- Email: [email protected]
Alcohol consumption and injury in the general population: from a national sample
I LSEVIER SCIENCF IRELANI>
Drug and Alcohol
Alcohol consumption
Dependence
34 (1994) 2 17-224
and injury in the general population: national sample?
from a
Cheryl J. Cherpitel Alcohol Research Group, C&fornia Pacific Medical Center Research Insritute. 2000 Heurst Armue. Berkeley. (‘A Y47W
(Accepted
USA
9 July 1993)
Abstract
The association of alcohol consumption with injuries is well documented in the literature. The majority of data for non-fatal injuries has come from emergency room (ER) studies, however, and little is known of how representative ER samples are of injuries in the general population or of the actual risk at which drinking places the individual for accidental injury. Data were collected (1990) from a national probability household sample (N = 2058; weighted N = I1 50) on: the respondent’s most recent injury; whether treatment was obtained for that injury, and where; drinking prior to injury; quantity and frequency (Q-F) of usual drinking; frequency of drunkenness; experiences associated with alcohol dependence; and social consequences related to drinking - all during the last year. Injury for the last year was categorized as follows: without injury, untreated injury, injury treated in the ER, other treated injury. Males treated in the ER were significantly more likely to be heavy drinkers and were more likely to report alcohol dependence experiences and social consequences related to drinking than those without injuries, while females treated in the ER were more likely to report social consequences related to drinking compared to those without injuries. Age (OR = 0.87) and Q-F (OR = 1.31) were found to be predictive of reporting an injury during the last year. Among injured none of these variables were predictive of reporting treatment. When the interaction terms of Q-F by age and Q-F by gender were entered into the logistic regressions, only Q-F was predictive of an injury (OR = 1.54), while the interaction of Q-F by gender was significant for reporting a treated injury, with Q-F predictive among males (OR = 1.78), but not among females (OR = 1.26). The region of the country was not a significant predictor of injury or of treatment for an injury. Key words: Alcohol; Injury; General population
1. Introduction
While a large literature exists which documents the association between alcohol consumption and non-fatal casualties (reviewed in Roizen, 1982) much of the data supporting this association have come from studies carried out in hospital emergency rooms (ERs). While these emergency room studies have consistently found the injured more likely than the non-injured to have positive breathalyzer readings at the time of ER admission, and to report drinking prior to the event, as well as to report more frequent heavy drinking and more alcohol-related
tPresented at the Research San Antonio, Texas, June
Society
on Alcoholism
Annual
19-24, 1993.
0376-8716/94/$07.00 0 1994 Elsevier Science Ireland SSDI 0376-8716(93)00950-7
Meeting,
problems (Cherpitel, 1993a), these studies are limited in their usefulness for assessing the risk at which drinking may place the individual for accidental injury. Only studies carried out in the general population can adequately address this issue. The purpose of this paper is to report the association of alcohol, injury and emergency room treatment found in a national sample of the general population. This is the first time such data have been collected on a representative sample of the general population of the U.S. Emergency room studies have, nevertheless, been important in providing a wealth of data on the association of alcohol and casualties, since ERs are the primary location for the treatment of injuries, which are relatively rare in the general population. Associations of alcohol and injury have been found to vary greatly from one
Ltd. All rights reserved
218
ER study to another, however, even when probability or representative samples of ER patients have been used (Cherpitel, 1993b). Much of this variation is due to the varying socio-demographic characteristics of the particular ER population studied. Other limitations of ER studies have also affected their usefulness for determining increased risk at which alcohol places the individual for accidental injury. In many of these studies the noninjured who received care in the same emergency room during the same period of time as the injured have been used as controls to assess the risk of injury. As mentioned above, the injured have been found to be more frequent heavy drinkers and to report more alcohol-related problems, including experiences associated with alcohol dependence and social consequences related to drinking, than the non-injured. Use of the non-injured for comparison with the injured, however, underestimates the association of alcohol with injury since the non-injured have been found to report higher rates for frequent heavy drinking and alcohol-related problems than that found in the general population from which they come (Cherpitel, 1993b). Other factors also limit the usefulness of ER studies. Many individuals do not reach the emergency room until well after the event, which affects the validity of estimated blood alcohol at the time of the event. We are also not aware from ER studies of those who seek care elsewhere or of those who seek no care, possibly because they were too drunk at the time to do so. Given these limitations of ER studies, general population studies of alcohol and injury can begin filling this gap in knowledge by providing data not only on the association of alcohol and injury, but also on injuries which are not treated, and on those which are treated in places other than emergency rooms. A limited amount of data on the association of alcohol, injury and treatment seeking has been reported from the general population of a northern California county. These data found similar associations of alcohol and injury among those sampled in the ERs of a health maintenance organization (HMO) in the county and those sampled in the general population of the county who belonged to the HMO and who reported an injury for which ER treatment was obtained (Cherpitel, 1992). While these data lend support for the generalizability of findings of associations of alcohol and casualties obtained among patients in ER studies to their counterparts in the general population, they tell us little about the actual risk at which alcohol places the individual for injury. Additional analyses of general population data on alcohol and injury in this county found those in the entire general population sample reported rates of usual drinking patterns and alcohol-related problems more similar to those in the HMO ER sample (both injured and noninjured) than those in a sample of ER patients from the county and community hospitals in the same county
C.J. Cherpitel
et al. /Drug
Alcohol
Depend.
