• Open Access

Identification of alcohol involvement in injury-related hospitalisations using routine data compared to medical record review

Authors


Correspondence to:
Kirsten McKenzie, National Centre for Health Information Research and Training, School of Public Health, Queensland University of Technology, Kelvin Grove, Queensland 4059, Australia. Fax: (07) 3138 5515; e-mail: k.mckenzie@qut.edu.au

Abstract

Objective: To quantify the extent that alcohol related injuries are adequately identified in hospitalisation data using ICD-10-AM codes indicative of alcohol involvement.

Method: A random sample of 4,373 injury-related hospital separations from 1 July 2002 to 30 June 2004 were obtained from a stratified random sample of 50 hospitals across four states in Australia. From this sample, cases were identified as involving alcohol if they contained an ICD-10-AM diagnosis or external cause code referring to alcohol, or if the text description extracted from the medical records mentioned alcohol involvement.

Results: Overall, identification of alcohol involvement using ICD codes detected 38% of the alcohol-related sample, while almost 94% of alcohol-related cases were identified through a search of the text extracted from the medical records. The resultant estimate of alcohol involvement in injury-related hospitalisations in this sample was 10%. Emergency department records were the most likely to identify whether the injury was alcohol-related with almost three-quarters of alcohol-related cases mentioning alcohol in the text abstracted from these records.

Conclusions and Implications: The current best estimates of the frequency of hospital admissions where alcohol is involved prior to the injury underestimate the burden by around 62%. This is a substantial underestimate that has major implications for public policy, and highlights the need for further work on improving the quality and completeness of routine administrative data sources for identification of alcohol-related injuries.

Alcohol-related injury is a global problem of substantial public health importance. Each year approximately 1.8 million people worldwide die from an alcohol-related condition, with around half of these from injury.1 Australian reports confirm the considerable burden of alcohol-related diseases and injuries in this country, with an estimated 500,000 hospitalisations attributable to alcohol between 1993 and 2001,2 and alcohol-related healthcare costs of around $2 billion annually.3 Alcohol has been declared an area for action in the current Australian National Injury Prevention and Safety Promotion Plan.4

Limited research has been conducted to validate data or to explore the potential of documentation contained in medical records as a means to identify the involvement of alcohol in injury. There are concerns that the quality of data available for ascertaining the nature and extent of the problem are not sufficient to support an adequate public health response. Estimates based on direct data collection from hospitals use a variety of methods and are conducted in different settings/institutions leading to variability in the enumeration of the extent of alcohol involvement. This has led to a marked variation in reported prevalence of the problem. Hospitalisation data have been used in two ways to indicate alcohol involvement in injuries and other conditions: proportional (or fractional) attribution of certain case types, using literature reviews as the basis for determining the proportions,5 and direct assignment of all cases with certain conditions as being alcohol-related.6 Proportional attribution enables population-level estimation where a probabilistic relationship exists between an exposure and outcome of interest, where it is not possible to meaningfully measure exposure by counting cases. The approach is also used where case-counting could be done, but data are lacking. Direct assignment is equivalent to proportional assignment with a proportion of 1.0. In both instances, the validity of estimates of the burden of alcohol involvement in injury-related hospitalisations depends on the quality of ICD coding.7

Few Australian studies that examined the extent of alcohol involvement in hospitalisations and fewer have focused on injury-related admissions. A recent study in New South Wales that used routinely collected emergency department data for identification of cases, found that 4% of all emergency department presentations (both for diseases and injuries) could be identified as alcohol-related (using coded data and free text fields).8 Using the national hospital morbidity data, Roxburgh et al. found that alcohol-related diagnoses were present in the admitted patient data for 7,115 separations per million in 2004/05 (less than 1% of cases), however these data are reliant on both the documentation in the medical records by clinicians and coding by clinical coders of alcohol involvement and as such are likely to be an underestimate of the magnitude of the problem.7 Using alcohol attributable fractions, Ridolfo and Stevenson9 estimated that harmful/hazardous alcohol consumption contributed to 43,033 hospital separations in 1997/98 (total hospital separations in 1997/98 totalled 5,563,074,10 accounting for less than 1% of total separations. Alcohol attributable fraction calculations for Victoria show similar results with an estimated that 1% of hospitalisations attributable to harmful alcohol use.11 Estimates of alcohol involvement in hospital admissions from the US vary widely, with estimates ranging from 7 to 70% of general admitted patients.12

