Improving the Health and Safety of Transit Workers with Corresponding Impacts on the Bottom Line (2020)

Below is the uncorrected machine-read text of this chapter, intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text of each book. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.

A-1 A P P E N D I X A Case Study Identification This section summarizes the research activities and results related to data collection for case study site identification as of September 30, 2018. The Econometrica team developed and distributed a survey to agencies and unions that had participated in a prior TCRP project, the results of which are published in TCRP Report 169: Developing Best-Practice Guidelines for Improving Bus Operator Health and Retention and to agencies that responded to a query on the American Public Transportation Association (APTA) Listserv. Agencies selected from both sources reported tracking metrics on costs and benefits of their wellness programs. The project team next applied selection criteria including willingness to participate, charted out respondents in terms of representation by size and geography, and made telephone calls to gain information about the programs and to verify whether the agencies would be willing to contribute data to participate in the benefit-cost analysis. A.1 Case Study Identification: Local Health Promotion Programs That Measure Cost-Effectiveness To address the research question about tools and techniques that can be used for measuring program costs, benefits, and effectiveness, the project team conducted a series of case studies that featured a benefit-cost analysis. The case studies detail the features of a series of health and safety promotion programs at several transit agencies. Desirable sites had either already conducted some form of a benefit-cost analysis or had sufficient data to allow the project team to conduct one. A narrative on the programs at each of the selected sites augmented the quantitative analysis. The Econometrica team created a list of the agencies that had tracked metrics for programs or estimated a past or projected return on investment (ROI) based on information from a prior TCRP project. If an agency tracked a program or estimated the ROI, it was more likely to have data that allowed the team to measure program costs and benefits. This list served as the preliminary list of potential candidates for benefit-cost analysis case studies. A second source of case studies was developed based on responses to a query sent by the project team to the APTA’s Workforce Development Committee Listserv. The query asked for interest and responses to questions that allowed the project team to apply preliminary screening criteria. The query asked about transit agencies with health promotion programs that maintain performance metrics. The APTA Listserv recipients include more than 300 agency representatives (although multiple recipients may correspond to the same agency). The agencies and unions identified through these two sources constituted a preliminary list from which the project team could select sites for case studies. Information in the preliminary list

A-2 Improving the Health and Safety of Transit Workers with Corresponding Impacts on the Bottom Line A.2 Preliminary Screening This section lists the program characteristics that were required to pass the preliminary screening for consideration as one of the benefit-cost analysis case studies. Much of the information needed for the preliminary screening was found by reviewing the information provided in TCRP Report 169 and the email responses to the APTA query. 1. Program is located in the United States. As the data analysis was to be based on U.S. data, the list for potential case studies excluded Canada. 2. Program targeted the frontline worker population, at a minimum. Additionally, it was possible to isolate the frontline workers from other participants. 3. Agency/union had at least 20 participants in its program. This criterion provided for a benefit-cost analysis with sufficient inferential power. 4. Program used one of the following metrics to measure cost-effectiveness: absenteeism, sick days, turnover, worker disability, workers’ compensation, smoking cessation, weight/Body Mass Index (BMI) reduction. 5. Willingness to participate. Agencies were encouraged to participate through potential gain from the results of the study: These agencies had already undertaken some benefit- cost analysis, and by participating in this research project, they would benefit from a more detailed analysis of their program. Potential case studies that met the criteria of this initial screening were subjected to a secondary screening that focused on data availability. A.3 Secondary Screening With input from the project panel, the Econometrica team developed a list of questions for secondary screening based on data availability. Each agency that passed the preliminary screening was asked questions to determine the availability of data in the following categories: 1. Occupational categories of targeted employees; 2. Program type (comprehensive health promotion, smoking cessation, weight management, nutrition, cardiovascular disease prevention, ergonomics, etc.); 3. Specific program activities; 4. Number of participants in the program; 5. Length of the program; 6. Existing program benefit-cost analysis (conducted by program staff or external researchers); 7. Program costs (administrative costs, opportunity costs of workers’ time, incentive payments); 8. Program benefits (healthcare claims costs, workers’ compensation claims cost, time lost/absenteeism, disability, productivity, turnover, etc.); and 9. Program impact on other operational areas (scheduling, hiring, safety, training, work assignment/work accommodation, procurement, etc.). included agency name, size, region, and type of service. The Econometrica team then submitted the preliminary list to two rounds of evaluation criteria.

