Measures
Client characteristics. Client characteristics included sociodemographic characteristics (age, race, marital status, education, employment, and income), living arrangement, community adjustment, and clinical status (mental and physical health). Clients' responses to two items asking the number of nights spent during the previous 60 days in various living arrangements and the presence of others in the home were crossed to create a five-level categorical living arrangement variable: alone, in own place with family, in own place with non-kin, in someone else's place, and in a halfway house.
Clinical status items included psychiatric diagnoses, symptoms, medication, lifetime psychiatric hospitalization, substance abuse, and physical health. Primary psychiatric diagnoses were abstracted from clients' charts. The intensity of psychiatric symptoms was rated by trained interviewers using the Brief Psychiatric Rating Scale (BPRS) (
20). Subjective distress was measured with 33 items from the SCL-90 (
21). Use of alcohol and illicit drugs was assessed by using composite indexes from the Addiction Severity Index (ASI) (
22). Clients rated their physical health on a 5-point scale (
19) and identified the total number of physical health problems from a list of 13 conditions, with responses including 0, no problem, 1, had problem but received no treatment, and 2, had problem and received treatment. Thus the medical problems scale ranged from 0 to 26 points.
Adjustment to community living was measured by the size of social support networks and by lifetime incarceration. Clients were asked how many people they felt close to in each of nine relationship categories—for example, parents, siblings, friends, and health care providers. A continuous social support variable was computed by summing the number of persons in each of these nine relationship categories, indicating the total number of persons to whom the client felt close.
Neighborhood characteristics. Population and housing characteristics of the 233 census-tract block-group neighborhoods in which clients resided at the time of the interview were downloaded from the Census 2000 Web site (www.census.gov). On the basis of these data, ten neighborhood variables were obtained: median household income, percentage of persons with incomes above the poverty level, home ownership rate, median gross rent (contract rent plus cost of utilities), percentage of one-bedroom rental units with a gross monthly rent of at least $500, percentage of rental units with gross rent below 20 percent of household income, percentage of household income spent on gross rent, and percentage of housing units with complete plumbing facilities, kitchen facilities, and both plumbing and kitchen facilities.
A factor analysis of these ten variables (varimax rotation) produced a three-component solution in which nine of ten factors had loading scores of at least .75. These three components were composed by neighborhood income (four of the ten variables), plumbing and kitchen facilities (three variables), and affordability (two variables). Standardized scores for the items in each of these factors were averaged for use in multivariate regression analyses.
Subjective housing quality. An overall measure of subjective housing quality was constructed by using all 24 housing quality items and six subscales representing four housing quality dimensions (space and privacy, physical maintenance, affordability, and relationship with landlord) and two neighborhood quality dimensions (safety and cleanliness, and convenience). These measures were derived from two housing quality scales—one documenting the extent to which various types of problems were present in the residence and neighborhood (12 items, ranging from 0, no problem, to 2, a big problem) (
23) and another indicating the presence of various positive housing and neighborhood attributes (16 dichotomous items coded as 1, not present, or 2, present) (
18). A more detailed description of the six measures of subjective housing quality is available from the authors on request.