For hospitalisation data to be used to create a reliable indicator of alcohol involvement in injuries, a) information regarding alcohol involvement needs to be consistently documented by clinical staff in the medical records, b) the classification system needs to provide appropriate codes to capture this information, and c) clinical coders need to be provided with guidelines or standards regarding the assignment of these codes when coding medical records. Medical records are medico-legal documents requiring thorough documentation of issues of clinical relevance in the treatment of the patient. Prior alcohol use is an important factor for consideration by clinicians as it affects patient prescription and treatment plan decisions.12 Despite the importance of documenting a full alcohol history in medical records, previous research examining the extent of documentation of alcohol involvement has found that of all identified alcohol-related admissions determined via inpatient interviewing by a researcher, only 57% of cases had documentation in the respective medical record to indicate alcohol involvement.12 A recent study in New South Wales found that only 24% of cases where alcohol was involved prior to the emergency department presentation were identified from the coded data and 76% of cases were identified by an automated search of the text descriptions collected by the triage nurse in the emergency department information system.8

If information recorded within medical documentation is to be efficiently accessed, analysed and applied to injury surveillance, it must be classified using standardised codes and definitions.13 Globally, morbidity data is coded using the International Statistical Classification of Diseases (ICD). All Australian hospitals use a modification of this classification, ICD-10-AM as the standard morbidity classification system.14 Clinical coders employed in hospital settings are responsible for translating the descriptions of medical diagnoses and procedures from medical records into coded data to enable analysis and comparison of Australian morbidity data over time and across institutions. ICD-10-AM coded data capture both diagnostic information about the diseases/injuries being treated as well as injury causation information such as the ‘intent’ (e.g. unintentional injury, assault, intentional self-harm), causal mechanism (e.g. transport event, fall, collision with person) and object (e.g. sharp knife, firearm, wall, etc). All these data items are represented within a single external cause code. The place of occurrence (e.g. home, school), and the activity at the time of the injury (e.g. working for an income, playing sport) are separately coded. ICD-10-AM codes may be three, four or five characters long depending on the specificity of the code. At the complete five character level, the codes are structured as three characters, then a point followed by two additional characters.

Codes describing the involvement of alcohol are scattered throughout the classification in the diagnoses codes (alcohol use disorders, toxic effects of alcohol), external causes (unintentional/intentional poisoning by alcohol), supplementary external causes (evidence of alcohol by blood alcohol level, evidence of alcohol by intoxication level), and factors affecting health status codes (alcohol rehabilitation, history of alcohol use). The two most specific ICD codes available for coding current alcohol involvement are the supplementary external cause codes: Y90- Evidence of alcohol involvement determined by blood alcohol level (BAL)); and, Y91- Evidence of alcohol involvement determined by level of intoxication. Y90 codes are specified at the four character level by Blood Alcohol Level (e.g. BAL < 20 mg/100 mL, BAL 20–39 mg/100 mL etc), and Y91 codes are specified at the four character level by the degree of intoxication as determined by the clinician (e.g. Mild alcohol intoxication [smell of alcohol on breath, slight behavioural disturbances etc]; Moderate alcohol intoxication [smell of alcohol on breath, moderate behavioural disturbances, etc]).15 However, the ICD-10-AM 6th Edition, in use for coding hospital discharges from July 2008 to June 2010, includes an extension to the specialty coding standard ‘0503 Drug, Alcohol and Tobacco Use’ to advise coders that:

  • Y90 should only be assigned if there is both documentation of the blood alcohol level, and a code for intoxication (F10.0), harmful use (F10.1) or dependence syndrome (F10.2) is also assigned to the same hospitalisation; and
  • Y91 should not be assigned for any inpatient morbidity coding.

