CSAT Publications -- TAP 25, The Impact of Substance Abuse Treatment on Employment Outcomes Among AFDC Clients in Washington State

Chapter II - Methods, Data, and Analytic Procedures

a. Sample Selection Criteria and Description of the Study Population

b. Treatment Modalities Analyzed and Assignment of Clients to Treatment and Comparison groups

c. Data Sources and Construction of Analytic Files

d. Measures

e. Descriptive and Statistical Analytical Procedures

The study sample consisted of 5,664 individuals who (1) were admitted to substance abuse treatment in Washington State from July 1994 to June 1996; and (2) were receiving financial assistance from AFDC at the time of treatment admission.
The client treatment database maintained by the Washington State Division of Alcohol and Substance Abuse was used to identify all persons admitted to substance abuse treatment from July 1994 through June 1996. Substance abuse treatment units serving publicly funded (Federal or State) clients in Washington State are required to report to this database, and approximately half of all treatment programs do so. Once individuals admitted to substance abuse treatment from July 1994 to June 1996 were identified, the database was searched again to identify individuals who were receiving financial support from AFDC at time of their admission to treatment.

This two-step procedure resulted in the initial identification of 6,140 individuals in Washington State, admitted to substance abuse treatment from July 1994 through June 1996 and receiving financial assistance from AFDC at the time of their admission. Subsequently, 476 (7.7 percent) of these 6,140 clients were dropped from the study because either (1) they had a dual diagnosis (mental illness and substance abuse); or (2) data on key variables (e.g., discharge date or treatment modality) were missing for the individuals.

The final study population, therefore, included 5,664 individuals admitted to substance abuse treatment from July 1994 through June 1996 who were receiving financial aid from AFDC at the time of their admission. Selected characteristics of the study population are shown in Table II-1.

Table II-1. Selected Characteristics of the AFDC Study Population in Washington State (N = 5,664)*
Characteristics	Percent of total study population or mean  (Standard deviation)
Children in household
0 - 1	55.6%
2 - 3	37.9%
$ 4	6.4%
Self-reported household monthly income (including AFDC income)		$474 ($366)
Age at admission	31.1 (7.0)
Race
White	72.2%
Black	12.3%
Native American	8.1%
Asian and other (includes a small percentage of Hispanic clients)	7.3%
Female	75.7%
Single, divorced, widowed, or separated	79.7%
Education
No high school degree	37.5%
High school degree only	52.7%
Vocational training or post-high school education	7.7%
Other	2.1%
Primary substance of abuse
Heroin	11.2%
Cocaine	15.7%
Methamphetamines	12.2%
Alcohol	48.6%
Other drugs	12.2%
Frequency of drug/alcohol use
Daily use	31.9%
Weekly use	35.4%
Less than weekly use	32.6%

*All of the 5,664 AFDC clients in the sample were admitted to publicly funded substance abuse treatment in Washington State from July 1994 through June 1996 and were receiving financial assistance from AFDC at the time of their admission to treatment.

About 38 percent of the AFDC client households had two to three children, and 6.4 percent had four or more children. Slightly over half (56 percent) of the clients had one or no children (pregnant clients who meet income eligibility requirements qualify for AFDC). The self-reported monthly household income of AFDC clients averaged $474. Approximately 72 percent of the AFDC clients were white, 12 percent were black, 8 percent were Native American, and 7 percent were Asian, Hispanic, or other. More than one-third (38 percent) had no high school degree. Almost half reported alcohol as their primary substance of abuse; the others reported heroin, cocaine, methamphetamines or other drugs as their primary substance of abuse. About two-thirds reported abusing their primary substance on a daily or weekly basis.

B. Treatment Modalities Analyzed and Assignment of Clients to Treatment and Comparison Groups

The 5,664 AFDC recipients in the study population were admitted to a range of substance abuse treatment modalities, reflecting their different treatment needs (see Table II-2). 

Table II-2. Admissions of AFDC Clients to Substance Abuse  Treatment by Treatment Modality at Index Admission* (N = 5,664)
Treatment modality	Number of  admissions	Percent
Outpatient treatment	3,784	66.7%
Intensive inpatient (21- or 28-day non-hospital) treatment	706	12.5%
Detoxification	626	11.1%
Methadone maintenance	299	5.3%
Long-term residential treatment	131	2.3%
Extended care	60	1.1%
Recovery house	58	1.0%
Total	5,664	100.0%

*A client's initial admission to substance abuse treatment is referred to as the index admission.As shown, about two-thirds (3,784) of the study population were initially admitted to an outpatient substance treatment program. Although all 5,664 AFDC clients were admitted to treatment at least once from July 1994 to June 1996, nearly half (46 percent) of the clients had more than one admission to treatment during this period. Specifically, 24 percent (1,369) had two admissions, 12 percent (675) had three admissions, 6 percent (315) had four admissions, and the remaining 4 percent (224) had more than four admissions. Overall, study population clients accounted for 10,417 separate admissions to substance abuse treatment during the study period.