34 (1994)
217-224
(Cherpitel, 1993b). While these data from the general population and emergency rooms in the same county provide some interesting and important comparisons, we do not know how representative the general population of this county may be to a national sample of the household population. Reported here, then, are data from a national household probability sample of adults on demographic characteristics including region of the county, drinking patterns, alcohol dependence, social consequences related to drinking and injury during the last year. Comparisons are made between those reporting without injury, an untreated injury, an injury for which other than ER treatment was sought, and an injury treated in the ER. Logistic regression is then used to analyze the predictive value of alcohol and demographic variables on these injury outcomes. Such analyses in a national sample are important in understanding the role of alcohol in injury occurrence and in treatment for an injury, and the risk at which alcohol may place the individual for accidental injury. 2. Methods 2.1. The sample The data are from the Alcohol Research Group’s 1990 National Alcohol Survey. Field-work for the study was subcontracted to the Institute for Survey Research at Temple University. The sample consisted of adults 18 years and older living in households within the 48 contiguous states. A multistage area-probability sample was drawn using 100 primary sampling units, which provided 2058 completed interviews, representing a 70% completion rate. Non-interviews were due to refusals (13%) and incapacitation, language barriers and failure to relocate the respondent after the initial screening contact (17%). Data were weighted to reflect the number of adults living in a selected household and the interview completion rate in a given area. Data were also weighted to take into account the design effect inherent in the use of multistage cluster sampling, using an approach originally suggested by Kish (1965) for estimating a design effect average. The average design effect of 1.75 was calculated using 31 variables from five domains of the questionnaire which resulted in an effective sample size of 1150 for analysis (Greenfield et al., 1992). 2.2. Instruments Interviews were carried out with informed consent in the respondent’s own home by trained interviewers using structured interview schedules of about 75 min in length. Among other items, respondents were asked the quantity and frequency of usual drinking and frequency of drunkenness during the last year, a number of questions related to alcohol dependence and to social consequences of drinking and demographic characteristics,
C. J. Cherpirel et al. / Drug Alcohol Depend. 34 i 1994) 217-224
including region of the country where the respondent lived: Northeast (New England and Mid-Atlantic), Midwest (East North Central and West North Central), Pacific, South (East South Central and West South Central), Mountain. Respondents were also asked in what year they most recently had had an injury, whether or not they received treatment for that injury, and if so, where. Since recall is best for more recently occurring events, the last year was taken as the time frame for an injury event. A quantity-frequency (Q-F) typology was developed so the association of patterns of use with injury could be analyzed. This variable has been found to be predictive of injury in both emergency room studies and in the general population (Cherpitel, 1993a). Frequency of drinking was defined as: low (
219
items can be found elsewhere (Caetano and Room, 1991). Social consequences related to drinking during the last year were measured by 21 items which have been used in previous general population surveys in the U.S. (Room, 1977). These items included the following problems related to drinking: problems with spouse, partner or other personal relationships; work: the police or other authorities; physical health; getting hurt or contributing to an accident; getting into a tight or heated argument. Both experiences related to alcohol dependence and social consequences of drinking were asked only of those who ever reported having at least 5 drinks on one occasion in their lifetime. Analysis of the prevalence of alcohol dependence and social consequences of drinking in an earlier general population survey which used many of these same alcohol-related problem items found that only 2% of those who never consumed 5 or more drinks on one occasion reported three or more dependence experiences and 1% reported more than one of the social consequences related to drinking (Midanik and Clark, 1993). 2.3. Data analysis Comparisons are made among the following groups: (i) those without injury; (ii) those with an untreated injury; (iii) those with an injury for which other than ER treatment was obtained; (iv) those with an injury for which ER treatment was obtained. Since our purpose here is to describe the association of alcohol with injury and with injury treatment, the proportion of those in each category of injured are compared, separately, to the proportion who reported no injury during the last year, in relation to quantity and frequency of usual drinking, frequency of drunkenness, alcohol dependence and social consequences. Logistic regression analysis (SPSS, 1990) was used to predict, separately, the following outcomes during the last year: (i) no injury vs. an injury; (ii) without injury or no treated injury vs. a treated injury; (iii) an untreated injury vs. a treated injury; (iv) a non-ER treated injury vs. an ER treated injury; (v) an untreated injury or a non-ER treated injury vs. an ER treated injury. Age, gender, region of the country, quantity-frequency of usual drinking, frequency of drunkenness, alcohol dependence and social consequences were entered simultaneously in each regression along with the interaction terms of Q-F by age and Q-F by gender. Among predictor variables, the following variables were coded as dichotomous: gender (males = 0, females = l), alcohol dependence ( < 3 = 0, 2 3 = l), social consequences (< 2 = 0, 2 2 = 1). Ordinal variables were coded as follows: age (18-29 = 0, 30-39 = 1, 40-49 = 2, 50-59 = 3, 60-69 = 4, 70+ = 5), Q-F (abstainer = 0, infrequent = 1, light = 2, moderate = 3,
C.J. Cherpitel et al. /Drug Alcohol Depend. 34 (1994) 217-224
220
heavy = 4, frequency of drunkenness (not in last year = 0,
Table 1 compares demographic characteristics among the injury categories. The injured who sought treatment in the ER were significantly more likely to be male and to be under 30 than those reporting without injury. The age distribution was analyzed separately for males and females within injury categories (not shown). Males seeking treatment in the ER were significantly more likely to be under 30 (50%) than those without injury (25%), while this difference was not significant among females (36% vs. 25%).
2.4. Sample characteristics Of the 1150 respondents, data were not available on reports of an injury during the last year for six subjects, leaving 1144 cases for analyses. Of these, 16% (N = 184) reported an injury during the last year. Males were more likely than females to report an injury (19% vs. 13%).
3. Results
Table 1 Demographic
Characteristics
Table 2 shows the quantity and frequency of usual drinking during the last year by injury category. Males who were injured and were treated in the ER were less likely to be abstainers and more likely to be heavy
(in percent) Total
No injury
Injury no treatment
Injury other treatment
Injury ER treatment
1144
960
44
61
79
Gender Male Female
48 52
46 54
60 40
54 46
58* 42*
Age 18-29 30-39 40-49 so-59 60-69 70+
27 23 16 11 I2 II
25 22 16 I2 13 12
31 27 I5 13 IO 4
26 35* 15 6 11
44* 23 IO 9 8 6*
14 14 8 3 1
13 15 8 4 1
88* 6* 4 2 0*
87* 8 4 0*
70 17 8 1* 4
10 14 39 19 19
6 14 38 18 18
1* 6* 48 23 21
2* 6* 42 25 25
5 8 36 30* 16
63
64
64
61
54
3 8 7 19 6
3 8 8 I8 5
4 7 2* 23 6
4 6 6 24 6
5 II 3* 27 6
Race White Black Latin0 Asia Other
7
Education 5 8th grade Some high school High school graduate Some college College graduate Marital status Married/marriage-like relationship Separated Divorced Widowed Never married Unemployed lP < 0.05, comparison
of proportion
in each category
of injury,
to non-injured.
