Prevalence of injecting drug use in Estonia 2010–2015: a capture-recapture study

Abstract Background It has been observed in an earlier study that the number of people who inject drugs (PWID) in Estonia is declining. We provide nationwide estimates of the number of PWID in Estonia for years 2010–2015 and compare different modelling strategies to minimise over-coverage-induced bias in capture-recapture estimates. Methods We obtained data from the Estonian Causes of Death Registry (DR) for opioid-related deaths, the Estonian Health Insurance Fund (HIF) for opioid-related overdose and drug dependence treatment episodes, and the Estonian Police and Border Guard Board (PB) drug-related misdemeanours. Datasets were linked by identifier based on sex, date of birth, and initials; a capture-recapture method was used to estimate the number of PWID aged 15 or more, each year from 2010 to 2015. Log-linear regression maximum likelihood (ML) and Bayesian methods were used; over-coverage of police data was accounted for. Results The annual population size estimates of the number of PWID (aged 15 and over) varied from 6000 to 17,300 (ML estimates not accounting for over-coverage of PB) to 1500–2300 (Bayesian estimates accounting for over-coverage). Bayesian estimates indicated a slight decrease in the number of PWID, and the median estimates were > 2000 in years 2010–2012 and < 1800 in years 2013–2015. Conclusions Over-coverage of a registry can have a great impact on the estimates of the size of the target population. Bayesian estimates accounting for this over-coverage may provide better estimates of the target population size