Article Text

Download PDFPDF

Number of medications and road traffic crashes in senior Swedish drivers: a population-based matched case-control study
  1. Joel Monárrez-Espino,
  2. Lucie Laflamme,
  3. Berty Elling,
  4. Jette Möller
  1. Department of Public Health Sciences, Karolinska Institutet, Stockholm, Sweden
  1. Correspondence to Dr Joel Monárrez-Espino, Department of Public Health Sciences, Karolinska Institutet, Widerströmska huset, 4th floor, Tomtebodavägen 18A, Stockholm SE-17177, Sweden; Joel.Monarrez-Espino{at} jette.moller{at}


Objective This study investigated the relationship between the number of different medications dispensed (NDMD) to senior drivers and the risk of injurious road traffic crashes (RTCs).

Design A matched case-control study with data from various population-based national registers was conducted. Cases were drivers aged 50–80 years involved in a crash in Sweden between 2005 and 2009. Only the first non-alcohol-related RTC was studied. Controls were residents with a valid license who did not crash. Four controls were matched by sex, age (year and month of birth), and place of residence. Exposure to NDMD prior to the crash date was assessed using four time periods: 1–8, 1–15, 1–30 and 1–90 days. Conditional logistic regression was used and analyses adjusted for civil status, occupation and dispensation of medications affecting the cardiovascular or nervous systems (C/N).

Results ORs (95% CI) increased progressively with the NDMD. For 1–8 days the OR ranged from 1.15 (1.10 to 1.20) for 1–2 medications to 1.27 (1.13 to 1.42) for five or more medications. The magnitude of the effect declined gradually with longer exposure periods, but remained when five or more medications were used. Adjusting for C/N medications resulted in slightly higher effects; for 1–8 days it ranged from 1.16 (1.10 to 1.23) for 1–2 medications to 1.35 (1.17 to 1.56) for five or more, with similar trends by exposure period. The highest effects were seen for single crashes and for drivers aged 66–80 years.

Conclusions The NDMD was linked to the likelihood of a senior driver being involved in an injurious RTC. The strength of the association steadily increased with increased NDMD, especially when medications were taken closer to the index date, or when more than five medications were dispensed.

Statistics from


The good health and longevity of elderly people has allowed an increasing number of individuals to continue driving up to a very advanced age.1 Older drivers are considered a risk group, as they seem to be over-represented in some road traffic crashes (RTCs) resulting in injuries or death, and have therefore become a concern in several countries.2 ,3 This has been attributed to impaired motor, sensory and cognitive faculties that may affect driving performance4–6 and crash experience.2 ,7 ,8 As a consequence, efforts to identify preventable risk factors related to the occurrence of RTCs in senior drivers have become a research priority.

From among those factors, the use of medicinal drugs has received particular attention,9 ,10 and the need for more research has been stressed.11 ,12 Thus far, however, evidence rests mostly on studies dealing with specific medical substances or drug classes expected to impair the driving competence, in particular due to adverse effects on the nervous system.13 ,14 Benzodiazepines are one such group of drugs showing consistent effects on driver involvement in RTCs.14 ,15 However, little attention has been paid to the potential adverse outcomes of the concurrent consumption of several medications, especially those apparently innocuous.

We are aware of three studies presenting risks for RTCs by number of medications used by senior drivers. A study from the early 1990s with drivers aged 65 years or more who were prescribed two or more psychoactive medications within 60 days of the reference date reported a significant increased risk for injurious RTCs (OR 2.0; 95% CI 1.0 to 4.0).16 A more recent register-based study from France also looked at drivers responsible for injurious RTCs showing an increasing risk trend by number of medications (OR, 95% CI one 1.14, 1.06 to 1.22; two 1.20, 1.17 to 1.43; three 1.86, 1.59 to 2.16; and four or more 1.88, 1.58 to 2.25).17 However, the analyses in both studies focused only on a selection of high-risk medications that could affect or are known to affect driving ability, and did not consider all kinds of medications. The only study with senior drivers relating the number of any type of prescribed medications with the occurrence of vehicle crashes was a small cohort of 174 individuals that showed non-significant increased risks (RR, 95% CI one or two 1.24, 0.71 to 2.18; three or more 1.58, 0.70 to 3.59).18

