Family medicine residents' risk of adverse motor vehicle events: a comparison between rural and urban placements.

BACKGROUND
Sleep deprivation and fatigue are associated with long and irregular work hours. These work patterns are common to medical residents. Motor vehicle crashes (MVCs) are a leading cause of injury related deaths in Canada, with MVC fatality rates in rural areas up to three times higher than in urban areas.


OBJECTIVES
To: 1) examine the number of adverse motor vehicle events (AMVEs) in family medicine residents in Canada; 2) assess whether residents with rural placements are at greater risk of experiencing AMVEs than urban residents; and 3) determine if family medicine residency programs across Canada have travel policies in place.


METHODOLOGY
A prospective, cross-sectional study, using a national survey of second-year family medicine residents.


RESULTS
A higher percentage of rural residents reported AMVEs than urban residents. The trend was for rural residents to be involved in more MVCs during residency, while urban residents were more likely to be involved in close calls. The majority of Canadian medical schools do not have resident travel policies in place.


CONCLUSION
AMVEs are common in family medicine residents, with a trend for the number of MVCs to be greater for rural residents. These data support the need for development and incorporation of travel policies by medical schools.


Introduction
Motor vehicle crashes (MVCs) are a leading cause of injury related deaths in Canada. 1 In addition to factors such as amount driven, speed, and seat belt use, driver fatigue and/or sleepiness are known risk factors for adverse motor vehicle events (AMVE), including crashes. [2][3][4][5] Medical residents may be especially susceptible to MVCs after a night on call due to lack of sleep. 2,5 In a survey by Steier and colleagues, 6 physicians who had been in a MVC within the previous year reported that they remembered being sleepy immediately before the MVC. Forty-nine percent of physicians in another study 3 reported that they had fallen asleep at the wheel of their vehicle, with 90% of these events occurring after a night on call. In a study of emergency medicine residents, Steele et al. 7 found that 74% of MVCs and 80% of near-crashes were reported after working a night shift compared to after working day and evening shifts.
In a number of studies, the effects of sleep deprivation on performance have been compared to alcohol intoxication. [8][9][10][11] For example, in an examination of male drivers aged 20-50 years, Fairclough and Graham 12 found that sleep-deprived participants made critical safety errors (i.e., errors that would be likely to cause a crash, such as lane crossing), similar to individuals with a blood alcohol content (BAC) of between 0.08-0.10% during simulated driving performance. Arendt and colleagues 13 compared the neurobehavioral performance of medical residents during a heavy call rotation (80-90 hours per week with night call every fourth night) to those during a light call rotation (44 hours per week, with night call only if the on-call resident became ill). Results indicated that the neurobehavioral performance of residents after a heavy call rotation was similar to that of residents with a BAC of 0.04-0.05% after a light call rotation, suggesting that the judgment and thinking abilities are impaired by sleep deprivation as much as by alcohol consumption.
Fatigue is not the only factor that can affect MVC risk. There have been several studies that have shown that MVC risk 14,15 and fatalities [14][15][16] are significantly higher in rural than in urban areas, [14][15][16] with the pattern of findings consistent whether using national data, 17 province-specific data, 15 or data from special populations. 15,17 Zwerling and colleagues 14 found that fatal MVC incidence density (number of fatal crashes/number of miles per 100 million miles driven) in the United States was 2.86 in rural areas compared to 1.28 in urban areas, and the crash injury rate (number of crashes with an injury/number of all crashes per 1,000 crashes) was 372.25 in rural areas compared to 331.55 for those in urban areas. Zwerling offered several explanations for these differences including the increased severity of MVCs in rural versus urban areas, a higher preponderance of people not wearing seat belts in rural areas compared to those in urban areas, and increased medical transport time to care in rural areas. Kmet and Macarthur 15 found similar results in a study of fatal MVCs in Alberta, Canada. In that study, the fatal MVC rate per 100,000 individuals was 11.3 in rural areas, a rate that was nearly six times higher than in urban areas of the province (1.9 per 100,000 individuals). Anecdotal evidence from the Director of the Rural Medicine Program at the University of Alberta in Edmonton, Canada, suggests that medical residents completing their rural rotation may be at higher risk of AMVEs than their urban counterparts. If true, the findings have important safety implications for residents completing rural residency rotations. As such, information on the risks of driving while sleep deprived could be incorporated into medical residency training programs and policies could be developed to minimize medical residents driving while sleepy in both rural and urban residency rotations.
The purpose of this study was threefold: 1) to examine the number of AMVEs in family medicine residents in Canada; 2) to assess whether family medicine residents with rural placements are at greater risk of experiencing AMVEs than their urban counterparts; and 3) to determine if family medicine residency programs across Canada have policies in place related to resident travel during residency. e30

