Evaluation of a peer support program to reduce substance abuse within a large manufacturing company.

ABSTRACT

Substance abuse in the workforce represents a significant problem for employers and employees. One promising prevention and early intervention strategy is a peer support program.

This paper describes a prospective cohort study designed to test the impact of a worksite peer support program on (1) mental health and EAP utilization, (2) substance abuse and worksite safety statistics, (3) health benefit utilization and costs, and (4) productivity. The study was conducted at a large manufacturing company and included 8 intervention worksites and 14 matched control sites. Outcome measures included drug use (from employer drug-testing results), substance abuse benefit utilization, overall health-care costs, workers' compensation claims, and use of short- and long-term disability leave.

None of the worksite-level analyses of the outcome variables produced statistically significant results, and in only one case was the difference in pre-/post-change outcomes statistically significant between groups when using an employee-level analysis (short-term disability days increased significantly in the control group compared to the Intervention group). Despite the negative results, peer helper programs merit further study given their relatively low cost and ability to build on the natural trust workers have in their respected peers.

KEY WORDS: Peer support, substance abuse, health care, employee assistance program (EAP).

Introduction

Substance abuse in the workforce represents a significant problem for employers and employees. Of the $1.85 billion in estimated economic costs of alcohol abuse in the United States. 70 percent (or $1.34 billion) is attributable to lost productivity (Harwood 2000). Substance abuse affects safety, productivity, absenteeism, and turnover in the workplace (Harwood, Fountain, and Livermore 1998).

Estimates of illicit substance abuse in the public sector and among different industries vary widely, ranging from 3 percent in the service industries to more than 20 percent in the mining and construction industries (Bureau of Labor Statistics 1989). One survey found that 9 percent of employees and 12 percent of applicants tested positive for some illicit substance (Bureau of Labor Statistics 1989).

However, simply identifying and treating or removing persons with frank substance diagnoses cannot address the effects of substance abuse in the workplace. This is because the negative effects of alcohol abuse on workplace performance usually are not attributable to people with alcohol dependence (Mangione, Howland, and Lee 1998). Most are due to occasional abuse, such as drinking at work or hangover effects from episodic binge drinking at night. Such drinking behavior is unlikely to be detected by traditional for-cause screening or to come to the attention of EAPs. In addition, substance abuse is related to worksite culture, such as attitudes toward drinking, workplace policies, the accessibility of alcohol, the degree of worker autonomy, and workers' sense of control (Ames and Janes 1992; Trice and Sonnenstuhl 1988).

Data concerning employment outcomes for workers who use illicit substances often show conflicting results. One prospective study examined the relationship between pre-employment drug testing and subsequent employment outcomes. New hires testing positive for pre-employment drug use were followed over time and found to be associated with higher adverse employment outcomes, including higher turnover rates and more accidents, injuries, and disciplinary actions than those with negative pre-employment test results (Zwerling 1990). However, the magnitude of the differences between groups was small (relative risk <2). In a similar study, no relationship was found between drug use and job performance (Zwerling 1990).

Drug testing after occupational injury is practiced widely in the transportation industry, but little is known about the association between illicit substance use and occupational injury in other occupations (Zwerling 1993). One report suggests that alcohol and prescription drugs are a more frequent factor in workplace fatalities than are illicit substances (Lewis and Cooper 1989).

Results from a recent corporate drinking survey show that the answer is not as simple as identifying workers with diagnosable substance abuse problems. In that study, three in five incidents of workplace tardiness, absenteeism, and poor performance were caused by alcohol use in light to moderate drinkers (Mangione 1998).

To address more effectively the problem of substance abuse in the workplace, strategies are needed to change workplace culture, detect hazardous substance use earlier, and provide interventions that penetrate the organization more broadly One promising prevention and early intervention strategy is a peer support program. Peer support programs got their start in schools and among other adolescent populations. While some programs focused simply on coalescing peer influence, more complex programs involved teaching communication and persuasion skills to peer helpers (McAlister 1980). A meta-analysis of outcome results from 143 adolescent prevention programs concluded that peer programs were superior to other interventions (Tolber 1986).

Various industries have begun to harness the influence of peer support to help manage substance abuse problems. The railroad industry, for one, has a long history of using peer intervention programs (e.g., Operation Red Block, Union Pacific; Operation Stop, Burlington Northern). Union Pacific has attributed a number of perceived benefits to peer intervention programs (Eichler 1988), including reduced workplace substance use or abuse, decreased tolerance of substance use by other workers in the workplace, increased effectiveness of supervisory personnel, improved union-management relations, enhanced safety, fewer personal injuries, reduced utilization of medical benefits, decreased workplace accident investigations, and less money spent on legal proceedings.