In addition, the most recent advice for clinical coders from the National Centre for Classification in Health stated that ‘Z72.1 Alcohol use’ should be coded if alcohol has been documented as being involved in an injury admission.16

There has been limited research examining the sensitivity of Y90 and Y91 codes for identifying all cases of alcohol involvement. One Australian study examining alcohol involvement in work-related injuries found an under-usage of Y91 codes, with this code present in only 0.22% of cases flagged as being attributable to alcohol (as calculated by alcohol attributable proportions).17 Previous research, where clinical staff were provided guidance in the use of the Y91 codes, found that the identification of patients who were either not intoxicated or very severely intoxicated (Y91.3) was accurate, while the differentiation of mild, moderate and severe intoxication levels was weak.1,18 Thus, while the use of Y91 levels of intoxication may be unreliable (i.e. how intoxicated the patient was from mild to severe), Y91 codes could be used to exclude cases where there was no alcohol involved in the injury event. However, this would require a change in the Australian Coding Standard to direct coders in when and how to use the Y91 code in morbidity reporting.

Some other ICD-10 codes could be used to identify cases as ‘alcohol- related injury’, but there is no accepted standard to enable comparisons of results between studies, and a considerable lack of evidence on the proportion of cases with a given code that are likely to be alcohol-related. There is a limited understanding of the nature of alcohol-related history recorded in medical documentation and few reports of how this information is typically captured by clinical coders. Consequently, the extent of underestimation of alcohol as a cause of hospitalised injury has not been quantified. The aims of this study are to:

  • 1Identify the most common documentation sources for information about alcohol involvement in medical records of patients admitted to hospital for treatment of an injury.
  • 2Identify the proportion of cases for which alcohol involvement is documented in the medical record of patients admitted for treatment of an injury.
  • 3Identify a set of ICD-10-AM codes suitable for identification of cases where alcohol was involved in an injury event.
  • 4To quantify level of agreement between the identification of alcohol-related injury cases using: i) ICD-10-AM alcohol codes; and, ii) narrative medical record history of alcohol involvement.

Methodology

A retrospective on-site medical record review was used to examine the reliability, consistency and comprehensiveness of information regarding alcohol involvement for injury-related hospital admissions. This was a sub-study to a larger project that examined the quality of external cause coding and the comprehensiveness of injury information for hospitalisation data in Australia.

Study sample

Sample selection followed a two stage procedure designed to ensure a dispersion of cases across locality and hospital caseloads, while approximating a probability based result. The sampling process has been described in detail elsewhere.19 This study included a stratified random sample of 50 hospitals across four states in Australia (Queensland, New South Wales, Victoria and South Australia). All public hospitals within the four states were stratified by locality (urban, regional, remote) and injury caseload (large >2,500 injury cases/year, medium 1,000–2,499 injury cases/year, small 200–1,000 injury cases/year, or very small <200 injury cases/year).

Data collection process

A random sample of cases was extracted by the respective state health department for each selected hospital. The sample was drawn from all acute care admissions between 1 July 2002 and 30 June 2004 with an injury-related principal diagnosis (S00-T79) and at least one assigned external cause code (V01-Y98). Medical record numbers for each selected case were supplied to the relevant hospital, and the Health Information Manager at the hospital provided these records to the Auditor who attended each site.

The Auditor (a qualified Health Information Manager) reviewed the full medical records for each case and recorded text descriptions for any information available regarding the circumstances of the injury (including information such as intent, mechanism, alcohol involvement, etc). Information was extracted from all medical record documentation sources including ambulance reports, emergency department notes, clinical notes, progress notes, discharge summaries, specialist reports and allied health reports. Detailed text descriptions of the injury event and circumstances were recorded from each documentation source, identifying the source to which the text description belonged.

