Motor coach driver fatigue is an issue of paramount importance to both policy makers and motor coach management. This paper describes the application of the Commercial Motor Vehicle (CMV) Driver Fatigue Model presented at the 1999 International Large Truck and Bus Safety Symposium, to the motor coach industry . This model identifies the antecedents of commercial motor driver fatigue as measured by the frequency of close calls due to fatigue, perceptions of the prevalence of fatigue among drivers, and actual crash involvement. Since the original formulation of this conceptual model, subsequent empirical work has been undertaken and led to some modifications in the model . The latest rendition of the model (see Figure 1) specifies three general categories of fatigue antecedents, or factors hypothesized to affect driver fatigue: CMV Driving Environments, Economic Pressures, and Carrier Support for Driving Safety. CMV Driving Environments and Economic Pressures are each comprised of three constructs. Carrier Support for Driving Safety is a driver fatigue moderating factor and a "stand-alone" construct.

The original CMV Driver Fatigue Model was based on a literature review of 55 studies, few of which dealt with fatigue among motor coach drivers. Empirical work that refined the model was limited to truck drivers. Thus, the purpose of this study is to assess the robustness of the Driver Fatigue Model among a key part of the CMV driver population, motor coach drivers.

Prior to collecting data to test the model, more specific information concerning the operating situations and issues confronting the motor coach industry was deemed necessary. This information was secured from eight focus group sessions conducted for the Bus Driver Fatigue and Stress Issues Study . Focus groups were held in California, Illinois, Mississippi, New Jersey, Tennessee, and Virginia and entailed 154 participants representing the relevant operational areas within motor coach operations, namely, owners, operation managers, safety directors, drivers, and travel/tour planners and coordinators. While many of the expressed issues and concerns mirrored those of the truck drivers, some unique differences were observed:

• Motor coach seats do not accommodate comfortable positions for quality rest or sleep, and motor coaches do not easily accommodate acceptable "sleeper berth" areas for drivers. With most charter and tour trips, however, drivers almost always sleep in hotel beds.



• Driver/passenger interface - because the driving area is not physically isolated from the passenger area, drivers may be distracted by passenger activities.



• Schedules and itineraries - inverted duty/sleep cycles can occur because of group itineraries; also, itineraries may be "spontaneously" altered, disrupting the driver's planned schedule.



• Non-driving work - drivers must tend to passengers' needs (e.g., luggage), take tickets, and perform other tasks, adding to their work time and possibly increasing stress.



• Having numerous people observing driver behavior may produce an incentive for more diligence and professionalism on the part of a driver, but may also cause stress and fatigue



• Seasonality - during the peak seasons, the need for driver services increases and there are fewer extended rest periods for drivers

The major differences between motor coach and truck driving thus pertain to differences in customers (i.e., tour organizers and passengers rather than shippers and brokers) and in potential schedule irregularities. The CMV Fatigue Model was revised to reflect these differences. For example, within the category of CMV Driving Environments, the trip control element was modified to reflect issues concerning trip planning difficulties induced by passenger demands through a category called schedule delays. Under the category of Economic Pressures, the element of scheduling demands of commerce incorporates elements concerning irregular (inverted) driving schedules and the impact of business associated with tour organizers.

Turning to the outcome variables in the model, the category of Fatigue and Crash Outcomes is the same for both industries. There is little consensus in the literature regarding how driver fatigue should be viewed and measured. Numerous indicators of perceived driver fatigue are possible, but care must be taken to obtain these estimates in ways that minimize self-incrimination and elicit accurate responses. Williamson et al., (1992) note that while many drivers will acknowledge that fatigue is an industry-wide problem, fewer may admit that fatigue is a problem for them personally . Accordingly, a broad array of direct and indirect fatigue indicators were retained. Frequency of driving "tired" is the first indicator and it has been used in prior truck , and more importantly, prior motor coach research . Harris and Mackie also developed other fatigue and crash rate indicators germane to this study. They include the number of close calls experienced by the driver because of less-than-full alertness, estimates of the frequency other company drivers drive when they are tired, and the number of reportable and chargeable crashes a driver has had over some defined time/mileage period . Harris and Mackie, found that bus drivers had significantly more reportable accidents than truck drivers, supporting the importance of a better understanding of motor coach crash rates.

