General setting. In 2002 Ontario had a population of about 12 million people, 8.4 million licensed drivers, and 9,000 primary care physicians. During that year 200,000 crashes occurred, resulting in 873 fatalities.23 Fewer than 0.1% of Ontario drivers per year had their license reviewed for medical conditions.24 Conditions requiring notification were uniform in all regions as set out in section 203[1] of the Highway Traffic Act (first enacted in 1968), and pertain to generalist and specialist physicians. The regulations underwent no amendments during the study and were endorsed by local and national medical regulating bodies. Similar to other mandatory reporting laws, physicians were not expected to be accurate or precise in diagnoses but were obliged to notify licensing agencies so that a review by a separate board could be conducted.25 Physicians reporting in Ontario were protected against legal liability for the associated breach of patient confidentiality.26

Included patients. We identified consecutive trauma patients admitted to Canada’s largest trauma centre, Sunnybrook Health Sciences Centre, because this hospital receives trauma patients from crashes throughout Ontario (based on Injury Severity Scores of 13 or higher) and regardless of chronic conditions.27 The accrual interval spanned 30 June 1996 to 30 June 2001 to provide sufficient sample size and observation time before the day of the crash. We included only individuals in the driver’s position at impact because passengers and pedestrians would not be directly targeted by policies regarding restricted licenses. Drivers of motorcycles were included; drivers of bicycles, snowmobiles, and recreational vehicles were excluded since they are not the usual consideration of regulations for unfit drivers. All screening was conducted blind to at-fault considerations such as excessive speed, driver inattention, improper lane change, or other non-medical fallibilities.

Individual characteristics. We obtained baseline data for each patient by chart review at the trauma center using a structured explicit standardized approach with inbuilt software checks, real-time iteration and regular random audits.28-31 The time and location of the crash were based on paramedic reports. We used admission, progress, and discharge records (including nursing, social work, and other professionals) for ascertainment of alcohol abuse as evidenced by patient report, collateral family history, findings on examination, or positive testing.32-33 Similarly, clinical evaluation at the trauma centre provided evidence of antecedent cardiovascular diseases (unstable angina, chronic arrhythmia, stroke, or extensive vascular disease) and neurological disorders (dementia, seizure, mental retardation, or active psychiatric disorders).34-35 We did not include sleep disorders, metabolic diseases, vision or hearing limitations, musculoskeletal disabilities, and general debility because these are difficult to determine in the aftermath of a crash.36 Throughout, we counted only conditions judged by the responsible trauma physician as chronic and contributing to the patient’s injury.

Database linkages. We linked patients to their Ontario Health Insurance Plan records by using an encrypted version of their individual health card number.37 Matches were checked by verifying birth date and included spot checks to reconcile inconsistencies in name (e.g., “William” and “Bill”). Faulty or missing matches were addressed by manual searches according to first name, last name, birth date, and sex. Similar methods were used to link each patient to Ministry of Transportation records connecting to driver licences (with reconciliations denoted as “probable” and “possible” matches according to government protocols housed at the Ministry of Transportation). This project was approved by the research ethics committee of Sunnybrook Health Science Centre and conducted according to confidentiality protocols of the Institute for Clinical Evaluative Studies in Ontario.

Previous health care. We searched each person’s records backward in time from their crash until 1 July 1991 (start date of the health care database) to identify previous outpatient visits to a physician anywhere in Ontario. If the patient had multiple prior visits we also analyzed the visit that most closely preceded the crash. For the prior visit we recorded the date, diagnostic category, fee code, and a unique identifier for the physician providing the visit. Finally, because trauma charts can be fallible, we also conducted a comprehensive analysis of all past outpatient and inpatient records for any mention of alcohol abuse, cardiac disease, or neurological disorders.

Previous driving reports. Past reports on driving notifications, restrictions, and suspensions were obtained by contacting the Driver Improvement Office of the Medical Review Section of the Ontario Ministry of Transportation. Special attention was again directed to avoid faulty or missing matches using the standardized in-house protocols of the Ministry of Transportation. Governmental transportation records were then searched backward from the crash date for any reported notification of the individual patient by a physician. These analyses were conducted blind to whether the patient had a reportable condition when examined in the trauma centre. For each notification we extracted information about the interval between the previous notification and the subsequent crash. We did not have data on either the process or the outcome of cases reviewed by the Ministry of Transportation.

