Estimating the Burden of Alcohol on Ambulance Callouts through Development and Validation of an Algorithm Using Electronic Patient Records

Abstract

Background: Alcohol consumption places a significant burden on emergency services, including ambulance services, which often represent patients’ first, and sometimes only, contact with health services. We aimed to (1) improve the assessment of this burden on ambulance services in Scotland using a low-cost and easy to implement algorithm to screen free-text in electronic patient record forms (ePRFs), and (2) present estimates on the burden of alcohol on ambulance callouts in Scotland. Methods: Two paramedics manually reviewed 5416 ePRFs to make a professional judgement of whether they were alcohol-related, establishing a gold standard for assessing our algorithm performance. They also extracted all words or phrases relating to alcohol. An automatic algorithm to identify alcohol-related callouts using free-text in EPRs was developed using these extracts. Results: Our algorithm had a specificity of 0.941 and a sensitivity of 0.996 in detecting alcohol-related callouts. Applying the algorithm to all callout records in Scotland in 2019, we identified 86,780 (16.2%) as alcohol-related. At weekends, this percentage was 18.5%. Conclusions: Alcohol-related callouts constitute a significant burden on the Scottish Ambulance Service. Our algorithm is significantly more sensitive than previous methods used to identify alcohol-related ambulance callouts. This approach and the resulting data have potential for the evaluation of alcohol policy interventions as well as for conducting wider epidemiological research.

Lockdown and licensed premises: COVID‐19 lessons for alcohol policy

“…Both studies capture professional perspectives but there is also a need for in‐depth qualitative research to understand consumer experiences, including their view of possible long‐term outcomes of changes in alcohol consumption. Considering the quantitative ambulance callout data, while the algorithm used to identify alcohol‐related callouts has been found to be highly accurate [ 30 ], it may have over or underestimated alcohol‐related callouts. People may be less likely to call an ambulance for domestic incidents and it may be more difficult to identify alcohol as a factor in a callout to a home compared to callouts to licensed premises, which may be a source of bias.…”

“…Alcohol‐related ambulance callouts were defined as those identified using an algorithm that makes use of free text notes completed by ambulance staff in electronic patient records for each callout as well as an alcohol ‘flag’—a field in the electronic patient record allowing ambulance clinicians to indicate if a callout is alcohol‐related. The algorithm was validated and performed well with 98% accuracy [ 30 ]. Our analyses utilised both a descriptive and an inferential approach.…”

“…Much UK public discourse has centred on hardships for hospitality businesses arising from public health measures with surprisingly little discussion of reduced pressures on emergency services [ 28 , 29 ]. The burden of alcohol on health services is well documented, including recent work estimating that 16.2% of all ambulance call‐outs in Scotland are alcohol‐related [ 30 ]. This discourse contrasts with examples elsewhere of governments placing or maintaining restrictions on alcohol explicitly to reduce this burden on front‐line services [ 31 , 32 ].…”

Ambulance attendance for substance and/or alcohol use in a pandemic: Interrupted time series analysis of incidents

“…The splines were plotted against the time in days since lockdown, giving a polynomial function that describes the curve. Across the time length of 497 days, a knot was fixed every 7 days chosen given the well‐documented weekly patterns in AOD‐involved ambulance attendances [ 19 , 20 ].…”

“…This seasonality effect was controlled by including a term for the month of year. Day of the week was also adjusted for in the model, given variations in alcohol consumption patterns across the week [ 23 ], and ambulance attendances [ 19 , 20 ] being higher at the weekend.…”