Adverse Motor Vehicle Events (AMVEs)
Survey methodology was used for data collection relative to the first two objectives. Recruitment of PGY2 residents in family medicine was done through their respective departments of family medicine in universities across Canada. The program secretary of each residency program acted as the primary contact for residents within her/his respective program. Documents outlining the purpose of the study and the study procedure were provided to each program secretary. The program secretary was also provided with documents for the residents, with a request to forward the documents to each of the PGY2 residents. That contact occurred during the last two months of their residency.
Residents were asked to provide informed consent and complete a structured, web-based questionnaire by following an electronic link that was sent in the email. To ensure anonymity, the survey software automatically assigned each resident a computergenerated study identification number. A reminder notice was sent out to all residents after two weeks, asking those individuals who had not completed the questionnaire to do so. This process was repeated after a further 2 weeks.

Policies related to family medicine resident travel during residency
To determine if family medicine residency programs across Canada have residency travel policies in place, the following steps were taken. First, a search on the website of each school that has a family medicine program was performed. Second, if a travel policy could not be found through a search, each school was contacted and asked whether a travel policy was in place, and if so, the travel policy document was requested from that school (see Appendix A for an example from one school). The travel policies were examined in terms of their content and level of detail.
Ethical approval for both phases of the research was obtained from the University of Alberta's Health Research Ethics Board (Panel B) and from each of the individual medical schools when its own institutional approval was required.

Adverse Motor Vehicle Events
Descriptive statistics (means, standard deviations, frequencies) were used to describe the sample, with t-tests/ANOVAs (as appropriate) and chi-square analyses used to test for differences between groups on demographic and primary outcome measures. A liberal alpha value of 0.10 was chosen given the exploratory nature of the study and the fact that a Type II error (i.e., saying there is no difference between the rural and urban residents on the AMV measures when in fact there is) was of a more important concern in the present study, with a potential to inform driving policy in medical schools. Increasing alpha from 0.05 to 0.10 decreases the chances of making a Type II error. 18 The data were analyzed in aggregate form (overall and urban vs. rural) to ensure participant anonymity. Due to low response rates in some provinces, data analyses by province were not performed.

Policy related to family medicine resident travel during residency
Descriptive statistics (frequencies) were used to analyze the data related to family medicine resident travel policy.