Other than RTCs, there is some evidence from a recent study showing that the simultaneous prescription of two or more medications of any type is associated with increased risk of fall injuries (OR, 95% CI 2.5, 1.3 to 4.8),19 a finding that raised the attention of a major medical journal.20

Apart from this limited evidence, no study has been published aimed at systematically investigating this phenomenon in drivers, and even more so among senior drivers where the combination of medications is common, as shown by a national survey of American adults revealing that nearly 40% of those aged 65 years or older used five or more different medications per week.21 Research has also shown that taking various medications at the same time implies a higher number of potentially inappropriate prescriptions, which in turn may lead to harmful side effects, some attributed to interactions between substances.22 ,23

Against this background, this study was embarked upon to shed more light on this question by studying the association between the number of different medications dispensed (NDMD) to senior drivers and the risk of them being involved in injurious RTCs.


Study design

A matched case-control study was adopted using data from various population-based registers linked through the unique personal identity number assigned to all residents in Sweden. The study focused on RTCs that occurred in Sweden between 1 July 2005 and 31 December 2009 among people aged 50–80 years.

The date and place, passenger position, main consequence and other characteristics related to the RTCs were extracted from the Swedish Traffic Accident Data Acquisition register,24 which contains data routinely collected by the police covering all RTCs occurring in the entire road traffic system resulting in death or significant injury requiring medical care.

The Swedish Prescribed Drug Register was used to extract the dispensation date and the type of medication based on the Anatomical Therapeutic Chemical (ATC) classification;25 this register captures electronically all medications dispensed to individual patients at Swedish pharmacies. The age and sex were extracted from the Population Register; the license issue dates and vehicle endorsement from the National Driver's License Register; and the civil status (categorised as married, widowed, divorced, unmarried), place of residence (divided up into eight geographic areas), and occupation (10 occupational groups categorised into professionals, technicians and skilled workers) from the Longitudinal Integration Database for Health Insurance and Labour Market Studies. The study was approved by the Regional Ethical Review Committee in Stockholm, Sweden.

Selection of cases and controls

Cases were defined as drivers aged 50–80 years who were involved in a car, truck or bus crash during the study period. Only the first event was studied, and drivers reported as suspected of being under the influence of alcohol determined by the police at the time of the crash were excluded, as this is known to be a major determinant for RTCs.26 ,27

Eligible as controls were Swedish residents holding a valid driving license, who were not known to have been involved as a driver in a RTC for the duration of the study (01.07.2005–31.12.2009). Four random population controls were then selected and matched to each case by age (year and month of birth), sex and place of residence. Once controls were assigned to a case, they were censored so that they could no longer be assigned to any other case of equal characteristics. The number of medications were those dispensed for the specified time period prior to the index date for both cases and controls.

Exposure definitions

The NDMD for the following exposure periods defined in number of days prior to the index date were calculated based on the five-level ATC code: 1–8, 1–15, 1–30 and 1–90 days. These periods were all set before the index date, as the RTC itself could have led to the dispensing of medications (eg, analgesics or sedatives). The dispensation date was excluded on the assumption that the medication was not taken until the following day. For example, if the index date was 10 April 2006, then the exposure assessment for the 1–8-day period included the total NDMD between April 2 and 9 for both, cases and controls.

Statistical analyses

Frequencies were used to describe RTCs by drivers’ age and to compare cases and controls in terms of matched variables and potential confounders. For the matching procedure, place of residence was matched using the eight geographic areas, but the results were summarised into three regions for presentation purposes. Mean and SDs were compared between cases and controls for the NDMD by drivers’ age group for the exposure periods defined.

