In general practices, there are various tests GPs use to assess whether their patients are at high risk of health harms because of the amount of alcohol they drink. If an individual’s health starts to deteriorate because of their alcohol use the two major approaches to treatment available to GPs (or other health professionals) are to prescribe medications and/or to provide brief advice. However, evidence on what works is conflicting. The overall aim of this project is to examine how well do tests and treatments for risky drinking of alcohol work in a general practice setting and whether they provide value for money.
Analysing large datasets like CPRD provides an opportunity for finding out what interventions and treatments work and has some specific benefits, such as finding out how well things work in real-world circumstances. In this research we will study the effects of different tests and treatments for risky drinking by taking advantage of GPs tendencies to use the same testing tools or to prescribe the same treatment to their patients who have similar problems. For example, when there are a few options available for medicines to help reduce risky drinking, one GP may tend to prescribe a particular medicine more often than another.
We will calculate how much it costs to provide these tests and treatments and compare these costs against the likely health benefits over the short- and long-term. This will allow us to see which tests and treatments provide the best value for money.
In this cohort study we will ascertain two cohorts of individuals and follow them up to quantify risk of future outcomes (primary care visits, hospitalisations and deaths) and to estimate remaining (quality adjusted) life expectancy.
Cohorts:
Cohort 1 (secondary prevention population): individuals at risk of alcohol harm (defined by screening tests such as AUDIT and FAST; self-reported ‘high’ level of alcohol consumption; other alcohol consumption related Read/Snomed codes)
Cohort 2 (tertiary prevention population): individuals hospitalised for any alcohol related condition (defined by ICD codes)
Screening test / treatment variables:
Alcohol use screening tests; Alcohol brief interventions; Pharmacological interventions (e.g. Acamprosate, Disulfiram, Nalmefene, Naltroxone)
Outcome variables:
Primary care record of alcohol use disorder; hospitalisations (alcohol intoxication / harmful use; alcohol dependency; alcoholic liver disease; liver disease (all)) and deaths (same categories as hospitalisations and all cause deaths)
Other variables (not mentioned above):
Patient’s age, patient’s sex, socio-economic deprivation (area-based), lab test results, comorbidity measured by Read/Snomed/ICD codes (e.g Charlson index)
Statistical methods:
Comparative effectiveness analyses for head-to-head comparisons of screening tests, alcohol brief interventions and pharmacological interventions will be carried out using multivariable logistic/Cox regressions, propensity score adjusted logistic/Cox regressions and instrumental variables adjusted logistic/Cox regressions. The latter will use physician’s prescribing preferences and general practice preferences as instruments.
Decision analytic model / economic evaluation tool:
The outcomes from the statistical analyses will be used to populate a decision analytic model which will extrapolate the outcomes for the cohorts ascertained above using parametric survival modelling, validating against external data sources (e.g. national life tables). The different cohorts will allow economic evaluation of ‘secondary’ and ‘tertiary’ prevention strategies.
Sensitivity analyses:
As uncertainty exists in all aspects above, pre-specified sensitivity analyses will be undertaken.
• Primary care record of alcohol use disorder (a. identified by AUDIT PC > 4; b. identified by AUDIT C > 4; c. identified by AUDIT C > 10; d. identified by FAST > 2; e. identified by ‘Single question alcohol use test’ (M-SASQ) > 1)
• Hospitalisation for alcohol intoxication / harmful use (AIH)
• Hospitalisation for alcohol dependency (AD)
• Hospitalisation for alcoholic liver disease
• Hospitalisation for liver disease (all)
• Death (caused by AIH)
• Death (caused by AD)
• Death (caused by alcoholic liver disease)
• Death (caused by liver disease – all)
• Death (all cause)
Note: the rationale for using alcohol use disorders (intoxication / harmful use / dependency) and alcoholic liver disease is that they make up a quarter of alcohol-attributable mortality, and are 100% alcohol attributable [1]
For developing the decision analytic model, we require linked data to all hospitalisation records.
Jim Lewsey - Chief Investigator - University of Glasgow
Jim Lewsey - Corresponding Applicant - University of Glasgow
Bhautesh Jani - Collaborator - University of Glasgow
Claudia Geue - Collaborator - University of Glasgow
Eileen Kaner - Collaborator - Newcastle University
Elise Whitley - Collaborator - University of Glasgow
Francesco Manca - Collaborator - University of Glasgow
Frederick Ho - Collaborator - University of Glasgow
Janet Bouttell - Collaborator - University of Glasgow
Linsay Gray - Collaborator - University of Glasgow
Lisong Zhang - Collaborator - University of Glasgow
Srinivasa Vittal Katikireddi - Collaborator - University of Glasgow
HES Admitted Patient Care;Mental Health Services Data Set (MHSDS);ONS Death Registration Data;Patient Level Index of Multiple Deprivation;Practice Level Index of Multiple Deprivation