C J. Cherpitel et 01. /Drug
Table 2 Quantity
and frequency
221
Alcohol Depend. 34 (1994) 217-224
of usual drinking
during
the last year by gender
(in percentld
Total
No Injury
Injury no treatment
Injury other treatment
-___ Injury ER treatment
Quantity/frequency Abstainer Infrequent Light Moderate Heavy
(1140) 35 17 23 23 2
(959) 38 I7 23 21 I
(44) 20* 20 24 34% 2
(61) 27 14 2s 30 4
(77) 18’ 16 21 35’ II*
Males Quantity/frequency Abstainer Infrequent Light Moderate Heavy
(544) 29 12 22 33 4
(442) 32 I3 23 30 3
(26) 23 I3 20 42 2
(331 20 7 19 48 8
(43) 9* I3 I4 46 19’
(596) 41 21 25 13 I
(517) 42 21 24 I3 I
117) I7 31 29 22 2
(28) 37 22 33 9 0
(34) 30 20 30 21 0
Females Quantity/frequency Abstainer Infrequent Light Moderate Heavy
“This table excludes 4 respondents on whom quantity-frequency data were not obtained. *P < 0.05, comparison of proportion in each category of injury to non-injured.
drinkers compared to the non-injured. They were also more likely to be heavy drinkers than those who had untreated injuries and they also appear to be more likely to be heavier drinkers than those in the other injury categories as well, although this difference was not significant. Males who sought treatment for an injury in the ER were also less likely to report no drunkenness during the last year compared to those reporting without injury (Table 3). Table 4 shows the proportion of those within each injury category reporting alcohol dependence and social consequences related to drinking during the last year. Those treated in the ER were more likely to report social consequences of drinking than those without injuries. While both males and females treated in the ER were more likely to report social consequences of drinking than those in the other injury categories, and males treated in the ER were also more likely to report alcohol dependence than those in the other categories, the number of respondents in each of these categories is small, and therefore these comparisons did not reach significance. The injured were also asked if they had been drinking within six hours of the injury event. While the number of respondents in each of the injury categories is too small for statistical significance to be reached, 13% of
the males who sought ER treatment reported drinking during this time compared to 8% of those with no treatment and 3% of those reporting other sources of treatment. Females treated in the ER, however, were less likely to report drinking prior to the event (3%) than those reporting no treatment (19O/,) or those reporting treatment from another source (4%). A number of logistic regressions were then used to estimate the predictive value on injury outcome of the following variables: gender, age, region, Q-F, drunkenness, dependence, consequences. Age (OR = 0.87) and Q-F (OR = 1.3 1) were found to be predictive of reporting an injury compared to no injury. When the interaction terms of Q-F by gender and Q-F by age were included in the regression, only Q-F was predictive of an injury (OR = 1.54) while the interaction of Q-F by gender was predictive of a treated injury (compared to no injury or an untreated injury) (Table 5). Using the logistic coefficient of the interaction term, odds ratios for males and females were calculated separately for each unit change in Q-F. For each level increase in Q-F, males were over three-quarters again as likely to report an injury (OR = 1.78, P c .Ol) while females were only a quarter again as likely to report an injury (OR = 1.26, C.I. = 0.90-1.75). None of these variables was predictive of reporting treatment among those injured, possibly due to the
C.J. Cherpitel et al. /Drug Alcohol Depend. 34 (1994) 217-224
222
Table 3 Frequency
of drunkenness
during
the last year among
drinkers
by gender
(in percent)”
Total
No Injury
Injury no treatment
Injury other treatment
Injury ER treatment
(743) 10 37 53
(599) 9 35 56
(35) 8 52 40
(44) 13 40 41
(65) 16 45 39*
Frequency of drunkenness B Monthly
(388) 13 40 48
(301) I2 37 52
(20) 2 63 35
(27) 22 46 33
(41) 20 49 32*
Females Frequency of drunkenness 2 Monthly
(354) 7 34 59
(298) 6 34 60
(15) I6 38 46
(18) 0 32 68
(24) IO 39 51
Frequency of drunkenness z Monthly
1This table excludes 2 drinkers on whom drunkenness data were not obtained. *P < 0.05, comparison of proportion in each category of injury to non-injured.
small number of respondents (less than 200) on whom the regressions were carried out. Region of the country was not a significant predictor in any of the regressions. 4. Discussion Analyses reported here are important in two ways which contribute to a better understanding of the association of alcohol and casualties. These data from the general population of a national sample provide the opportunity to assess the actual risk at which drinking may
Table 4 Alcohol dependence
and social consequences
of drinking
during
place the individual for accidental injury. Additionally, these data also provide the opportunity of assessing how those treated for injuries in the emergency room may be different in drinking characteristics from those with injuries who received other kinds of treatment or no treatment. Those who reported treatment in the ER for an injury were more likely to be male and to be younger than those without injuries. Males treated in the ER were more likely to be younger, to be heavy drinkers, and to report more frequent drunkenness than those in the
the last year among
drinkers
by gender
(in percent)
Total
No Injury
Injury no treatment
Injury other treatment
Injury ER treatment
(745)
(600)
(35)
(44)
(66)
5
5
4
6
IO
13
I2
8
15
23+
Males Three or more Dependence items Two or more Consequence items
(389)
(301)
(20)
(27)
(42)
8
7
0
7
I4
16
I5
9
20
28
Females Three or more Dependence items Two or more Consequence items
(355)
(299)
(15)
(18)
(24)
3
3
IO
3
2
9
9
6
6
I5
Three or more Dependence items Two or more Consequence items
*P < 0.05, comparison
of proportion
in each category
of injury
to non-injured.
C.J. Cherpitel et al. /Drug Alcohol Depend. 34 (1994) 217-224
Table 5 Logistic regression coefficients drinking variables on injury injury = 1) (N = 1140)
and odds ratios for demographics and status (no treated injury = 0, treated
Logistic coefficient
S.E.
Odds ratio
Wald statistic
0.4543 -0.0343 -0.1032 0.0212 -0.1293 0.0682 0.5785
0.3465 0.0928 0.2780 0.3250 0.2602 0 4001 0 1535
1.58 0.97 0.90 I .02 0.87 1.07 1.78
14.20**
Drunkenness Dependence
-0.0552 -0.0630
0.2089 0.4729
0.95 1.06
Consequences Q-F by gender Q-F by age
0.0148 -0.343 I -0.0770
0.3392 0. I628 0.0510
1.01 0.71 0.93
Variable
Gender Age Midwest Pacific South Mountain Quantity-frequency
(Q-F)
*P <0.05, (x*with **P < 0.01.