From the various confounding factors for the association between medication use and RTC in senior drivers, illness is probably the most relevant. Since we were unable to assess the drivers’ medical conditions, we decided to adjust the main analyses for medications affecting the cardiovascular and nervous systems (C/N), as surrogates for comorbidity. Also, when comparing available demographic and socioeconomic characteristics between cases and controls that could potentially be associated with both the use of medications and the risk of crashing, such as civil status and occupation, we observed differential distributions and effects across the categories compared, and were therefore adjusted for in the analyses.

Analyses were conducted to statistically assess the bivariate association between civil status (married, used as reference group), occupation (professionals, as reference group), and the dichotomous dispensation of any C/N medication for the four exposure periods defined, with the risk of injurious RTCs.

Conditional logistic regression was used to compute adjusted ORs with 95% CIs. The dependent variable was defined according to the case-control status. Regression models were built using NDMD as the main independent categorical variable coded as 0, 1–2, 3–4 and 5 or more different medications. The reference group was the group for which no medication was recorded. Models were adjusted for civil status, occupation, and for the dispensation of C/N medications for the corresponding exposure period analysed. Results were stratified by type of crash (all vs single-vehicle crashes) and age group (50–65 and 66–80 years).

Data management, merging and statistical analyses were performed using SAS software V.9.2 (SAS Institute, North Carolina, USA).


A total of 30 845 RTCs involving drivers aged 50–80 years were registered by the police. The majority of crashes involved two or more vehicles (85.6%), occurred in areas with a speed limit under 60 km/h (58%), during daylight (71.3%), and in clear weather conditions (78.8%). In 2.4% of the crashes, at least one person died, and in 16% someone was reported severely injured (ie, required hospital care).

More than two-thirds of the RTCs involved male drivers. Nearly 90% of the events occurred in central and southern Sweden. Drivers aged 66–80 years made up 28.5% of all those involved in RTCs. A higher proportion of cases was married at the time of the crash compared with the controls (55.6% vs 63.3%) and had lower-ranking occupations. More cases were dispensed at least one medication within 90 days prior to the index date compared with the controls (60.8 vs 57.2%), and particularly five or more medications (19.7 vs 16.2%). Also, the proportion of cases who received N medications during 90 days before the index date was significantly higher than the controls (22.0 vs 17.4%; p<0.001), but it was similar for dispensation of C medications (30.7 vs 30.2%; p=0.13) (table 1).

Table 1

Distribution of sociodemographic characteristics of cases and controls; study on medications and injurious road traffic crashes in senior drivers, Sweden 2005–2009

The mean NDMD was higher among the cases throughout the different periods evaluated, although the difference tended to decrease as the length of exposure periods increased. For instance, while cases received 20% more dispensations on average than the controls within 1–8 days prior to the index date, they obtained 16.2% more when the exposure period was 1–90 days. Elderly drivers were dispensed more medications on average than those aged 50–65 years over all the periods of exposure, but the relative difference in the mean NDMD between the cases and controls was similar across age groups (table 2).

Table 2

Differences in the mean number of medications dispensed between cases and controls by period of exposure prior to the index date; study on medications and injurious road traffic crashes among senior drivers, Sweden 2005–2009

Table 3 presents ORs for injurious RTCs by drivers’ civil status, occupation and dispensation of C/N system medications. Unmarried drivers, but particularly those divorced and widowed, showed significantly higher risks of injurious RTCs compared with the married (ORs 1.17, 1.51 and 1.56, respectively; p<0.05). Skilled workers also had an increased risk compared with professionals (OR 1.32; 95% CI 1.28 to 1.37). With respect to the dispensation of one or more nervous system medications prior to the index date, all exposure periods showed statistically significant increased ORs, but with a decreasing trend the longer the period. No significant associations were seen for cardiovascular medications.