The American Flight Attendants Association (AFAA) also uses peer helpers, who are trained to refer their fellow union members to EAP services. The AFAA reports that 3.6 percent of its members used EAP services from 1983 to 1985. Of those, more than 30 percent were referred by peer helpers and 17 percent by supervisors, while 51 percent were self-referrals (American Flight Attendants Association 1997).

In the only controlled trial of workplace peer support intervention, 3M used peer support along with supervisor training and employee education programs to reduce substance use, risky behaviors, and negative consequences of alcohol use at work (Stoltzfus and Benson 1994). Self-report data collected from one experimental and one control site revealed less alcohol consumption and fewer negative consequences of alcohol use at work among members of the experimental group. The authors concluded that employee participation and ownership of the program might have helped change the normative culture of workplace alcohol use.

Study Overview

In 1996, management of a large manufacturing firm wanted to strengthen the company's substance abuse efforts and implemented a pilot peer support program at three locations. The primary goal of the program was to offer voluntary prevention and early intervention services to reduce employee impairment arising from substance abuse, workplace stress, family stress, and similar causes. At the core of the program were a group of employees called peer helpers who were trained to listen to and assess co-workers' needs and recommend appropriate referrals to the EAP before there was an effect on workplace performance. Training focused on listening and support skills as well as understanding and promoting EAP services.

The estimated costs of a program with 10 peer helpers totaled approximately $22,000 for start-up and implementation and $11,000 for annual maintenance (Table 1). Estimated costs associated with employees losing time from their regular jobs, based on a 10-person design team and 10 peer helpers, totaled roughly $18,000.

Methods

Research design. The peer support program was evaluated using a community-based prevention design. Sites were selected for participation based on a pre-selected group of criteria. Sites agreeing to participate, called self-selected intervention sites, were matched with comparable control sites. The self-selected sites comprised the "community" in which peer support services were provided. (A randomized selection of sites was not possible due to conflicting business needs of the company; instead, a hierarchically matched control group was selected for each intervention site.)

Selection process. The peer helper program was implemented in a large manufacturing company with diverse plants located in 46 states as well as an international division. Total employment at the company in 1998 was 32,000. Eight worksites with some 2,000 employees participated in the peer support program.

Worksites were invited to self-select using the following criteria:

* Non-intrusive but supportive management, union, and human resources representatives;

* A strong and involved EAP;

* Adequate and dedicated developmental resources;

* Use of "teaming" systems (i.e., employees are empowered to exercise greater decision-making authority);

* A trusting work environment; and

* A stable work environment (e.g., labor bargaining activity). Control sites were chosen based on characteristics most likely to influence or confound the outcomes associated with the peer support program and/or utilization of a managed health care delivery system. Matching of control sites with self-selected sites was based on union status, product line, geographic region, site size, ethnicity of the worksite, and manufacturer environment.

Once a site accepted program implementation, a design team of management, human resources, union, and employee representatives was tasked to tailor and communicate the program to the worksite. Peer helpers were chosen by one of three strategies: (1) nomination by co-workers. (2) conscription by the union, or (3) self-nomination. Nomination by a co-worker appeared to result in faster program start-up in that peer helpers were identified more quickly, had already been functioning as informal "helpers" in the workplace, and often had a history of keeping co-workers' confidences.

Training of peer helpers. Training of peer helpers occurred in a two-day program conducted by two EAP counselors with the assistance of the onsite EAP counselor. A one-day follow-up training was conducted after three months, followed by annual one-day refresher training. All training focused on developing active listening skills, reflective paraphrasing, explaining how to give and receive feedback, and role-playing. Peer helper interventions were both passive (when a co-worker sought assistance from a peer helper) and active (when a peer helper contacted a co-worker with a concern). Peer helpers received supervision in monthly or quarterly meetings with the onsite EAP counselor.

The program began with three pilot sites in mid-1997: three programs started in 1998 and two in 1999. The sites ranged in size from 120 to 470 full-time employees (FTEs), with 8 to 23 peer helpers (comprising 3.0 to 10.8 percent of the workforce) per site. The number of peer helpers depended largely on the physical size of the plant and the number of work shifts.

Sample. Across all sites, just over 80 percent of the employees were male, and roughly three-quarters were Caucasian. The average age was slightly older than 40 years and average tenure about 12 years in both groups. Nineteen percent of employees in the intervention group and 11 percent in the control group belonged to unions. There were no statistically significant differences between the intervention and control groups on control variables (Table 2).