4.44*
I df).
other injury categories. They also appeared to be more likely to report alcohol dependence experiences and social consequences of drinking but these differences did not reach significance. While differences in injury categories were not significant among females, those who reported being in the ER for treatment of an injury during the last year appeared to be more likely to report social consequences of drinking than those in the other injury categories. It should be noted here that multiple comparisons were made among the injury categories which increases the probability of detecting significant differences between categories by chance alone. These data suggest that both drinking patterns and alcohol-related problems are associated with reporting an injury, especially among males, and that alcohol consumption variables may also be associated with seeking treatment for an injury in the ER as compared to seeking no treatment or to seeking treatment elsewhere. This may be explained in that alcohol has been found to be associated with severity of injury (Shepherd et al., 1988) and more specifically, with certain types of injury but not with others (Honkanen and Smith, 1990). It seems safe to assume that injuries for which treatment is obtained in an emergency room, compared to injuies for which no treatment or treatment of a less urgent nature is obtained, may be considered more severe. These data emphasized the caveat that data from emergency room studies cannot be generalized to represent all injuries. It is also possible that some bias may have been operating in recall of an injury. Of those who reported an in-
73-3 -._.
jury during the last year, 43% reported an injury which was treated in the ER compared to 33”/0who reported an injury treated elsewhere and 24% who reported no treatment. It may be that those who sustained an injury serious enough to be treated in an emergency room would be more likely to remember the injury than those with injuries treated elsewhere, and similarly, those with injuries treated elsewhere may be more likely to remember the injury than those who had injuries requiring no treatment. It is possible that these differences in injury recall may have affected study findings reported here. The question remains however, are young males just more likely to become involved in situations in which they become injured and require ER care, and unrelatedly, are they also more likely to be heavier drinkers with more alcohol problems than others? Or is frequent, heavy and problem drinking of young males somehow more than coincidentally related to injury occurrence requiring ER treatment? The logistic regression analyses sheds some light on this issue. Quantity and frequency of drinking and age were, separately, both significantly predictive of reporting an injury, with heavier younger drinkers most likely to do so. When the interaction term of quantity-frequency by gender and by age were included, quantity-frequency continued to be predictive of reporting an injury, while quantity-frequency was predictive of reporting a treated injury only among males. The data suggest, then, that usual heavy drinking (but not frequency of drunkenness, alcohol dependence or social consequences of drinking) is positively predictive of injury with age and gender controlled, and is also positively predictive of a treated injury among males, regardless of age. When interaction terms are used in logistic regression a larger N is needed to detect significant differences from 1 in an odds ratio. Power analysis determined, however. that the number of respondents in these regressions (1140) was sufficient to detect a significant difference from 1 in an odds ratio of 1.7 (P = 0.05, 0 = 0.80, one-tailed) when interaction terms were included (Hsieh, 1989). Regressions carried out to analyze the predictors of treatment seeking among the injured, while powerful enough (using the above criteria) to detect significance in odds ratios of 1.5 when interaction terms were not used, had considerably decreased power when interaction terms were entered into the equations. It should also be noted that data here provide conservative estimates because the design effects weight used was derived from a large number of variables rather than only from variables used in these analyses (Greenlield et al., 1992). Although demographic and drinking characteristics have been found to vary greatly from one emergency room study to another (Cherpitel, 1993b), it is reassuring to note here that those who reported ER treatment
224
for an injury during the last year are similar in demographic characteristics and drinking patterns to those injured sampled in a representative sample of all of those seeking emergency room treatment for an injury in one Northern California county (Cherpitel, in press). While these data are an important first step in understanding the association of alcohol and non-fatal casualties in the general population, and how these associations may differ from those found in ER samples, other factors may also affect associations found. For example risk taking and sensation seeking behaviors have been found to be associated with both drinking and with activities which might put one at high risk for accidental injury, such as sky diving or skiing. These activities are more common among young males as well. Prior analyses of data from this national survey, however, found no independent effect of risk taking, impulsivity, sensation seeking or risk perception on reporting a treated injury (Cherpitel, 1993~). Additionally, data were not analyzed here on injury by type or cause and one might expect associations of alcohol, injury and treatment to vary by the kind of injury sustained, regardless of severity. Future analyses to address concerns raised here would require larger data sets of nationally representative samples which would provide the power for determining differences in associations of alcohol, injury and treatment for various kinds of casualties within gender- and agespecific categories. While no differences were found in the association of alcohol and injury by region of the country here, it would not be unlikely that future analyses with larger data sets may, indeed, find that these associations vary by region, since drinking patterns are known to vary from one region to another, and treatment seeking for injuries may vary as well.
5. Acknowledgements
Supported by a National Alcohol Research Center Grant (AA-05595) from the U.S. Institute on Alcohol Abuse and Alcoholism.
C. J. Chrrpitel
et al. /Drug
Alcohol
Depend.
34 ( 1994)
217-224
6. References American Psychiatric Association (1987). Diagnostic and Statistical Manual of Mental Disorders. Washington, DC: American Psychiatric Association. Caetano, R. and Room, R. (1991). Alcohol dependence in the 1990 U.S. national alcohol survey: operationalizing and comparing two nosological systems. Berkeley, Alcohol Research Group. Cherpitel, C.J. (1992). Drinking patterns and problems: a comparison of ER patients in an HMO and in the general population. Alcoholism: Clin. Exp. Res. 16(6), 1104-l 109. Cherpitel, C.J. (1993). Alcohol and casualties: comparison of countywide emergency room data with the county general population Addiction, in press. Cherpitel, C.J. (1993a). Alcohol and injuries: a review of international emergency room studies. Addiction 88. 651-665. Cherpitel, C.J. (1993b). Alcohol consumption among emergency room patients: comparison of county/community hospitals and an HMO. J. Stud. Alcohol 54, 432-440. Cherpitel. C.J. (1993~). Alcohol, injury and risk taking behavior: Data from a national sample. Alcohol. Clin. Exp. Res. 17. Fleiss, J.L. (1981). Statistical methods for rates and proportions. John Wiley & Sons. New York. Greenfield, T.K.. Hudes. E.S. and Krotke. K.P. (1992). Practical sampling design issues in national alcohol surveys. Presented at the Alcohol Epidemiology Symposium, Kettil Bruun Society for Social and Epidemiological Research on Alcohol, Toronto. Canada, June 1-5. Honkanen, R. and Smith, G.S. (1990). impact of acute alcohol intoxication on the severity of injury: a cause-specific analysis of nonfatal trauma. Injury 21. 353-357. Hsieh. F.Y. (1989). Sample size tables for logistic regression. Stat. Med. 8, 795-802. Kish. L. (1965). Survey sampling. John Wiley & Sons, New York. Midanik. L.T. and Clark, W.B. (1993). Changes in drinking problems in the U.S. from 1984 to 1990: results of two national alcohol surveys. Presented at the Research Society on Alcoholism Scientific Conference, San Antonio. TX, June. Roizen, J. (1982). Estimating alcohol involvement in serious events. In: Alcohol consumption and related problems, (U.S. Department of Health and Human: Services eds.). pp. 179-219. NIAAA. Rockville. MD. Room, R. (1977). Measurement and distribution of drinking patterns and problems in general populations. In: Alcohol-related disabilities, (Edwards, G. Gross, M.M. Keller, M. Moser, J. and Room, R., eds.). pp. 61-87. Offset Publication No. 32. World Health Organization. Shepherd, J.. Irish, M., Scully, C. and Leslie, 1. (1988). Alcohol intoxication and severity of injury in victims of assault. Br. Med. J. 296. 1299. SPSS. (1990). SPSS Reference Guide. SPSS. Inc., Chicago
Drug and Alcohol
Alcohol consumption
Dependence
34 (1994) 2 17-224
and injury in the general population: national sample?