Three substance abuse treatment modalities were selected for in-depth analysis: (1) outpatient treatment; (2) intensive inpatient (21- or 28-day non-hospital) treatment; and (3) methadone maintenance. Complicating the process of selecting specific treatment episodes for analysis, some clients in the study population were admitted to different modalities at different times. For example, a client could have an initial admission to outpatient treatment and a subsequentadmission to methadone maintenance. To address this situation, a series of decision rules was established to define treatment episodes for specific modalities.

• Outpatient treatment episodes. Treatment episodes for outpatient treatment were limited to clients who received outpatient treatment at their index (first) admission. This rule ensured that the analysis focused on outpatient treatment provided as the primary modality, excluding outpatient treatment provided as followup to earlier inpatient treatment.
  
  
• Intensive inpatient and methadone maintenance treatment episodes. For intensive inpatient treatment and methadone maintenance, treatment episodes were selected for analysis regardless of whether they represented the index admission. Thus, for example, a client receiving detoxification initially who was later admitted to intensive inpatient treatment would be classified as an inpatient client.

• Treatment and comparison groups for outpatient treatment. Clients staying in outpatient treatment a minimum of 90 days during their index admission formed the treatment group. Clients remaining in outpatient treatment fewer than 21 days during their index admission formed the comparison group.
  
  
• Treatment and comparison groups for intensive inpatient (21- or 28-day non-hospital) treatment. Clients whose inpatient discharge record indicated "treatment completed" for either a 21- or 28-day treatment regimen formed the treatment group. It was not feasible to construct a comparison group for the inpatient analysis consisting of clients with minimal inpatient treatment. Instead, detoxification clients who had three or fewer detoxification treatments and no other treatment were used for the comparison group.
  
  
• Treatment and comparison groups for methadone treatment. Methadone maintenance clients staying in treatment for a minimum of 120 days formed the methadone maintenance treatment group. Methadone clients staying in treatment fewer than 60 days formed the comparison group.

The numbers of AFDC clients in the substance abuse treatment and comparison groups for each treatment modality are shown in Table II-3 below. On average, AFDC clients in the outpatient treatment group stayed in treatment 148 days; their counterparts in the outpatient comparison group stayed 9 days. AFDC clients in the methadone maintenance treatment group stayed intreatment for 262 days; their counterparts in the methadone maintenance comparison group stayed 29 days. 

C. Data Sources and Construction of Analytic Files
 

Data for the 5,664 AFDC clients treated for substance abuse from July 1994 through June 1996 were obtained from three computer databases maintained by Washington State agencies:

• Data for clients receiving substance abuse treatment. The Washington State Division of Alcohol and Substance Abuse client database has information on the characteristics of individuals receiving substance abuse treatment and on their admission to treatment, discharge, and treatment activity. It includes data on age, gender, race, education, primary substance of abuse, frequency of use, past mental health treatment, probationary status, referral source, living arrangement, number of children in household, marital status, disability status, general health status, types of crimes committed, and household income.
  
  
• Data for employment and earnings. A database maintained by the Washington State Department of Employment Security includes wage data submitted by employers in the State each quarter.
  
  
• Data for welfare payments. A database maintained by the Economic Services Administration within the Washington State Department of Social and Health Services has data on monthly cash amounts distributed to welfare recipients in the State.

The Division of Alcohol and Substance Abuse client database was used to obtain activity data for individuals in outpatient or methadone treatment, including date and type of service provided. The Department of Employment Security database was used to obtain data on quarterly earnings for each of the AFDC clients in the study for the period from July 1993 through September 1998. The Economic Services Administration database was used to obtain data on the monthly cash grant amounts paid to each of the AFDC clients in the study over the 48-month period from July 1993 to June 1997.

These data from the three State administrative databases were entered into separate Access databases and then uploaded into SPSS data files. The individual SPSS data files were then combined into a master file that contained client, treatment, employment, and welfare payment data. The master file used for this study contained 10,417 records representing each treatment episode, or admission, for the 5,664 AFDC clients in the study population.

D. Measures1. Dependent VariablesKey dependent study measures were employment and welfare payment measures constructed using data from the Washington State administrative databases described above. The monthly welfare payment data were aggregated to reflect quarterly data to facilitate the analysis.

