Article Text
Abstract
Background The burden of harm associated with alcohol misuse is disproportionately high in rural areas of Australia, and a considerable proportion of this burden is borne by the health system. The health impact of alcohol in rural areas has been measured in terms of the contribution of alcohol to hospital inpatient admissions and mortality rates, despite many more alcohol-related cases presenting to emergency departments (EDs). This study aims to estimate the proportion of presentations to EDs in rural Australia that are alcohol-related and to identify the associated patient and presentation characteristics.
Methods Patients aged ≥14 years presenting to four EDs in rural NSW were assessed on two measures: (1) Clinician judgement of alcohol consumption, and (2) patient self-report of alcohol consumption in the 6 h preceding the onset of their condition.
Results Preliminary analyses revealed sample selection biases in two of the EDs, and these samples were consequently excluded from further analyses. In the two remaining EDs, 46% of presentations were assessed, of which 9% were identified as alcohol-related. Presentations for mental disorders, those with more urgent triage categories and those occurring on weekends or at night were more often alcohol-related.
Conclusions The prevalence of alcohol-related ED presentations observed was at the lower end of the documented range, probably due to methodological differences and limitations, as well as geographic variation. Despite this, alcohol-related presentations were associated with a substantial impact on the ED. Policies and programs to reduce the impact of alcohol on rural emergency departments are needed.
- Alcohol
- Emergency Department
- rural
- Australia
- emergency care systems
- emergency departments
- remote and rural medicine
- mental health
- alcohol abuse
Statistics from Altmetric.com
- Alcohol
- Emergency Department
- rural
- Australia
- emergency care systems
- emergency departments
- remote and rural medicine
- mental health
- alcohol abuse
There is a substantial burden of harm associated with alcohol misuse, with alcohol accounting for 2.3% of the total burden of disease in Australia in 2003, equivalent to 1084 deaths and 61 091 disability adjusted life years.1 In Australia, this burden of harm is disproportionately high in rural areas: between 1992 and 2001 rates of alcohol-caused deaths were higher than in metropolitan areas (2.24 vs 1.67 per 10 000), as were alcohol-caused hospitalisations (48.4 vs 37.0 per 10 000).2 This finding most likely reflects that rural residents are more likely to engage in drinking that places them at risk of acute and chronic harm and that the resources with which to address these harms are limited in rural areas: compared with metropolitan areas, the availability of hospital beds is reduced, as is accessibility to general practitioners, specialists, registered nurses and allied health professionals.3
A considerable proportion of this burden is borne by the health system: the chronic effects of alcohol, such as alcoholic liver cirrhosis, alcohol dependence, cardiovascular disease and cancer, accounted for 10 081 deaths and 146 916 hospital admissions in Australia between 1992 and 2001; while the acute effects of alcohol, such as road crash injuries, attempted suicide and assaults, accounted for 16 756 deaths and 39 283 hospitalisations in the same period.2 The majority of the data that are available on alcohol-related health issues in Australia describe the contribution of alcohol to hospital inpatient admissions and mortality rates, despite many more alcohol-related cases presenting to the emergency department (ED).4–6
Despite the disproportionately high burden of harm associated with alcohol in rural areas and on EDs, the impact of alcohol on rural Australian EDs has been measured in only one pilot study, which found 5% of presentations to the ED in Broken Hill, a regional centre in far west New South Wales (NSW), were alcohol-related.7 Although an impressive example of local collaboration between services, the identification of alcohol-related cases was based, in part, on retrospective review of medical record content, which has been shown to detect only 60% of presentations identified as alcohol-related by a range of other methods.8
The aims of this study are to estimate the proportion of alcohol-related presentations to four rural EDs in Australia and to identify associated patient and presentation characteristics.
Method
Sample
The current study was conducted in the EDs of four rural communities in NSW, Australia, involved in the experimental arm of the Alcohol Action in Rural Communities project. Twenty communities were selected on the basis that they had a population between 5000 and 20 000 (according to 2001 Census),9 were at least 100 km away from a major urban centre and were not known to be currently involved in any other large-scale project aimed to reduce alcohol-related harm. Communities were placed in matched pairs, and within each pair, communities were randomly allocated to the experimental or control condition. Of the 10 experimental communities, five had EDs with electronic medical records, and the four of these with the largest throughput were invited to participate in the current study. Approval to conduct the study was granted by the University of New South Wales ethics committee and the ethics committees that oversee research at the participating hospitals.
As can be seen in table 1, the population served by each of the participating hospitals ranged from 13 590 to 29 140, according to the 2006 Census postal area of usual residence10 and the number of beds in each hospital ranged from 41 to 105.11
The dates of data collection in each of the hospitals are listed in table 1. In order to avoid potential seasonal effects, efforts were made to spread the data collection across the year. All individuals aged 14 years and older who were treated at the participating EDs during the data collection periods were eligible for assessment.