Table 3

Matched ORs for injurious road traffic crashes among Swedish senior drivers by drivers’ civil status, occupation and dispensation of nervous and cardiovascular system medications

Table 4 shows that ORs increased progressively with NDMD, though the magnitude of the effect declined gradually with longer exposure periods, except for the dispensation of five or more medications where the effect remained or increased. The inclusion of C/N medications resulted in similar or even slightly higher effects; for instance, when only 1–2 medications were dispensed within the 1–8 exposure period, the OR was 1.15 (95% CI 1.10 to 1.20), but after controlling for C/N medications the adjusted OR was 1.16 (1.10 to 1.23). Further, ORs were notably higher when the analyses were constrained to single-vehicle crashes, but followed the same pattern as seen when all crashes were accounted for. Following the previous example for the 1–8-day period, adjusted ORs for C/N medications ranged from 1.26 (1.09 to 1.44) with 1 to 2 medications to 1.63 (1.17 to 2.27) with five or more medications.

Table 4

Adjusted ORs for injurious road traffic crashes among Swedish senior drivers depending on the number of different medications dispensed (full 5-level ATC code), stratified by type of crash and period of exposure prior the index date

Adjusted models stratified by the drivers’ age for all crashes showed higher odds among drivers aged 66–80 years when 1–2 and 3–4 medications were dispensed, with a gradual decline the longer the exposure period. However, when drivers received five or more medications, the adjusted ORs were higher among those aged 50–65 years, and remained relatively stable across the exposure periods. For instance, when five or more medications were dispensed within 1–8 days prior to the index date, the adjusted ORs were 1.43 (1.18 to 1.73) and 1.32 (1.06 to 1.64) for those aged 50–65 and 66–80 years, respectively, compared with 1.38 (1.30 to 1.48) and 1.28 (1.16 to 1.41) for the 1–90-day exposure period (table 5).

Table 5

Adjusted ORs for injurious road traffic crashes among Swedish senior drivers depending on the number of different medications dispensed (full 5-level ATC code) stratified by driver's age and period of exposure prior the index date


This is the first study presenting data for an association between the number of any kind of medications dispensed to senior drivers and the occurrence of RTCs. The main finding is that the NDMD relates to the likelihood of senior drivers to be involved in an injurious RTC, and that the strength of this association increases gradually with increasing number of medications. This, in turn, echoes the findings reported before16 ,17 for specific medicinal groups considered as high risk for driving impairment, but now in this study for all types of medications.

We excluded drivers suspected of alcohol use, as determined by the police at the crash location (∼2% of eligible cases after all other exclusion criteria were applied, n=657), because we were interested in restricting the sample to examine the contribution of NDMD rather than in replicating the finding of increased risk when combining medicines with alcohol.28–30

The analyses for the exposure to nervous system medications and the risk of crashing showed significantly elevated risks with a decreasing trend as the period of exposure increased, evidencing their potential role as confounders for the association between NDMD and the risk of crashing. This emphasises the importance of adjusting for type of medication when studying polypharmacy.

Surprisingly, when the analyses were adjusted by C/N medications the adjusted ORs resulted in similar or even slightly higher associations. This finding suggests an independent effect of number of medications on the risk of injurious RTCs.

Although it can be argued that adjusting for other unsafe medications could have resulted in loss of effect, this alone could not explain the increasing trends seen with NDMD. In fact, when we adjusted the models for all 14 ATC-groups (first letter), the positive effects and trends remained (data not shown). However, we decided to present the models adjusting only for C/N medications; the former was used as a proxy for cardiovascular disease (though it did not show any significant association), and the latter because that group of medications has shown more consistent elevated risks for RTCs.14

Although the type, frequency and dosage of medications dispensed to older people to some extent relate to the number, type and severity of the illnesses,31 ,32 some medications may also be prescribed incorrectly, leading to both expected and unexpected adverse effects,22 ,23 ,33 some of which being susceptible to act as a trigger for driver involvement in RTCs. It is also of note that even in cases where combinations of medications can be regarded as appropriate or safe, most potential side effects resulting from combining medications are yet to be documented. An additional preoccupation regarding the combination of medications is the fact that even for known adverse consequences, primary physicians can demonstrate insufficient knowledge regarding appropriate prescribing.34

In Sweden, 38% of individuals aged 70–79 years were dispensed five of more medications within a 3-month period in 2006,35 and the reported prevalence of inappropriate drug use among those aged ≥75 years was 17% in 2005 with potentially serious drug interactions accounting for 4.1% of all prescribed drugs.36 To the best of our knowledge, we have not been able to find prior studies that have looked at the association between the combined use of medications with established clinical interactions, and the risk of involvement in injurious RTCs. The increasing risk of crashing with the NDMD seen here could relate to the effect of such interactions and should prompt further studies.

