Validity of self-reports of drinking before injury by cause of injury and societal context among emergency department patients

Abstract

Introduction and Aims.

Validity of self-reported alcohol consumption is a topic of continued interest with mixed findings. Validity of self-reports prior to injury is analysed among emergency department (ED) patients by cause of injury, blood alcohol concentration (BAC) and societal context.

Design and Methods.

Data are on 16 196 injured patients from 23 countries in the International Collaborative Alcohol and Injury Project. The proportion of those reporting drinking within 6 h prior to injury among the 2658 BAC positive patients was analysed by injury cause (traffic, violence, falls, other), BAC level (0.01–0.079 mg%, ≥ 0.08 mg%), detrimental drinking pattern (DDP) and the International Alcohol Policy and Injury Index (IAPII).

Results.

Validity averaged 92%, ranging from 85% for drivers to 95% for violence-related injuries. Validity was significantly lower at lower BAC levels for all injuries (P < 0.01) and for violence and falls. Countries with a high DDP and those with lower IAPII scores (less restrictive alcohol policy) also appeared more likely to deny drinking which was significant for DDP for falls (P < 0.01) and for IAPII for passengers/pedestrians (P < 0.05) and violence (P < 0.05). Validity was highest for high IAPII countries, ranging from 92% to 99%.

Discussion and Conclusions.

Validity of self-report compared to breathalyzer readings at the time of admission to the ED was high, but varied by cause of injury, BAC level, country-level detrimental drinking patterns and the restrictiveness of alcohol control policy. Findings may support the utility or self-reports as an alternative to a quantifiable measure of BAC in the ED setting.

Introduction

The validity of self-reported alcohol consumption prior to injury compared to an objective physiological measure of blood alcohol concentration (BAC) among patients in hospital emergency departments (EDs) has been reasonably well-documented in the literature, but with somewhat mixed findings [1–5]. Rates of denying drinking within 6 h prior to the injury while registering positive on breath analysis soon after admission to the ED have ranged from 0.5% [2] to 3.6% [1] in two early studies, while more recent analysis across 16 countries found an average of 12% who denied drinking while registering positive for BAC [4] and a study in a Swiss ED found 11% who denied drinking [5]. In all of these studies self-report was obtained after the patient had been breathalysed, which could conceivably boost rates of positive self-reports (the bogus pipeline effect) [6]. To evaluate this, one ED study obtained the self-report prior to obtaining a BAC estimate on part of the sample, and following the BAC estimate on the remainder of the sample, with no effects observed in the validity of self-report [3].

While these and other non-ED studies have found verbal reports of drinking are generally valid when using structured interviews [7–9], validity has varied across populations studied. The 16-country study noted above found validity ranged from 33% in one country to 100% in six countries [4]. Societal contextual factors may be important in partially explaining this heterogeneity in the validity of self-reports found across countries. The over-all drinking pattern in a country, which reflects to some degree the integration of alcohol in society, might be an important factor in the validity of self-reported consumption following an injury, as well as alcohol control policy in a country or region.

Coupled with this, cause of injury may also be a factor in the heterogeneity of validity of self-reports in ED populations. Validity of self-reports among those arrested for drunk driving has not been good [10,11] and among ED patients, those injured in motor vehicle crashes or violence-related events have been found more likely to deny drinking than those injured from other causes [2], although the opposite was found in another ED study [4]. Denial of drinking following a motor vehicle crash might be more likely in those societies with strong drinking and driving legislation, especially among drivers in the event compared to passengers or pedestrians, but this has not been examined. On the other hand, validity of self-report might be expected to be higher among those with injuries related to violence in societies in which alcohol use is less well integrated into the culture and where drinking and drunkenness are used as excuse factors for otherwise socially unacceptable behaviour [12].

Analysed here is the probability of a positive self-report given a positive BAC at the time of ED admission among those arriving at the ED within 6 h of the event, by cause of injury, BAC level (0.01–0.079 mg% and ≥ 0.08 mg%), country-level drinking pattern and alcohol control policy. We hypothesise that self-report may be less valid for traffic injuries (especially for drivers), those with a lower BAC, and in countries where alcohol is less integrated into society and countries with more restrictive alcohol control policies.

