1 Addiction Research Institute (IVO), Heemraadssingel 194, 3021 DM Rotterdam, The Netherlands, 2 Department of Internal Medicine, Meander Medical Center, Amersfoort, The Netherlands and 3 Tilburg University, Faculty of Social and Behavioural Sciences, Tilburg, The Netherlands
* Author to whom correspondence should be addressed at: Tel.: +31 10 4253366; Fax: +31 10 2763988; E-mail: vitale{at}ivo.nl
(Received 23 March 2005; first review notified 13 May 2005; in revised form 28 June 2005; accepted 5 July 2005)
![]() |
ABSTRACT |
---|
![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() |
---|
![]() |
INTRODUCTION |
---|
![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() |
---|
Studies outside the emergency room focusing on alcohol use and injuries are conducted by surveys among the general population (Cherpitel, 1999) or within specific groups, especially among people involved in car or motor accidents (Del Rio et al., 2002
; Weber et al., 2002
; Kurzthaler et al., 2003
). Results from a household survey among the general population show that alcohol consumption was predictive for emergency treatment, and that heavier drinking was associated with treatment for an injury and lighter drinking with treatment for an illness (Cherpitel, 1999
). However, because household surveys make it difficult to assess a causal relationship between alcohol use and the injury, the emergency room is the most obvious place to study a possible causality. Such emergency room studies mostly employ patient self-reported alcohol use, blood alcohol concentrations, or breathalysers. Another, less frequently employed, measure of assessing alcohol use by patients is to use clinical or staff assessment/recognition. Self-report and tested blood alcohol concentration both have advantages and disadvantages, and it remains debatable whether to use self-reported drinking or to test for blood alcohol concentration (Treno et al., 1998
). No single instrument or method stands out as the gold standard; a combination of clinical, self-report and biochemical markers is considered optimal, but will vary according to the purpose of the measurement (El-Guebaly et al., 1998
).
Numerous studies have compared self-reported alcohol use with blood alcohol concentration measures. In contrast, studies comparing differences in self-reported alcohol use among emergency room samples are scarce. Therefore, this study focuses on this topic in order to identify factors that influence found prevalence rates of self-reported alcohol use among emergency room patients. The two main issues explored in this study are factors related to differences resulting from study procedure and resulting from sample selection bias. Several studies conclude that self-report is sufficiently valid to measure alcohol use prior to the injury event (Cherpitel, 1993; Treno et al., 1998
). Although self-report of alcohol consumption shows adequate reliability and validity, social context factors, respondent characteristics, and task attributes can influence response validity (Del Roca and Darkes, 2003
). In addition to differences resulting from alcohol measures and cultural drinking patterns, sample selection bias in the emergency room should not be underestimated and can lead to difficulties in that the emergency room sample cannot truly represent the general population of injured (Treno et al., 1998
).
In this study, the aspects of sample selection bias and study procedure related to alcohol prevalence rates are examined more closely by studying different methods of self-report in an emergency room population of one hospital. The aim is to compare three different methods of data collection using patient self-report among an emergency room population in order to identify variations in study results. The main questions to be addressed are
![]() |
METHODS |
---|
![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() |
---|
Method 1 (M1)
In the first method emergency room personnel (nurses and administrative personnel) were responsible for data collection (M1). The period of data collection was from July 2003 to January 2004. In this method, patients were approached in two ways: (i) Patients with minor injuries/illness meeting the inclusion criteria were handed a questionnaire by the administrative staff shortly after entering the emergency room; the questionnaire was then completed in the waiting room. (ii) Patients with more serious/severe injuries/illness were approached in the treatment room by the nurses before or shortly after the treatment; the patients then filled in the questionnaire.
Method 2 (M2)
The second method consisted of research staff handing out the questionnaire (M2). This method was used for 2 weeks in May 2004. Patients with minor injuries/illness were invited to participate by the research staff while waiting for treatment (i.e. shortly after entering the emergency room). Patients with more severe injuries/illness were approached after their consultation with the emergency room personnel, by the research staff before or shortly after treatment; the patient then filled in the questionnaire.
Method 3 (M3)
The third method collected data retrospectively via a postal questionnaire (M3). All patients aged 12 years and older visiting the emergency room in April 2004 were approached. At the end of the emergency room visit, each eligible patient was given a letter to take home with information about the study and the questionnaire. After this (710 days after the visit) all patients received a mailed questionnaire at home, accompanied by a second letter with information about the study.