Variable definition

Definition of an ‘alcohol-related’ condition using ICD-10-AM codes

To identify the range of ICD-10-AM codes, a search was conducted of all ICD-10-AM index terms for diseases, external causes and drugs mentioning ‘alcohol’, ‘alcoholic’, or ‘alcohol-induced’. A liberal approach was taken to the inclusion of codes as potential flags for alcohol-related events to include both codes that signified specific involvement of alcohol in the event of interest (such as T51 Toxic effects of alcohol) and codes that signified a long-term alcohol abuse problem (such as K70 Alcoholic liver disease). The rationale for this approach was that for any alcohol codes (acute or chronic) to be coded as comorbidities in the medical records the patient needed to have been treated for and/or had their hospital stay extended due to the condition that was coded. Given that the principal diagnosis of the patients in our sample was for treatment of an injury, the presence of these codes may be a useful indicator of a potentially hazardous alcohol problem. Additionally, a free text search of ICD-10-AM 3rd edition (the edition in use for 2002–2004 hospitalisation data) was performed to identify codes that could indicate alcohol involvement. Table 1 shows the list of ICD codes identified.

Table 1.  ICD-10-AM codes with mention of alcohol in description
Diagnosis codesCode descriptionComment
E24.4Alcohol-induced pseudo-Cushing's syndromeIncluded
E52Niacin deficiency [pellagra]Included
F10Mental and behavioural disorders due to use of alcoholIncluded
F15Mental and behavioural disorders due to use of other stimulants, incuding caffeineExcluded as even though alcohol is mentioned the inclusion notes also note a wide range of other stimulants and hence inclusion of this code could include too many non-alcohol cases.
G31.2Degeneration of nervous system due to alcoholIncluded
G62.1Alcoholic polyneuropathyIncluded
G72.1Alcoholic myopathyIncluded
I42.6Alcoholic cardiomyopathyIncluded
K29.2Alcoholic gastritisIncluded
K70Alcoholic liver diseaseIncluded
K85.2Alcohol-induced acute pancreatitisIncluded
K86.0Alcohol-induced chronic pancreatitisIncluded
O35.4Maternal care care for (suspected) damage to fetus from alcoholIncluded
P04.3Fetus and newborn affected by maternal use of alcoholThis code was excluded as it does not pertain specifically to the patient being coded and therefore considered outside the scope of the current study.
Q86.0Fetal alcohol syndromeThis code was excluded from the selection algorithm as it does not pertain specifically to the patient being coded and therefore considered outside the scope of the current study.
R78.0Finding of alcohol in bloodIncluded
T51Toxic effect of alcoholIncluded
Z04.0Blood-alcohol and blood-drug testIncluded
Z50.2Alcohol rehabilitationIncluded
Z71.4Counselling and surveillance for alcohol use disorderIncluded
Z72.1Alcohol useIncluded
Z81.1Family history of alcohol use disorderThis code was excluded as it does not pertain specifically to the patient being coded and therefore considered outside the scope of the current study.
Z86.41Personal history of alcohol use disorderIncluded
External Cause CodesCode DescriptionComment
X45Accidental poisoning by and exposure to alcoholIncluded
X65Intentional self-poisoning by and exposure to alcoholIncluded
Y15Poisoning by and exposure to alcohol, undetermined intentIncluded
Y90Evidence of alcohol involvement determined by blood alcohol levelIncluded
Y91Evidence of alcohol involvement determined by level of intoxicationIncluded

Identification of ‘documented’ cases using text description

The identification of documented cases of alcohol involvement using the detailed text description extracted from the medical records was conducted in two steps. The first step involved a manual review and the second step involved an automated search algorithm for identification of cases of alcohol involvement.

Step One: Text descriptions were reviewed by an independent Research Assistant (blinded to the ICD coded data) and were flagged for inclusion in the alcohol-related sample if there was mention of the patient consuming alcohol prior to the injury event. All selected cases were independently reviewed and accepted by the first author (KM) to ensure appropriate identification as alcohol-related.