These findings are important for a better understanding of those factors which many affect the validity of self-reported alcohol consumption, which is the most commonly used measure of individual-level alcohol consumption in epidemiological studies. These findings are also important for further documenting the utility of self-reports as an alternative to a quantifiable measure of BAC in the ED setting.

Methods

Samples

Individual-level data were analysed from 16 196 injury patients with BAC and self-report data who were treated between 2000 and 2016 in 47 EDs from 27 studies in 23 countries included in the International Collaborative Alcohol and Injury Study (ICAIS) [13], which is comprised of ED data on alcohol and injury from four international collaborative research projects, all using the same instrumentation and protocols [14]: the Emergency Room Collaborative Alcohol Analysis Project (ERCAAP), and three other international collaborative studies conducted by the World Health Organization (WHO), the Pan American Health Organization (PAHO) and the National Institute on Alcohol Abuse and Alcoholism (NIAAA). All studies were reviewed by the appropriate ethics committee and conducted in accordance with the ethical standards laid down by the Declaration of Helsinki.

Probability samples of patients 18 years and older arriving within 6 h of injury were selected, based on consecutive arrival to the ED, including those arriving by ambulance. Samples provided equal representation of each shift for each day of the week during the study period, ensuring generalisability of the sample to the patient population of those injured in the respective ED. Following informed consent patients were breathalysed and administered a structured interview of about 25 min in length that included questions about the cause of the injury and drinking within 6 h prior to the injury event. Completion rates averaged 87% across the studies. Reasons for non-interviews were primarily due to refusing, but also included incapacitation, leaving prior to completing the interview, being in police custody and language barriers. Patients who were too severely injured to be approached in the ED were followed into the hospital and interviewed once their condition had stabilised.

Individual-level measures

Breathalyzer and self-reported drinking.

An estimate of BAC was obtained, using a hand-held breathalyser, as soon as possible after the patient had been admitted to the ED. These devices have been found to provide estimates which are highly correlated with chemical analysis of blood [15]. Patients were asked, ‘In the six hours before your injury, did you have any alcoholic beverages to drink – even one drink? This includes wine, beer, hard liquor and any other beverages containing alcohol?’

Cause of injury.

Patients were then asked about the cause of the injury that brought them to the ED, which was categorised as: traffic, violence, fall and other causes (e.g. near drownings, burns, animal and insect bites, lacerations, suicide attempts and poisonings). Traffic-related injuries were further divided into those in which the patient was the driver and those in which the patient was either a passenger or pedestrian.

Aggregate-level measures

Detrimental drinking pattern (DDP).

DDP is a measure of the ‘detrimental impact’ on health and other harmful consequences at a given level of alcohol consumption, and was derived from country-level survey and key informant data across 51 countries, based on indicators of alcohol’s integration in society (e.g. drinking in public places, drinking to intoxication, drinking with meals) [16,17]. DDP scores range from 1 (the least detrimental pattern of drinking) to 4 (the most detrimental). In analysis here, DDP was categorised as low (DDP of 1 or 2) and high (DDP of 3 or 4).

Alcohol control policy.

The International Alcohol Policy and Injury Index (IAPII) was used as the alcohol policy measure and includes four regulatory domains: physical availability (legal minimum drinking age, government monopoly of retail sales, restrictions on outlet density, restrictions on outlet hours and days of operation); vehicular (random breath testing, legal BAC limits, penalties for exceeding the maximum BAC); advertising/promotion (a composite measure of restrictions on the majority of media advertisements); drinking context (community mobilisation programs, mandatory server training). The IAPII was developed to predict injury on an international basis [18], and was found to predict variation in rates of alcohol-related injury among ED patients in 28 countries [13]. The IAPII generates a score ranging from 1 (least restrictive) to 100 (most restrictive) which is item weighted according to effectiveness, stringency and enforcement of policies. The IAPII score for each study reflects the same period during which the ED data were collected. For analysis here, the IAPII was categorised as low (0–65) and high (66–100) based on a similar number of studies in each category.

