Alcohol use, alcohol disorders, and the use of health services : Results from a population survey

Data from a population survey were used to explore relationships among drinking levels/patterns, alcohol dependence or abuse, and the use of emergency services, hospital admissions, and frequent visits to general practitioners in the past year. For both males and females, self-reported hospital admissions were less common among daily moderate drinkers than among lifetime abstainers. Among males, drinkers with no history of alcohol dependence or abuse were less likely to report being in hospital in the last year than lifetime abstainers. For females, some groups defined by drinking patterns/levels and current drinkers without symptoms of alcohol abuse or dependence were more likely to report using emergency services than lifetime abstainers.

Keywords: Alcohol and gender; Alcohol and health; Drinking survey; Health service utilization

Many studies have shown that heavy drinking is associated with a variety of health problems, and there is good evidence that heavy drinkers make more use of health services than others. Zook et al. [1], Seale et al. [2], Callahan and Tierney [3], and Kristenson et al. [4] found that heavy drinking was associated with increased risk of hospitalization, while Browne et al. [5], Westhead [6], and Rush [7] showed that heavy drinkers make more visits to general practitioners than others. More recently Virgo et al. [8] found that Vietnam veterans who had alcohol or drug problems made more use of Veterans Affairs health services than others, while Cryer et al. [9] found that heavy alcohol users in the general population were disproportionate users of acute medical services. The influence of heavy drinking on emergency room use has also been clearly demonstrated [9], [10], [11].

There is, however, increasing evidence that certain types of drinkers make less use of health service than abstainers, and this is consistent with evidence showing that the occasional use of alcohol can protect the drinker against coronary heart disease and can have other health benefits [12], [13].

Armstrong et al. [14] examined the relation between alcohol use and utilization of health services during a 9-year period in a sample of 4264 adults in a survey of members of a health maintenance organization (HMO). Respondents were categorized as abstainers (no drinks in the past year), lighter drinkers (less than 7 drinks/week), moderate drinkers (7 to 13 drinks/week), and heavier drinkers (greater than or equal to 14 drinks/week). Controlling for age, race, and health plan membership, the mean number of outpatient visits was inversely related to the amount of alcohol consumed. Also, compared with the three drinker groups, abstainers had more inpatient admissions.

Similar results were reported by Rice et al. [15] using data from a telephone survey of members of a large HMO. These data showed that current nondrinkers with no past history of drinking had higher rates of outpatient visits and hospitalizations than current drinkers. Among current drinkers, medical care use declined slightly as drinking levels increased. Among nondrinkers, those with a drinking history exhibited significantly higher use of outpatient visits and hospital care than nondrinkers with no drinking history and current drinkers. The results also suggested that former drinkers used more services than others because they are sicker than other lifetime abstainers, nondrinkers, or current drinkers.

Reduced rates of hospital admissions among moderate drinkers relative to abstainers have also been reported by Longnecker and MacMahon [16] and Haapanen-Niemi et al. [17] using data from population surveys. Kunz [18], who also used data from a population survey, found that drinkers reported fewer visits to health service providers than abstainers, and that the number of visits to health professionals decreased with increased daily alcohol consumption. Rice and Duncan [19] found that alcohol consumption was negatively associated with physician visits using data from the 1990 U.S. National Health Interview Survey. Cryer et al. [9] found, in a U.K. population health and lifestyle survey, that both heavy drinkers and abstainers were disproportionate users of acute medical services, but relative underusers of preventive medical care services.

Current research on alcohol problems and service utilization is limited in several respects. Some studies of the influence of drinking on health services use have not controlled for potentially confounding variables such as age, gender, marital status, education, income, ethnicity, drug use, and smoking. Others have not distinguished between lifetime abstainers and former drinkers. This distinction is important because some people stop drinking for health reasons, and this may increase the risk of health service use. Finally, there is a need for more research on the influence of drinking on the use of different types of health services. Callahan and Tierney [3] found that, in an elderly population (age 60+), those with evidence of alcoholism were more likely than others to be hospitalized over a 2-year period, but not more likely to use ambulatory or emergency room services. However, Reid et al. [20] found no consistent relationship between veterans' alcohol use and the use of health services. Cherpitel et al. [21] found that, in a general population, heavier drinking increased the use of emergency services, but not the use of primary care services. Cherpitel [21] also showed that the associations among drinking, drug use, and the use of emergency room and primary care services vary with gender and ethnicity. Further research is this area is important because services differ in cost, and the differential impact of drinking on the use of different services will need to be considered in calculations of the health care costs of alcohol use.