• Employment outcome measures. Employment outcome data covering nearly 5 years (July 1993 to September 1998) allowed the construction of outcome measures representing 8 quarters of post-treatment followup. Three employment outcome variables were examined: (1) a binary variable set equal to 1 if a client had any positive earnings (> $0) during the 8 quarters following treatment and 0 otherwise; (2) a binary variable set equal to 1 if a client had earnings of at least $1,500 or more in at least 1 of the 8 quarters following treatment (the approximate equivalent of having a half-time job for a quarter at an hourly rate of $6) and 0 otherwise; and (3) total aggregate earnings in the 8 quarters following treatment, which reflected the combined effects of the number of hours worked and the hourly rate of pay.
  
  
• Welfare payment outcome measures. Welfare payment outcome measures for clients in the study included quarterly welfare payment amount and number of quarters during which welfare payments were received. Welfare payment data covered a shorter time period (July 1993 to June 1997) than the employment data, necessitating a followup period for welfare payments of only 6 quarters. Consequently, the welfare payment analysis was limited to AFDC clients (n = 4,320) discharged from substance abuse treatment by January 1996.

Information on selected employment and welfare payment measures for AFDC clients in the study population is shown in Table II- 4. 

2. Independent Variables

Because assignment to treatment and comparison groups was not random, differences between the two groups of AFDC clients were controlled through statistical analysis. Independent client variables included (1) demographic variables (age, gender, race, years of education, marital status, number of children in household); (2) substance abuse variables (primary substance of abuse; frequency of use, severity of dependence, mode of drug intake-needle use versus other modes, number of previous admissions to substance abuse treatment before July 1994, number of admissions to substance abuse treatment from July 1994 to June 1996, and quarter of discharge), (3) general health variables (disabled, past hospitalization for mental illness, under current treatment for mental health problem, and pregnant); and (4) employment and earnings variables (unemployed but seeking employment at time of admission, and baseline earnings-average quarterly earnings during the 4 quarters before substance abuse treatment).

E. Descriptive and Statistical Analytical Procedures

Both descriptive and multivariate statistical analyses of the data were performed, as described below.

1. Descriptive Analysis

The descriptive analysis identified patterns and trends in the 5,664 AFDC clients' earnings and welfare payments over time, both before and after substance abuse treatment. It also examined the pattern of employment and earnings among welfare recipients in the first year following TANF's implementation in Washington State

2. Statistical Analysis

The statistical analysis determined the effects of substance abuse treatment on AFDC clients' employment outcomes, making statistical adjustments for client characteristics (age, sex, race, education, primary substance of abuse, frequency of use, disability status, past mental health treatment, marital status, number of children in household, number of admissions prior to July 1994, number of treatment episodes during the study period, and baseline earnings) that might affect these outcomes. A significance level of p = .05 was used for the analysis based on one-tail tests. Earnings and welfare payment data analyzed represented nominal dollars.

In the statistical analysis, both logistic and multiple linear regression models were used:

• Logistic regression was used to evaluate employment measures for AFDC clients in binary form (e.g., any post-treatment employment versus no employment). This type of regression model provides information on the odds (likelihood) of the occurrence of an event in relation to some independent variable of interest. For example, in the context of the current study, there is interest in knowing whether AFDC clients who stayed in outpatient treatment for more than 90 days were more likely to become employed than clients who stayed less than 21 days. Logistic regression provides information to address this question. But it is also important to know whether treatment had some effect on the level of earnings for clients who became employed.
  
  
• Ordinary least squares (OLS) regression was used to determine the effects of treatment on the amount of earnings for clients who became employed at least part time. Because the earnings data exhibited some skewness, the OLS regressions were re-estimated using log-transformed data. The estimates were unchanged, so the original results (based on the untransformed data) are reported.

A separate subanalysis was performed to determine the effects of substance abuse treatment on post-treatment earnings among AFDC clients who had been on welfare for different lengths of time during the pretreatment period. Because of a larger number of cases were required, only outpatient clients were analyzed. In this analysis, the pretreatment baseline period was extended from 4 to 6 quarters by limiting the period of admission to substance abuse treatment to 1 ½ years (January 1995 to June 1996). Although this approach reduced the number of AFDC clients that could be analyzed, it made it possible to create a better stratification of pretreatment time on welfare.

For the subanalysis, AFDC outpatient clients in the study population were divided into two groups depending on the number of quarters they received welfare payments during the 6-quarter baseline period. Clients who had received welfare payments for 4 or more quarters were included in one group, and clients who had received welfare payments for 3 or fewer quarters were included the second group. A binary variable representing the two groups was created, and this variable was then interacted with the treatment variable. Including this interactive term in the regression models made it possible to examine the effects of treatment for clients with different levels of time on welfare.
   

For further information, contact the CSAT staff at 240-276-2750.