An observational study of alcohol-related readmission dynamics in young people

Abstract

BackgroundAlcohol-related harm among young people in England is a substantial public health concern. Hospital Episode Statistics records for the NHS in England are a powerful yet flawed source of information about the development of problematic drinking careers of young people. The aim of this study was to assess whether these careers could be summarised and investigated after novel processing techniques had been applied to these records. These techniques attempted to account for problems in the data, specifically coding errors and splitting of relevant admission records between multiple annual datasets. MethodsA cohort of 15 203 young people aged 10–25 years in England who had an index alcohol-related emergency admission between April, 2003, and March, 2004, were tracked for subsequent alcohol-related readmission through the use of a pseudo-anonymised identification number for 3·75 years (data collection ended March, 2008). International Classification of Diseases 10 codes used in initial sample selection focused on those that indicated the immediate effects of excessive alcohol consumption (F10.0, F10.1, R78.0, T51.-, X45.-, X65.-, Y15.-, Y90.6, Y90.7, Y90.8, Y91.1, Y91.2, Y91.3, and Z72.1). Mean age of the cohort was 18·4 years (56·5% male). For those for whom ethnicity was known, just over 95% were white, although missing data makes this finding uncertain. Novel processing techniques were used to prepare data for analysis, whereby algorithms undertook record-by-record processing, with the intention of quality checking and appropriately combining incomplete or duplicate admission entries. The algorithms accomplished this preparation by applying standard data quality checks to each record, checking for repetition or incomplete or faulty records. By stepping through the dataset line-by-line, the algorithms were able to analyse a record in the context of adjacent, associated records. This method eliminated much uncertainly in the dataset and allowed a detailed picture of subsequent drinking careers to emerge. Confounding factors are inevitable in a cohort of this nature (in particular, dual diagnosis relating to mental health, self-harm, injury, assault, and other substance use) and were accounted for by incorporating them into the analysis as independent variables, along with other patient demographic variables. The resulting analysis dataset was subject to analysis in an attempt to distinguish the determinants of readmission. FindingsAlmost 13% (1950 of 15 203) of the cohort was readmitted during the following 3·75 years, and almost 5% (707 of 15 203) of the cohort had two or more alcohol-related readmissions. The characteristics of each return were investigated. On average, each subsequent admission featured a lengthened hospital stay, was progressively more likely to occur on a non-traditional drinking day, and occurred after a progressively smaller number of days since previous admission. The likelihood increased for a patient, as they cycled through subsequent readmissions, to be living in a neighbourhood of a lower average wealth than that where they lived during their index admission. InterpretationThe use of an automated method of summarising admission sequences proved feasible. The resulting findings can be used to inform the process of hospital screening for referral to alcohol treatment. Additionally, the deprivation finding can be interpreted as showing that poverty is both a part cause of, as well as being a result of, a problematic drinking career. Potential limitations of the study include (1) data quality problems that may have persisted despite the algorithms; (2) an absence of data about actual level of alcohol consumption; and (3) a focus on data that only capture people at specific periods when their drinking (often briefly) emerges in a clinical setting. Associated with this last point, an alternative explanation of the results is that the cohort offers an unrepresentative picture of problem drinking among young people, because the methodology excludes a group of unknown size who develop drinking problems, but who are never admitted (or readmitted) into hospital. FundingNone.