Outcome measures and research hypotheses. The effect of the peer support program was evaluated using outcome indicators available in administrative databases. We expected that the presence of a peer support network in the workplace would result in lower rates of positive random and for-cause drug tests. We also predicted that the peer support program would result in lower rates of substance abuse and mental health treatment utilization, lower cost per covered life for substance abuse treatment, and lower workers' compensation-related medical and time-loss payments.

Statistical testing. The analysis of quantitative dependent variables was conducted using the worksite as the unit of analysis. Pre-intervention data were aggregated at the worksite level and converted to an annual rate per employee (for example, the average annual cost of mental health benefits per employee). Post-intervention data were then aggregated in a similar way and a change score computed (post-intervention minus pre-intervention).

Unpaired T-tests were conducted to test for the significance of the difference in change scores between intervention and comparison groups. A sensitivity analysis was conducted with and without large cases to test and adjust for non-parametric outcomes. We found no differences in reported outcomes with and without large cases.

Validity of the intervention. The project included a substantial process evaluation to measure the extent to which the peer support intervention was implemented (effect size and dose effect) and to identify lessons that might be applied to future efforts. Levels of peer helper and EAP activity (e.g., number of contacts and referrals) were tracked using quarterly reports obtained from the peer support program coordinators and EAP administrators. In addition, the corporate drug testing coordinator conducted key informant interviews. Finally, three study team members independently ranked the peer support programs on a number of dimensions, including management support of program, degree of buy-in by employees, employee knowledge of the program, strength/cohesiveness of the peer helper team, and the overall quality of the program.

Results

Process findings. The number of peer helper contacts at each intervention worksite ranged from 30 to 156 per quarter, for a rate of 11 to 87 contacts per 100 employees per quarter. The number of EAP contacts was substantially lower, ranging from 6 to 36 per quarter, for a contact rate of 4 to 10 per 100 employees per quarter. The problems encountered most frequently by peer helpers were related to work and marriage/family, with about 17 peer helper contacts per 100 employees.

Structured interviews were conducted with key informants at seven of the intervention sites. Regarding program focus, the vast majority of respondents said the purpose of the program was to provide an additional avenue of support for employees with problems. When asked to identify key areas of need for the peer support program to address, respondents overwhelmingly cited work stress resulting from a variety of factors, including downsizing, work system changes, new product lines, and layoffs.

The individual structured interview ratings revealed three clear groupings of the intervention worksites: (1) a high-rated group (two sites) with an overall average ranking of 4.7 (on a 5-point scale) and a quality rating of 5.0; (2) an intermediate group (three sites) with an overall average ranking of 3.5 and a quality rating of 4.0; and (3) a low-rated group (two sites) with an overall average ranking of 2.2 and a quality rating of 2.0. This grouping of programs into three levels, based on the ratings, roughly confirmed the results from the key informant interviews.

Respondents at the high- and intermediate-rated sites reported respect for the peer helpers, a belief that interactions with peer helpers were confidential, and increasing trust and confidence in the program over time. The two low-rated sites reported a lack of stakeholder commitment, lack of ownership, and ineffective program coordination.

Outcome findings. None of the worksite-level analyses of the outcome variables produced statistically significant results, and in only one case was the difference in the pre-/post-change statistically significant between groups when using an employee-level analysis (short-term disability days increased significantly in the control group compared to the intervention group,.

In most other instances, the modal pattern was a decline in both the intervention and control groups. For example, substance abuse treatment costs per employee per year declined from $2.36 to $0.36 in the intervention group and from $22.24 to $1.56 in the control group. An exception to the downward trend in expenditures was mental health treatment costs in the intervention group, which increased from $11.84 to $27.88. Positive test rates from random drug testing declined from 2.8 percent to 0.7 percent in the intervention group and from 3.3 percent to 0.5 percent in the control group. Changes in random drug-testing rates were driven primarily by one intervention site.

Discussion

As with other community-level studies, this study included relatively few units of analysis and, consequently, has limited statistical power. This problem was compounded by our inability to obtain data (e.g., EAP and health benefit utilization data) from some of the sites. Therefore, it is difficult to say whether the peer support programs had no impact on outcomes or whether we were unable to detect the impact given the limited number of sites in the study

However, while the study was inconclusive in terms of the main quantitative outcome analyses, we think there are sound business reasons to consider implementing peer helper interventions based on the qualitative results. The peer helpers were able to reach a larger percentage of the workforce than the EAP We believe this reflects the capacity of the peer helpers to detect recipient problems and intervene earlier in the process of problem development. The qualitative data suggest that peer helpers are seen as a highly acceptable, available, and respected alternative source of support for workers within a large manufacturing company.

