Modelling patterns of polydrug use in the population of Great Britain
: A latent class approach

  • Gillian W Shorter

Student thesis: Doctoral Thesis


Substance use and their consequences account for substantial public
spending in Great Britain and Europe. However, alcohol and drug policy change
is often based on the harms attributed to a single drug, when polydrug use (the
use of more than one drug) is common and may increase associated harm. This
research has employed advances in statistical analysis to create
typologies of polydrug use including alcohol and illicit drugs at a general
population level in the 2000 National Psychiatric Morbidity Survey to better inform risks associated with these substances. Alcohol and drug use patterns were determined separately and then in an integrated model using latent class analysis. The relationship between the classes, demographic, and psychological status was assessed through multinomial logistic regression. Three drug classes, wide range, moderate range and no drug use represented illicit polydrug use. Six
alcohol classes were found, most of which followed a continuum of increased
consumption relating to increased related problems, with one exception who experienced problems with low alcohol consumption. Any drug use elevated the
risk for psychological conditions compared with the no drug use class, and
increased levels of alcohol consumption were related to increased risk of
generalised anxiety disorder and lifetime suicide attempts. Two integrated
alcohol and illicit drug polydrug use models were proposed, one with eight and
one with 18 classes. Further investigation of these models found two main
conclusions. First, illicit polydrug classes change when measured with alcohol
use, and the three classes were now ‘no drug’ use, ‘cannabis only’ and
‘polydrug users’. This in turn affected relationship with demographic variables
and psychological status. Second, whilst alcohol use patterns do not change
dramatically when measured in the presence of illicit drug use, there were
changes in relationships of alcohol with current psychological status as a
consequence accounting for illicit drug use. Risk of poorer psychological health
is elevated in heavier patterns of alcohol use who also consume illicit drugs
compared to the same alcohol pattern without illicit drug use.
In conclusion, latent class analysis is a useful way in which to model population
level polydrug use. Through this methodology the consequences of separating
alcohol and illicit drug use in research have been shown. Research into
either alcohol or illicit drug use should consider modelling both in the same
model. The presentation of polydrug use in this context may be able to
show some of the ambiguities in the literature regarding demographic
differences and risk relating to poorer psychological health.
Date of Award1 Jul 2009
Original languageEnglish
SponsorsHSC R&D Public Health Agency
SupervisorMark Shevlin (Supervisor) & Brendan Bunting (Supervisor)


  • polydrug use
  • substance use
  • alcohol
  • drug use
  • mental health
  • statistical modelling
  • epidemiology
  • latent class analysis
  • latent variable analysis
  • population data

Cite this