Work Outcomes After Benefits Counseling Among Veterans Applying for Service Connection for a Psychiatric Condition

Veterans were identified by three criteria: they were scheduled to be evaluated by a psychologist or a psychiatrist for a service-connected disability, they were not already receiving disability benefits for a psychological condition, and they were receiving no more than 30% of the full benefit for a general medical condition. At the appointment for the evaluation, veterans were referred by posted advertisements or by clinic staff to meet with a research assistant for study screening. Veterans who answered “yes” to any of the questions on the 36-Item Short-Form Health Survey (SF-36) (23) indicating that emotional problems had interfered with work in the past 28 days were enrolled in the study.

The study was approved by the Yale and Veterans Affairs (VA) Connecticut Healthcare System institutional review boards. After a brief assessment was conducted to determine eligibility, the study was described to the veterans, and those who agreed to participate provided written informed consent. For assessment purposes, all of the veterans completed a comprehensive questionnaire at baseline that assessed demographic characteristics and history and incorporated material from other sources (23–28). Psychiatric diagnoses were obtained subsequently from electronic medical record review.

Veterans were selected at random to receive four sessions of benefits counseling or four sessions of VA orientation. Participants were then contacted by a counselor who scheduled the sessions and explained the assigned counseling. Follow-up assessments were completed one, three, and six months after randomization by a research assistant who was blind to group assignment. Veterans in both conditions were paid up to $260 in divided amounts for attending data collection visits.

Benefits counseling consists of four 50-minute sessions of individual counseling using a motivational interviewing framework and techniques to increase veterans’ desire to engage in work and related activities (29). The counselor begins the first session by discussing the veteran’s experience of the claims process, the possible impact of working on the claim, and the veteran’s relative valuation of work vis-à-vis other concerns. In the second session, the counselor directs discussion toward the veteran’s feelings about work and other vocational activities, exploring the veteran’s ambivalence with structured exercises, such as listing the pros and cons of working. The third session focuses on the financial implications of working (or not); for veterans receptive to suggestions, counselors provide information about work, vocational rehabilitation, and education opportunities that are available. The first three sessions were conducted weekly when possible; the fourth session was conducted after the disability determination to allow the veteran and the counselor to discuss the impact of the claim decision on the veteran’s plans.

The control condition was four sessions of VA orientation that taught veterans about the VA health care system generally and about the specific general medical and mental health services available to veterans.

Four master’s-level clinicians delivered both interventions (a crossed design). Fidelity to the interventions was verified by the use of therapy manuals, veterans’ confidential ratings of topics covered during counseling (data not reported), audiotapes of selected sessions, and supervision. The counseling manuals for benefits counseling and VA orientation and the nonproprietary measures that were developed for this study are posted on our group’s Web site (www.behaviorchange.yale.edu) and are briefly described below.

The primary outcome measure was based on a timeline follow-back calendar describing whether the veteran was involved in paid work, volunteer work, school, and vocational or rehabilitative activities on each day. The timeline follow-back calendar has been a standard way to assess recent employment (30) and has been a sufficiently sensitive outcome measure of between-group differences in other clinical trials of employment-focused interventions (31). A secondary outcome was the number of days spent working for pay in the past 28 days. Veterans were also asked how much income they had earned in the past 28 days.

To test hypotheses that benefits counseling fostered working by convincing veterans that receipt of benefits was compatible with working, veterans were asked to rate their agreement with three statements asserting a relationship between work and receipt of disability payments (32). Agreement was rated on a 4-point, forced-choice Likert scale anchored by strongly disagree, somewhat disagree, somewhat agree, and strongly agree. The meaning of work to the participant, a second possible mediator of work history, was assessed with the Meaning of Work Scale (33), which yields a summed score (range 12–48) reflecting the importance of various aspects of work to the respondent.

To test the hypothesis that benefits counseling acted by fostering engagement in VA mental health services, participants’ use of VA services was extracted from national databases by using codes to identify in-person outpatient mental health or substance abuse treatment encounters. Each week during the 26 weeks before or after randomization was assigned a dichotomous value (yes or no) reflecting treatment attendance.

Data analysis proceeded in several stages. First, t tests and chi square tests were conducted to determine if there were any differences in baseline characteristics of the veterans who had been randomly assigned to the benefits counseling or the control group.

The primary outcome measure was the number of days in the past 28 spent engaged in any kind of work or related activity (including going to school or vocational rehabilitation), as derived from the timeline follow-back calendar, and a secondary outcome was number of days of paid work. The effect of group assignment (benefits counseling or VA orientation) on the primary and secondary outcomes over time was the primary effect of interest. This interaction was included in a hierarchical, mixed Poisson regression model with fixed effects, including baseline value of the dependent variable, month, and group assignment.

