OUP user menu


Zsusza Varvasovszky, Martin McKee
DOI: http://dx.doi.org/10.1093/alcalc/35.6.574 574-579 First published online: 1 November 2000


Alcohol is an important determinant of the overall burden of disease in Eastern Europe. It is a particularly important problem in Hungary, where death rates from cirrhosis have increased rapidly to levels much higher than in neighbouring countries. This study sought to describe the prevalence of problem drinking among the hospitalized population in Hungary by means of a survey of self-reported alcohol consumption and of the prevalence of current or lifetime problem drinking among hospitalized patients in the four teaching hospitals in Hungary. A survey was conducted of all patients (n = 3140) admitted to medical, surgical, trauma, psychiatric and neurological wards over a 2-week period in 1997 who stayed in hospital for at least 24 h, using a survey instrument based on the Alcohol Use Disorders Identification Test (AUDIT) and Luebeck Alcoholism Screening Test (LAST) instruments designed to screen for current and lifetime problem drinking respectively. In all, 23.5% of men and over 53.5% of women reported never drinking alcohol. Of those who did drink, about one in eight men and less than 1% of women reported drinking 5 or more drinks on a day when they drank. Whether defined by LAST or AUDIT, the prevalence of problem drinking was ~19% among men and 2% among women, although this rose to 32–35% among men aged 35–44 years. The rate did not vary significantly with employment or education, but was higher among those who were divorced. These high rates of problem drinking indicate the need for coherent policies on alcohol in Hungary.


Alcohol is now recognized to be one of the leading causes of premature mortality in Hungary where death rates from cirrhosis increased from 13 per 100 000 population in 1970 to 80 per 100 000 in 1995, a rate of increase substantially greater than in neighbouring countries (Varvasovszky et al., 1997). On the basis of American attribution fractions for different causes of death (Department of Health and Human Services, 1990), 11.4% of all deaths in Hungary can be considered to be due to alcohol (Varvasovszky, 1998), a figure that is much higher than has been recorded elsewhere (Department of Health and Human Services, 1990). Thus, it is surprising that this is the first systematic attempt to quantify the prevalence of prob-lem drinking in the hospital population in Hungary. There has been some earlier research but, in general, either the methods were unclear or the samples were small. A study based on autopsies undertaken in the University Hospital in Budapest in 1984 reported that 5% of autopsies were on patients described in their case notes as chronic alcoholics, although this was not defined (Konyàr et al., 1984). A study based on medical records reported that, in 1982–1984, 6.1% of hospital admissions were of chronic alcoholics, again undefined (Simonyi, 1987/8). This had risen to 13% in 1986. A survey of those injured in traffic accidents who required hospital treatment, in Vas county in 1988–1989, found that 6.2% were drunk, although this was based on the judgement of the admitting physician (Kazàr et al., 1993). Of admissions to a psychiatric unit in 1991–1992, 27% were estimated to be alcohol-related (Környei and Siska, 1994). Other studies have examined the prevalence of problem drinking, typically without standardized measurement, among patients with a particular diagnosis.

It has been recognized for many years that the prevalence of problem drinking is especially high among hospitalized patients (Jarman and Kellet, 1979), although reported levels vary widely depending on the population studied and the method of screening used (Pearson, 1962; Green, 1965; Nolan, 1965; Barcha et al., 1968; Moore, 1971; McCusker et al., 1975; Jariwalla et al., 1979; Barrison et al., 1982; Taylor et al., 1986; Mansoor and Edwards, 1991; Umbricht-Schneiter et al., 1991). Outside specific surveys, however, the scale of the problem is often overlooked and many patients with alcohol-related problems are unrecognized by their physicians (Moore et al., 1989). This is unfortunate, because a hospital admission offers potential to intervene and there is some evidence that brief interventions, including patient assessment and education, counselling, goal setting, and monitoring of liver enzyme levels can be effective in reducing alcohol consumption (Kahan et al., 1995).