from a
Cheryl J. Cherpitel Alcohol Research Group, C&fornia Pacific Medical Center Research Insritute. 2000 Heurst Armue. Berkeley. (‘A Y47W
(Accepted
USA
9 July 1993)
Abstract
The association of alcohol consumption with injuries is well documented in the literature. The majority of data for non-fatal injuries has come from emergency room (ER) studies, however, and little is known of how representative ER samples are of injuries in the general population or of the actual risk at which drinking places the individual for accidental injury. Data were collected (1990) from a national probability household sample (N = 2058; weighted N = I1 50) on: the respondent’s most recent injury; whether treatment was obtained for that injury, and where; drinking prior to injury; quantity and frequency (Q-F) of usual drinking; frequency of drunkenness; experiences associated with alcohol dependence; and social consequences related to drinking - all during the last year. Injury for the last year was categorized as follows: without injury, untreated injury, injury treated in the ER, other treated injury. Males treated in the ER were significantly more likely to be heavy drinkers and were more likely to report alcohol dependence experiences and social consequences related to drinking than those without injuries, while females treated in the ER were more likely to report social consequences related to drinking compared to those without injuries. Age (OR = 0.87) and Q-F (OR = 1.31) were found to be predictive of reporting an injury during the last year. Among injured none of these variables were predictive of reporting treatment. When the interaction terms of Q-F by age and Q-F by gender were entered into the logistic regressions, only Q-F was predictive of an injury (OR = 1.54), while the interaction of Q-F by gender was significant for reporting a treated injury, with Q-F predictive among males (OR = 1.78), but not among females (OR = 1.26). The region of the country was not a significant predictor of injury or of treatment for an injury. Key words: Alcohol; Injury; General population
1. Introduction
While a large literature exists which documents the association between alcohol consumption and non-fatal casualties (reviewed in Roizen, 1982) much of the data supporting this association have come from studies carried out in hospital emergency rooms (ERs). While these emergency room studies have consistently found the injured more likely than the non-injured to have positive breathalyzer readings at the time of ER admission, and to report drinking prior to the event, as well as to report more frequent heavy drinking and more alcohol-related
tPresented at the Research San Antonio, Texas, June
Society
on Alcoholism
Annual
19-24, 1993.
0376-8716/94/$07.00 0 1994 Elsevier Science Ireland SSDI 0376-8716(93)00950-7
Meeting,
problems (Cherpitel, 1993a), these studies are limited in their usefulness for assessing the risk at which drinking may place the individual for accidental injury. Only studies carried out in the general population can adequately address this issue. The purpose of this paper is to report the association of alcohol, injury and emergency room treatment found in a national sample of the general population. This is the first time such data have been collected on a representative sample of the general population of the U.S. Emergency room studies have, nevertheless, been important in providing a wealth of data on the association of alcohol and casualties, since ERs are the primary location for the treatment of injuries, which are relatively rare in the general population. Associations of alcohol and injury have been found to vary greatly from one
Ltd. All rights reserved
218
ER study to another, however, even when probability or representative samples of ER patients have been used (Cherpitel, 1993b). Much of this variation is due to the varying socio-demographic characteristics of the particular ER population studied. Other limitations of ER studies have also affected their usefulness for determining increased risk at which alcohol places the individual for accidental injury. In many of these studies the noninjured who received care in the same emergency room during the same period of time as the injured have been used as controls to assess the risk of injury. As mentioned above, the injured have been found to be more frequent heavy drinkers and to report more alcohol-related problems, including experiences associated with alcohol dependence and social consequences related to drinking, than the non-injured. Use of the non-injured for comparison with the injured, however, underestimates the association of alcohol with injury since the non-injured have been found to report higher rates for frequent heavy drinking and alcohol-related problems than that found in the general population from which they come (Cherpitel, 1993b). Other factors also limit the usefulness of ER studies. Many individuals do not reach the emergency room until well after the event, which affects the validity of estimated blood alcohol at the time of the event. We are also not aware from ER studies of those who seek care elsewhere or of those who seek no care, possibly because they were too drunk at the time to do so. Given these limitations of ER studies, general population studies of alcohol and injury can begin filling this gap in knowledge by providing data not only on the association of alcohol and injury, but also on injuries which are not treated, and on those which are treated in places other than emergency rooms. A limited amount of data on the association of alcohol, injury and treatment seeking has been reported from the general population of a northern California county. These data found similar associations of alcohol and injury among those sampled in the ERs of a health maintenance organization (HMO) in the county and those sampled in the general population of the county who belonged to the HMO and who reported an injury for which ER treatment was obtained (Cherpitel, 1992). While these data lend support for the generalizability of findings of associations of alcohol and casualties obtained among patients in ER studies to their counterparts in the general population, they tell us little about the actual risk at which alcohol places the individual for injury. Additional analyses of general population data on alcohol and injury in this county found those in the entire general population sample reported rates of usual drinking patterns and alcohol-related problems more similar to those in the HMO ER sample (both injured and noninjured) than those in a sample of ER patients from the county and community hospitals in the same county
C.J. Cherpitel
et al. /Drug
Alcohol
Depend.