Statistical analysis. The primary outcome was the prevalence of previous reporting to the Ministry of Transportation among patients with reportable conditions. No adjustment was made for misclassification of patients who might have been incompletely documented at the trauma centre and incorrectly classified as having no reportable condition. Categorical data were compared using the chi-square test, continuous data using the t-test, and time intervals using the Kaplan-Meier technique. All statistics were two-tailed estimates and based on all available data.

General overview. A total of 4,208 patients were admitted to the trauma centre during the 5-year interval, of whom 1,605 were drivers involved in motor vehicle crashes and included for analysis. No repeat hospital admissions were observed in the study cohort during the study interval. Validated outpatient health care record linkages could be verified for 97% as a definite match (unmatched = 3%). Validated driver record linkages could be verified for 80% as a definite match, 10% as a probable match, and 2% as a possible match (unmatched = 8%). In total, 596 individuals had reportable conditions documented at the trauma centre and 1,009 did not have a reportable condition. The most common reportable condition was alcohol abuse (72%), followed by cardiac disease (23%) and neurological disorders (5%).

Injury comparisons. The average driver with a reportable condition was marginally older and more likely to be male than those who did not have a reportable condition (Table 1). The two groups required a similar number of units of blood transfused on average (2.1 vs 2.4, respectively), corresponding to a relative difference of about 12% (95% CI -4 to 29) and further suggesting that injury severity was substantial in both groups. Patients with reportable conditions averaged marginally fewer trips to the operating room (0.9 vs 1.1, respectively) but a similar median length of stay (14 days vs 14 days, respectively). The patients with reportable conditions accounted for a total of 53 deaths, 551 surgeries, 1252 units of blood, and 8440 hospital days (corresponding to about $3 million in hospital costs).38

Previous visits. About 98% of all patients (n = 1537) had previously visited a doctor, of whom 83% had visited in the previous year and 14% in the previous week. The total number of visits in the last 5 years for any reason among those with reportable conditions was 20,505, reflecting contact with 2,332 distinct physicians. More than 80% of all patients had at least one mention of a reportable condition when based on database linkage rather than trauma chart review (Table 2). Almost half of all patients had 5 or more visits in the year immediately before the crash. The most common reasons for the most recent previous visit for all patients were respiratory tract infections, minor injury or general care. The most common professional code for all visits, an “intermediate assessment,” did not differ between the two groups and corresponded to a physician fee of $24.80.

Mandatory reporting. A total of 28 of the 1,605 patients were reported by physicians to the Ministry of Transportation prior to their crash. The overall frequency of reporting for those with reportable conditions was only marginally higher than the baseline frequency for those without reportable conditions (3% vs 1%, p = 0.003). This non-zero baseline frequency was explained by routine reporting of those with commercial driver licenses (for whom mandatory reports must be filed). The highest frequency of reporting was for patients diagnosed with neurological conditions (6/29, 21%) and the lowest frequency of reporting was for patients diagnosed with alcohol abuse (7/429, 2%). No other patient characteristic (Table 1) was independently associated with rates of reporting. The median lag between the prior report and the patient’s subsequent crash was 3.5 years (95% CI 2.4–4.5).

Table 1. Patient characteristics [view]

Table 2. Reasons for previous physician visits (excludes those with no previous visits) [view]

We studied drivers involved in life-threatening motor vehicle crashes to assess physicians’ compliance with mandatory reporting in real-world circumstances. We found that about one-third of trauma patients had a prior medical condition that might have impaired their ability to operate a motor vehicle, may have contributed to injury, and was potentially reportable to licensing authorities. Almost all these drivers had seen a physician before the crash. In contrast to the prevailing law, almost none of these drivers were reported by physicians to licensing authorities (and the few reports were made more than 3 years prior to the crash). Alcohol abuse was the most frequent reportable condition contributing to a serious crash (with a rate similar to that reported in past literature39) and the least frequent reason for notification.