Measures
Data were collected using a combination of self-report and emergency room data. In the first two studies (M1 and M2), emergency room and research personnel were able to score reasons for non-response (physical status and/or mental status, refusal, without sufficient command of the Dutch language) and give their judgement on the patient's alcohol consumption at the time of the emergency room visit. This was done for patients not able to fill in the questionnaire owing to their medical condition.
All three methods used an identical questionnaire which addressed the following topics: reason for the emergency room visit (traffic accident, accident, injury/illness, aggression/violence, suicide attempt, or self-mutilation), location of accident or illness (home, other people's home, public place, catering establishment, work, school, or street), demographic data (cultural background, work, and living situation), alcohol use, location of alcohol consumption, licit drug use, illicit drug use, and location of illicit drug use. Alcohol use, 24 and 6 h prior to the visit was asked, as was general alcohol consumption pattern (number of drinking days in the weekend and during the week, average number of consumptions on a drinking day in the weekend and during the week). Based on alcohol consumption the patient was classified as abstainer, moderate drinker, occasional excessive drinker, and frequent excessive drinker (Table 1). This classification has been used by Lahaut et al. (2002), based on the classification of Garretsen (1983)
. This classification was used among emergency room patients because data on the relationship between injuries and occasional/frequent excessive drinking are valuable from a prevention point of view, i.e. to identify which alcohol consumers are at risk for injuries.
|
Statistical analysis
The response and non-response populations for each method were compared regarding demographics, emergency room characteristics and alcohol use, using bivariate cross-tabulation. Chi-square tests were conducted to determine whether included and excluded patients per method differed significantly. The categories of the various variables were tested separately. A Bonferroni correction was used to decrease the possibility of false positives. Only mean age was compared using a t-test. The different response populations in the three methods were compared using bivariate cross-tabulation. Chi-square tests were conducted to determine whether included patients in the three methods differed significantly. All results were regarded significant at P < 0.05, except for those variables with more than two categories, where the Bonferroni correction was applied in which case results were significant at P < 0.05/n (=number of variable categories).
![]() |
RESULTS |
---|
![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() |
---|
|
|
Sample selection bias
Table 4 presents the characteristics of the included and excluded patients according to the three study methods. Patients included in the study population of M1 differed significantly from those excluded with respect to gender, mean age, age category, time of emergency room visit, and referral to emergency room. In M2 the included patients differed significantly from the excluded patients only with respect to the part of the week. Significant differences were found between included and excluded patients in M3 with respect to gender sex, mean age, age category, and referral to the emergency room.
|
To summarize the results, M1 resulted in a study sample (those included) that has relatively more males, is younger, less patients arriving between 0:008:00 h and less patients frequenting the emergency room by ambulance compared with the excluded population in this method. In M2 the included patient sample only resulted in relatively more patients frequenting the emergency room during weekends compared with those patients excluded. M3 results in an older sample with more females and more referrals by ambulance, but with no variation in the part of the week and time of emergency room visit compared with the excluded population. The three methods compared showed that least differences occurred concerning included population between M2 and M3.
Sample characteristics
Table 5 presents the characteristics of the study samples in the three methods. M2 included significantly less patients with a Dutch cultural background than M1 and M3. Concerning living situation no differences were found between the three methods concerning patients who live alone and those who live with others. The occupational status and reason for emergency room visit did not differ between M1 and M2, with the exception of M2 including more students compared with M1. M3 included more patients who were retired compared with M1 and M2. Also, more patients in M3 seek emergency treatment as a result of an illness compared with M1 and M2. Among injured patients, less injuries were caused by accidents in M3 compared with M1 and M2.
|
![]() |
DISCUSSION |
---|
![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() |
---|
Do different self-report methods result in different alcohol prevalence rates?