Step Two: An automated search algorithm (using SPSS) was applied to the text information extracted from the medical records. Similar to the approach used by Indig and colleagues,8 this automated algorithm searched alcohol abbreviations (e.g. etoh, ethol), alcohol type (e.g. beer, wine, rum), alcohol consumption (e.g. drinking, drank) and intoxication (e.g. intoxicated, inebriated, drunk). The second search identified negation or denial of alcohol involvement (e.g. no/denies-alcohol/etoh/drinking). Cases were flagged as potential alcohol involvement cases if they were identified in the first search and not in the second. These cases were then reviewed manually by the first author (KM) for inclusion in the final sample.

Analysis

Cases were included in the alcohol-related sample if they contained either: a) an ICD-10-AM alcohol-related diagnosis or external cause code; or, b) had mention of alcohol involvement through the text description review. Descriptive analyses were conducted to identify the number of cases identified through the coded data compared to the text description review, the external cause of injury for each identification method, the consistency of text descriptions of alcohol involvement, and the main documentation sources for alcohol-related information.

Results

Identification of ‘alcohol-related’ cases

The manual review of all text descriptions by the Research Assistant identified 409 cases, all of which were accepted as being alcohol-related cases by the first author [KM] upon review. The automated search identified 475 cases. Upon review of these cases, 64 were rejected as being not alcohol related for the following reasons:

  • The terms ‘drank/drunk/drink’ referred to the consumption of a different liquid (i.e. poisoning) or the term ‘drink’ was in relation to burns from a hot drink (n=25).
  • The term was part of another word such as d‘rum’ (n=14).
  • The text referred to previous alcohol abuse problems, but did not mention alcohol as a precursor to the current event (n=7).
  • The text stated that the person was cut by an ‘alcoholic bottle/wine glass,’ etc, without mention of alcohol consumption occurring prior to the event (n=6).
  • A different negation term was used beyond that which was searched for in the algorithm (n=5).
  • The text mentioned alcohol in relation to its affect on another person, not the patient (n=3).
  • The text referred to a work-related injury at a ‘wine'ry (n=2).
  • The text referred to ‘intoxication’ of a different substance other than alcohol (n=2).

There were six cases identified by the automated search that were missed by the Research Assistant and these were included in the final cases for selection. There were four cases identified by the Research Assistant that were missed by the automated search for the following reasons:

  • A specific alcoholic beverage or receptacle that wasn't included in the search was mentioned in the text (i.e. port, stubbies) (n=3).
  • The text was a misspelling that wasn't accounted for in the automated search (i.e. ‘indoxicated’) (n=1).

Cases were selected as alcohol-related if they fulfilled any of the following criteria:

  • 1Coded with alcohol code in diagnosis string or external cause string;
  • 2Identified by Research Assistant and/or automated search and accepted after manual review as cases of alcohol involvement.

Of the 4,373 cases sampled in the larger study, in total 442 cases (10.1% of total cases) were identified as being alcohol-related. Of these, 168 cases (38.0% of alcohol-related cases) could be identified using alcohol-related diagnosis or external cause codes, and 415 cases could be identified from the text descriptions. These two percentages were not mutually exclusive as some cases were identified by both code and text. Notably, 274 cases were only identifiable through text review (61.9% of alcohol-related cases).

Use of codes to identify alcohol involvement

Of the 38.0% of cases identified using ICD codes, the most common codes for identification were F10 Mental and behavioural disorders due to use of alcohol (n=106, 63.1% of cases with alcohol code), and T51 Toxic effect of alcohol (n=52, 30.9%). The remaining codes included Z72.1 Alcohol use (n=6), X45 Accidental poisoning by and exposure to alcohol (n=2), R78 Findings of drugs or other substances not normally found in blood (n=1) and Z86.41 Personal history of alcohol use disorder (n=1). Two of the cases with an F10 Mental and behavioural disorders due to use of alcohol code and one of the case with a T51 Toxic effect of alcohol code also had a comorbidity ‘K70 Alcoholic liver disease’ code assigned. Fifty-four of the cases with an alcohol-related diagnosis code also had an alcohol-related external cause code as follows: X65 Intentional self poisoning by and exposure to alcohol (n=33), X45 Accidental poisoning by and exposure to alcohol (n=18) and Y15 Poisoning by and exposure to alcohol, undetermined intent (n=3). Despite this sample including all cases where alcohol was identified as being involved, there were no cases in this dataset coded with either a Y90 (blood alcohol level) or Y91 (intoxication level) external cause code.