Case Study Identification A-3 communications from the project team to collect additional data for the narrative and benefit-cost analysis. Next, the selected case studies were analyzed for representation among categories of interest. The goal was at least five case studies representing a diversity of program types, regions, and size and type of agency. The case studies would include benefit-cost analysis if data on program participation and absenteeism and/or workers’ compensation were available for individual transit workers. Five case studies were examined, of which four provided individual-level data that could be used for statistical modeling. A.4 Case Study Narratives The project team compiled a narrative for each site selected as a case study. The narrative described the general landscape of the agency and the types of workforce health and safety programs available to transit workers. The narratives were based on information gained from communications with each agency, including a phone interview and program literature from the agency’s website or other materials provided by the agency. Developing the narratives allowed the project team to explore some of the outcomes of health conditions that have costs that are difficult to measure directly. Because such costs are difficult to quantify, they could not be used as dependent variables in the benefit-cost modeling; however, they added to the discussion in the literature review. The analysis of the various site characteristics and methods used to measure performance, strengths, and weaknesses is provided in the body of this report. A.5 Case Study Benefit-Cost Analysis The goal of the benefit-cost analysis was to produce estimates of the impact of programs administered by transit agencies on transit worker health and overall safety. The first step of the analysis was to identify those factors that may affect transit worker health or safety. Based on the literature review and the process discussed above, the project team identified variables and used as many variables as were available at each site. In general, demographic (age, race, and gender) and occupational (operator, maintenance) variables were included. The models produced estimates of the effects of specific programs on health or safety (insofar as it was possible to capture these data). The models could isolate the potential effects of both a program and a specific demographic characteristic. For example, a program might be effective at decreasing some adverse health issues for persons under age 50, but not for older persons. The analysis was designed to identify those distinctions. The analysis was constructed as a series of regressions that were run for each year (annually) and in aggregate (for all years pooled together). A single observation was created from data at the worker level and included demographic, economic, geographic, program, and health variables. The models were designed to evaluate changes in health over time, then ascribe the changes to either demographic or structural changes; thus, a hypothetical regression analysis controlling for race, gender, income, and rural location might estimate that 20 hours spent on a specific program decreases sick days by 6%. The models also allowed for sensitivity analysis of changes in population components or estimated effects—that is, the estimated effects could be varied over a range to determine if small variations had an outsized influence on the results. Preliminary case study identification and data collection occurred during the secondary screening. Candidate agencies whose programs passed the second screening then received follow-up

A-4 Improving the Health and Safety of Transit Workers with Corresponding Impacts on the Bottom Line Measures of program outcomes were typically the dependent variables (e.g., sick days, disability, wages lost), whereas the demographic and occupational variables were the independent variables. This structure helped control for the effects of the program-related independent variables in the model. Health claims costs or health costs were not used as a dependent variable because, for various reasons, it is difficult to isolate and compare health-related costs. Employers may not have access to all sources of employee health-related payments (e.g., Medicaid, Medicare, employee out-of-pocket payments) to generate accurate health costs for use in analyses. Furthermore, definitions of inputs and outputs vary, which means the costs as collected and reported by different sources may be difficult to compare. Table A-1 shows a list of potential dependent variables. Table A-1. Potential dependent variables. Dependent Variables Value Absent Days Count Sick Days Count Disability Days Count Workers’ Compensation Indemnity Payments Dollars Table A-2 shows a list of potential independent variables. Table A-2. Potential independent variables. Independent Variables Value Duration of Length Participant Spent in Program Hours Type of Program = 1 if general = 2 if smoking = 3 if weight Occupational Category = 1 if operator = 2 if maintenance = 3 if other Years in Occupation Count Age Years Race = 1 if Black = 2 if White = 3 if Asian Gender = 1 if female Program Participant = 1 if participant Participation × Age Interaction term To produce empirical estimates that provided evidence on the impact of health programs administered by transit agencies and their effect on transit worker health and overall safety of transit agencies, the project team used a series of regressions. For count or ratio variables, ordinary least squares (OLS) models were run. The dependent variables depended on what was available for each agency (e.g., data on absent days, lost wages, number of sick days and workers’ compensation payments.) The following equation provides an example of the analysis:

Case Study Identification A-5 The results allowed for an analysis of the characteristics of programs that produced measures of cost savings or net positive benefits that could be quantified, isolated, and potentially applied successfully in other situations. Although past studies have included profiles of successful strategies to promote the health and safety of public transit workers, there have been limited examples of strategies that provided net benefits (benefits minus costs) through empirical testing, such as the testing performed through the benefit-cost analysis in this research. A.6 Screening Survey Design and Distribution The Econometrica team used two sources to generate a list of agencies and programs for case studies. First, the team examined agency and union survey responses in TCRP Report 169 to identify agencies that, at the time of the study, reported having tracked metrics for programs or estimated past or projected ROIs. The rationale behind this criterion was that if an agency had tracked a program or ROI, it would be more likely to have the data that would allow the team to measure program costs and benefits. Data from the agency and union survey responses were extracted and filtered to identify locations and contacts that met the project team’s preliminary criteria. Based on this initial analysis, 16 programs reported by agencies fit the criteria. Additional candidate agencies were identified from a query to the APTA Listserv, which also had to meet the preliminary screening criteria. The project team then drafted and edited initial (opening) emails to be sent to the selected candidate agencies. Tailored emails were sent to candidate agencies identified from TCRP Report 169 and a more generic email was sent to candidate agencies identified from responses to the APTA Listserv query. Both groups received the same set of survey questions, which were entered into SurveyGizmo (an online survey engine) and provided by a link in the emails. The estimated time to respond to the questions in the survey/screening tool was about 5 minutes. Many follow-up emails were sent to the 16 agencies from the prior TCRP project, either to the initial agency or union contact or to other contacts within these agencies who could encourage a response. Five union respondents indicated one or more perceived improvements in metrics. In most cases, however, the local unions did not have access to the type of detailed data required by this study, except for a joint trust fund program. Based on these union leads, the research team reached out to contacts in senior management, joint trust fund administrators, and wellness specialists at these locations to obtain the necessary screening data. Simultaneously, the team requested that APTA send out an email to the Workforce Development Committee Listserv. A week later, APTA sent a reminder email to the Listserv. A.7 Survey Responses After deleting the “ghost responses,” for which no agency name was provided and only one or two questions were answered, the team determined that a total of 29 responses were received. Nine responses were considered partial responses—the respondent filled out the agency name and answered the question “Does your agency have a workplace health protection and promotion [WHPP] program or carry out related activities that target frontline operations and/or maintenance employees?” These responses did not contain enough information to determine the availability of