This article reports on analyses of data from an Ontario, Canada, population survey that included items on drinking, drinking problems, and health service use, as well as items for several other factors that have been shown to influence both drinking and health service use. The analysis was designed to explore the unique effects on service use of alcohol consumption and of alcohol abuse or dependence.

Data for this study were obtained from the 1990–1991 Ontario Mental Health Supplement (OHSSUP) [22], a stratified, multistage, area probability sample of the Ontario household population aged 15 years and over. Excluded were residents of First Nations Peoples' reserves, inmates, foreign service personnel, and residents of remote areas. The sample involved 9953 individuals randomly selected from households participating in the third and forth data collection quarters of the 1990 Ontario Health Survey (OHS), a general health survey of over 60,000 Ontario residents. Eligible respondents for the OHSSUP numbered 35,690, corresponding to a total of 14,478 eligible households. Participating households numbered 13,002, with one individual randomly selected from each. Of this number, 9953 agreed to participate, a response rate of 76.5%. Respondents and nonrespondents were very similar with respect to key measures of health status, employment, income, and marital status. However, males, those born outside Canada, and urban dwellers were slightly underrepresented among respondents [23].

The sample for the present study consisted of 8116 individuals aged 15 to 64 years. Older respondents were excluded because they were not asked all questions required to make a diagnosis of alcohol abuse or dependence or other mental health problems. To yield meaningful tests of significance, sample weights were applied to the data and then rescaled to equal the actual number of cases in the sample. These weights accounted for all relevant features of the sampling design, including unequal probabilities of selection, nonresponse (person and household level), and an adjustment to align the age and sex distribution of the sample with the age and sex distribution of the population.

Respondents who reported ever drinking at least 12 drinks in a given year were asked to indicate the largest number of drinks ever consumed in 1 day and then to indicate how often this occurred during the time they were drinking most. Response options for number of drinks were (a) 12 or more, (b) 8–11 (c) 5–7, and (d) 1–4. Response options for frequency were (a) once a year, (b) twice a year, (c) 3–6 times a year, (d) 7–11 times a year, (e) 1–3 times a month, (f) 1–2 times a week, (g) 3–4 times a week, and (h) nearly every day. Respondents were also asked if they had continued to drink the same amount in the last year and, if not, to indicate how often they had consumed 1–4 drinks and how often they had consumed 5 or more drinks in the past year.

Responses to these items were combined to create six groups with different levels and patterns of drinking:

• Heavy drinkers—those who indicated that they drank 5 or more drinks at least once a week in the past year. Most of these (79%) also indicated that they drank 4 drinks at least once a week.
• Daily moderate drinkers—those who indicated that they drank 1–4 drinks daily or almost daily in the past year, but who did not report drinking more than this more than 1–3 days a month.
• Other regular drinkers—those who drank more than once a month in the past year except for "heavy drinkers" and "daily moderate drinkers" as defined above.
• Infrequent drinkers— those who reported drinking less than once a month in the past year.
• Ex-drinkers—those who reported drinking at some time in their lives, but not in the last year.
• Lifetime abstainers—those who never drank or who had never had more than 12 in their life.

Information about alcohol disorders was obtained using a modified version of the World Health Organization's Composite International Diagnostic Interview [24] The modified version of the Composite International Diagnostic Interview is a structured diagnostic interview for field interviewers without any formal clinical training that is designed to generate DSM-III-R and ICD-10 diagnoses. Reliability and validity analysis of the Composite International Diagnostic Interview has indicated good test-retest and interrater reliability and acceptable levels of validity for most DSM-III-R lifetime diagnoses, including drug disorders [25], [26]. Diagnoses of alcohol dependence or abuse were combined to create four groups for the analysis:

• Lifetime abstainers
• Current or former drinkers with no history of alcohol dependence or abuse
• Current or former drinkers with dependence or abuse in the past, but not in the last 12 months
• Current or former drinkers with dependence or abuse in the last 12 months

Three variables concerning the use of health services were considered:

• Visiting an emergency department in the past year
• Being admitted to a hospital in the past year
• Making more than the median number of visits to a general practitioner (GP) in the past year

Analyses using total number of GP visits gave results that were generally consistent with those reported below.