Identification of alcohol involvement by external cause category

Table 2 shows the major external cause category associated with each alcohol code and the number of cases identified through relevant code and text searches. Alcohol involvement was most commonly associated with an external cause of poisoning at 30% of cases, with almost half these cases coded as toxic poisonings due to alcohol consumption. Alcohol was recorded as being involved in 26% of cases where injuries were coded as due to being struck by or colliding with another person, and approximately 16% of cases with injuries coded as being due to cutting or piercing objects.

Table 2.  Number of cases identified by code and text searches by external cause.
Major External Cause CategoryCases identified by alcohol-related codeCases identified by text search of medical recordsNumber (%) of cases where alcohol mentionedTotal number of cases reviewed
Intent
  Unintentional84232247 (6.52%)3,786
  Intentional self-harm528388 (34.92%)252
  Assault288895 (36.40%)261
  Undetermined intent499 (32.14%)28
  Other032 (5.13%)42
Mechanism
  Transport events155458 (9.06%)640
  Falls3797104 (6.18%)1,683
  Drowning and submersion011 (10%)10
  Other threats to breathing122(13.33%)15
  Smoke, fire, flames011 (4.16%)24
  Hot object or substance000 (0.00%)63
  Poisoning68107113 (30.05%)376
  Firearm000 (0.00%)3
  Cutting, piercing object154647 (15.98%)294
  Animal related011 (0.91%)110
  Machinery in operation000 (0.00%)121
  Electricity000 (0.00%)5
  Hot and cold conditions000 (0.00%)3
  Struck by or collision person175661 (26.41%)231
  Struck by or collision object61516 (7.47%)214
  Other external cause93538 (6.58%)577
  (missing n=4)168415442 (10.11%)4,373

Use of medical record text to identify alcohol involvement

Almost 94% (n=415) of alcohol-related cases were identified through a search of the text extracted from the medical records, with 62% (n=274) of alcohol-related cases only identified through the text search. A range of terms were searched for and the number of cases with a mention of one or more of these terms is shown in Table 3 (note: some cases included mention of several of these terms, therefore the column total exceeds the number of cases).

Table 3.  Number of cases with each alcohol-related search term identified in abstracted medical record text.
Alcohol-related search termNumber of cases identified through text search% of cases identified through text search (n=415)
etoh20048.19
drinking12730.60
alcohol11928.67
intox11627.95
beer6816.39
wine286.75
drinks286.75
drank245.78
drunk174.10
bourbon112.65
rum81.93
whisky71.69
scotch51.20
vodka40.96
inebriated20.48
ethol10.24
pints10.24

Documentation sources for terms relating to alcohol

The text description review identified whether documentation indicated if the injury was alcohol-related, as well as revealing which of the five main sources (i.e. ambulance report form, emergency department record, clinical progress notes, discharge summary, other documents) in the medical records contained this information. Table 4 shows the number of records containing text of any sort in each of the documentation sources, and the number and proportion of these cases that contained information to identify the case as alcohol-related. Emergency department records were the most likely to identify whether the injury was alcohol-related with almost three-quarters of alcohol-related cases where there was text present in the emergency department notes mentioning alcohol in the abstracted text. Discharge summaries were the least likely sources to identify whether the injury was alcohol-related with only a third of alcohol-related cases where there was text present in the discharge summary mentioning alcohol in the abstracted text. Table 4 shows the number of cases from each documentation source that mentioned alcohol within the text.

Table 4.  Documentation sources for alcohol-related information.
Documentation sourceNumber of cases containing any textNumber of cases containing alcohol-related textCases with alcohol-related text/any text
 nn%
  1. Notes: a) Includes: allied health notes, specialist reports and administration notes.