34 (1994)
217-224
(Cherpitel, 1993b). While these data from the general population and emergency rooms in the same county provide some interesting and important comparisons, we do not know how representative the general population of this county may be to a national sample of the household population. Reported here, then, are data from a national household probability sample of adults on demographic characteristics including region of the county, drinking patterns, alcohol dependence, social consequences related to drinking and injury during the last year. Comparisons are made between those reporting without injury, an untreated injury, an injury for which other than ER treatment was sought, and an injury treated in the ER. Logistic regression is then used to analyze the predictive value of alcohol and demographic variables on these injury outcomes. Such analyses in a national sample are important in understanding the role of alcohol in injury occurrence and in treatment for an injury, and the risk at which alcohol may place the individual for accidental injury. 2. Methods 2.1. The sample The data are from the Alcohol Research Group’s 1990 National Alcohol Survey. Field-work for the study was subcontracted to the Institute for Survey Research at Temple University. The sample consisted of adults 18 years and older living in households within the 48 contiguous states. A multistage area-probability sample was drawn using 100 primary sampling units, which provided 2058 completed interviews, representing a 70% completion rate. Non-interviews were due to refusals (13%) and incapacitation, language barriers and failure to relocate the respondent after the initial screening contact (17%). Data were weighted to reflect the number of adults living in a selected household and the interview completion rate in a given area. Data were also weighted to take into account the design effect inherent in the use of multistage cluster sampling, using an approach originally suggested by Kish (1965) for estimating a design effect average. The average design effect of 1.75 was calculated using 31 variables from five domains of the questionnaire which resulted in an effective sample size of 1150 for analysis (Greenfield et al., 1992). 2.2. Instruments Interviews were carried out with informed consent in the respondent’s own home by trained interviewers using structured interview schedules of about 75 min in length. Among other items, respondents were asked the quantity and frequency of usual drinking and frequency of drunkenness during the last year, a number of questions related to alcohol dependence and to social consequences of drinking and demographic characteristics,
C. J. Cherpirel et al. / Drug Alcohol Depend. 34 i 1994) 217-224
including region of the country where the respondent lived: Northeast (New England and Mid-Atlantic), Midwest (East North Central and West North Central), Pacific, South (East South Central and West South Central), Mountain. Respondents were also asked in what year they most recently had had an injury, whether or not they received treatment for that injury, and if so, where. Since recall is best for more recently occurring events, the last year was taken as the time frame for an injury event. A quantity-frequency (Q-F) typology was developed so the association of patterns of use with injury could be analyzed. This variable has been found to be predictive of injury in both emergency room studies and in the general population (Cherpitel, 1993a). Frequency of drinking was defined as: low (
219
items can be found elsewhere (Caetano and Room, 1991). Social consequences related to drinking during the last year were measured by 21 items which have been used in previous general population surveys in the U.S. (Room, 1977). These items included the following problems related to drinking: problems with spouse, partner or other personal relationships; work: the police or other authorities; physical health; getting hurt or contributing to an accident; getting into a tight or heated argument. Both experiences related to alcohol dependence and social consequences of drinking were asked only of those who ever reported having at least 5 drinks on one occasion in their lifetime. Analysis of the prevalence of alcohol dependence and social consequences of drinking in an earlier general population survey which used many of these same alcohol-related problem items found that only 2% of those who never consumed 5 or more drinks on one occasion reported three or more dependence experiences and 1% reported more than one of the social consequences related to drinking (Midanik and Clark, 1993). 2.3. Data analysis Comparisons are made among the following groups: (i) those without injury; (ii) those with an untreated injury; (iii) those with an injury for which other than ER treatment was obtained; (iv) those with an injury for which ER treatment was obtained. Since our purpose here is to describe the association of alcohol with injury and with injury treatment, the proportion of those in each category of injured are compared, separately, to the proportion who reported no injury during the last year, in relation to quantity and frequency of usual drinking, frequency of drunkenness, alcohol dependence and social consequences. Logistic regression analysis (SPSS, 1990) was used to predict, separately, the following outcomes during the last year: (i) no injury vs. an injury; (ii) without injury or no treated injury vs. a treated injury; (iii) an untreated injury vs. a treated injury; (iv) a non-ER treated injury vs. an ER treated injury; (v) an untreated injury or a non-ER treated injury vs. an ER treated injury. Age, gender, region of the country, quantity-frequency of usual drinking, frequency of drunkenness, alcohol dependence and social consequences were entered simultaneously in each regression along with the interaction terms of Q-F by age and Q-F by gender. Among predictor variables, the following variables were coded as dichotomous: gender (males = 0, females = l), alcohol dependence ( < 3 = 0, 2 3 = l), social consequences (< 2 = 0, 2 2 = 1). Ordinal variables were coded as follows: age (18-29 = 0, 30-39 = 1, 40-49 = 2, 50-59 = 3, 60-69 = 4, 70+ = 5), Q-F (abstainer = 0, infrequent = 1, light = 2, moderate = 3,
C.J. Cherpitel et al. /Drug Alcohol Depend. 34 (1994) 217-224
220
heavy = 4, frequency of drunkenness (not in last year = 0,
Table 1 compares demographic characteristics among the injury categories. The injured who sought treatment in the ER were significantly more likely to be male and to be under 30 than those reporting without injury. The age distribution was analyzed separately for males and females within injury categories (not shown). Males seeking treatment in the ER were significantly more likely to be under 30 (50%) than those without injury (25%), while this difference was not significant among females (36% vs. 25%).
2.4. Sample characteristics Of the 1150 respondents, data were not available on reports of an injury during the last year for six subjects, leaving 1144 cases for analyses. Of these, 16% (N = 184) reported an injury during the last year. Males were more likely than females to report an injury (19% vs. 13%).
3. Results
Table 1 Demographic
Characteristics
Table 2 shows the quantity and frequency of usual drinking during the last year by injury category. Males who were injured and were treated in the ER were less likely to be abstainers and more likely to be heavy
(in percent) Total
No injury
Injury no treatment
Injury other treatment
Injury ER treatment
1144
960
44
61
79
Gender Male Female
48 52
46 54
60 40
54 46
58* 42*
Age 18-29 30-39 40-49 so-59 60-69 70+
27 23 16 11 I2 II
25 22 16 I2 13 12
31 27 I5 13 IO 4
26 35* 15 6 11
44* 23 IO 9 8 6*
14 14 8 3 1
13 15 8 4 1
88* 6* 4 2 0*
87* 8 4 0*
70 17 8 1* 4
10 14 39 19 19
6 14 38 18 18
1* 6* 48 23 21
2* 6* 42 25 25
5 8 36 30* 16
63
64
64
61
54
3 8 7 19 6
3 8 8 I8 5
4 7 2* 23 6
4 6 6 24 6
5 II 3* 27 6
Race White Black Latin0 Asia Other
7
Education 5 8th grade Some high school High school graduate Some college College graduate Marital status Married/marriage-like relationship Separated Divorced Widowed Never married Unemployed lP < 0.05, comparison
of proportion
in each category
of injury,
to non-injured.