The use of different health services for groups defined by drinking patterns/levels or by alcohol dependence/abuse were first examined using bivariate cross tabulations. Logistic regression analyses was then performed to determine the odds of being in the higher service use category for those in different groups defined by drinking levels/patterns and those defined by dependence or abuse. Other variables that have been shown to influence health service use in general populations or in populations with addiction or mental health problems [27], [28], [29], [30] were also included in these analyses. These variables were age, gender, mental health (no diagnosis last year,[1]* one diagnosis last year, two diagnoses last year, three or more diagnoses last year); drug dependence or abuse[2]† (never, last year, previous); marital status (married/common law, single, widowed, separated/ divorced); urban residence (yes/no); income (high, medium, low, not reported); education (primary, secondary completed, some secondary, some postsecondary, completed postsecondary); self-identified sociocultural group (French, Asian, southern European, Canadian, other). Given the complex nature of the OHSSUP design (i.e., multistage sampling), all standard errors for proportions and odds ratios were adjusted using the Taylor series linearization method offered in the SUDAAN software package [31].

Table 1 shows the number and percentage of cases in each of the drinking level/patterns and dependence groups and the relationships among these groups. As expected, those classified as heavy drinkers were the most likely to qualify for a diagnosis of alcohol dependence or abuse in the past 12 months (28% vs. <8% for any other group).

Table 1. Relationships Between Groups Defined by Drinking Levels/Patterns by Dependence/Abuse

969 Percentages may not sum to 100 due to rounding.

Simple cross tabulations (not shown) and chi-square tests showed that the relationships among the three service use variables were all positive and statistically significant for both males and females.

Tables 2 and 3 show the relationships for males and females, respectively, among drinking levels/patterns, alcohol dependence/abuse, and the reported use of specific health services. Because the number of former drinkers with a history of dependence was too small for analyses involving gender, all former drinkers were excluded from Tables 2 and 3 and from subsequent analyses of the influence of abuse or dependence on service use.

Table 2. Drinking Levels/Patterns, Alcohol, Dependence/Abuse, and the Use of Specific Health Services: Males

970 aExcludes former drinkers; see text.

Table 3. Drinking Levels/Patterns, Alcohol, Dependence/Abuse, and the Use of Specific Health Services: Females

971 aExcludes former drinkers; see text.

Some of the cells in Tables 2 and 3 have few cases, and this limits their interpretation. However, Table 2 shows that, for males, former drinkers and abstainers were more than twice as likely than other regular drinkers and infrequent drinkers to report being in the hospital in the last year. Table 3 shows that, among females, heavy drinkers were more than twice as likely than abstainers to report use of emergency services. Table 3 also shows female moderate daily drinkers as less likely to report frequent visits to GPs than those with other drinking levels/patterns, and other regular drinkers were less likely to report being in the hospital than those with other drinking levels/patterns.

Results of the logistic regression analyses are summarized in Tables 4 and 5. In all cases, the reference group comprised abstainers. Dependence/abuse (current, previous vs. never) was included in the analyses of the influence of drinking patterns. However, drinking patterns were excluded from the analyses of the influence of dependence (current, previous, never vs. abstainer) because the design matrix would otherwise have a redundant category (the abstainer contrast for groups defined by drinking pattern/levels). Also, as previously noted, former drinkers were excluded from these analyses. Additional analyses involving drinking pattern/levels, but without controlling for alcohol dependence or abuse, gave essentially the same results for the influence of drinking patterns/levels. These analyses, therefore, are not reported in detail.

Table 4. Odds Ratios and 95% Confidence Limits for the Use of Services by Different Groups of Drinkers Compared with Lifetime Abstainers

972 Controlling for age, gender, mental health status, alcohol dependence or abuse, drug dependence or abuse, marital status, urban residence, income, education, and self-identified sociocultural group. (See text for details.) ap <. 01. bp <. 001.

Table 5. Odds Ratios and 95% Confidence Limits for the Use of Services by Drinkers with or without a History of Alcohol Dependence or Abuse Compared with Lifetime Abstainers

973 Controlling for age, gender, mental health status, drug dependence or abuse last year, marital status, urban residence, income, education, and self-identified sociocultural group. (See text for details.) ap <. 01. bp <. 001.

The five odds ratios in each row of Table 4 were always tested simultaneously, and thus there was a fivefold increase in the probability that any one odds ratio would be significant by chance. Only those that reached an adjusted level of significance (0.05/5 = 0.01) were considered statistically significant in this instance [32]. Using this adjusted p level for statistical significance, Table 4 shows that hospitalizations were less likely to be reported by male and by female daily drinkers and by males classified as infrequent drinkers when compared with gender-specific lifetime abstainers. There were no statistically significant differences with respect to gender-specific drinking patterns and visits to GPs. Visits to emergency departments were more likely to be reported by females who were heavy or infrequent drinkers when compared with female lifetime abstainers. The rates of use of emergency departments among male drinkers did not differ from the comparable rate for male abstainers.