Locally conducted surveys of the frequency of problem drinking among hospitalized patients can be used to draw attention to the scale of the problem and to act as a baseline against which to monitor future trends. Despite the high burden of disease attributable to alcohol, such surveys have not previously been undertaken in Hungary. The present paper addresses these issues.



The study was undertaken in the four Hungarian university hospitals (Budapest, Debrecen, Pécs and Szeged). The study population consisted of all patients admitted to medical, surgical, trauma, psychiatric and neurological wards over a 2-week period in March, 1997 and who stayed in hospital for at least 24 h. It was not possible to undertake interviews in some wards on the scheduled date, so these were surveyed in July and August 1997, on the same day of the week on which the original interview was scheduled. This accounted for only 60 patients (1.9%).

An instrument was designed and piloted that incorporated three sets of questions: demographic characteristics, screening for hazardous and problem drinking, and details on cause of admission and on alcohol consumption in the previous week (or, for those recording no consumption in the previous week, consumption in the previous month). Data were collected by medical students, typically in the day following admission, using interviews with patients and extraction of data from case notes. The medical students underwent two training periods, each lasting half a day. The first session covered the aims of the survey and clarified their understanding of the meaning of each question. The second involved observation of actual interviews with subjects, followed by feedback and discussion. The training emphasized the importance of a consistent administration of the interview.

Survey instruments

Several methods have been used to identify excessive or problem-drinking among hospitalized patients (Chick, 1994). They include questionnaires on consumption, ‘drinking diaries’, physical examination, and biological markers, any of which may be used alone or in combination (Centre for Reviews and Dissemination, 1993). Survey instruments include interviews and self-completed questionnaires. Most are designed to detect alcoholism (ICD-8) or alcohol dependence and alcohol abuse (ICD-9; DSM-IIIR criteria).

LAST (Luebeck Alcoholism Screening Test) is a 7-item questionnaire [two questions derived from CAGE (Ewing, 1984) and five from MAST (Seltzer, 1971)] that has been shown to have a higher sensitivity than CAGE or MAST among hospitalized patients (Rumpf et al., 1997). Although LAST is essentially a questionnaire to detect alcohol misuse and dependence, it differs from AUDIT in that it asks about lifetime problems. The cut-off score of ≥2 has been used in this study.

AUDIT (Alcohol Use Disorders Identification Test) was developed by a World Health Organization working group (Babor et al., 1987) to be a brief, culturally generalizable screening tool for early identification of problem drinking, rather than for identification of alcohol dependence. The original AUDIT instrument consisted of a questionnaire combined with a clinical examination. A subsequent development process produced a 10-item questionnaire. This has been designed specifically to facilitate international comparisons (Saunders et al., 1993a). The initial validation was undertaken in Australia, Bulgaria, Kenya, Mexico, Norway and the USA (Saunders et al., 1993b). The threshold for AUDIT above which problem drinking is defined has been the subject of discussion, as there is, inevitably, a trade-off between sensitivity and specificity. However, a score of ≥8 is now considered to be optimal (Cherpitel, 1995; Conigrave et al., 1995) and this was adopted in the present study.

LAST and AUDIT were chosen for their superior performance and international transferability. Both instruments were used because, when seeking information on the relationship between disease and alcohol consumption, it is relevant to consider both current and lifetime problem drinking. LAST asks questions about lifetime drinking, whereas AUDIT asks only about the previous year. The instruments were translated into Hungarian, with translation being checked through blind back translation.

Although it was not possible to undertake a comprehensive validation of the instruments in the Hungarian population, the performance of the two instruments was compared, reliability was examined using Cronbach's alpha test, and the results were compared with both physician and self-reports of alcohol consumption and related problems. Analyses were undertaken using the Statistical Package for Social Sciences.

The sample consisted of 3140 patients: 33.7% were in the University Hospital in Debrecen, 19.7% each in Pécs and Szeged, and 26.8% in Budapest; 52.4% were female; 42% were under the care of specialists in internal medicine; 28% surgery; 11% traumatology; 10% neurology, 5% respiratory medicine; and 4% psychiatry.