Measures
Blood alcohol concentration at the time of arrival at the ED, and self-report of alcohol consumption before the onset of one's illness, are the most commonly used methods of identifying alcohol-related ED presentations.12 Given that self-report can be subject to recall bias13 and response bias14 and that a measure based on blood alcohol concentration can overlook some alcohol-related cases because of the inherent delay between the onset of an illness and arrival at the ED,13 15 it is recommended that these two methods be used in combination.16 In the current study, clinician judgement of alcohol consumption was selected as an alternative to more objective measures of blood alcohol concentration, such as blood-, breath- or saliva tests, to maximise the likelihood that the measure would be incorporated into the routine activities of triage nurses. Specifically, alcohol-related ED presentations were identified using the following two methods:
Clinician judgement of intoxication involved the triage nurse noting if there were any clinical signs of alcohol intoxication during routine patient triage. This included slurred speech, poor balance and smelling alcohol on the patient's breath. Only yes or no responses were allowed.
Patient self-report required patients to respond with a yes or no when asked by the triage nurse “did you consume any alcohol in the 6 h preceding your injury or the onset of your condition?”
Nurses were free to administer these measures at any time during the triage of a patient. They were instructed, however, to administer the clinician judgement measure before the patient self-report to avoid the former being influenced by the latter.
The method of prompting triage nurses to administer these measures and record the results varied across hospitals due to differences in the recording of patient information. In ED1 and ED3, where intake information is collected and entered directly into an electronic database, signs were displayed on the computers reminding nurses to administer the measures and enter the responses into the relevant field of the database using a free text format. In ED2 and ED4, where patient information is recorded on paper forms, stickers with the two measures and the possible responses were attached to intake forms before the data collection period. Nurses circled the relevant response, and hospital administration staff transcribed this information into the electronic database along with the patient's other intake information. These procedures were explained to nurses by one of the authors (AH) at the beginning of the data collection period.
For all ED presentations occurring during the data collection periods, de-identified results on clinician judgement and patient self-report were extracted from the database of medical records, along with information on patient and presentation characteristics. The following variables were selected for extraction because they have been demonstrated to be associated with alcohol in previous ED studies: patient sex,17–21 age,17 18 21day of presentation,17 20 21 time of presentation,17 19–23 presenting condition,17 18 20–22 triage category24 and departure status.17 22 25 Presenting conditions were grouped according to the International Classification of Diseases 9th revision (ICD-9) code allocated as the primary diagnosis.26 Conditions defined as wholly attributable to alcohol27were categorised as alcohol specific, presentations with ICD-9 codes between 290 and 319.99, other than those included in the alcohol specific category, were categorised as mental disorders, codes of 800 to 959.99 were categorised as injuries and all remaining conditions were categorised as other illnesses.
Statistical analysis
Before the analysis of main outcome variables, an examination of potential biases in sample selection was conducted. Presentations that were assessed on the clinician judgement measure and/or the patient self-report measure were compared to those that were not assessed with respect to patient and presentation characteristics. Comparisons were made using the Pearson Chi-squared test, and for comparisons on variables with more than two categories, significant Chi-squared tests were followed by examination of standardised residuals to determine which categories accounted for the significant difference. For EDs in which assessed and unassessed presentations differed significantly on a number of key variables, the sample was deemed biased and excluded from further analyses.
The proportion of presentations that were alcohol-related was calculated according to clinician judgement alone, patient self-report alone and according to a combined measure, in which presentations with positive on either clinician judgement or self-report were considered alcohol-related.
In order to identify which patient and presentation characteristics were independently associated with alcohol-related ED presentations, a multivariate logistic regression was conducted. The alcohol-related status of presentations, according to the combined measure, was entered as the dependent variable. The categorical variables of age, sex, presenting condition, day of presentation, time of presentation, triage category and departure status were entered as independent variables.
All analyses were performed using SPSS for Windows V.15.0,28 and significance levels were set at 0.05.
Results
Response rates
There were a total of 7406 presentations by patients aged 14 years and older to the four EDs during the data collection periods. Administration of the measures varied substantially between hospitals, with information from at least one measure available for 8–87% of eligible presentations (see table 2). Given the need to minimise time demands on triage staff and maximise the likelihood that the measures would be incorporated into the routine of triage nurses, reasons for missing information were not collected.