Data analysis

The validity of self-reported drinking was evaluated by examining, among injury patients who were BAC positive (≥0.01 mg%) (n = 2658), the proportion who reporting drinking within 6 h before the injury event. Comparisons were made across injury causes, and for a given cause of injury as well as total injuries combined, between BAC levels and by DDP and alcohol control policies. χ² tests were performed to test equality of proportions across categories. When comparing across-causes of injury, two different tests were performed, one across four categories (traffic, violence, fall and others) and the other across five categories with traffic further broken into drivers and passengers/pedestrians. Weights were assigned to several studies to adjust for unequal sampling such as overrepresentation of ED shifts in weekend evenings. To account for sample clustering and the potential within-study correlation, the analyses were performed adjusting for the design effect using STATA (StataCorp., 2017) survey command treating ED study as the primary sampling unit (PSU).

Results

Table 1 describes the 23 countries included in the analysis, and the proportions with a positive BACs and a positive self-report.

Table 1.

Prevalence of positive BAC and self-report of drinking before injury from injury patients across 47 EDs from 23 countries

Region	Country	City/study, year	#EDs	Total N	N valid BAC	% BAC ≥0.01%	% self-report alcohol
Africa	Mozambique	Maputo, 2001	1	459	448	16.1	17.1
	South Africa	Cape Town, 2001	1	464	464	45.5	46.4
	Tanzania	Moshi, 2013	1	516	511	28.8	28.0
Asia/Pacific	China	Changsha, 2001	1	533	500	10.2	18.8
		Five cities, 2009a	5	2540	2445	13.4	16.5
	India	Bangalore, 2001	1	544	544	22.8	21.7
	Korea	Five cities, 2007–2009b	6	2107	1934	17.4	24.7
	New Zealand	Auckland, 2000	1	153	138	27.6	38.5
		Auckland, 2015–2016	1	484	373	16.6	21.1
	Taiwan	Taipei, 2009–2010	2	1035	927	4.7	6.5
Europe	Belarus	Minsk, 2001	1	457	457	32.2	30.0
	Czch Republic	Prague, 2001	1	510	462	4.3	7.8
	Sweden	Malmö, 2001	1	497	380	12.1	15.1
	Switzerland	Lausanne, 2006–2007	1	325	230	12.6	25.4
North America	Canada	Orangeville (ON), 2002	1	222	206	4.0	6.3
		Vancouver, 2009	2	249	199	14.1	22.2
		Vancouver/Victoria, 2013–2015	3	1191	1076	9.5	14.7
	Mexico	Mexico City, 2002	1	456	420	15.2	17.2
Central/South America	Argentina	Mar Del Plata, 2001	1	452	450	16.1	21.3
	Brazil	São Paulo, 2001	1	496	485	11.8	12.8
	Costa Rica	San Jose, 2012	2	1013	936	8.5	8.7
	Dom. Republic	Santo Domingo, 2010	1	501	494	8.7	19.3
	Guatamala	Guatamala City, 2011	1	513	508	20.3	21.1
	Guyana	Georgetown, 2010	1	485	477	17.0	21.0
	Nicaragua	Managua, 2010	2	518	500	22.4	21.5
	Panama	Three cities, 2010c	3	490	468	17.1	20.8
	Trinidad/Tobago	Four cities, 2015d	4	252	164	13.4	20.6
Total			47	17 462	16 196	15.5	18.9

^aBeijing, Hangzhou, Chengdu, Hengyang and Changsha.

^bKyonggi, Seoul, Suwon, Chuncheon and Donggu.

^cLa Chorrera, Colon and Vearaguas.

^dMount Hope, San Fernando, Port-of-Spain and Scarborough.

Table 2 shows the proportion who reported drinking prior to injury among those who had a positive breathalyser test for all positive and for those with a BAC between 0.01–0.079 mg% and those with a BAC ≥0.08 mg%. Overall self-report averaged 92%, with a range from 85% for drivers to 95% for those with violence-related injuries. Drivers were less likely to report drinking than passengers/pedestrians (94%) involved in a traffic crash. This same pattern held regardless of BAC level. The likelihood of self-reported drinking was significantly greater at higher BAC levels than at lower levels for all injuries combined and for those related to violence and falls (P < 0.01). No significant difference was found in self-report by BAC level for traffic injuries, although both drivers and passengers/pedestrians were more likely to report drinking at higher BAC levels, with the difference for drivers smaller than that for passengers/pedestrians.

Table 2.