The overall participation rate was 81.1%, with 6.3% refusing to participate, 6.2% unable to do so because of their clinical condition, and 3.5% having been discharged or transferred before they could be interviewed. A variety of reasons accounted for the remaining 2.9%. Those not participating did not differ from those participating in terms of gender but were slightly older (57.3 vs 55.0 years, P = 0.006). Non-participation was approximately twice as frequent in Debrecen and in Budapest as in the other two hospitals.

Performance of survey instruments

Within each instrument, responses were consistent, with Cronbach's alpha scores of 0.76 for LAST and 0.83 for AUDIT (values in excess of 0.70 are generally taken as indicating reliability (Nunally, 1978). Item-to-total correlations were also calculated. Within LAST, the highest correlation was for ever feeling that one should cut down drinking (r = 0.77) and the lowest was for not always being able to stop when wanting to (r = 0.49). For AUDIT, the highest correlation was for frequency of heavy drinking (r = 0.79) and the lowest for someone being injured by one's drinking (r = 0.49).

Quantity and frequency of drinking

The self-reported quantity and frequency of drinking is shown in Table 1, which also shows the percentage of individuals in each quantity/frequency category who were identified by AUDIT as problem drinkers. As shown, 23.5% of men and 53.5% of women reported never drinking alcohol. Of those who did drink, about one in 10 men and less than 1% of women reported drinking 5 or more drinks on a day when they drank.

View this table:
Table 1.

Patients' self-reported alcohol consumption

No. of drinks (%)
Percentage in each quantity/frequency category (values in parentheses are percentages of cases in each cell identified as cases using AUDIT).
Males (n = 1488)
Never23.5 (0)23.5
Less than monthly19.4 (0)3.9 (2)1.2 (21)0.3 (33)0.4 (80)25.2
Monthly8.4 (4)3.2 (16)1.2 (57)0.6 (100)0.3 (100)13.6
Weekly8.5 (8)3.7 (45)1.9 (74)0.4 (100)0.2 (100)14.7
Daily/almost daily10.1 (18)6.8 (56)4.0 (72)1.0 (92)1.1 (100)23.0
Females (n = 1644)
Never53.5 (0)53.5
Less than monthly32.0 (1)1.4 (11)0.2 (50)33.6
Monthly5.5 (1)0.7 (22)6.2
Weekly3.3 (5)0.3 (25)0.2 (100)3.8
Daily/almost daily2.0 (12)0.6 (57)0.3 (100)0.1 (100)2.9

Prevalence and socio-demographic characteristics of those defined as cases

In all, 13.8% of the sample were defined as cases by one or both instruments, including 25.3% of men and 3% of women. However, only 12.4% of men and 1.5% of women were defined as cases by both instruments. The prevalence of cases defined by LAST and AUDIT, in each age group and in different socio-demographic groups, is shown in Table 2.

View this table:
Table 2.