C J. Cherpitel et 01. /Drug
Table 2 Quantity
and frequency
221
Alcohol Depend. 34 (1994) 217-224
of usual drinking
during
the last year by gender
(in percentld
Total
No Injury
Injury no treatment
Injury other treatment
-___ Injury ER treatment
Quantity/frequency Abstainer Infrequent Light Moderate Heavy
(1140) 35 17 23 23 2
(959) 38 I7 23 21 I
(44) 20* 20 24 34% 2
(61) 27 14 2s 30 4
(77) 18’ 16 21 35’ II*
Males Quantity/frequency Abstainer Infrequent Light Moderate Heavy
(544) 29 12 22 33 4
(442) 32 I3 23 30 3
(26) 23 I3 20 42 2
(331 20 7 19 48 8
(43) 9* I3 I4 46 19’
(596) 41 21 25 13 I
(517) 42 21 24 I3 I
117) I7 31 29 22 2
(28) 37 22 33 9 0
(34) 30 20 30 21 0
Females Quantity/frequency Abstainer Infrequent Light Moderate Heavy
“This table excludes 4 respondents on whom quantity-frequency data were not obtained. *P < 0.05, comparison of proportion in each category of injury to non-injured.
drinkers compared to the non-injured. They were also more likely to be heavy drinkers than those who had untreated injuries and they also appear to be more likely to be heavier drinkers than those in the other injury categories as well, although this difference was not significant. Males who sought treatment for an injury in the ER were also less likely to report no drunkenness during the last year compared to those reporting without injury (Table 3). Table 4 shows the proportion of those within each injury category reporting alcohol dependence and social consequences related to drinking during the last year. Those treated in the ER were more likely to report social consequences of drinking than those without injuries. While both males and females treated in the ER were more likely to report social consequences of drinking than those in the other injury categories, and males treated in the ER were also more likely to report alcohol dependence than those in the other categories, the number of respondents in each of these categories is small, and therefore these comparisons did not reach significance. The injured were also asked if they had been drinking within six hours of the injury event. While the number of respondents in each of the injury categories is too small for statistical significance to be reached, 13% of
the males who sought ER treatment reported drinking during this time compared to 8% of those with no treatment and 3% of those reporting other sources of treatment. Females treated in the ER, however, were less likely to report drinking prior to the event (3%) than those reporting no treatment (19O/,) or those reporting treatment from another source (4%). A number of logistic regressions were then used to estimate the predictive value on injury outcome of the following variables: gender, age, region, Q-F, drunkenness, dependence, consequences. Age (OR = 0.87) and Q-F (OR = 1.3 1) were found to be predictive of reporting an injury compared to no injury. When the interaction terms of Q-F by gender and Q-F by age were included in the regression, only Q-F was predictive of an injury (OR = 1.54) while the interaction of Q-F by gender was predictive of a treated injury (compared to no injury or an untreated injury) (Table 5). Using the logistic coefficient of the interaction term, odds ratios for males and females were calculated separately for each unit change in Q-F. For each level increase in Q-F, males were over three-quarters again as likely to report an injury (OR = 1.78, P c .Ol) while females were only a quarter again as likely to report an injury (OR = 1.26, C.I. = 0.90-1.75). None of these variables was predictive of reporting treatment among those injured, possibly due to the
C.J. Cherpitel et al. /Drug Alcohol Depend. 34 (1994) 217-224
222
Table 3 Frequency
of drunkenness
during
the last year among
drinkers
by gender
(in percent)”
Total
No Injury
Injury no treatment
Injury other treatment
Injury ER treatment
(743) 10 37 53
(599) 9 35 56
(35) 8 52 40
(44) 13 40 41
(65) 16 45 39*
Frequency of drunkenness B Monthly
(388) 13 40 48
(301) I2 37 52
(20) 2 63 35
(27) 22 46 33
(41) 20 49 32*
Females Frequency of drunkenness 2 Monthly
(354) 7 34 59
(298) 6 34 60
(15) I6 38 46
(18) 0 32 68
(24) IO 39 51
Frequency of drunkenness z Monthly
1This table excludes 2 drinkers on whom drunkenness data were not obtained. *P < 0.05, comparison of proportion in each category of injury to non-injured.
small number of respondents (less than 200) on whom the regressions were carried out. Region of the country was not a significant predictor in any of the regressions. 4. Discussion Analyses reported here are important in two ways which contribute to a better understanding of the association of alcohol and casualties. These data from the general population of a national sample provide the opportunity to assess the actual risk at which drinking may
Table 4 Alcohol dependence
and social consequences
of drinking
during
place the individual for accidental injury. Additionally, these data also provide the opportunity of assessing how those treated for injuries in the emergency room may be different in drinking characteristics from those with injuries who received other kinds of treatment or no treatment. Those who reported treatment in the ER for an injury were more likely to be male and to be younger than those without injuries. Males treated in the ER were more likely to be younger, to be heavy drinkers, and to report more frequent drunkenness than those in the
the last year among
drinkers
by gender
(in percent)
Total
No Injury
Injury no treatment
Injury other treatment
Injury ER treatment
(745)
(600)
(35)
(44)
(66)
5
5
4
6
IO
13
I2
8
15
23+
Males Three or more Dependence items Two or more Consequence items
(389)
(301)
(20)
(27)
(42)
8
7
0
7
I4
16
I5
9
20
28
Females Three or more Dependence items Two or more Consequence items
(355)
(299)
(15)
(18)
(24)
3
3
IO
3
2
9
9
6
6
I5
Three or more Dependence items Two or more Consequence items
*P < 0.05, comparison
of proportion
in each category
of injury
to non-injured.
C.J. Cherpitel et al. /Drug Alcohol Depend. 34 (1994) 217-224
Table 5 Logistic regression coefficients drinking variables on injury injury = 1) (N = 1140)
and odds ratios for demographics and status (no treated injury = 0, treated
Logistic coefficient
S.E.
Odds ratio
Wald statistic
0.4543 -0.0343 -0.1032 0.0212 -0.1293 0.0682 0.5785
0.3465 0.0928 0.2780 0.3250 0.2602 0 4001 0 1535
1.58 0.97 0.90 I .02 0.87 1.07 1.78
14.20**
Drunkenness Dependence
-0.0552 -0.0630
0.2089 0.4729
0.95 1.06
Consequences Q-F by gender Q-F by age
0.0148 -0.343 I -0.0770
0.3392 0. I628 0.0510
1.01 0.71 0.93
Variable
Gender Age Midwest Pacific South Mountain Quantity-frequency
(Q-F)
*P <0.05, (x*with **P < 0.01.
4.44*
I df).
other injury categories. They also appeared to be more likely to report alcohol dependence experiences and social consequences of drinking but these differences did not reach significance. While differences in injury categories were not significant among females, those who reported being in the ER for treatment of an injury during the last year appeared to be more likely to report social consequences of drinking than those in the other injury categories. It should be noted here that multiple comparisons were made among the injury categories which increases the probability of detecting significant differences between categories by chance alone. These data suggest that both drinking patterns and alcohol-related problems are associated with reporting an injury, especially among males, and that alcohol consumption variables may also be associated with seeking treatment for an injury in the ER as compared to seeking no treatment or to seeking treatment elsewhere. This may be explained in that alcohol has been found to be associated with severity of injury (Shepherd et al., 1988) and more specifically, with certain types of injury but not with others (Honkanen and Smith, 1990). It seems safe to assume that injuries for which treatment is obtained in an emergency room, compared to injuies for which no treatment or treatment of a less urgent nature is obtained, may be considered more severe. These data emphasized the caveat that data from emergency room studies cannot be generalized to represent all injuries. It is also possible that some bias may have been operating in recall of an injury. Of those who reported an in-
73-3 -._.