The study had several important limitations. Despite considerable effort, we were unable to compile data during several quarters, especially EAP and peer activity data. The worksite staff responsible for reporting the data had many competing priorities and few incentives to complete data forms or comply with research needs. Health benefits data were impossible to obtain from several insurers due to federal standards restricting the release of mental health and substance abuse information.

We had planned to conduct an employee survey and considered assessing several possible outcomes, such as perceptions of the program and actual drinking behavior. Ultimately, we abandoned these plans due to concerns about the confidentiality of the data, the intrusiveness of directly surveying workers. and fears that these factors might damage the nascent support workers had for the peer program.

Recommendations

For businesses considering peer helper programs, our experience underscores the importance of soliciting buy-in from key stakeholders, including management and union leadership, even before the commencement of the program* A major corollary of this principle is to ensure confidentiality in the peer helper relationship* Workers' confidence in the program seemed to depend on whether they felt they could trust peer helpers with personal information. Respect for and trust in peer helpers may have hinged on how peer helpers were chosen. Nominations of peer helpers by co-workers appeared to be the most effective means of selection in terms of ensuring respect and trust.

For researchers interested in studying peer helper programs in workplace settings, we have a number of recommendations. First, do not underestimate the speed and power of large-scale market forces and business changes--such as revising health benefits programs, increasing employee access to health care provider networks, and switching EAP providers--to influence key outcome variables. These secular changes in health care utilization may have overshadowed any smaller impact the peer helper program had on health benefits or other quantitative outcome data. Dependent variables and the time frame for outcome analyses should be based in part on anticipated changes in the health care market. Longer periods of study may be necessary to allow for maturity of the program and effect size.

Conclusion

Substance use and abuse in the workplace are important and costly problems that probably are not addressed adequately by drug testing and professional help (e.g., by EAPs). Peer helper programs have been used to address psychosocial and substance abuse problems in schools as well as in a number of industries. Peer helping also aligns with a workforce trend of teaming systems and worker empowerment to improve the overall workplace culture.

Our attempt at implementing peer helper programs within a large manufacturing company was reasonably successful at six of eight sites. However, the goal of demonstrating the efficacy of peer helpers was not met due to a variety of methodological limitations, operational challenges, and secular changes in the health care financing landscape.

Peer helper programs remain attractive for their relatively low cost, the ability of peers to broadly penetrate the workforce, and because they build upon the natural trust that develops between workers. Peer helper programs merit further study, especially in conjunction with programs designed to make grassroots changes in workplace culture regarding alcohol and drug use.

Table 1 Costs of Peer Helper Program: Program Implementation and
Maintenance Costs

Start-up and Implementation Costs

Design/training meeting (includes             $4,200
meeting room, food and travel)

Two trainers (includes transportation,       $10,300
meals and training)

Communications (includes materials,           $2,650
mailings and contractors)

EAP costs for design meeting                  $4,800

Peer helper recognition costs                   $500

Total                                        $22,450

Maintenance Costs

Design/training meeting (includes             $2,340
meeting room, food and travel)

One trainer (includes transportation,         $4,160
meals and training)

Communications (includes materials,           $2,650
mailings and contractors)

EAP costs for maintenance phase               $1,600

Peer helper recognition costs                   $500

Total                                        $11,250

Table 2 Demographic Characteristics of
Intervention and Control Groups

Characteristic                 Intervention   Control

# of worksites                            8        14

# of full-time employees              1,837     3,047

% of male employees                    80.7      81.8

% Caucasian                            77.1      73.5
% Black                                15.5      22.9
% Hispanic                              6.3       2.4
% Asian/Pacific Islander                0.3       0.4
% Native American                       0.8       0.8

% union members                        19.1      11.2

% married                              77.2      72.7

Mean age in years                      39.6      40.2

Mean years tenure                       9.0       8.8

NOTE: No significant differences were found between
intervention and control on any of the variables shown.