For the results in Table 5, the adjusted level for statistical significance was. 16 (.05/3) because each row shows three odds ratios for which significance was tested simultaneously. The table shows only two differences that were significant at this level: First, hospitalizations were less likely to be reported by male drinkers with no history of alcohol abuse and dependence than by male lifetime abstainers; second, use of emergency departments was more likely to be reported by female drinkers with no history of dependence or abuse than by females who were lifetime abstainers.

The results show that, for both males and females, rates of self-reported hospitalization in the past year were significantly lower for daily moderate drinkers than for lifetime abstainers. The results also show that no group of drinkers or former drinkers had a higher rate of self-reported hospitalization than lifetime abstainers. Among males, infrequent drinkers and those with no history of alcohol dependence or abuse were also significantly less likely to report being in the hospital than lifetime abstainers. Although not directly comparable, these results are consistent with other reports in the literature, and they support the view of moderate drinking as being associated with a reduced risk of health problems for both males and females. However, the data are not sufficient to indicate how service use and alcohol use were related. Alcohol usage may have been quite variable over the 1-year period considered and may have both influenced, and been influenced, by health problems and the use of health services. Further research is needed to show if this is the case.

The results concerning the relationships between drinking and frequent visits to GPs do not support the view that moderate drinking is associated with health benefits. No group of drinkers had a significantly lower rate of frequent visits to GPs than abstainers of the same gender. These differences in the results pertaining to alcohol use, hospital admissions, and frequent visits to GPs may reflect differences in the relationships between health needs and hospital admissions or visits to GPs. Hospital admissions are controlled by specialists and usually only occur in cases of serious illness. However, visits to GPs are largely self-initiated and influenced by subjective health status and a variety of non-heath-related factors [33]. Mackenbach et al. [34] found that moderate drinking was associated with a subjective sense of good health, and it is possible that this is also associated with an increased use of preventive health services, including annual check-ups by a GP. This would tend to reduce any differences in the rates of use of primary health services between moderate drinkers and abstainers who may actually forgo the health benefits of moderate drinking and thus have greater real health needs relative to moderate drinkers. It is of interest that Cryer et al. [9] found that "safe-limit" drinkers made less use than abstainers of preventive services that typically involve self-initiated referrals (dental appointments, eye tests, routine mammograms, and cervical smear tests). It would thus be important in further studies of this kind to distinguish between the use of services that typically involve self-referrals and those that require a prior medical assessment.

Reports of the use of emergency services were not significantly greater for any group of male drinkers than for male lifetime abstainers. However, among females with different drinking patterns/levels or histories of dependence/abuse, three groups were significantly more likely to report using emergency services than lifetime abstainers: heavy drinkers, infrequent drinkers, and drinkers with no history of dependence/abuse. This suggests that, among females, even moderate alcohol use is associated with increased risk of problems needing emergency treatment. One reason might be a greater use of tranquilizers and sleeping pills by females than by males [35]. In combination with alcohol, these drugs can have very serious effects that require immediate medical attention. However, it is also possible that, in the present survey, the wording of the question about the use of emergency services contributed to an overestimation of emergency service use by some women. The question was "During the past 12 months, did you use an emergency service room at a hospital?" Because the question did not specify the use of an emergency service use for personal health reasons, it is possible that some reported visits were for the benefits of children and not for the woman herself. This needs to be considered in future studies of this kind.

The present analysis has a number of limitations. Although large and carefully selected, the sample was slightly biased against males, those born outside Canada, and those living in urban areas. It is also likely that, as with other population surveys, heavy drinkers were underrepresented due to their involvement in treatment, incapacity, or lack of cooperation. The survey questions concerning alcohol consumption are also atypical of those used in other surveys, and there are no data concerning the reliability of responses. The lack of information on the use of tobacco is also a significant limitation.

The analyses were limited to self-reported data, which are vulnerable to a number of biases. These include deliberate or innocent distortions in the recall and reporting of events and behaviors, and such distortions may be particularly large for respondents with significant alcohol-related problems [36]. The survey methods included features designed to minimize these distortions, including the use of experienced and closely supervised interviewers and assurances of confidentiality. However, the extent of bias is unknown, and further research is needed to show if the present results are robust across different populations and using different research methods.