How does the primary care EMR recording of alcohol consumption compare to self-reported data in the UKB cohort in Scotland?
Problem
According to the Scottish Public Health Observatory, in 2016 the proportion of adults who drink alcohol beyond 14 units per week were 30% of men and 16% of women. Evidence suggests that even low-level regular alcohol use could harm. Reliable estimates of health-related behaviours are required to formulate and evaluate policies, and for alcohol research. One of the main problems when evaluating the effects of alcohol is accurately measuring how much people drink. The use of routinely collected data (RCD) and its linkage from different sources, such as health and administrative, are increasingly used in different fields and offer an unrivalled opportunity for research in biomedical sciences. In Scotland GPs often ask patients about their alcohol consumption. Plus, it has been one of the Quality and Outcome Framework (QoF) indicators from 2004-16. Although there are numerous codes in GP databases indicating alcohol consumption and multiple records per individual, this information is difficult to use for research or quality purposes and has not been validated.
Approach
From the UKBiobank cohort we selected and extracted those who answered to alcohol consumption question. We recoded these variables to establish, amount of alcohol drunk per week and then stratified these participants in four consumption categories. This dataset was linked to their GP’s records. GP data were processed using Read Codes V2 and V3 identification, using UKB guidance and compiling a list of all codes related to alcohol consumption. We performed the analysis only in participants from Scotland.
Findings
We included 35,769 participants in our analysis (those who had a recorded alcohol status in UKB questionnaire (99.7% of Scottish participants). After reclassification of the weekly alcohol consumption variables, we detected 8.38% of Non-drinkers ,56.74%Sensible drinkers (less than 14 Units per week) ,13.56% of Moderate drinkers (between 15 and 21 Units/week), 12.33% of Hazardous drinkers (between 22 and 35 Units/week and 8.99% of Harmful drinkers (more than 35 Units). GP data preliminary analysis detected alcohol records for 24,409(68.08%) distinct participants. Those participants included 133,726 coded records related to alcohol consumption (136.) with 93,057 associated recorded values.
Consequences
We found that 68.1% of Scottish UKB participants had recorded alcohol consumption in Primary Care data, considering that UKB had only released GP data for half of the cohort.Preliminary results show a poor correlation between UKB questionnaires and GP records. It is difficult to assess how it is related to real changes in alcohol consumption and if it is associated with underreportingIt’s crucial to improve data collection and proper codification for effective data research and epidemiologic surveillance. If properly validated, GP records could represent a great opportunity to have populational data on alcohol consumption.