Assessment of sample selection bias
Statistical comparisons between presentations that were assessed on at least one of the measures and those not assessed at all are reported in table 3. In ED1 and ED2, statistically significant differences between assessed and unassessed presentations on a number of key variables were identified. In ED1, relative to those not assessed, the proportion of assessed patients that were men was significantly lower (χ2 (1)=5.38, p<0.05), as was the proportion of presentations for mental disorders and alcohol specific conditions (χ2 (3)=76.39, p<0.05), the proportion of weekend presentations (χ2 (1)=64.99, p<0.05) and the proportion of visits occurring in the early hours of the morning (χ2 (3)=43.23, p < 0.05). In ED 2, there was a statistically significantly higher proportion of presentations for injuries and alcohol specific conditions (χ2 (3)=18.94, p<0.05), presentations occurring during the early hours of the morning (χ2 (3)=9.34, p<0.05), and presentations by 14- to 17-year-olds (χ2 (2)=9.91, p<0.05) among assessed presentations relative to those not assessed. Consequently, data from ED1 and ED2 were excluded from further analyses.
The samples obtained in ED3 and ED4 were for the most part, representative of other presentations occurring during the data collection periods. The only exception was that in ED4, the proportion of patients who departed before treatment occurred was lower among the assessed relative to those not assessed (χ2 (3)=9.61, p<0.05).
Sample characteristics
The demographic characteristics of patients and factors related to their ED presentations can be seen in table 3. In ED3 and ED4, the majority of assessed patients were aged at least 30 (median age 42 years), and 51% were men. One per cent of presentations were for alcohol-specific conditions, 2% were for mental disorders and 19% of presentations were for an injury. Two thirds of presentations occurred over the weekend. Over one third of presentations occurred in the morning, another third in the afternoon, one fifth in the evening and 6% in the early hours of the morning. Almost half of the presentations were allocated a triage category of potentially life threatening, imminently life threatening or immediately life threatening (triage categories 1 to 3).29 Seventeen per cent of patients were admitted to hospital, and 5% left before treatment occurred.
Prevalence of alcohol-related ED presentations
Table 4 shows that 9% of ED presentations were positive on at least one of the clinician judgement or self-report measures. According to clinician judgement alone, 6% of presentations were alcohol-related, and according to patient self-report alone, 8% were alcohol-related.
Characteristics of alcohol-related ED presentations
The proportion of presentations that were identified as alcohol-related within each category of patient and presentation characteristics are reported in table 5, as are the results of the logistic regression model comparing the characteristics of alcohol-related and non-alcohol-related ED presentations. The odds of a presentation being related to alcohol were significantly higher among presentations for alcohol-specific conditions, of which 83% were alcohol-related (OR=65.91, 95% CI 12.89 to 337.10), injuries (14%; OR=2.46, 95% CI 1.61 to 3.76) and mental disorders (18%; OR=2.99, 95% CI 1.14 to 7.83), relative to presentations for other conditions (7%). Other significant predictors of alcohol relatedness were weekend presentations (OR=1.58, 95% CI 1.07 to 2.32), evening (OR=2.33, 95% CI 1.44 to 3.76) and early morning presentations (OR=2.91, 95% CI 1.52 to 5.57), presentations with triage categories ranging from potentially life threatening to immediately life threatening (OR=1.70, 95% CI 1.13 to 2.56) and presentations in which the patient left before receiving treatment (OR=7.63, 95% CI 3.12 to 18.64).
Discussion
The current study estimated that 9% of presentations to two EDs in rural NSW were related to alcohol. Consistency with previous findings is difficult to judge, as the only other study of presentations to an ED in rural Australia employed a different method to identify alcohol-related ED presentations. The 5% of presentations identified as alcohol-related in this study7 is likely to be an underestimate, as cases were identified through retrospective review of medical records, which detects only 60% of presentations identified as alcohol-related by other methods.8
With the exception of research using convenience samples of ED patients, only two previous studies have identified alcohol-related ED presentations using a combination of self-report and clinician judgement. These studies produced widely varying prevalence estimates: 11% of presentations to three EDs in The Netherlands20 and 25% of presentations to an ED in the United Kingdom22 were found to be alcohol-related. Although some of the variation in the prevalence of alcohol-related ED presentations could be accounted for by methodological differences, such as dissimilar eligibility criteria or the collection of data during different seasons of the year, it is likely that much of it is due to the different locations of the emergency departments under study. Studies using identical methods in multiple sites have demonstrated that the association between alcohol consumption and ED presentations varies between countries30 and between regions within countries.20 31–33
This geographical variation is generally postulated to arise through differences in the demographic composition of the population served by the EDs under study and the alcohol consumption patterns of these communities.15 Differences in patterns of health service utilisation might also contribute to this variation. In Australia, treatment at the emergency department of a public hospital is free, while co-payment is often required for general practitioner consultations. This might result in greater use of the ED for ailments that would usually be treated by a general practitioner and proportionately less for emergencies.