The modifications made to the model to reflect the motor coach industry are shown in Figure 1 as italicized entries. Operationalizations used to test the model are described in Table 1.

The FMCSA's Motor Carrier Safety Status Measurement System (SafeStat, Version 6.1) database was used to obtain the population of motor coach firms. Candidate firms for inclusion in this study had to have accurate census data in the SafeStat database detailing their location, safety performance record, and a sufficient number of drivers to provide a reliable driver perspective. SafeStat has safety data for 136,745 firms. Census data could be matched with 78,621 CMV firms in the SafeStat database. Of these 78,621 firms, only 1198 were exclusively carriers of passengers. The 207 firms that carried both passengers and freight were excluded as were firms with three or fewer motor coach drivers. The subsequent universe consisted of 282 motor coach firms.

This universe was stratified on the basis of safety performance prior to drawing the sample, in order to assure sufficient variation among the sample carriers on the dependent variables specified in the model (i.e., there needs to be some variance in safety performance). Consequently, universe carriers were grouped into three safety performance rating categories (i.e., first quartile--poor, middle two quartiles--average, and fourth quartile--top) and sample carriers were selected randomly from each category. An effort was made to sample an equal number of carriers from each category. However, the percentages of firms agreeing to participate more closely resembled a normal distribution, with nearly equal numbers of top and poor performers.

The data collection methodology involved telephone calls to the safety director at each of the selected carriers to solicit the firm's voluntary participation in the study. Carriers that chose not to participate were replaced with firms selected at random from the appropriate safety performance group, except the poor performance category was expanded to include carriers at the top of the second quartile when the first quartile was exhausted.

At each motor coach company, the safety director was asked to distribute surveys customized for different audiences -- one each for a top executive, safety director, and dispatcher, and two for drivers. The safety director was instructed to select "typical" dispatchers and drivers. Each respondent was instructed to put her/his completed survey into the provided envelope, seal it, and return it to the safety director who would return the entire packet to the researchers.

Of the 161 companies contacted, 150 (93.2%) agreed to participate in the project. The percentage of companies agreeing to participate by company performance level ranged from 89.6% (poor performers) to 97.7% (top performers). Of the 150 companies who agreed to participate in the study 66 (44.0%) returned usable survey sets. This response rate is much higher than is typical for mailed surveys, especially in view of the methodology asking for surveys from different occupational categories within each company. Response rates by company performance level ranged from 34.9% (15 poor performers) to 52.3% (34 average performers).

Approximately 70% of the 66 companies were charter/tour operators. This is important since this type of operation would be more susceptible to the pressures from tour group organizers and from passenger pressures during trips. Half of all drivers employed by these companies were full-time drivers. Respondent firms employed an average of 60 drivers and approximately 80% of the drivers were non-union. Eighty (80) percent of the companies pay drivers principally by the hour; but, 41% also pay by the mile - meaning some use a combination.

The average bus trip for all respondents was 250 miles in length with a range of 50 to 1,200 miles. Drivers averaged 1,200 miles per week (with a range of 375 to 2,700) and 48 hours per week (with a range of 5 to 75). Almost 70% of respondent firms employed a Safety Director, but only one-third of these was a full-time position. The safety directors reported an average of two reportable accidents and two chargeable accidents during the past two years. The range of reportable accidents was 0 to 40, while the range of chargeable accidents was 0 to 85.

A total of 122 drivers completed surveys. The average age of the drivers responding to the survey was 53 years, with the range being from 28 through 68. Most driver respondents (i.e., 85%) were regular full-time employees and male (i.e., 88%). Seventy-one (71) percent worked for charter/tour operations with the remainder working for scheduled route operations.

Drivers reported driving an average 1,200 miles per week, with a range of 200 to 2,500. The average number of miles reported per assignment was 300. Drivers reported working an average of 40 hours per week and the average number of stops per day was four (4). Three-fourths of drivers indicated that they were paid by the hour, while 47% indicated they were paid by the mile; again suggesting some firms use a combination of pay. About half of the driver respondents reported being subject to inverted duty/sleep cycles to some extent or to a very large extent. However, this response is tempered by the average number of inverted duty/sleep cycles per trip being just one. Sixty-one (61) percent reported one or two inverted duty/sleep cycles per trip. Only 10% of the drivers experienced an average of three (3) inverted duty/sleep cycles per trip, while 23% of the drivers experienced no inverted duty sleep cycles per trip.