The largest limitation of our research is the fallible manner of ascertaining chronic medical conditions in trauma patients. The Ontario law is written using vague language (“a condition that may make it dangerous”), which could be interpreted to include minor issues of any severity. Our specific list of reportable conditions was based on the Canadian Medical Association handbook34 but did not include all diseases (e.g., sleep apnea). Our methodology relied on chart review, which is prone to false negatives. The specific charts reviewed were from a trauma service, a setting in which attention may not focus on chronic conditions and where detailed medical ascertainment is sometimes not feasible. Together, these limitations imply that our study is conservative in design and underestimates the extent of under-reporting of drivers to licensing authorities.

The data could have underestimated physicians’ compliance with the law if multiple other unobserved patients visited a physician during the same interval, were reported to licensing authorities, and subsequently renounced driving. However, this possibility is unlikely for three reasons. First, the total frequency of reporting is low, far below the prevalence of conditions in the population. Second, alcohol abuse is the most common reportable condition we observed in drivers, yet it is notoriously difficult to detect in primary care and notoriously linked to poor compliance with restrictions. Third, the relative risk of a crash from unreported cardiac and neurological conditions would need to be unrealistically high (odds-ratios beyond 100) to have these conditions both commonly present yet rarely reported in unobserved non-trauma patients (Appendix).

Our research suggests that opportunities to prevent serious crashes are missed by physicians even in communities with mandatory reporting (seven provinces and all territories in Canada have such laws). Physician under-reporting in this study could have several explanations. An intent to preserve the patient’s quality of life can explain some cases, but would not explain the extent of under-reporting given that the driver is the most common person disabled in a crash in North America.1,40 A lack of attention is another explanation, since patients rarely seek care for a driving assessment.41 The failure to report unfit drivers also indicates that reporting requirements are not straightforward (similar to the interpretation needed when reporting aviators, air traffic controllers, and selected railway workers).42,43

Our findings agree with anecdotal reports, editorials, and depositions that physicians are unsure in many cases about when to report medically unfit drivers and that mandatory laws do not eliminate this uncertainty.44-47 Moreover, physicians don’t have all the scientific evidence: the contribution of many diseases to driver risk is speculative, risk will vary if the condition is inactive or under control, and the Ministry’s own reporting form provides little data for guidance.48 Physicians are often unaware of each patient’s driving abilities (e.g., attention), impairments (e.g., alcoholism), or attitudes (e.g., speeding) and thereby cannot form a judgment of dangerousness.49,50 A crackdown to increase reporting by physicians, moreover, might deter patients from seeking medical care.51 When it comes to a patient’s driving habits, for example, doctors are less knowledgeable than family members in many ways.

Reporting benefits have limitations even if clinicians are perfectly knowledgeable about their patients, equipped with solid scientific evidence, and fully compliant with the law: healthy young adults still represent the majority of crashes and have no reportable conditions.52-54 Drivers who have reportable conditions may over-contribute to crashes relative to the general population, yet the associations are not overwhelming. Further, drivers who are reported to authorities are not always compliant with restrictions.55-57 Also, licensing authorities can become overworked and may make fallible judgments about exemptions and reinstatements. Finally, mandatory reporting laws impose affirmative duty on physicians using broad language and thereby conceivably extend to almost every patient seen in practice.

Our study also underscores how difficult it is to conduct research to test whether legislation has achieved its intended effect. Evaluations cannot be designed as randomized blinded trials, since this would violate the principle of uniform justice. Many years need to elapse between the enactment of a law and subsequent evaluation to provide sufficient sample size and eliminate chance findings. Lawmakers who enact legislation also have little incentive to follow up and check on anticipated consequences or adverse effects. No study applies universally due to the subtleties of culture that vary across regions. Finally, the indirect deterrent effect upon other individuals, the alternative uses for finite legal resources, and the theoretical effectiveness under ideal circumstances are inevitable questions that are difficult to study in a quantitative manner.

In summary, our data suggest that mandatory reporting in Ontario does not achieve its stated purpose; however, the data do not indicate whether regulations should be reduced or expanded. Reducing regulations would decrease the liability for physicians and might encourage attention toward alternative safety policies such as roadside police or periodic driver testing. Expanding regulations would extend societal participation—analogous to suspected child abuse laws—by obliging others in the community to notify authorities regarding mishaps (e.g., hospitals, vehicle insurers or mechanics). Less radical changes are possible that might focus on screening, detection, and counselling with respect to alcoholism.58, 59 In the absence of policy changes, physicians seem mostly inclined toward resuscitation and treatment in the aftermath of a crash.

Appendix. [view]