Adverse Motor Vehicle Events
Sample as a whole One hundred and forty-one PGY2 residents completed the survey. From a national perspective, percentages of residents who completed the survey were the highest in Alberta and Manitoba (67% and 45%, respectively), whereas the lowest percentages of residents were observed in Quebec and Saskatchewan (16% and 11%, respectively) ( Table 1). Residents were asked the reason for their AMVE. Thirty-four percent cited fatigue, 25% cited driver inattention, 24% cited the weather, 12% cited road conditions, and 5% cited a wildlife encounter as the reason for the AMVE.
Rural vs. Urban The mean age of both rural and urban respondents was 32 years. For both locations, a higher percentage of females responded to the survey (68% urban and 59% rural). The difference in response rate as a function of gender and location was not statistically significant (p = 0.35). Rural residents had driven fewer years on average (12.48 years vs. 14.05 years for urban residents), but the difference was not statistically significant (p = 0.16). Total kilometres driven during the first 22 months of residency and kilometres driven specifically for residency purposes differed significantly between the two groups, with rural residents driving on average a greater number of kilometres (37,103 vs. 22,528, respectively) (p < 0.001) and more than double the number of kilometres on average for residency purposes (28,804 vs. 12,663) (p < 0.001).  † The data were converted to create two groups (residents with no crash and residents with at least one crash) for rural and urban residents, with associations tested using chi-square test. ‡ The data were converted to create two groups (residents with no close call and residents with at least one close call) for rural and urban residents, with associations tested using chi-square test. § Adverse motor vehicle events defined as a crash or a close call. || Data presented represent the number of times each reason was cited. Not all respondents provided a reason for their AMVE. Some provided multiple reasons for a single event.  Table 3). As noted previously, two of the MVCs resulted in injury (one urban and one rural resident), with the rural resident's crash resulting in the need for medical attention ( Table 2).
Fifty-seven of the 112 urban residents (50.9%) had a total of 157 close calls (Mean = 1.40 close calls/resident) compared to 19 of the 29 rural residents (65.5%) who had a total of 49 close calls (Mean = 1.69 close calls/resident) ( Table 2). However, the differences in the percentages of residents involved in close calls as well as the number of close calls per resident for the two groups of residents were both determined not significant (p = 0.21 and p = 0.42, respectively). The average of work-related close calls when adjusted for exposure (number of close calls/10,000 kilometres driven) was higher for urban residents (2.74) than for rural residents (0.62), though this was not statistically significant (p = 0.29) ( Table 3). Urban residents reported more close calls after work (60%) while rural residents reported that most of their close calls were unrelated to work (45%). However, rural residents still experienced a high rate of close calls after work (35%). Finally, a greater percentage of rural residents reported having received a citation (45%) versus 36% for urban residents, although the difference was not statistically significant (p = 0.39) ( Table 2).
Residents were also asked what they thought was the potential cause of their AMVE (driver inattention, fatigue, weather, wildlife, road conditions, or other). A higher percentage of urban residents identified fatigue as the cause for the AMVE (40%) followed by inattention (27%) and weather conditions (24%), while rural residents were most likely to blame the weather (23%), fatigue (21%), and inattention (21%) as causes of AMVEs. These differences were not statistically significant. Not surprisingly, compared to urban residents, rural residents were more likely to identify 'wild animals on the road' as the cause of an AMVE (19% vs. 0%) (p < 0.001). Rural residents also were more likely than urban residents to identify [poor] road conditions as the cause of an AMVE (17% vs. 9%, respectively) (p = 0.03).
Finally, given a somewhat lower than expected response rate and the fact that the observed mean differences for the four outcome variables were determined not to be statistically significant (see Table 3), we performed power analysis in an attempt to explain non-significant results. However, due to the lack of research in this area (i.e., AMV events among medical residents), it was difficult to make any hypothesis with respect to the effect sizes to be expected. At the same time, we had no grounds to expect high effect sizes, and thus, the power analysis was based on the observed effect sizes and the liberal alpha level of 0.10. Increasing the alpha level from 0.05 to 0.10 increases statistical power because the null hypothesis (i.e., no difference) will be rejected more often, and consequently, the true alternative hypothesis (i.e., there is a difference) will have a greater chance of being accepted (i.e., power). 18 The observed average effect size (Cohen's d) for the four outcome variables was 0.20 (Table 3). Based on this effect size, and with a power of at least 80%, 310 participants would be required for each group to obtain significant results at the chosen alpha level of 0.10. This could potentially have been achieved with a higher response rate and, most importantly, the full participation of medical schools e35 in the present study. If all the medical schools in Canada had participated in the study, the total number of family residents available for surveying would have been 1049, and depending on how each medical school defines their rural and urban residencies, in total between 250 and 350 rural residents could have been expected during our survey period.