With respect to safety performance, 81% of the drivers reported having no accidents during the past two years, and 99% of the drivers had two (2) or fewer accidents during the past two years. Finally, 84% of the drivers had no chargeable accidents during the past two years, and all drivers responding had two (2) or fewer chargeable accidents during the past two years.

Because this study is the first attempt to assess empirically motor coach driver fatigue, no attempt to limit potential fatigue indicators was made prior to data collection. Rather, a systematic method for reducing the number of measures was employed after data collection. Each measure was subjected to several assessments. Factor analysis was used to establish multiple-item measures of indicators and subsequent Cronbach alpha (a) measures of internal consistency reliability were calculated. The measures had to achieve an alpha of at least .7 to justify retention (unless otherwise noted). Each indicator was also evaluated to assure that it yielded sufficient variability to be of interest. Within a broad construct (e.g., scheduling demands of commerce), indicators were evaluated for "redundancy." However, no indicators were found to demonstrate excessive multicollinearity (i.e., > .7) and had to be eliminated on this basis.

Finally, since the overarching goal is the identification of factors predictive of fatigue and crashes, an indicator's association with these outcomes (i.e., frequency of close calls, perceptions of fatigue, and crash involvement) was deemed relevant in determining indicator retention. Each indicator was assessed in combination with others in the same construct relative to ability to account for variation in fatigue and crash outcomes (i.e., all of the indicators within a construct were simultaneously entered into a regression model seeking to explain each of the three outcome measures). Each indicator that exhibited a statistically significant relationship (i.e., standardized beta weight) with at least one outcome (p < .10) became a final part of the amended model. The study variables are identified in Table 1 and described in more detail in the following sections.

Three dependent variables are included in the model, measures of close calls, driver fatigue and safety performance (i.e., crash involvement). Drivers were asked to report the frequency of close calls or near misses due to fatigue at four separate locations (e.g., at terminals) using a 1 (never) to 5 (very frequently) response framework. Responses were summed and yielded a reliable measure (a = .72). Self and others perceptions of fatigue was measured using five statements to which drivers reported the frequency of fatigue occurring (e.g., nodding off while driving) using the same response framework. Responses were summed and generated a reliable measure (a = .85). Crash involvement was assessed by asking drivers the number of reportable and preventable accidents they had while working the last two years. Responses were summed, normalized for driving exposure, and demonstrated adequate reliability (a = .87). The normalized crash involvement measure range is slightly inflated by the inclusion of a single driver reporting 9.62 crashes per 100,000 miles. The next highest number of crashes was 3.85.

Regularity of time. This construct is concerned with the opportunity for drivers to establish a routine and with schedules that run counter to the natural circadian rhythms of drivers. Indicators employed were (1) the frequency a driver reported normally driving the same daily hours and (2) how driving time is distributed over the 24-hour day.

Trip control. Trip control reflects the ability of drivers to plan trips and how closely trips conform to what was expected. Five single-item indicators measured trip control. They were: regularity of route, freedom to choose one's routes, percent of work time consumed by scheduling delays, difficulty in finding a place to rest; and the average number of stops made daily.

Quality of rest. This construct captures when drivers are able to obtain uninterrupted sleep and the quality of their rest. It was represented by four indicators: the extent to which drivers sleep at nighttime, the amount of uninterrupted sleep a driver gets when working, the frequency with which drivers get home, and frequency of starting the workweek tired.

In order to determine whether these indicators should be retained in the model, all eleven indicators were regressed on each of the three fatigue and crash outcome indicators. Three indicators were retained. From regularity of time, driving the same hours (b= -.18, p < .10) was related to perceptions of fatigue. From quality of rest, uninterrupted hours of sleep was associated with close calls (b= .19, p < .05), and starting the workweek tired was associated with both close calls (b= .37, p < .001) and perception of fatigue (b= .50, p < .001).