Despite finding that the prevalence of alcohol-related ED presentations in rural NSW is at the lower end of the documented range, these presentations impose a substantial burden on the health system. Presentations allocated triage categories ranging from potentially life threatening to immediately life threatening were more often alcohol-related, as were presentations for mental disorders, which are more resource intensive.34 35 Alcohol-related presentations were also more likely to occur on weekends and at night, when staffing levels are generally lower. In addition to the implications these presentations have for the planning of acute care services and their cost to the health system, which has been shown to be disproportionately high,36 alcohol-related presentations are associated with poorer clinical outcomes37 38 and greater risk of future harm.39 40
The finding that almost 1 in 10 presentations to rural EDs are related to alcohol and that these presentations are associated with a substantial impact on the ED raises the need for policies and programs to reduce this impact. One possibility is the widespread implementation of ED-based interventions for heavy-drinking patients, which have been successful in reducing self-reported alcohol-related injuries by almost 50%.41 Future research should first focus on establishing the impact of these programs on alcohol-related ED presentations, as research thus far has produced varying results.42–44
In designing programs aimed at reducing alcohol-related ED presentations, it should be borne in mind that the presentations identified as alcohol-related in the current study were not necessarily caused by the patient's alcohol use. It is, however, reasonable to assume some degree of causality. Studies using case-control designs, in which the recent alcohol consumption of ED patients is compared to the recent consumption of general community members, suggest there is a causal relationship,45–47 as do studies in which ED patients are asked whether they would attribute their condition to their consumption of alcohol.19 48 49 Conversely, the current study did not identify ED presentations for illnesses that are a consequence of chronic risky drinking, previously estimated to be 20% of alcohol-related ED presentations.50 This estimate arises from research conducted internationally, but given the high rates of risky alcohol consumption in rural Australia,3 it will be important to investigate the additional impact on rural EDs from this type of drinking.
Limitations
Although the use of clinician judgement in place of an objective measure of blood alcohol concentration might be considered a threat to the validity of the results, previous research suggests it is unlikely to have a dramatic impact. In a study examining the concordance between clinician judgement and breath test, only 1% of patients with a blood alcohol concentration greater than 100 mg/100 mL (0.10) were identified as not intoxicated, and 0.5% of patients with negative blood alcohol concentrations were identified as intoxicated.45
The low response rates of 53% and 33% achieved in the two EDs included in the final analyses can reasonably be considered a limitation to the current study. Although the samples obtained were otherwise representative of patients who presented in the same period, there was an underrepresentation of patients who left before receiving treatment in the sample from ED4. Given that patients who departed before the treatment were more likely to have alcohol-related presentations, an underestimate in the rate of alcohol-related presentations generated by this study is possible. Biases in terms of other unmeasured variables are also possible. The assessment of representativeness was limited to the variables available in the routinely collected data, with information on other characteristics known to be associated with alcohol use, such as Aboriginal and Torres Strait Islander status, not available.
The lack of comprehensive administration of the measures in the current study also has implications for the generalisability of the results. Although the intention was to present data from four rural EDs, the samples obtained in two of the EDs were not representative of other ED presentations occurring in the same period. Presentations with characteristics that have previously been demonstrated to be associated with alcohol were underrepresented in ED1 and overrepresented in ED2. Consequently, data from these two EDs were not analysed, reducing the generalisability of the findings to other rural EDs.
Similarly, the intention was to spread data collection in the four EDs across the year to avoid potential seasonal effects, but exclusion of data from two of the EDs meant that only presentations occurring during autumn and winter were included in the analyses. This is likely to have resulted in an underestimate of the rate of alcohol-related ED presentations, which has previously been observed to double in warmer weather.7
Given their low and inconsistent administration, and the resulting biases in the data from two of the EDs, future use of triage nurse-administered clinician judgement and patient self-report is not recommended. Although the higher response rates in ED1 and ED3 might suggest that incorporating data collection into existing electronic systems improves the rate of administration, this does not necessarily enhance the representativeness of the assessed sample. The highest response rate was achieved in ED1; yet, biases in sample selection were evident. Basing research staff in the ED could improve assessment rates, but for the current study, this was considered uneconomical due to the low presentation rates in rural EDs. Measures based on routinely collected ED data, such as alcohol-specific ICD codes, nursing triage text or surrogate measures are more feasible. Although not appropriate for providing an accurate absolute measure of alcohol-related ED presentations, they can be used to detect changes in alcohol-related ED presentations over time.25 51
Acknowledgments
The authors would like to thank the managerial, clinical and administrative staff at the participating hospitals, and the staff at associated data centres, for their valuable contribution to the collection and extraction of the data.
References
Footnotes
Funding Alcohol Education Rehabilitation Foundation.
Competing interests None.
Ethics approval This study was conducted with the approval of the University of New South Wales Human Research Ethics Committee, Greater Southern Area Health Service Human Research Ethics Committee, Hunter New England Human Research Ethics Committee and North Coast Area Health Service Human Research Ethics Committee.
Provenance and peer review Not commissioned; externally peer reviewed.