Policy related to family medicine resident travel during residency
To determine if family medicine residency programs across Canada have policies in place related to resident travel during residency, a national survey was conducted, with all 14 English speaking schools contacted by email. Follow-up phone calls were made to schools not responding to emails. Of the 14 schools that were sent emails or follow-up phone calls, 12 responded representing an 86% response rate. The majority of the responding schools (7, or 58%) did not have a policy in place as of August, 2011 (Table 4).

Memorial University
No policy in place * University did not respond to requests for information.

Discussion
In our national survey of second year family medicine residents, the risk of AMVEs overall was high for both urban and rural residents, with rural residents significantly more likely to be involved in a MVC. There were also significantly more MVCs per rural resident than urban. After adjusting for exposure, rural residents also had a higher number of crashes and number of work related crashes per 100,000 kilometres driven, but those differences failed to reach statistical significance. Finally, urban residents reported more close calls when adjusted for exposure compared to rural residents.
The observed trend for rural residents to be involved in more MVCs and for urban residents to be involved in more close calls, when adjusted for exposure, is unexpected. The design of rural roads (e.g., narrow, more curves, faded markings, etc.) may be less safe than urban roads, potentially leading to higher accident rates. 19 Conversely, higher vehicle volume and increased prevalence of other road users in urban locations (e.g., pedestrians, cyclists) may account for the higher number of close calls reported by urban residents. When compared to other studies of medical residents, a higher percentage of residents in the current study experienced MVCs (22% vs. 8%, 20 8%, 7 and 13% 6 ) with more MVCs per resident in the current study (0.28 vs. 0.14 2 ). Methodological differences may account for these findings.
Landrigan and colleagues, 20 using prospective methodology, had pediatric residents from three large pediatric training programs in Boston, Stanford, and Washington complete daily logs on hours worked, hours of sleep, as well as MVCs and near misses. The data were collected the spring before and after the introduction of work hour limits for residents. It may be that the completion of daily logs documenting the number of hours worked and amount of sleep heighted the residents' awareness of the effects of fatigue on routine activities such as driving, resulting in a modification of behaviour. A shorter study time (12 months vs. 22 months for the current investigation) also helps to explain the differences in MVC rate between the two studies. Prospective methodology was also used in the study of first year residents conducted by Barger  , 38 physicians, 37 nurses, and 40 hi-tech workers were asked to self-report MVCs in the last year. The overall reported crash rate (13%) was for the sample as a whole. It is reasonable to assume that nurses and hi-tech workers may not be as 'at-risk' for MVCs due to shorter and more regular work hours compared to their physician counterparts. Thus, the inclusion of two groups of participants with a lower risk of MVCs, as well as the shorter study period, may have resulted in a lower risk of MVCs than that found in the current investigation. Finally, none of the studies investigating medical residents' risk of MVCs have stratified their sample by urban and rural placement.
Interestingly, in our study, a higher percentage of AMVEs were 'unrelated to work' for rural residents. It may be that the nature of rural residency is such that the need for travel and distances travelled for 'after work' activities (e.g., shopping, entertainment, etc.) are greater for rural residents than for urban residents. It is also the case that travel that is 'unrelated to work' (e.g., travelling into urban centres on days off) is greater for rural residents than for their urban counterparts, increasing the opportunity for an AMVE. Irrespective of time of occurrence, the higher percentage of AMVEs for rural residents is cause for concern and needs to be explored in future research. Future research should also include residents at all schools with family medicine programs in order to increase the sample size. It also would be useful to examine AMVEs of residents before driving policies were implemented at schools and then after implementation in order to see if the changes in policies resulted in change.
Despite the documented relationship between sleepiness and AMVEs in medical residents, it appears that few medical schools offer advice on the role of sleepiness and driving. No published studies on this topic were found in Canadian schools. There is, however, published literature from the United Kingdom, with one study indicating that only 6 medical schools offered students advice on how to avoid MVCs. 21 Our survey results from programs in Canada indicate that the majority of programs do not have travel policies in place for residents (Table  4). For those programs that do have policies, the policies vary in detail and the circumstances covered.