Scheduling demands of commerce. This construct reflects the external pressures that are brought to bear on CMV firms by the expectations and requirements of the tour organizers and passengers. It was represented by the frequency drivers experience inverted schedules, the percent of business from tour organizers, and percent of time spent on non-driving activities.

All were regressed on each of the three fatigue and crash outcome indicators. Only the frequency of inverted duty rest cycles was significantly related to close calls (ß= .29, p < .01) and perceptions of fatigue (ß= .28, p < .01) and thus met the standard for retention.

Driver economic or personal factors. This component was intended to capture practices and circumstances that encourage certain driving behaviors. Four categories are recognized: drivers' personal motivations to continue driving even when tired, rewards or penalties for on-time and late arrivals, rewards for safe driving, and personal pride in on-time arrivals.

The regression results revealed that strong personal motivations to continue driving when tired was predictive of close calls (ß= .50, p < .001) and perceptions of fatigue (ß= .60, p < .001). Penalties for late arrivals was significantly related to perceptions of fatigue (ß= .19, p < .05). These variables were retained in the model.

Carrier economic factors. The third component of Economic Pressures refers to the pressures perceived by various personnel to be economically successful. It also entails the policies and practices adopted by carriers to promote economic outcomes, which may sometimes come at the expense of maximizing safety outcomes. Three general areas were investigated: the extent to which carriers emphasize financial performance over safety performance (including perceived pressure on drivers to accept trips when drivers are tired or out of hours, pressure to bend safety rules, and driver perceptions that dispatchers valued scheduling over safety), the extent to which there are rewards or penalties for dispatchers based on operating efficiency, and the extent to which there are rewards or penalties for dispatchers for safe driving.

Drivers' perceptions that they have to bend safety rules to get the job done was positively associated with fatigue perceptions (ß= .38, p < .001). Safety Directors' perceptions regarding pressure by their companies to accept trips even when they have no drivers with remaining hours was significantly related to perceptions of fatigue (ß = .29, p < .01). Pressure to ask drivers to overlook rest requirements was significantly related to crash involvement (ß= .30, p < .05). These four variables were retained in the model.

Carrier support for driving safety. As shown in Figure 1, carrier support for driving safety is positioned as a moderating variable. The extent to which firms institute safety practices possibly may affect the extent to which CMV driving environments and economic pressure impact fatigue and crash outcomes. Four areas of possible carrier support were investigated: safe driving culture, safety training and meetings (to include, voluntary attendance and paid for attendance), company orientation toward driver tiredness, and company policies to minimize nighttime driving.

Two variables in the safety construct were retained in the model. Drivers' perceptions of a safe driving culture was significantly related to close calls (ß= -.39, p < .01) and fatigue perceptions (ß= -.61, p < .001). Company policies minimizing nighttime driving was significantly related to crash involvement (ß= .24, p < .05).

Figure 2 contains the retained indicators for the CMV Driver Fatigue Model Amended for the Motor Coach Industry. The extent to which operational scheduling requirements (i.e., driving environments and economic pressures) and safety affect fatigue and crash outcomes was assessed via regression analysis. A hierarchical regression was conducted using the ten indicators of operating requirements as independent variables in Step 1 followed by the two safety indicators in Step 2 on the three outcome measures of fatigue and crash involvement. Because this research represents initial inquiry into the determinants of fatigue and crashes, a conservative significance level of p < .10 was selected for evaluating both the models and specific indicators.

As shown in Table 2, 23% of the variance in close calls (p < .001) was accounted for by operating requirements. Most of this explained variation is attributable to the pressure perceived by drivers on them to accept trips (ß= .28, p < .05). The addition of safety practices to the model did not add to the prediction of close calls. With adjustments made for the inclusion of additional independent variables, the amount of explained variation decreased to 22%. The findings suggest that the effects of safety culture, while not statistically significant, may help offset the pressure on drivers to accept trips.