On the other hand, if the number of medications truly reflects the person's health status, then those receiving more medications could have the higher risks of crashing attributable to their illnesses. Studies have documented that elderly drivers may die just before crashing due to a health-related condition, mainly cardiovascular disease.37 ,38 Some estimates suggest that the proportion of drivers who die due to an acute disease can range from 10% to 20% of all crashes among those aged ≥65 years.39 ,40 A significant increased risk of RTC has also been associated with having had medical contact within 1 month prior to the collision.41 Adjusting for cardiovascular medications aimed at partially controlling for this potential confounder.

Following this line of reasoning, our results suggest that NDMD could, in part, be reflecting the drivers’ medical conditions, but perhaps also the use of potentially harmful combinations of medications, particularly if those with well-established side effects are used.

Besides the association observed between the NDMD to senior drivers and the increased probability of being involved in RTCs, another important finding relates to the differential risks seen by the drivers’ age. As expected, we observed that older drivers (66–80 years) received more medications on average than those aged 50–65 years old. The results also showed that the risk for injurious RCTs with increasing NDMD tended to be somewhat higher in those aged 66–80 years, which seems plausible, since individuals in this age group are more frail (ie, more likely to be involved and injured in a RTC). The effect of age in itself and the comorbidity linked to it could explain this relative difference, as it could the types, doses and combination of medications dispensed, which are likely to be associated with the type and severity of the individuals’ illnesses too. However, it is worth noting that the risk for injurious RTCs was very similar between the groups (actually slightly higher in those aged 50–65 years) when surpassing the threshold of five of more NDMDs. One plausible explanation could be a higher tolerance to multiple medications among the oldest drivers, as they have been exposed to polypharmacy for a longer period. Looking at the effect of combinations of medications would also shed more light on these results.

The trends observed by exposure period indicated higher risks with shorter periods, as there was a higher chance that the medicines were actually taken and that their effect was present during the crash, which is consistent with studies on specific medications14 though the mechanisms remain to be determined. Yet, for five or more medications dispensed, the risks remained stable across the four exposure periods studied, suggesting a polypharmacy threshold regardless of the drivers’ age. Future studies could address the effects of chronic use of multiple medications, as other authors have done with that of single medicinal drugs,42 although defining the exposure period for chronic intake of combinations of medications is much more challenging.

More research is also needed to better understand the extent to which sociodemographic variables, such as civil status and occupation, could have modified the association observed in the current study.43 Herein, we observed that risks increased in persons with lower occupational status (eg, machinery operators) compared with highly qualified individuals (eg, managers and professionals), and we also found that unmarried individuals and, above all, the divorced and widowed, had higher risks compared with those married, as has been reported previously.43

Limitations inherent to this study design ought to be underlined. First, dispensation was used as a proxy for intake, as no data concerning actual consumption was available in the registers; differential non-compliance between cases and controls in the actual use of dispensed medications (ie, differential misclassification of exposure) could have led to biased estimates, particularly for shorter exposure periods where fewer dispensing events are accounted for.44 Second, we relied on crash involvement rather than responsibility; however, when we limited the analyses to single-vehicle crashes where only one driver can be assumed liable, the effects were stronger, and the patterns remained. Finally, we used driving license as a proxy for exposure to driving among the controls, but we were unable to compare the driving frequency and duration between cases and controls.