Alcohol prevalence rates vary between the different self-report methods. Despite these differences, some similarities between the methods are also seen. Alcohol use 6 h prior to the emergency room visit was reported by 5% of the patients in both M1 and M3. Prevalence of self-reported alcohol use in M2 was 9.1%. When additionally using emergency room and research staff judgements on the patient's alcohol use, prevalence rates are 16.2% (M1) and 6.8% (M2), respectively. Despite the lower proportion of self-reported alcohol use 6 h prior to the visit in M1 compared with M2, results show that emergency room staff identifies relatively more patients under the influence of alcohol compared with the research staff, eventually leading to higher prevalence rates. This is probably attributable to the selection bias resulting from this method. Using emergency room staff leads to a small sample; however, staff seem to have selected a biased sample with a relatively large proportion of alcohol positive patients. This is illustrated by the large differences in alcohol judgement between emergency room staff and research staff. In the present study, using only patient self-report would have resulted in an underestimation of alcohol prevalence. The mail survey leads to almost the same proportion of self-reported alcohol use (4.8%) as M1 and approximately half the proportion of self-reported alcohol use 6 h prior to the visit compared with M2. However, with M3 no judgement of the patient's alcohol use by emergency room or research staff is possible. The only other Dutch emergency room study on self-reported alcohol use combined with emergency room staff recognition of alcohol use among victims of traffic accidents reported a prevalence rate of 8% (Kingma and Klasen, 1994
). In our study self-reported alcohol use combined with staff recognition of alcohol use was almost twice as high (16.2%). This can be explained by the fact that our study also included all emergency room patients and not only victims of traffic accidents. Emergency room staff was able to identify significantly more patients who were positive for alcohol at the time of the emergency room visit compared with research staff. Alcohol consumption rates show no differences between M1 and M2, but the retrospective study reported a higher proportion of abstainers and less frequent excessive drinkers. In our study another explanation for the lower alcohol consumption rates in M3 are the characteristics of the sample population selected in M3, which included relatively old people.
Are the differences in prevalence rates the result of sample selection bias?
Two of the three methods resulted in sample selection bias on various demographic and emergency room characteristics. Only M2 (with research staff available 24 h) resulted in almost no sample selection bias. Comparing sample selection biases, the methods using emergency room staff (M1) and research staff (M2) show more similarities on demographic variables regarding the included emergency room population compared with the retrospective method (M3), with M1 and M2 including relatively more males and more patients aged 1860 years. Emergency room staff seem to include more alcohol suspect patients: males and patients aged 1835 years identified by previous studies (Cherpitel, 1993).
Data collection through emergency room staff (M1) also differed with respect to patients included in the study population compared with the other two methods (M2 and M3): with respect to emergency room variables, more patients visited the emergency room between 8:0016:00 h and less patients arrived at the emergency room by ambulance. The first result can be explained by the fact that during the day more emergency room staff are present. For the second finding the explanation lies in the fact that these are more seriously injured patients, who need direct treatment for injury or illness. Therefore, there is less time (or it is not possible) to approach these patients to participate in the study, because they are less eligible to be interviewed, as also pointed out by Treno et al. (1998).
Are there other explanations for these differences in alcohol prevalence rates?
In addition to sample selection bias, differences in alcohol prevalence rates can be explained by characteristics of the included population. In contrast to the two research methods using emergency room staff and research staff, use of a retrospective mail survey among an emergency room population seems to select a different emergency room population, including significantly more older people with an age-related referral pattern (more ambulance and GP). Consequently, the M3 sample population includes more patients who are retired and who visit the emergency room owing to an illness compared with the other two methods. Previous research among the general Dutch population via postal questionnaires show that the response among elderly individuals is higher (Van de Mheen, 1998). It can be assumed that older people and people with an illness are less likely to drink (excessive amounts of) alcohol.
![]() |
STUDY LIMITATIONS |
---|
![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() |
---|
Recommendations for further research
Directions for further research on self-report of alcohol use among emergency room populations are the following. Research should focus on the possibility to combine different methods in order to limit sample selection bias; for example, data collection through emergency room staff for specific groups of patients (e.g. seriously injured) combined with research staff for other groups (e.g. milder injured patients).
