Associations between heavy episodic drinking and alcohol related injuries: a case control study

Sample of injured patients

A total of 2118 adult patients (18 years or above) admitted to the emergency department of a large hospital in Oslo, Norway because of injuries from accidents, assault, or self-poisoning, were asked to participate in a study of substance use and injuries. Of these, 7% refused to participate and 4% were unable to provide blood samples (for alcohol and drug testing). Thus, the sample of injured patients comprised a total of 1882 patients (see [18] for further details). The data collection was undertaken during a 12 months period from December 2007 to December 2008. This part of the study was approved by the Regional Ethics Committee in Norway and the Norwegian Data Inspectorate. The sample comprised 58.5% men and the mean age was 51.4 years (SD = 22.6).

General population sample

In 2012 a general population sample was contacted by Statistics Norway and asked to participate in a telephone survey on alcohol and drug related topics. The sample aimed to reflect a national representative sample of the adult population (16–79 years) in Norway and the net sample comprised 1947 respondents (response rate 53.3%). As hospital admissions due to injuries occur seldom (2 per 1000 adult inhabitants per year in Norway) and those related to alcohol occur even more seldom, we assumed that there were close to zero persons in the general population sample that had experienced a hospital admission due to alcohol related injury in the preceding year. The age group 21–30 years was over-sampled and data were thus weighted to account for design effect. The sample comprised 50.8% men and the mean age was 42.1 years (SD = 18.1).

Measures

Blood samples were collected from the injured patients and enzymatic dehydrogenase method was used to determine blood alcohol concentration (BAC) [19], and the injury was categorized as alcohol related if the BAC level of the blood sample exceeded 0.01% or/and the patient stated that s/he had consumed alcohol within six hours prior to the injury (see [18, 20] for further details). Other drugs were analysed for in whole blood using liquid and gas chromatography with mass spectroscopy. The rationale for applying self-report in addition to the objective BAC level, was that many of the patients arrived in the hospital many hours after the injury occurred, and in these cases the injury could be alcohol related but not detected in the blood alcohol determination/analysis. A previous study has demonstrated good accordance between BAC and self report of alcohol use prior to injury among patients who arrived in hospital within six hours after the injury [20].

In Norway, admission to a hospital requires referral from primary health services or ambulance services with a tentative diagnosis. Based on this diagnosis, the injuries were categorized into three broad categories; violence, self-poisonings, and accidental injuries. The violence injuries included a broad range with respect to type and severity, the self-poisonings were often overdoses of paracetamol or alcohol or other intoxicants, and the accidents included road traffic accidents, falls, and other accidents at home or in the work place.

Statistical analyses

Only those who reported any consumption of alcohol in the preceding year or had alcohol in their blood were included in the analyses of the association between HED and risk of alcohol related injury. In the case-control design, the outcome variable is being a case or control and the cases represent being at risk of injury, whereas the controls represent not at risk. The risk curve for the association between HED and alcohol related injuries was assessed in logistic regression models, entering frequency of intoxication as categorical (with the lowest frequency category as reference group) and gender and age as co-variates. Thus the risk function was assessed as odds ratios (which correspond closely to relative risk estimates for low incident outcomes). Moreover, similar model specifications were applied to assess the association between intoxication frequency and the risk of specific types of alcohol related injury. Finally, as a previous study [15] demonstrated, some gender differences in the alcohol – injury associations such possible moderating effect was also explored in this study by including a multiplicative interaction term in the logistic regression models.

The categorization of high risk drinkers was based on the HED frequency distribution in the general population sample and a cut-off was chosen which obtained the upper 5 to 10% of the drinkers in terms of frequent heavy drinking occasions. The fraction of alcohol related injuries attributable to high risk consumers was obtained directly from the distribution of the intoxication frequency variable among the patients with alcohol related injuries.

Sensitivity analyses

As the general population sample included respondents from all parts of Norway whereas the injured patients most probably resided in hospital’s catchment area, comprising the capital of Oslo and the surrounding county, the analyses were re-run applying data on cases and controls only from the capital area. The 2012 survey included only 380 respondents from this area, which compromised the test power for this analysis. Therefore, another data set of controls was added. This supplement data set stemmed from a general population survey in 2007 addressing alcohol and other substance use. Identical questions on frequencies of past year drinking and heavy episodic drinking were posed to these respondents and the response categories were roughly the same. This data set provided an additional 238 respondents and thus the total number of controls from the hospitals’ catchment area was 618. All analyses were re-run applying data on cases and controls from only the capital area; i.e. the city of Oslo and the surrounding county.