Percentages (95% confidence intervals^a) of self-reporting drinking before injury among BAC positive groups by cause of injury

	BAC ≥0.01% (N = 2658)	BAC 0.01–0.079% (N = 1231)	BAC ≥0.08% (N = 1427)
Traffic (N = 454)	90.2 (83.6, 94.8)	88.9 (81.7, 94.0)	91.6 (81.2, 97.3)
Drivers (N = 188)	85.2 (75.6, 92.1)	84.3 (72.0, 92.7)	86.1 (73.2, 94.3)
Passengers or pedestrians (N = 266)	93.8 (88.6, 97.1)	91.9 (86.0, 95.9)	95.9 (88.1, 99.2)
Violence (N = 978)	94.9 (91.6, 97.2)	92.3 (87.8, 95.5)	97.3 (94.3, 99.0)*
Fall (N = 602)	90.3 (82.8, 95.3)	85.7 (74.5, 93.2)	93.5 (87.1, 97.3)*
Others (N = 624)	88.1 (78.2, 94.6)	83.6 (71.0, 92.3)	92.9 (80.7, 98.5)**
Total	91.5 (86.6, 95.0)	88.2 (81.7, 93.0)	94.4 (89.9, 97.3)*
χ2 test across four cause categories	P = 0.019	P = 0.020	P = 0.176
χ2 test across five cause categories	P = 0.003	P = 0.009	P = 0.041

^*P < 0.01

^**P < 0.1

χ² test between BAC 0.01–0.079% and BAC ≥0.08% for each injury cause separately.

^aExact (Clopper–Pearson) confidence intervals.

Table 3 shows the proportion who reported drinking among those BAC positive at high and low country-level DDP and high and low levels of the IAPII. Countries with a high DDP were more likely to deny drinking prior to the event than those countries lower on DDP, although this difference was significant (P < 0.01) only for violence-related injuries. Countries with a lower IAPII score were more likely to deny drinking than those with a higher IAPII score, and this was significant (P < 0.05) for passengers/pedestrians and those with violence-related injuries. The proportion reporting drinking for countries with high IAPII scores was quite large, ranging from 92% to 99%, with drivers now in the middle of the range.

Table 3.

Percentages (95% confidence intervals^a) of self-reporting drinking before injury among BAC positive (≥0.01%) by cause of injury, DDP and alcohol control policy

	DDP		IAPII
	1 or 2 (N = 972)	3 or 4 (N = 1686)	Low (N = 1808)	High (N = 850)
Traffic	93.5 (78.1, 99.2)	89.0 (80.8, 94.5)	88.9 (81.5, 94.1)	98.1 (86.4, 99.99)***
Drivers	89.7 (68.7, 98.6)	83.1 (71.2, 91.6)	83.3 (74.5, 91.1)	96.3 (72.9, 99.99)
Passengers or pedestrians	97.0 (88.3, 99.7)	92.8 (86.0, 96.9)	92.9 (86.9, 96.7)	99.4 (94.8, 100.00)**
Violence	99.1 (95.7, 99.96)	93.5 (89.1, 96.5)*	94.0 (89.7, 96.9)	98.8 (94.3, 99.9)**
Fall	91.9 (79.9, 97.9)	88.8 (75.2, 96.5)	88.4 (76.8, 95.5)	92.5 (78.2, 98.7)
Others	91.6 (75.6, 98.6)	85.9 (69.9, 95.4)	85.4 (71.7, 94.2)	91.7 (75.1, 98.7)
Total	94.1 (85.3, 98.4)	90.2 (83.3, 94.9)	90.3 (84.0, 94.8)	94.2 (84.4, 98.7)
χ2 test across four cause categories	P = 0.091	P = 0.115	P = 0.025	P = 0.121
χ2 test across five cause categories	P = 0.061	P = 0.042	P = 0.005	P = 0.128

^*P < 0.01

^**P < 0.05

^***P < 0.1

χ² test between DDP = 1/2 versus DDP = 3/4 and between IAPII = low versus IAPII = high for each injury cause separately.

^aExact (Clopper–Pearson) confidence intervals.

The validity of self-report between lower and higher BAC levels (0.01–0.079 vs. ≥0.08 mg%) was also examined, stratified by DDP and IAPII (not shown). Larger proportions of self-report at higher BAC levels compared to lower BAC levels were consistently observed across DDP and IAPII categories for traffic, violence, fall and other injuries, separately. The only exception was, among traffic injuries where a smaller proportion of self-report was found at higher BAC levels for countries with high IAPII scores (sample size was small for this sub-group analysis, how-ever; n = 65).