The analyses tested whether three processes targeted by benefits counseling mediated the relationship between benefits counseling and paid work: the belief that work is compatible with receiving disability benefits, increased valuation of employment, and attending more weeks of in-person mental health or substance abuse treatment at VA facilities, thus improving veterans’ employment readiness. The effects of change in the proposed mediators on the treatment-work relationship were tested in latent growth curve mediation models (34), estimating the indirect effects of benefits counseling on work through the mediators.

Recruitment was conducted from December 2008 to December 2011. Of the 425 veterans who expressed interest in participating, 195 were eligible to participate. As shown in the CONSORT diagram (Figure 1), 92 were randomly assigned to one of the interventions, but only 84 (42 in each group) completed at least one follow-up assessment. The statistical analyses apply to the participants who completed at least one follow-up assessment. Both groups participated in roughly the same number of treatment encounters (2.6±1.1, benefits counseling, versus 2.2±1.1, control group; Mann-Whitney U test=740, p=.19). The groups did not differ significantly in the number of follow-up assessments completed.

A majority of veterans were male, were Caucasian, and had had some treatment for a psychiatric or substance use disorder at the VA within a year of enrollment (Table 1). There were no significant differences in any demographic characteristics between the veterans assigned to benefits counseling and the control group (Table 1).

At baseline, veterans strongly agreed that working would lead to loss of benefits, and most agreed that they would turn down a job if it entailed a loss of disability payments (Table 1). Despite the importance of retaining disability payments, veterans’ ratings of work’s importance were relatively high. Work was ranked the second most important activity, after family, among five choices (leisure, community, work, religion, and family) (2.9±1.0 on a 5-point scale). The mean±SD score for the sum of all scales on the Meaning of Work Scale was 33.8±7.1. Both the ranking of work’s importance and the mean score on the Meaning of Work Scale reflected that work was at least moderately important to most participants.

Veterans assigned to the control group were engaged in work-related activities during 8.7 of the 28 days preceding the baseline assessment and during 10.5 days of the 28 days preceding the six-month follow-up. Among the veterans assigned to benefits counseling, days of work-related activity in the past 28 days increased from 8.3 at baseline to 12.7 at month 6 (Table 2). Although the standardized effect size of the group × time interaction was .89, the group × time effect was not statistically significant (Table 2).

Veterans who received benefits counseling showed a large increase in days worked for pay in the past 28 days, from an average of 6.6 days at baseline to 10.5 days six months later (Table 2), compared with negligible changes in days worked for pay during the same period among veterans in the control group (Table 2). The effect of time on days worked for pay was significantly more positive for the benefits counseling group (group × time effect, p=.01). The effect size after accounting for covariates (.69) reflected an average of .5 more days of paid work for every month that veterans received benefits counseling versus the VA orientation intervention.

The greater number of days worked by veterans who received benefits counseling was largely accounted for by an increase in the proportion who worked rather than by a greater number of hours of work among those who worked at all. The percentage of veterans who were assigned to benefits counseling and who worked at all increased from 43% at baseline to 66% at month 6, whereas the percentage of veterans in the control group who worked did not change significantly from the baseline level (52%).

Veterans spent relatively few days in activities other than competitive work. Veterans assigned to benefits counseling spent more days in vocational rehabilitation over time compared with the control group, although the absolute number of days of vocational rehabilitation was small, equal to having attended less than one day of vocational rehabilitation in the past 28.

The difference in earnings between the two groups were in the expected direction (more among benefits counseling participants), but they were not statistically significant.

To characterize veterans who stopped coming to the data assessment visits, a proportional-hazards regression was conducted. Dropouts were significantly more likely to have been younger, unemployed at baseline, and assigned to benefits counseling. However, only two recipients of benefits counseling who did not complete follow-ups had been unemployed at baseline, compared with three of five VA orientation participants who dropped out.

The main outcome analysis was rerun with multiple imputation of missing values for days worked by using the data augmentation algorithm in NORM 2.03 software (35). The measures used to model missing values were days worked (at nonmissing timepoints), time (days) spent in other ways (volunteer work, classes, or vocational rehabilitation), and measures associated with missing values (benefits counseling group assignment and unemployment at baseline). After imputation, the effect of benefits counseling in increasing paid work over time remained significant (p<.01).

These analyses suggest that selective loss to follow-up did not account for the greater number of days of paid work among veterans assigned to benefits counseling.

Beliefs about work or benefits (Table 1) were stable, with minimal changes over time and no differences between the two treatment groups at baseline or over time.

Benefits counseling participants attended in-person mental health or substance abuse treatment an average of 25%±30% of the weeks after randomization, compared with 19%±27% weeks for the control group. In a longitudinal model, a significant phase × group interaction reflected that the proportion of veterans who attended treatment after randomization increased more among recipients of benefits counseling than among members of the control group (β=8.44, p<.01).

None of the three putative mediators of the effect of benefits counseling on working were supported by the mediation models.