The influence of operational scheduling variables on fatigue is quite impressive, with these requirements accounting for 56% of the variation in perceptions of fatigue. Multiple factors appear to play a role. The extent to which motor coach drivers can drive the same hours reduces fatigue (ß= -.16, p < .10) while not unexpectedly, starting work tired (ß= .25, p < .05) contributes to fatigue. Pressure on drivers to accept trips (ß= .31, p < .01) and "bend the rules" (ß= .18, p < .10) also add to fatigue. However, safety practices did not appear to have an additive impact on fatigue as safety increased explained variation by only 1% (to 57%) and was not statistically significant. As was the case with close calls, safety culture did appear to account for some of the variation in fatigue, as the effects of the operational requirements are diminished.

The results associated with crash involvement were disappointing. None of the individual elements in the model were statistically significant nor were either of the overall models statistically significant. As noted below, these results are not only a function of the small sample size but the restriction in range evident in the crash measures.

The CMV Driver Fatigue Model explains a significant percentage of the variation in close calls and fatigue outcomes among the driver respondents. A number of carrier scheduling and safety practices proved to be good predictors of at least one measure of driver fatigue. Not surprisingly, the model performed less well in predicting crash involvement. Two likely explanations for this are: (1) this safety measure suffered from restriction in range for the sample drivers and (2) crash involvement is affected by several factors not addressed in this study.

The results of the study depict situations that are mostly controlled by the individual driver yet significantly influenced by the company. Obtaining adequate rest and recovery time to begin the work refreshed requires personal responsibility and time management by drivers during their off-duty periods. It also requires adequate amounts of recovery time to be provided by the company. The pressure felt by drivers to drive tired to make a good income and to accept trips or bend rules indicate a belief by drivers that if they do not respond to passenger or company demands, their incomes will suffer. It also reflects a reluctance on the part of drivers to say "no", possibly because of their personal needs for earning income. The driver shortage of the 1990's, in conjunction with high demand for motor coach services, may also contribute to these outcomes.

The only significant predictor of both close calls and fatigue was a carrier economic pressure, the drivers' perceptions of pressure by dispatchers to accept trips even when tired or nearly out of hours. Dispatchers likely exert this pressure in order to meet customer demands and perhaps in response to company pressures upon them (e.g., dispatcher reward structure). Motor coach companies should be aware of the fatigue implications of dispatcher pressure and take steps to mitigate it. However, detection of such subtle pressure is a challenge. Fortunately, the driver respondents report that dispatcher pressure is not a prevalent occurrence in the sample firms.

Customer pressures have a bearing on fatigue as well. Drivers and other members of companies feel the pressure to respond to customer requests. These demands, and the pressures felt by the companies and drivers to meet them, create the atmosphere within which drivers' perceptions related to driving tired and accepting trips or bending rules is formed.

While company safety culture and company policies that minimize nighttime driving were not statistically significant in the final testing of the model, they were important in the initial screening of indicators. Carrier support for driving safety is important in developing necessary communications among customers, management, and drivers on safety and operational issues. The quality of this support will be reflected through the effectiveness in managing customer demands and driver assignments. Carrier support systems are the conduit through which customer and driver perceptions of service and safety are formed. Carrier support is also an important channel for assisting drivers in their personal time and life style management.

In spite of these pressures, drivers apparently like what they do. Respondents were clearly stable in their employment situations (i.e., the average driver had 20 years of CMV experience). Thus, drivers apparently find many more positive aspects to their jobs than negative.

In conclusion, the results of the motor coach company study provide substantial support for the CMV Driver Fatigue Model. This study provides useful insights for individual and company management on the underlying factors related to carrier scheduling practices that influence driver fatigue. These practices should be the focus of efforts to minimize driver fatigue, including effective education for individual employees which should include open discussion of carrier practices and their impacts on individual drivers, personal time and fatigue management, and customer services practices as they relate to driver fatigue. Educational efforts should also be aimed at customers to help them better understand both the capabilities and limitations of services provided by motor coach companies, regulatory requirements, and potential fatigue impacts of customer demands during trips. Lastly, the carrier's ability to hire and retain sufficient drivers to reduce the individual pressures associated with healthy demand is critical. Having a sufficient number of drivers to operate a carrier's vehicle fleet for expected and planned demand is fundamentally important in reducing the scheduling pressures associated with driver fatigue.