Discussion

Findings here suggest that validity of self-report compared to breathalyzer readings at the time of admission to the ED for those arriving at the ED within 6 h of the event was high, with an average of 91%, but varied by cause of injury, BAC level, country-level detrimental drinking patterns and the restrictiveness of alcohol control policy in a country.

We hypothesised that self-report may be less valid for traffic injuries (especially drivers), those with a lower BAC, and in countries where alcohol is less integrated into society and countries with more restrictive alcohol control policies. Findings here suggest support for the hypothesis that self-reports were less valid for drivers in motor vehicle crashes who were more likely to deny drinking than those in the other injury categories were. Those with traffic-related injuries have been found more likely to provide valid self-reports of drinking prior to the event elsewhere, but that study did not examine drivers separately from passengers and pedestrians [4]. Those with violence-related injuries were found most likely to report drinking, possibly due to alcohol being used as an excuse for engaging in violence-related behaviours [12]. Findings also lend support for the hypothesis that those with a lower BAC would be more likely to deny drinking than those with a higher BAC, which may be due to either their recall being less accurate or because, since they had already been breathalysed, it would be more obvious that they had not told the truth. This difference was significant, but varied by cause of injury, with findings not significant for traffic injuries and the difference smallest for drivers.

Findings also lend support to the hypothesis that self-report is less valid in countries high on DDP, although this difference was significant only for violence-related injuries, and suggests that alcohol was not used as an excuse factor for engaging in such behaviour. Drinking appeared more likely to be denied in those countries where alcohol is less integrated into society and where drinking may be viewed as less acceptable. Additionally, those in high DDP countries tend to drink more heavily and may consequently deny drinking at moderate levels of consumption.

Findings, however, did not support the hypothesis that self-report is less valid in countries with more restrictive alcohol control policies and the opposite was found. Self-reports appeared to be more valid in those countries with more restrictive alcohol control policies and was significant for passengers/pedestrians and those with violence-related injury. Interestingly, while it might be expected that denial of drinking by drivers involved in a traffic crash might be more likely in those countries with stronger drinking and driving legislation, the data do not support this conjecture. It is possible that in those countries with more restrictive alcohol policy individuals are more cognizant of harms and health issues related to drinking and may consequently be more likely to report their drinking in the context of emergency care treatment.

BAC estimates were obtained before the patient was asked about drinking, which could presumably account for the overall high validity of self-report in this study, but as noted above, a sensitivity analysis was performed in an earlier ED study by obtaining the BAC reading before the self-report compared to after and no difference was observed in validity of self-report[3].

Other factors than those examined here may affect validity of self-reported consumption prior to injury. Those who arrive to the ED sooner have been found less likely to deny drinking than those arriving later, possibly due to less recall bias [4], but this was not taken into account. Additionally, those arriving later to the ED may have had more opportunity to drink following injury, resulting in conservative estimates of validity of self-report if these same patients did not drink before injury. While data on drinking following injury were not available in the data sets analysed here, other ED data have found that among those with a positive BAC who denied drinking prior to injury, 38% reported drinking following the injury and arrival in the ED [4].

Findings here may not be transferable to other settings eliciting self-reported consumption of alcohol, such as general population surveys. ED patients are providing self-reports for a finite period (6 h before injury) and in close proximity to their drinking (within 6 h). They may also be more likely to accurately report their drinking with the belief that this disclosure is important to their ED treatment. Although the amount consumed prior to injury was not included in these analyses, it might be assumed that the validity of such data may also be more accurate than similar data elicited in surveys obtaining self-reports over a longer period of time.

Drinking prior to injury is more likely to be denied by those with a lower BAC, and also appears to be more likely to be denied in those countries with more detrimental drinking patterns and those countries with less restrictive alcohol control policies. Those in high IAPII countries provide the most valid self-reports, regardless of cause of injury. These findings are important for a better understanding of those factors which many affect the validity of self-reported alcohol consumption, which is the most commonly used measure of individual-level alcohol consumption in epidemiological studies. Additionally, positive self-reports of drinking have been found to be strong motivators in changing drinking behaviours among injured ED patients in the context of a brief intervention [19]. Findings here may also further document the utility of self-reports as an alternative to a quantifiable measure of BAC in the ED setting as well [20].