Prevalence of problem drinking in different socio-demographic groups

% (95% CI)n% (95% CI)n
CI = confidence interval.
(a) Current problem drinking (defined by AUDIT)
Age group (years)
0–3421.4 (14.7–28.1)1451.3 (0–3.0)158
35–4431.5 (24.5–38.6)1684.6 (1.5–7.7)174
45–5423.0 (17.7–28.3)2394.1 (1.7–6.4)271
55–6415.3 (10.8–19.8)2480.9 (0–2.1)226
65–7413.1 (9.0–17.3)2510.4 (0–1.2)254
≥758.6 (3.5–13.7)1160.0172
Marital status
Married18.0 (15.3–20.6)8021.9 (0.9–2.9)682
Single19.7 (13.0–26.4)1372.7 (0–5.8)110
Divorced28.1 (18.8–37.4)892.5 (0–5.3)119
Widowed11.6 (5.1–18.0)951.3 (0.0–2.5)311
Co-habiting28.6 (14.9–42.2)423.2 (0–9.4)31
Employed21.5 (17.3–25.8)3621.6 (0.2–3.0)314
Unemployed17.9 (15.3–20.6)7922.0 (1.1–3.0)930
Highest educational level
Primary20.3 (16.0–24.6)3302.2 (0.9–3.4)508
Secondary16.1 (10.9–21.2)1931.9 (0.3–3.6)262
Specialist secondary21.1 (17.2–25.0)4211.8 (0.2–3.4)271
Higher15.6 (10.5–20.8)1921.7 (0–3.7)174
(b) Lifetime problem drinking (defined by LAST)
Age group (years)
0–3421.1 (14.5–27.7)1471.3 (0–3.0)160
35–4435.5 (28.3–42.6)1727.9 (3.9–11.8)178
45–5425.8 (20.3–31.3)2443.2 (1.2–5.3)277
55–6415.2 (10.8–19.6)2561.7 (0–3.4)233
65–7410.5 (6.8–14.3)2571.1 (0–2.4)266
≥756.8 (2.2–11.3)1180.6 (0–1.7)177
Marital status
Married17.3 (14.7–19.9)8211.9 (0.9–2.9)699
Single18.1 (11.7–24.5)1383.6 (0.1–7.1)111
Divorced36.3 (26.4–46.1)914.9 (1.1–8.7)123
Widowed14.4 (7.4–21.4)972.5 (0.8–4.2)324
Co-habiting33.3 (19.6–47.1)456.3 (0–14.6)32
Employed21.6 (17.4–25.8)3662.2 (0.6–3.8)319
Unemployed18.2 (15.5–20.8)8152.7 (1.7–3.7)961
Highest educational level
Primary17.6 (13.6–21.7)3401.9 (0.7–3.0)530
Secondary18.8 (13.3–24.2)1973.7 (1.4–5.9)272
Specialist secondary22.5 (18.5–26.4)4322.6 (0.7–4.5)272
Higher14.4 (9.5–19.4)1943.4 (0.7–6.1)177

If both current and lifetime problem drinking are taken together, then prevalences in each of the study hospitals were not significantly different for women but, among men, were highest in Debrecen (30%), followed by Budapest (26%), Pécs (23%) and then Szeged (21%) (χ2 = 7.9, P = 0.048).

Regardless of the instrument used, problem drinking was most common in the age band 35–44 years, in which about a third of men were so categorized. Problem drinking was especially common among divorced men and divorced and cohabiting females. There was no obvious association with either employment or education.

A total of 22.5% of men and 5.5% of women had consumed alcohol in the 48 h prior to admission; 8.3% of men and 3% of women reported that they considered alcohol to have contributed to their admission to hospital. Of these, among men, 80.6% thought that this had been due to their own consumption of alcohol, 6.5% to that of others, and 9.7% to both. The corresponding figures for women were 53.8%, 38.5% and 7.7%.

There was, as expected, a large difference in the percentage of patients defined as current or lifetime problem drinkers (subjects meeting criteria on either AUDIT or LAST or both) among those with different diagnoses. Because there were very few women categorized as problem drinkers, data are presented for men only (Table 3). The diagnostic categories used are, inevitably, somewhat arbitrary but have been designed to be clinically meaningful in the context of a study of alcohol use. The ‘other’ diagnoses was largely made up of patients undergoing elective surgery or with neurological or dermatological disorders. The categories used were derived from the case notes which were coded using ICD-10. Two-thirds of men with chronic liver disease were, or had been, problem drinkers. The prevalence of problem drinking was also high among those with tuberculosis and cardiomyopathy.

View this table:
Table 3.

Percentage of patients with various diagnoses identified as cases using LAST, AUDIT or both (males only)

Diagnosis% (95% CI)n
CI = confidence interval.
Alcoholic-related psychosis10023
Cancer of pancreas14 (5–23)23
Cardiomyopathy39 (21–57)12
Chronic liver disease69 (51–87)64
Diabetes29 (13–45)86
Hypertension20 (9–32)124
Ischaemic heart disease22 (13–32)197
Injuries28 (21–36)267
Lung cancer9 (0–20)56
Other cancer11 (0–23)228
Other gastrointestinal disease26 (19–32)448
Respiratory disease19 (9–29)144
Stroke18 (5–30)108
Tuberculosis56 (23–88)18
Upper aerodigestive cancer21 (5–37)49
Other diagnoses23 (19–27)1293