Table 3 Outcome indicators

                                              Peer Helper Sites

Outcome variable                              Pre-          Post-

                                                       N = 3

Random drug testing (1)                       2.8 percent   0.7 percent
                                              (812 tests)   (716 tests)

For-cause drug testing                        2.7 percent   0.0 percent
                                              (70 tests)     (34 tests)

Health benefits (cost per covered life) (2)            N = 6

Substance abuse treatment                           $2.36         $0.36
Mental health treatment                            $11.84        $27.88
Other medical treatment                         $1,054.48       $692.20

Workers' compensation (per FTE) (3)                    N = 6

Cost of workers' compensation claims              $173.60       $117.44

Short-term disability (per 100 FTE) (4)                N = 6

Number of STD claims                                  6.0           4.0

STD disability days                                 164.0         126.8

                                              Control Sites

Outcome variable                              Pre-          Post-

                                                   N = 4 (random)
                                                  N = 5 (for cause)

Random drug testing (1)                       3.3 percent   0.5 percent
                                              (314 tests)   (158 tests)

For-cause drug testing                        1.5 percent   2.2 percent
                                               (40 tests)    (20 tests)

Health benefits (cost per covered life) (2)            N = 13

Substance abuse treatment                          $22.24         $1.56
Mental health treatment                            $17.88        $10.72
Other medical treatment                         $1,217.92     $1,189.32

Workers' compensation (per FTE) (3)                    N = 13

Cost of workers' compensation claims              $209.12        $79.16

Short-term disability (per 100 FTE) (4)                N = 13

Number of STD claims                                  8.4           8.0

STD disability days                                 236.0         233.6

NOTES: None of the pre-/post-changes differed significantly between
intervention and control sites.

(1) Percent of tested employees with a positive test result

(2) Average of quarterly costs per covered employee for the indicated
category, at an annual rate

(3) Average of quarterly costs per full-time employee, at an annual
rate

(4) Number of claims and days per 100 full-time employees

ACKNOWLEDGEMENTS: This study is supported by a grant from the Center for Substance Abuse Prevention as part of the Workplace Managed Care initiative # U1K SP08139-03-2.

References

American Flight Attendants Association. 1997. Personal communication with authors.

Ames, G., and C. Janes. 1992. "A cultural approach to conceptualizing alcohol in the workplace." Alcohol Health and Research World 16: 112-119.

Bureau of Labor Statistics. "Survey of Employer Antidrug Programs." 1989. Washington. D.C.: U.S. Dept of Labor.

Eichler, S. 1988. Operation: Red Block. Rockville, Md.: Institute for Human Resources. Inc.

Harwood, H. 2000. "Updating estimates of the economic costs of alcohol abuse in the United States: Estimates, update methods and data." Bethesda, Md.: National Institute on Alcohol Abuse and Alcoholism.

Harwood, H., D. Fountain. and G. Livermore. 1998. "The economic costs of alcohol and drug abuse in the United States, 1992." Bethesda. Md.: National Institute on Drug Abuse.

Lewis, R., and S. Cooper. 1989 "Alcohol. other drugs, and fatal work-related injuries." Journal of Occupational Medicine 31 (1): 23-28.

Mangione, T., J. Howland, and M. Lee. 1999. "New perspectives for worksite alcohol strategies: Results from a corporate drinking study." Washington, D.C.: Robert Wood Johnson Foundation.

Mangione. T. 1988. "New perspectives for worksite alcohol strategies: Results from a corporate drinking study." Boston Mass.: JSI Research and Training Institute.

McAlister. A. 1980. "Pilot study of smoking, alcohol, and drug abuse prevention." American Journal of Public Health 70:719-721.

Parish, C. 1989. "Relation of the pre-employment drug testing result to employment status." Journal of General Internal Medicine 4 (1): 44-47

Stoltzfus, J. and R Benson. 1994. "The 3M alcohol and other drug prevention program: Description and evaluation." The Journal of Primary Prevention 15: 147-159.

Tolber. N. 1986. "Meta-analysis of 1,433 adolescent drug prevention programs: Quantitative outcome results of program participants compared to control or comparison group." Journal of Drug Issues 16 (4): 537-567.

Trice. H. and W Sonnenstuhl. 1988. "Drinking behavior and risk factors related to the workplace: Implications for research and prevention." Journal of Applied Behavioral Science 24: 327-346.

Zwerling, C. 1993. "Current practice and experience in drug and alcohol testing in the workplace." Bulletin on Narcotics XLV (2): 155-196.

--. 1990. "The efficacy of pre-employment drug screening for marijuana and cocaine in predicting employment outcome." Journal of the American Medical Association 264 (20): 2639-2643.

AUTHORS' AFFILIATIONS

Washington State Medical Assistance Agency

Jeffrey N. Thompson

University of Washington

Kelly B. Kyes

Alien D. Cheadle

Charles H. Bombardier

Weyerhaeuser Corporation

Robert Jacobsen 

Paula Stewart

Steve Hill 

Washington Mutual 

Karen Sorenson