A-6 Improving the Health and Safety of Transit Workers with Corresponding Impacts on the Bottom Line data and the agency’s willingness to participate. They did not provide a detailed program description, even though some did indicate the existence of a program. Twenty-one respondents completed the majority of the questions in the survey, and of those, 17 indicated that they had active programs. Three respondents indicated that they did not wish to participate in the study, leaving 14 good leads for potential sites for the next step. Of these, 4 respondents tracked only program outputs (participation rate) and non-quantitative outcomes such as employee feedback or behavioral changes. These agencies would not be strong candidates for the full benefit-cost analysis, but they were still pursued if they had some raw data that could be analyzed. Ten sites fully satisfied the initial screening criteria set out by the team (based in the United States, targeting frontline employees, with at least 20 participants, tracked program metrics, and willingness to participate). Three of the sites had calculated the ROI for their worksite health promotion programs. To ensure an adequate number of final case studies, the team decided to pursue all 14 “good” leads to discuss case study engagement. Given the design of the screening survey and the wording of the emails, respondents were highly likely to self-select based on the existence of their programs and metrics they tracked. The survey was not intended to provide a scan of the industry practices, but rather to identify enough locations for the case studies. Therefore, the number of responses generated was considered adequate. A.8 Site Selection Matrix A.8.1 Agency Size No official categorization of public transportation agencies by size is available from APTA or other sources, such as the FTA’s National Transit Database; however, many agencies are ranked by annual unlinked passenger trips (ridership). For the initial analysis, the research team defined the following agency size categories based on unlinked passenger trips in 2015: • Large: 100 million passenger trips or more, • Medium: 15–100 million passenger trips, or • Small: Less than 10 million passenger trips. Based on this categorization, the 14 potential sites included four large, four medium, and six small agencies. A.8.2 Service Modes The potential sites provided a variety of transit modes, from fixed-route buses and trolleybuses to various rail services to demand-response service and vanpools. A.8.3 Union Representation Two of the potential sites were non-union. Because the majority of the frontline workers in public transportation are organized, this distribution indicates a good representation of the industry. The 14 potential sites in Table A-3 were analyzed for representation among categories of interest including agency size, types of services provided (i.e., bus, heavy rail, and light rail), union representation, and region.

Case Study Identification A-7 A.8.4 Regions The proposal indicated the following six regions: Northeast, Southeast, Southwest, Pacific, Northwest, and Midwest. As an official map with these divisions was unavailable, the team decided to use the map from the U.S. Census with four divisions: West, Midwest, Northeast, and South. Table A-3. Potential sites for case studies (ranked by ridership). Agency Name A pp ro x. N um be r of P ar tic ip an ts 20 15 A ge nc y Ri de rs hi p (T ho us an ds ) Ag en cy C at eg or y Bu s He av y Ra il Li gh t R ai l Co m m ut er R ai l Hy br id R ai l St re et ca r Tr ol le yb us Bu s Ra pi d Tr an si t Co m m ut er B us De m an d- Re sp on se Va np oo l Un io n G eo gr ap hi c Ar ea – Ce ns us Los Angeles County Metropolitan Transportation Authority (LACMTA) 4,000 457,356 Large ü ü ü ü ü Y West Southeastern Pennsylvania Transportation Authority (SEPTA) 2,000 344,297 Large ü ü ü ü ü ü Y Northeast King County Transit 5,000 126,268 Large ü ü ü ü ü Y West Denver Regional Transportation District 1,000 102,250 Large ü ü ü Y West Metro Transit (Metropolitan Council) 2,100 85,832 Medium ü ü ü Y Midwest Utah Transit Authority (UTA) 738 46,722 Medium ü ü ü ü ü ü Y West Capital Metro 800 34,700 Medium ü ü ü ü ü Y South Regional Transit Service 650 17,292 Medium ü ü Y Northeast Transit Authority of River City 100 14,684 Small ü ü Y South OmniTrans 100 14,391 Small ü ü N West Champaign–Urbana Mass Transit District 100 13,536 Small ü ü N Midwest Indianapolis Public Transportation Corporation, Inc. 500 9,952 Small ü ü Y Midwest Intercity Transit 100 5,135 Small ü ü ü ü Y West Des Moines Area Regional Transit Agency 85 4,795 Small ü ü ü Y Midwest