In this study, the most influential factor in the variation in alcohol prevalence rates is sample selection bias as a result of the chosen research method; i.e. a retrospective mail survey among emergency room patients leads to a relatively older sample with more abstainers and therefore, lower alcohol consumption 6 h prior to the emergency room visit. Using the judgement of research staff, and especially using emergency room staff, results in higher alcohol prevalence rates compared with a retrospective mail survey, or to relying only on patient self-report. As a result of this, variations in alcohol prevalence rates among emergency room studies between and within countries not only result from consumption patterns in a culture or region, but can also be influenced by sample selection. Therefore, unless sample selection bias is controlled, comparing results between studies and countries should be done carefully. The identification of patients positive for alcohol is done more efficiently by emergency room staff, but results in a biased sample; including more alcohol suspect patients. In contrast, the method using research staff is less biased, but it is the most expensive method of data collection in this study. Future emergency room studies on patient's self-reported alcohol use should take into account that the selected method can influence the alcohol prevalence rates. Therefore, future studies should confirm if a combination of research staff handing out the questionnaire and emergency room staff judging patient's alcohol use results in the lowest sample bias and more accurate alcohol prevalence rates.
![]() |
ACKNOWLEDGEMENTS |
---|
![]() |
REFERENCES |
---|
![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() |
---|
Cherpitel, C. J. (1999) Drinking patterns and problems, drug use and health services utilization: a comparison of two regions in the US general population. Drug Alcohol and Dependence 53, 231237.[CrossRef]
Cherpitel, C. J. and Borges, G. (2002) Substance use among emergency room patients: An exploratory analysis by ethnicity and acculturation. American Journal of Drug and Alcohol Abuse 28, 287305.[CrossRef][ISI][Medline]
Del Rio, M. C., Gomez, J., Sacho, M. et al. (2002) Alcohol, illicit drugs and medicinal drugs in fatally injured drivers in Spain between 1991 and 2000. Forensic Science International 127, 6370.[CrossRef][ISI][Medline]
Del Roca, F. K. and Darkes, J. (2003) The validity of self-reports of alcohol consumption: state of the science and challenges for research. Addiction 98 (Suppl 2), 112.[CrossRef]
El-Guebaly, N., Armstrong, S. J. and Hodgins, D. C. (1998) Substance abuse and the emergency room: programmatic implications. Journal of Addictive Diseases 17, 2140.[ISI][Medline]
Garretsen, H. F. L. (1983) Probleemdrinken: prevalentiebepaling, beinvloedende factoren en preventiemogelijkheden., Dissertatie, Swets & Zeitlinger B.V., Lisse.
Kingma, J. and Klasen, H. L. (1994). Trend in alcoholgebruik bij verschillende categorieën verkeersslachtoffers in de periode 19701992. Tijdschrift voor Alcohol, Drugs en andere Psychotrope Stoffen 20, 210214 [In Dutch].
Kurzthaler, I., Wambacher, M., Golser, K. et al. (2003) Alcohol and/or benzodiazepine use in inured road users. Human Psychopharmacology 18, 361367.[CrossRef][ISI][Medline]
Lahaut, V. M., Jansen, H.A., van de Mheen, D. et al. (2002) Non-response bias in a sample survey on alcohol consumption. Alcohol and Alcoholism 37, 256260.
Monshouwer, K., Dorsselaer van, S., Gorter, A. et al. (2004). Jeugd en riskant gedrag. Kerngegevens uit het Peilstationonderzoek 2003. Utrecht: Trimbos-instituut [In Dutch].
Treno, A. J., Gruenewald, P. J. and Johnson, F. W. (1998) Sample selection bias in the emergency room: an examination of the role of alcohol in injury. Addiction 91, 113129.
Van de Mheen, D. (1998) Inequalities in health, to be continued? A life-course perspective on socio-demographic inequalities in health. Dissertation, Erasmus University Rotterdam.
Verdurmen, J., Monshouwer, K., Van Dorsselaer, S. (2003). Bovenmatig drinken in Nederland: uitkomsten van de Netherlands mental health survey and incidence study [NEMESIS]. Utrecht: Trimbos-instituut [In Dutch].
Weber, J. E., Maio, R. F., Blow, F. C. et al. (2002) Alcohol and/or other drug use among adult non-occupant motor vehicle crash victims. Alcohol and Alcoholism 37, 468471.
Wilsterman, M. E. F., Dors, N., Sprij, A. J. et al. (2004). Clinical characteristics and management of adolescents admitted to the emergency ward for alcohol intoxication in the region of The Hague during the period 19992001. Nederlands Tijdschrift voor Geneeskunde 148, 14961500 [In Dutch].[Medline]
|