Associations between psychoactive substance use and sensation seeking behavior among drivers in Norway

A cross-sectional design was applied to estimate the association between psychoactive substance use and sensation seeking behavior among Norwegian drivers. The drivers were recruited in two roadside surveys in cooperation with the local police and the Norwegian Mobile Police Service of Norway. The first of the two surveys was conducted in the arctic, rural county of Finnmark, the northernmost county of Norway, in the period from September 2014 to October 2015. The second of the surveys was conducted in five counties with both urban and rural areas in south-eastern Norway including Oslo, Akershus, Buskerud, Hedmark and Oppland from April 2016 to April 2017. In both surveys, drivers of cars, vans, motorcycles and mopeds were included. The six above mentioned counties have a total of approximately 2.0 million inhabitants, which accounts for about 38% of the Norwegian population.

Because rural areas in Norway are sparsely populated, roadside sampling could not be done completely random in neither of the two surveys. Hence, a three-stage cluster sampling procedure was used in both surveys.

The first stage of the sampling was to select major roads in the studied counties. For the survey in Finnmark, main roads with an annual average daily traffic of 500 motor vehicles or more were selected to achieve the desired number of participants. For the survey in south-eastern Norway, roadways included in the previous roadside survey performed in the same area in 2008–2009 formed the foundation when choosing roads for data collection [19, 20]. In the second stage, time intervals were systematically selected to cover all time periods of the week, and the third stage consisted of randomly stopping drivers in cooperation with the police to ask for participation. Further details on the sampling procedure of the roadside survey in Finnmark have been described by Jamt et al. [21] and of the roadside survey in south-eastern Norway by Furuhaugen et al. [19].

Participation in both surveys was voluntary and anonymous. Oral and written information was given to each driver. Informed consent was obtained verbally by all participants. A representative from the research project signed the consent with his or her own name on behalf of the participant to protect the anonymity of the participant. All drivers 18 years of age and older were eligible to participate, but professional drivers were not included in neither of the surveys. Both roadside surveys were approved by the Regional Committees for Medical and Health Research Ethics in Norway.

Based on numbers of participants in previous, comparable Norwegian roadside surveys and official statistics on traffic volume obtained from the Norwegian Public Road Administration it was estimated that 3000–5000 participants were needed to reach adequate statistical power [19, 21].

Whenever an informed consent was given, an oral fluid sample was collected for the analysis of psychoactive substances and a questionnaire was filled in. The questionnaire contained questions about the persons’ sex, age, citizenship, type of vehicle, geographical area, time of participation (2-h intervals), day of the week and month were also recorded. In addition, the drivers were asked to answer to what extent they agreed to four statements about sensation seeking (Brief Sensation Seeking Scale 4 (BSSS-4)) (Table 1) on a scale ranging from 0 (strongly disagree) to 4 (strongly agree). No personally identifiable data was collected.

Stephenson et al. has noted the potential drawbacks associated with the use of short measures such as the BSSS-4, as sensation seeking is apparent in a variety of behaviors and preferences [22]. More extensive measurements are advantageous with respect to the content validity and measurement errors. However, long measurements are not always the best suited option as in large-scale surveys. Stephenson et al. found that the BSSS-4 performed similarly to the more established measures of impulsive sensation seeking from the Zuckerman-Kuhlman Personality Questionnaire (r² = 0.81) and the eight-item brief sensation seeking scale (r² = 0.89) [22]. In the present study, all data was collected at the roadside which is a challenging place to gather data with respect to lengthy questionnaires. Because available space and time is particularly limited when collecting data at the roadside, we chose to use the BSSS-4.

Samples of oral fluid were collected using Quantisal™ Oral Fluid Collection Device (Immunalysis Corporation, Pomona, CA) [19, 21]. The collection pad was placed in the mouth beside the cheek or under the tongue until the indicator turned blue (when absorption reached 1 ml of oral fluid) or until 5 min had passed. The samples were analyzed for alcohol with an automated enzymatic method [24], and for medicinal- and illicit drugs using ultra-high-performance liquid chromatography with tandem mass spectrometry detection as previously described [19, 21].

Adjusted odds ratios (OR) were calculated using multivariable logistic regression. Five different models were estimated where the dependent variable was either alcohol, tetrahydrocannabinol (THC), benzodiazepines and/or z-hypnotics, stimulants or polydrug use (yes/no, with “yes” corresponding to concentrations equal to or above the analytical cut-of limits). Covariates were sex (with “male” as reference), age (in completed years, with 18 years as reference), geographical area (with the roadsidesurvey in Finnmark as reference) and the four items of the BSSS-4. The items were included as ordinal categorical variables with three levels. The originally five-level scales were collapsed down to three levels as follows: the scores “1” (disagree), “2” (neutral) and “3” (agree) were collapsed into one level (reported as “medium” in Table 2) and set as reference, score “0” (strongly disagree) corresponding to level 1 (reported as “low” in Table 2) and “4” (strongly agree) corresponding to level 2 (reported as “high” in Table 2). This resulted in a three-level-scale consisting of the levels “strongly disagree”, “neutral” and “strongly agree”. The rationale for this classification was to separate those scoring the extreme values in both directions from the rest. Another reason for this classification was to gain statistical power and to simplify the interpretation of the results from the multivariable logistic regression analysis [25, 26].