Emergency department record41230774.51
Ambulance report form28115956.58
Clinical progress notes22212556.31
Other documentsa1608351.88
Discharge summary2438133.33

Discussion

This study quantified the extent to which alcohol-related injuries are identified in hospitalisation data using ICD-10-AM codes indicative of alcohol involvement. Overall, it was found that the ICD codes identified alcohol involvement in less than 4% of the total injury cases and 38% of the alcohol-related sample. However, after reviewing the text extracted from the medical records, the identification of alcohol involvement increased to 9.5% of the total injury cases (accounting for 62% of the alcohol-related sample identified from the text review). The use of the narrative review increased the estimate of alcohol involvement for injury-related hospitalisations in this sample from 4% to 10.11%. Thus, relying on the current coded separation data alone vastly underestimates the extent of alcohol involvement in injury-related hospitalisations. As Australian estimates of the extent of alcohol involvement range from around 1% for hospital admissions7,9,11 to 4% for emergency department presentations,8 strengthening the quality and accuracy of the evidence base underpinning national estimates may help produce a more precise estimate of the proportion of alcohol-related injury presentations.

We have used the term ‘alcohol-related’ injury in this paper without explanation. The relationship between alcohol use and injury risk is diverse and complex. Appropriately, attention for injury prevention focuses on the immediate effects of ethanol intoxication on cognitive and physical performance in complex tasks such as driving, but risk also extends to chronic use at high levels, at least in the case of risk of intentional self-harm20 and probably to hangover periods.21 People other than, or in addition to, the person who is affected by alcohol are often injured. Another complication for measurement is the transience of alcohol intoxication. A person who may have been intoxicated when injured may not be when he or she arrives at hospital or reaches a ward. These and other aspects of the alcohol-injury relationship should be taken into consideration when assessing what can be collected in hospital records, and when interpreting findings of analyses of hospital records. In this study, for example, cases identified as ‘alcohol-related’ were all ones that the alcohol consumption related directly to the admitted person not to other parties. Other forms of study, combined with attributable proportions methods, may offer a way to obtain more complete estimates of alcohol-related hospitalisations for injury than can be obtained by direct assignment of codes that refer to alcohol. However, the quality and completeness of such estimates will be improved by making the best use of case-level information in hospital records when assigning ICD codes.

It is evident from this study that the medical records contain information regarding alcohol involvement beyond that which is currently being captured by the coded data. While several codes exist in the ICD-10-AM that could be used to identify alcohol involvement, only a portion of those were used (mainly those representing acute intoxication). The ICD codes that are, at face value, the most relevant, Y90 (blood alcohol content) and Y91 (intoxication level), were not used for any of the cases identified in this sample.

The cases in this sample were all admitted for the treatment of an injury. Where alcohol intoxication was coded as an additional diagnosis, the external cause that was assigned always reflected the direct cause of the injury being treated (i.e. fall, transport event). However, given the importance of alcohol as a risk factor for injury, it could be argued that alcohol is an equally important cause of injury; sufficient to warrant the assignment of an external case code to flag these cases where alcohol involvement is recorded in the medical records. Given the national interest in identifying alcohol related harm in the community, it is recommended that use of Y90 and Y91 be made mandatory where there is documentation of alcohol levels and/or alcohol intoxication in a medical record.

While we have identified the number of cases where the coded data has ‘missed’ documented cases of alcohol involvement, one of the limitations of this study is that we are unable to ascertain the true extent of alcohol involvement for injury related cases as we cannot identify cases where alcohol involvement was not documented. Hence, the results reported in our study regarding the extent of underestimation from the coded data is mostly likely a conservative estimate with up to 43% of records lacking documentation regarding alcohol involvement found previously.12 To address this issue, future research could examine the ‘information continuum’ by collecting information regarding alcohol involvement from the patient in the emergency department, examining what is documented in the medical record text, and analysing the coded data upon discharge to identify the extent of information loss and underestimation of the problem.

In summary, this study found that best estimates of the frequency of injury-related hospital admissions where alcohol is involved underestimate the burden by around 62%. This is a substantial underestimate with major implications for public policy, and highlights the need for further improvement of the quality and completeness of routine data sources, such as hospitalisation records, for more reliable identification of alcohol-related injuries.

Ancillary