Another sensitivity analysis pertained to the use of other psychoactive substances than alcohol (illegal drugs and prescription drugs), as these may also increase the risk of injury [21]. As such substance use tends to co-vary with drinking behavior [22–24], this was taken into account by re-running the analyses, excluding those patients who had any illicit drug or psychoactive prescription drug in blood (see [18] for details) from the group of alcohol related injuries.

Sensitivity analyses

The analyses were re-run applying data on control subjects residing only in the capital area. Among controls, respondents living in the capital area reported on average somewhat more frequent HED occasions than others. Consequently, these analyses obtained associations between HED frequency and risk of alcohol related injuries that were somewhat attenuated, but otherwise the results were fairly similar; the associations were of a straight linear kind and less than half of the alcohol related injuries were found among the group of high risk drinkers.

A substantial proportion of the patients with alcohol related injuries had also illegal drugs and/or psychoactive prescription drugs in their blood (34%). After exclusion of these cases from the analyses, we found that the association between HED frequency and risk of alcohol related injury was of a straight linear kind, but, compared to the main analyses, the association was clearly attenuated, and a slightly less proportion of these injuries were found among the group of high risk drinkers.

Strengths and limitations

The participation rate in the sample of injured patients was high, which is likely to reduce the risk of sampling selection bias and thus an underestimation of alcohol related injuries. Moreover, the use of blood sample analyses to assess whether the injury was alcohol related, provided an objective criterion, as compared to the subjective assessments of alcohol involvement which studies from survey data rely on [10, 13, 14].

On the other hand, the response rate in the general population sample was relatively low, as is generally the case in adult population surveys these days. This may well have implied that the heavier drinkers were under-represented and therefore biased the distribution of the intoxication frequency downwards. If so, the magnitude of the group of high risk drinkers will be underestimated and consequently the proportion of alcohol related injuries attributable to a small group of heavy drinkers may be overestimated.

While it may be argued that the two samples were relatively large, the breakdown into drinking behaviour categories and control for age and gender, implied fairly large confidence intervals for the point estimates and thus lessened the accuracy of the risk function, and particularly so with respect to the risk of alcohol related violence and accidents. Thus, the assessment of the risk function would have been more precise if larger samples of cases and controls had been available.

Study implications

Empirical support for the validity of the prevention paradox implies that population strategies may be tenable to prevent or curb alcohol related injuries. Population strategies to reduce heavy drinking occasions comprise control policies to reduce demand; i.e. restrictions on physical availability, price regulations (e.g. taxation), and restrictions on advertising and promotion. These strategies are shown to be effective in reducing alcohol consumption and related harms [25, 26]. Moreover, effective strategies to reduce intoxication or avoid alcohol consumption in risky situations comprise drinking driving regulations and enforcement that prevent traffic injuries [25–27] and integrated programs and enforcement in bars and pubs that prevent over-serving and violence [28, 29].

Beyond the importance of effective population strategies, it may also be argued that a significant proportion may be reached by selective and indicated strategies. Regarding the former, there is good evidence that screening for harmful consumption and brief intervention in primary health care is an effective, yet not widely used measure to reduce consumption among at-risk drinkers [30, 31]. Regarding indicative strategies, many patients treated for alcohol related injuries in the emergency department may also need follow up for their alcohol problems [32].

While selective and indicative strategies may not necessarily reduce injuries to any significant extent at the population level, its effect on alcohol intake and likelihood of improved health for the individual patient should also be taken into consideration [32, 33]. In this context, it may be argued that the motivation to change drinking habits may be better among injured drinkers, as they have experienced adverse effects of their drinking [34].

Finally, given the importance of establishing the validity of the prevention paradox for assessment of tenable prevention strategies to curb alcohol related harm, the scarcity of empirical studies in this area calls for further research along these lines. Future studies should, amongst others, address the relationship between injury severity and the use of alcohol alone and in combinations with other psychoactive substances.