From a public health perspective, as in the case of falls where multiple prescribed medications increased the injury risks,19 ,45 our results provide evidence that appropriate prescription of medications should be promoted, and awareness of the potentially negative consequences of combined medications should be the object of public and medical information measures. For now, it could perhaps be more practical to counsel senior drivers on self-refraining from driving after taking several medications simultaneously, regardless of the type, and to advise them about the risks involved when using specific drugs (eg, benzodiazepines) where higher risks have been identified. Patients seem to need this kind of pragmatic approach.46 ,47 There is evidence from the USA showing that individuals, especially women, can self-impose driving restrictions for medical reasons,43 but they are not yet aware of the possibility of doing so because of the number of medications consumed.

This does not mean though that individuals can decide to use fewer medications in order to drive safely, as such decision should ultimately rest on the prescribing physician based on individual judgment. In fact, failure to take properly prescribed medications by ill persons without physician involvement could lead to reduced driving performance, as has been seen with untreated or poorly controlled pain48 ,49 with potentially serious consequences. It does, however, raise awareness of the potential effect of polypharmacy among senior drivers on road safety.

What is already known on the subject

  • Specific medicinal groups expected to impair the driving competence, mostly due to adverse effects on the central nervous system (eg, benzodiazepines), have been associated with driver involvement in road traffic crashes (RTCs).

  • Polypharmacy has been associated with an increased risk of fall injuries, but this relationship remains to be documented for RTCs.

  • There is limited evidence reporting association between the number of medications used and the risk of injurious RTCs for medical drugs considered as high risk for driving impairment, but not for number of medications independent of type.

What this study adds

  • The number of different dispensed medications to senior drivers is associated with the likelihood of being involved in an injurious RTC; the strength of the association steadily increases with the number of medicines dispensed.

  • The number of medications could be reflecting the drivers’ medical conditions, but probably also the use of potentially harmful combinations of medications particularly if those with well-established side effects are used.

The open access argument

For those who still have an open mind on this topic, I urge that you visit the triple C: Communication, Capitalism & Critique, Open Access Journal for a Global Sustainable Information Society. Don't be put off by the title; go to the website and read two papers. The first is by Martin Parkerm ‘Becoming Editor: Or, Pinocchio finally notices the Strings’, and the second is by Jeffrey Beall ‘The Open-Access Movement is Not Really about Open Access’. Beall tracks Open Access journals that appear to be exploiting authors. (; the link courtesy of Sharon Shoshak, translator and editorial consultant.) (noted by IBP).

New device may help detect concussions

Checklight is headgear (a sort of liner) by Reebok intended to alert players, coaches, trainers and parents when an athlete receives a blow that could be serious. Worn under a helmet (hockey, skiing or football) three lights are visible on back of the neck. Green means all is well; yellow or red indicates that the built-in accelerometer and gyroscope have detected an impact that should be evaluated. Reebok cautions that the tool is not intended to be used to diagnose concussions, only to sound an alert. The Checklight has been on sale since June 2013 on the manufacturer's web site, at a cost of $C150 each. (noted by IBP).



  • Contributors JME (guarantor) conceived the study idea and designed the study, performed the statistical analyses and wrote the first draft of the manuscript; LL conceived the study idea and was involved in the study design, interpretation of the results and writing of the manuscript; BE built the dataset on which the analyses were performed, and took part in the data treatment and results interpretation; JM conceived the study idea and was involved in the study design, interpretation of the results and manuscript writing. All authors approved the final version of the manuscript.

  • Funding This study was funded with grants from the Petrus and Augusta Hedlund's Foundation, the Swedish Transport Agency and the Swedish Research Council. The study sponsors had no role in the study design, data analyses, interpretation of results, writing of the manuscript, or in the decision to submit the article for publication. Researchers had access to all relevant data, and were independent from the sponsors at all times.

  • Competing interests All authors declare that they have received support for the submitted work from the Petrus and Augusta Hedlund's Foundation, the Swedish Transport Agency and the Swedish Research Council, but had no financial relationships with any organisations that might have an interest in the submitted work in the previous 3 years or had other relationships or activities that could appear to have influenced the submitted work.

  • Ethics approval The study was approved by the Regional Ethical Review Committee in Stockholm, Sweden.

  • Provenance and peer review Not commissioned; externally peer reviewed.

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.