This study is subject to the usual limitations of surveys of alcohol consumption. First, it is necessary to recognize that instruments developed for one population do not necessarily perform in the same way in another (Cherpitel, 1998b). This problem cannot easily be overcome. Second, there is scope for participation bias. However, unlike some other surveys, this one did achieve a high level of participation, thus reducing the risk of participation bias. In addition, there is no reason to believe that admissions at the time of the survey were atypical. One concern is that university hospitals might be expected to admit a different case-mix than other hospitals. We believe that this is unlikely to be of importance here because, in Hungary, university hospitals draw their patients primarily from the county in which they are situated, with the exception of certain specialist units that serve a larger population. These specialist units were excluded from the study.

It is possible that the figures obtained may under-estimate the scale of the drinking in Hungary. First, although the medical students underwent training and the importance of consistency was stressed, they may have under-estimated the extent of problem drinking as their results were not validated against a gold standard. Second, especially among the elderly, there may, as in some other countries, be a tendency by some people claiming to be abstainers to discount consumption that is viewed as medicinal.

There are no directly comparable data on the quantity/frequency of drinking in the general population in Hungary. However, a Health Behaviour Survey undertaken in 1994, which provided a nationally representative sample of the population aged 15–64 years, reported frequencies of drinking that are similar to those reported in this hospitalized population (Central Statistical Office, 1996): 22% of men (23.5% in this survey) and 47.1% of women (53.5% in this survey) were abstainers; 24.9% of men (23%) and 3.2% of women (2.9%) drank daily. Unfortunately, only overall figures from this survey have been published, and so it is not possible to adjust for age. The hospitalized population is, however, somewhat older than the general population. In most societies, older age groups consume less alcohol than younger people. Thus, had it been possible to adjust for the different age distributions, the comparable population rates could be expected to be lower. The frequency of drinking thus seems to be higher among the hospitalized than the general population, but it is not possible to say by how much.

This study indicates the importance of using more than one instrument. The two used are intended to identify different things and, importantly, examine different periods of time. AUDIT has been found to be especially good at predicting alcohol-related social problems (Conigrave et al., 1995). LAST is intended to detect alcohol dependence or abuse at some time in the past, and thus to indicate the propensity to alcohol-related health problems that might be missed by AUDIT if problem drinking was in the past. The limited evaluation of the instruments is encouraging and concerning. The results obtained indicate that both instruments display a high level of internal consistency in the Hungarian setting. In addition, they have some apparent validity in that they correlate well with measures of quantity/frequency of drinking.

Comparison with similar studies is fraught with problems. First, the threshold for admission and the pattern of disease treated in hospital is likely to vary considerably within and between countries. Second, many studies have used instruments other than LAST or AUDIT. Third, many report on only a selected group of patients, such as those attending an emergency room (Cherpitel, 1998a).

One of the few almost comparable studies is from Belfast, which used the AUDIT Instrument. This found that 30.8% of male and 7% of female in-patients had AUDIT scores of ≥8 (Sharkey et al., 1996). The comparable figures in the present study were 18.9% and 1.9% respectively. These seem improbably low in view of the very high level of alcohol-related mortality in Hungary, although the patients served by the Belfast hospital were drawn from a very deprived inner city area with many social problems, and so may be less representative of the overall hospitalized population in Northern Ireland than is the present sample of the Hungarian population.

This study does establish that a substantial proportion of patients in Hungarian hospitals have alcohol-related problems, with, in some age groups, almost a third of hospitalized men so defined. Importantly, it provides a baseline for future measurement and contributes further support to those advocating a concerted policy on alcohol in Hungary (Varvasovszky and McKee, 1998).


This project was funded by the World Bank and the Hungarian Ministry of Welfare as part of the Close the Gap programme. However, neither the World Bank or Hungarian Government can accept responsibility for the views expressed. We are grateful to the deputy editor and three anonymous reviewers for valuable comments on earlier drafts.


  • * Author to whom correspondence should be addressed.


View Abstract