Send Orders for_Repjiilts}o- ieji,rints7ii5Jient/iUmscience.nei
130 Currellt Drug Abuse Reviews, 2013,6, 130-143
Methods for Estimating Incidence of Drug Use: A Review
Albert Sánchez-Niubo··I, Alessia Mammone2 and Gianpaolo Scalia Tomba2
I Drllg Abuse Epidemiology Research Grollp - IMIM, Barcelona, Spain
2Cenlro Inlerdiparlimenlale di Bioslalislica e Bioinformalica, Universila di Roma Tor Vergala, Rome, ltaly
Abstraet: Several methods for estimating drug use incidence, that have been used in the Jiterature or that could be used,
having been used in a different framework, are described and commented. The applicability of the different methods
depends on available data and knowledge about relevant parameters. The many similarities between drug use incidence
estimation and estimation of disease incidence are highlighted, but also the distinguishing aspects that make drug use
incidence estimation a challenge to standard statistical methods.
Keywords: Drug use, incidence, prevalence, problem drug use, statistical modelling.
1. INTRODUCTlON
The incidence of first drug use is an important
epidemiological concept that helps understanding the
diffusion of drug use in space and time. Incidence estimates
can be used to evaluate current and future needs for, and
effects of, services and interventions.
Incidence is defined as the number of new cases
occurring within a given time period, usually ayear.
Compared to prevalence, which is defined as the number of
existing cases at a certain point in time (or, for practical
reasons, in a given period; in studies using administrative
data this is often a calendar year), incidence gives more
direct infonnation on the recruitment of new cases. It is
therefore a more sensitive indicator of effects of prevention
efforts than prevalence. However, incidence is directly
related to prevalence: adding up the incidence over a number
of years, and subtracting the number of cases that have died
or ceased to be cases for other reasons, gives the change in
prevalence over the same time periodo Note that the
incidence of first drug use usually refers to the first use ever.
People who relapse after abstinence are not included.
Sometimes, the initial event of drug use is called "onset of
drug use", but there may be so me uncertainty about the exact
detinition of this term, in particular whether this should refer
to the "first use ever" or "first habitual use".
In this report we will outline some methods that are
currently applied or that could be applied for drug use
incidence estimation. In the description of the methods, it
will often be seen that they were developed with heroin use
in mind or for diseases that have the same character as heroin
use, namely the progression start of use/infection - continued
use/infected state - treatment or other type of
registration/diagnosed disease - cessation/cure or death. The
typical application in disease epidemiology has be en the
estimation of the size of the HIV epidemic assuming AIDS
as an observable endpoint. This focus on heroin use or
•Addresscorrespondence lo Ihisaulhoral Ihe IMIM-InslitulHospilaldel
Mard'InvesligacionsMediques,DoclorAiguader88.E-08003Barcelona.
Spain;Tel:+34933160753; Fax:+34933160797;
E-mail:asanchez@imim.es
1874-4745/13558.00+.00
equivalently the use of one dominating, strongly addictive
drug whose use almost invariably leads to request for
treatment or other notifiable events does not always translate
well to today's drug scene, where intennittent multi-drug use
or use of drugs that rarely lead to statistically observable
events has become more and more common. Thus, some of
the methods, that have been of common use in the past, now
secm dated or limited in scope. However, other methods,
such as the model based ones or the survey based ones, that
have not been much used in the past, may now become more
popular.
The described methods are:
General inclusion function
Retrospective estimation from periodic population
surveys
Lag correction method
Log-linear model of quasi-independence
Back-calculation
Multistate model with immigration
Parametric models
Incidence from prevalence
Snapshot samples
These methods are not altematives to each other. They
each have a certain range of applicability, depending on
available data. In the following pages the aim is to give some
basic understanding of how these methods work and what
their requirements, limitations and possibilities are.
2. GENERAL INCLUSION FUNCTlON
The tenn "general inc1usion function" (GIF) was
probably used for the first time in a paper by Nordt el al. [1].
In that paper, the authors show, using data from the case
register for substitution treatment in the canton of Zurich,
Switzerland, [rom the period 1992-2004, that the probability
of tinding a patient in therapy at a given time depends
essentially on the time elapsed since onset of heroin use and
not on when this onset occurred. The function rclating these
© 2013BenthamSciencePublishers