Both roadside surveys included drivers of passenger cars, vans, motorcycles and mopeds. Of the invited drivers, 94% agreed to participate in Finnmark and 91% in south-eastern Norway. However, eight drivers were not illegible because of low age. Altogether, this resulted in a total of 8053 participating drivers; 5026 drivers from the roadside survey in south-eastern Norway and 3027 from the roadside survey in Finnmark [19, 21].

Data on either the sex of the driver, age, month, day of the week and time of the day for participation were missing for four of the participants. Data on the four sensation seeking items were missing or incomplete for 27 of the participants in the survey in Finnmark and 109 of the participants in the survey in south-eastern Norway. All the cases with incomplete data were included in the bivariate analysis where possible (Table 2) but excluded from the multivariable analysis.

There were small but statistically significant differences in gender and age distributions between the two surveys (Table 2); 69.4% of the participants in the survey performed in Finnmark were men versus 67.2% in the survey performed in south-eastern Norway (p = 0.04). 40.0% of the participants in Finnmark were under the age of 40 years versus 37.2% in the south-eastern Norway (p = 0.012). There were also significantly more samples collected during weekend in south-eastern Norway (36.1%) compared to Finnmark (29.6%) (p < 0.001). There were significant differences in the distributions of BSSS-4 average sum scores between the surveys; the proportion of average sum scores on the lowest level was higher in Finnmark than in south-eastern Norway. The proportions on the highest level were approximately the same in both surveys. Table 3 shows the substance groups that were analyzed in oral fluid samples. The prevalence of the substance groups among sexes and age groups (under/over the age of 40) in both surveys is also presented in Table 3. Significant differences were found in the prevalence of THC and polydrug use among female drivers between the two roadside surveys; significantly more female drivers were positive for THC and polydrug use in south-eastern Norway compared to Finnmark.

Adjusted odds ratios for finding alcohol, THC, benzodiazepines/z-hypnotics or stimulants in oral fluid were calculated by considering four multivariable logistic regression models for the different substance groups.

We first estimated full models including as covariates: sex, age, geographical area (north or south-east) and the four subscales of the BSSS-4 (each subscale included as a categorical variable with five levels). Then with a backward stepwise procedure reduced models were selected. This procedure found the effect of the subscales “experience seeking” and “boredom susceptibility” and the variable “geographical area”” not to be statistically significant in the model with use of THC as dependent variable (F-test = 14.4 and p-value = 0.11). In the model with use of stimulants as dependent variable there were no significant associations with the subscale “boredom susceptibility” and the variable “geographical area”” (F-test = 0.97 and p-value = 0.97). The model with use of benzodiazepines and/or z-hypnotics as dependent variable, the procedure found the effect of the subscale “disinhibition” and “geographical area” not to be statistically significant (F-test = 5.08 and p-value = 0.41). None of the variables except for the “sex” variable in the model with the use of alcohol as dependent variable were significant.

Furthermore, new reduced models with the BSSS-4 subscales classified into three levels, as described, were estimated (Table 4). From the new models, statistically significant associations were found between scoring the lowest level on the “thrill and adventure seeking” subscale and the use of THC, and benzodiazepines and/or z-hypnotics. Statistically significant associations were also found between scoring the highest level on the “experience seeking” and “disinhibition” subscales and the use of stimulants.

An important strength of this study is that we could estimate the association between sensation seeking behavior and psychoactive substance by specific chromatographic analysis of oral fluid samples and not self-reported use of psychoactive substances. Another strength is that the participation rates in the roadside surveys were high, above 90%.

In our study we found no association between testing positive for alcohol use and sensation seeking behavior. A number of studies have demonstrated an association between alcohol impaired driving and sensation seeking behavior [11, 37, 38], hence it is reasonable to assume that such an association also may exist among Norwegian drivers. The lack of association is likely due to a low number of alcohol positive drivers in this study (n = 22) and the result might have turned out differently if our study sample had been larger. The same was the case for polydrug use; no association was found with sensation seeking behavior. Polydrug users have been found to be higher in sensation seeking than both non-users and depressants users [33, 39]. As with the alcohol positive drivers the number of drivers positive for more than one drug was rather low (n = 48). Because the participation in the roadside surveys was voluntary, it is likely that some drivers, who had recently used alcohol or drugs of abuse not revealed by the police, might have rejected to participate in fear of being prosecuted.

A well-known concern using questionnaires is the phenomenon of socially desirable response (SDR) bias; that is “the tendency for participants to present a favorable image of themselves” [40]. This may confound the results and lead to inaccurate conclusions.

As have been outlined above, lengthy questionnaires were practically impossible to conduct in the roadside surveys of the present study and so we have not been able to measure the degree of SDR bias in our results. Although the participation in the roadside surveys was anonymous, it might be that the context our data were collected in (i.e. the police were present) may have contributed to bias the results in the direction of lower scores on the sensation seeking.