jury during the last year, 43% reported an injury which was treated in the ER compared to 33”/0who reported an injury treated elsewhere and 24% who reported no treatment. It may be that those who sustained an injury serious enough to be treated in an emergency room would be more likely to remember the injury than those with injuries treated elsewhere, and similarly, those with injuries treated elsewhere may be more likely to remember the injury than those who had injuries requiring no treatment. It is possible that these differences in injury recall may have affected study findings reported here. The question remains however, are young males just more likely to become involved in situations in which they become injured and require ER care, and unrelatedly, are they also more likely to be heavier drinkers with more alcohol problems than others? Or is frequent, heavy and problem drinking of young males somehow more than coincidentally related to injury occurrence requiring ER treatment? The logistic regression analyses sheds some light on this issue. Quantity and frequency of drinking and age were, separately, both significantly predictive of reporting an injury, with heavier younger drinkers most likely to do so. When the interaction term of quantity-frequency by gender and by age were included, quantity-frequency continued to be predictive of reporting an injury, while quantity-frequency was predictive of reporting a treated injury only among males. The data suggest, then, that usual heavy drinking (but not frequency of drunkenness, alcohol dependence or social consequences of drinking) is positively predictive of injury with age and gender controlled, and is also positively predictive of a treated injury among males, regardless of age. When interaction terms are used in logistic regression a larger N is needed to detect significant differences from 1 in an odds ratio. Power analysis determined, however. that the number of respondents in these regressions (1140) was sufficient to detect a significant difference from 1 in an odds ratio of 1.7 (P = 0.05, 0 = 0.80, one-tailed) when interaction terms were included (Hsieh, 1989). Regressions carried out to analyze the predictors of treatment seeking among the injured, while powerful enough (using the above criteria) to detect significance in odds ratios of 1.5 when interaction terms were not used, had considerably decreased power when interaction terms were entered into the equations. It should also be noted that data here provide conservative estimates because the design effects weight used was derived from a large number of variables rather than only from variables used in these analyses (Greenlield et al., 1992). Although demographic and drinking characteristics have been found to vary greatly from one emergency room study to another (Cherpitel, 1993b), it is reassuring to note here that those who reported ER treatment
224
for an injury during the last year are similar in demographic characteristics and drinking patterns to those injured sampled in a representative sample of all of those seeking emergency room treatment for an injury in one Northern California county (Cherpitel, in press). While these data are an important first step in understanding the association of alcohol and non-fatal casualties in the general population, and how these associations may differ from those found in ER samples, other factors may also affect associations found. For example risk taking and sensation seeking behaviors have been found to be associated with both drinking and with activities which might put one at high risk for accidental injury, such as sky diving or skiing. These activities are more common among young males as well. Prior analyses of data from this national survey, however, found no independent effect of risk taking, impulsivity, sensation seeking or risk perception on reporting a treated injury (Cherpitel, 1993~). Additionally, data were not analyzed here on injury by type or cause and one might expect associations of alcohol, injury and treatment to vary by the kind of injury sustained, regardless of severity. Future analyses to address concerns raised here would require larger data sets of nationally representative samples which would provide the power for determining differences in associations of alcohol, injury and treatment for various kinds of casualties within gender- and agespecific categories. While no differences were found in the association of alcohol and injury by region of the country here, it would not be unlikely that future analyses with larger data sets may, indeed, find that these associations vary by region, since drinking patterns are known to vary from one region to another, and treatment seeking for injuries may vary as well.
5. Acknowledgements
Supported by a National Alcohol Research Center Grant (AA-05595) from the U.S. Institute on Alcohol Abuse and Alcoholism.
C. J. Chrrpitel
et al. /Drug
Alcohol
Depend.
34 ( 1994)
217-224
6. References American Psychiatric Association (1987). Diagnostic and Statistical Manual of Mental Disorders. Washington, DC: American Psychiatric Association. Caetano, R. and Room, R. (1991). Alcohol dependence in the 1990 U.S. national alcohol survey: operationalizing and comparing two nosological systems. Berkeley, Alcohol Research Group. Cherpitel, C.J. (1992). Drinking patterns and problems: a comparison of ER patients in an HMO and in the general population. Alcoholism: Clin. Exp. Res. 16(6), 1104-l 109. Cherpitel, C.J. (1993). Alcohol and casualties: comparison of countywide emergency room data with the county general population Addiction, in press. Cherpitel, C.J. (1993a). Alcohol and injuries: a review of international emergency room studies. Addiction 88. 651-665. Cherpitel, C.J. (1993b). Alcohol consumption among emergency room patients: comparison of county/community hospitals and an HMO. J. Stud. Alcohol 54, 432-440. Cherpitel. C.J. (1993~). Alcohol, injury and risk taking behavior: Data from a national sample. Alcohol. Clin. Exp. Res. 17. Fleiss, J.L. (1981). Statistical methods for rates and proportions. John Wiley & Sons. New York. Greenfield, T.K.. Hudes. E.S. and Krotke. K.P. (1992). Practical sampling design issues in national alcohol surveys. Presented at the Alcohol Epidemiology Symposium, Kettil Bruun Society for Social and Epidemiological Research on Alcohol, Toronto. Canada, June 1-5. Honkanen, R. and Smith, G.S. (1990). impact of acute alcohol intoxication on the severity of injury: a cause-specific analysis of nonfatal trauma. Injury 21. 353-357. Hsieh. F.Y. (1989). Sample size tables for logistic regression. Stat. Med. 8, 795-802. Kish. L. (1965). Survey sampling. John Wiley & Sons, New York. Midanik. L.T. and Clark, W.B. (1993). Changes in drinking problems in the U.S. from 1984 to 1990: results of two national alcohol surveys. Presented at the Research Society on Alcoholism Scientific Conference, San Antonio. TX, June. Roizen, J. (1982). Estimating alcohol involvement in serious events. In: Alcohol consumption and related problems, (U.S. Department of Health and Human: Services eds.). pp. 179-219. NIAAA. Rockville. MD. Room, R. (1977). Measurement and distribution of drinking patterns and problems in general populations. In: Alcohol-related disabilities, (Edwards, G. Gross, M.M. Keller, M. Moser, J. and Room, R., eds.). pp. 61-87. Offset Publication No. 32. World Health Organization. Shepherd, J.. Irish, M., Scully, C. and Leslie, 1. (1988). Alcohol intoxication and severity of injury in victims of assault. Br. Med. J. 296. 1299. SPSS. (1990). SPSS Reference Guide. SPSS. Inc., Chicago