, Dennis E, Toomer M, HollowayJ. The diversity of case management needs for the care of homeless persons. Public Health Rep. 1991; 106: Hibbs JR et al. Mortality in a cohort of homeless adults in Philadelphia. New EnglJ Med 1994; 331: Stewart AL, Greenfield S, Hays RD, Wells K, et al. Functional status and well-being of patients with chronic conditions. J Amer Med Assoc 1989; 262: Hennessy CH, Moriary DG, Zack MM, Scherr PA, Brackbill R. Measuring health-related quality of life for public health surveillance. Public Health Rep 1994; 109: Stewart AL, Hays RD, Ware JE. The MOS short-form general health survey: reliability and validity in a patient population. Med Care. 1988;26: Ware JE, Sherbourne CD, Davies AR, Stewart AL. The MOS Shortform General Health Survey: Development and Test in a General Population. Santa Monica, CA: The Rand Corp; Publication P Zolopa AR, Hahn JA, Gorter R, et al. HIV and tuberculosis infection in San Francisco's homeless adults: Prevalence and risk factors in a representative sample. J Amer Med Assoc 1994; 272: Burnham MA, Koegel P. Methodology for obtaining a representative sample of homeless persons: The Los Angeles Skid Row Study. Eval Rev 1988; 12:

12 282 JOURNAL OF COMMUNITY HEALTH 14. Ewing JA. Detecting alcoholism: the CAGE questionnaire. J Amer Med Assoc 1984; 252: National Center for Health Statistics (US DHHS). Design and estimation for the NHIS, Vital and Health Statistics Series 2, no. 110, Hennekins CH, Buring JE. Epidemiology in Medicine. Boston: Little, Brown & Co., Gardner MJ, Altman DG. Statistics with Confidence. London; British Medical Journal, Hosmer DW, Lemeshow SL. Applied Logistic Regression. New York; Wiley & Sons, Annis HM. Self-report reliability of skid-row alcoholics. BrJ Psychiatiy. 1979;134: Shanks NJ. Consistency of data collected from inmates of a common lodging house. J Epidemiol Community Health. 1981;35: Ritchey FJ, LaGory M, Mullis J. Gender differences in health risks and physical symptoms among the homeless. J Health Soc Beh 1991; 32: Gelberg L, Linn LS, Usatine RP, Smith MH. Health, homelessness, and poverty. Arch Intern Med 1990; 150: Ropers RH, Boyer R. Perceived health status among the new urban homeless. Soc Sci Med 1987; 24: Health-related quality of life measuresunited States, Morbid Mortal Weekly Rep 1995; 44: Gottlieb DJ, Beiser AS, O'Connor GT. Poverty, race, and medication use are correlates of asthma hospitalization rates. Chest 1995; 108:28-35.