Most of the policies state that residents should not drive in inclement weather and should not be on call prior to driving a long distance. At the Department of Family medicine at the University of Alberta, we have instituted a travel policy to account for safe driving conditions. The travel policy applies to the rural stream of the family medicine residency program (Appendix A). The travel policy applies to rural residents only, and not to urban residents, as it documents policy on driving long distances in poor weather conditions, an issue not concerning the urban program. For example, urban program residents are excused from mandatory academic activities if they are situated more than 50 e37 kilometers from the city. Note, however, that in the rural program this distance is considerably farther at 350 kilometers which increases the amount of time rural residents spend on the road. Based on the results from our survey and a review of current practices regarding driving policies for residents, our recommendation would be for programs to include a formal policy related to residency travel within their departments.
It is of interest that recent guidelines regarding resident duty hours have been published by the Accreditation Council for Graduate Medical Education (ACGME) in the United States. 22 The guidelines dictate the number of consecutive and weekly hours that residents are permitted to work as a means of managing the adverse effects of sleep deprivation. Specifically, in the ACGME guidelines, it states that residents can work no more than 80 hours per week averaged over a 4 week period and that all residents will be assigned a minimum of one day free of duty per week when averaged over 4 weeks. In addition, duty hours cannot exceed 16 hours for PGY1 residents and 24 hours for residents in PGY2 and above. The use of 'strategic napping' also is recommended for overnight shifts. 22 Currently, there are no consensus guidelines for Canadian residents and restrictions can vary by province. For example, in Manitoba, residents are limited to working 89 hours per week, 23 whereas in the Maritime Provinces, the limit for resident work is 90 hours per week 24 (both averaged over a 4 week period). Residents in Quebec are limited to 78 hours of work per week over a 28-day rotation and no more than 16 hours per shift. 25 Even with these guidelines, it may be that residents are working in excess of the recommended maximum, potentially increasing the danger to patients via medical errors and to themselves via increased crash risk following an extended work shift. Notably, the Canadian Association of Internes and Residents (CAIR) has recently released a position paper on resident duty hours. In that paper CAIR "calls on all PGME departments, employers, governments, and other relevant stakeholders to ensure…that physicians' duty hours must be managed such that they do not in any way endanger their health or the health of patients". 26 It is important to acknowledge the limitations of our study. First, the data are based on self-report which has the potential to influence the accuracy of the data. Previous research on AMVEs using self-report indicates that the events often are underreported. 27 Thus, the data presented here may actually underrepresent the scope of the problem. On the other hand, sampling bias may account for our findings in that residents experiencing AMVEs may have been more likely to complete the survey. Unfortunately, we have no way of determining whether sampling bias was present in our research. Lastly, the overall response rate from residents, based on participating schools, was 34%, a response rate that is consistent with rates reported for web-surveys. 28 The somewhat low response rate was in part dictated by the school's policy of access to residents (Table 1). Despite the limitations, the strength of this research is that, to our knowledge, this study is the first attempt to survey, on a national level, the frequency of AMVEs for family medicine residents in Canada.
In order to decrease the probability that AMVEs occur, it is important to take driving safety into consideration when planning educational activities. This research has the potential to inform on policies related to safety issues for family medicine residency programs in both rural and urban locations across Canada. The research also helps to increase awareness of factors that may lead to AMVEs (e.g., fatigue, distraction, etc.), which in turn could lead to future enhancement of driving safety through behaviour change.
Finally, as family medicine program expansion incorporates more learners into the rural environment, driving safety and travel policies become a more important consideration. Driving safety may be a motivational force to explore other avenues for structured learning such as web-based interfaces and video-conferencing. The latter are not without disadvantages and require support from the respective Information Technology (IT) departments at each university. Sometimes IT departments are slow to meet these challenges. However, if there is a concern for resident safety, then there is reason to make the support of alternative methods for distributed learning a priority.