European Journal of Public Health, Vol. 19, No. 1, 32 37 ß The Author 2008. Published by Oxford University Press on behalf of the European Public Health Association. All rights reserved. doi:10.1093/eurpub/ckn126 Advance Access published on 26 December 2008... Capture recapture estimates of the local prevalence of problem drug use in six French cities Laure Vaissade 1, Stéphane Legleye 1,2 Background: General population surveys do not provide reliable estimates of problem drug users, which are essential for drug use prevention and treatment policies. Indirect estimation methods, especially capture recapture method, are preferred. Problem drug use receives now a growing interest on a European level. Very few studies using this method have been carried out in France especially on a multicentric level. Method: Six three-sample capture recapture analyses were carried out in urban areas of Lille, Lyon, Marseille, Metz, Rennes, Toulouse. Data were collected in 2006 from different sources including treatment data, harm reduction and social data and law enforcement data. Users of opiates, cocaine/ crack, stimulants and/or hallucinogens aged 15 64 were identified. Log-linear models were fitted to the data in order to test different interactions between the data sources representing potential dependencies. The simplest model was selected on the basis of its Akaike Information Criterion (AIC) and log-likelihood ratio tests. Confidence intervals were estimated using the likelihood interval method. Results: The total number of problem drug users was estimated to be 7900 (95% CI 6300 10 200) in Lille, 8400 (6300 11 800) in Lyon, 5600 (4200 7700) in Marseille, 2300 (1700 3200) in Metz, 1500 (1100 2300) in Rennes and 5400 (4300 6900) in Toulouse (estimates are rounded to nearest 100). These figures correspond to the following prevalence rates of the population aged 15 64 years: 10.8 per thousand (8.6 14.0) in Lille, 10.6 (8.0 15.0) in Lyon, 10.3 (7.7 14.2) in Marseille, 10.8 (8.0 15.0) in Metz, 7.6 (5.6 11.7) in Rennes, 10.1 (8.0 12.9) in Toulouse. Conclusions: Although the confidence intervals are wide, the method provides valuable information on the extent of problem drug use. Keywords: capture recapture, problem drug use, multicentric, France.... Introduction he general population surveys do not provide reliable Testimates of problem drug users, which are essential for drug use prevention and treatment policies. 1 Indirect estimation methods, especially capture recapture method, are preferred. 2 The capture recapture technique, originally developed to estimate the size of animal populations, has been adapted by epidemiologists, 3 with two or more data sources. 4,5 The application of the method requires that the captured (i.e. observed) population is an unbiased sample of the whole target population: therefore, data sources should belong to the widest range of activities related to the problem drug use, from the health, social, to criminal aspects. 6,7 Very few studies using this method have been carried out in France: the first one took place in the city of Toulouse in 1995, 8 whereas the last date from the late 1990s. 9 Within the framework of a partnership developed by a European Union agency, the European Monitoring Centre for Drugs and Drug Addiction, France should be in a position to provide results concerning estimated problem drug use. An operational definition could consider that an individual is a problem drug user when his consumption of a substance leads to health, social and/or legal problems for himself or others. Numerous towns across Europe and worldwide possess local estimates of the prevalence of problem drug use. 10 24 1 French Monitoring Centre for Drugs and Drug Addiction (OFDT), Saint-Denis La Plaine, France. 2 National Institute for Health and Medical Research (Inserm), Paris XI University, Paris, France. Correspondence: Stéphane Legleye, OFDT, 3 avenue du stade de France, 93218 Saint-Denis la Plaine, France, tel: +33 1 41 62 77 37, fax: +33 1 41 62 77 00, e-mail: stleg@ofdt.fr In 2006, the OFDT (French Monitoring Centre for Drugs and Drug Addiction) carried out a capture recapture study in six French urban areas, in which the feasibility of the study was guaranteed: Lille, Lyon, Marseille, Metz, Rennes and Toulouse. Methods Capture recapture We carried out six standard three-sample capture recapture analyses in order to estimate the number of problem drug users in Lille, Lyon, Marseille, Metz, Rennes and Toulouse. Methods have been described in detail elsewhere. 25 27 Participating cities The sample of cities is a convenient one. Choice was made according to the presence of a local network of agencies working in the field of drug use monitoring. Paris was not being included because of its size and the number of social and health agencies working there. In each city, the geographical area was defined taking into account the accuracy and relevance of the already implemented data sources and policies: the urban community of Lille (north of France), with almost 1.1 million inhabitants; Great Lyon (centre south-west) comprising 57 districts with more than 1.15 million inhabitants; inner-marseille (Mediterranean seaside) with almost 800 000 inhabitants (the districts outside Marseille were not included because very few social and health organizations are to be found there and because their inhabitants rarely visit the specialized centres for drug users in inner-marseille); the Metz (east of France) area covering 61 districts with just over 313 000 habitants; the Rennes (west) area comprising 10 districts, with almost 273 000 inhabitants
Capture recapture drug users in France 33 Table 1 Combination of data sources in three samples and numbers of records Sample A Sample B Sample C Number of records (%) Number of records (%) Number of records (%) Lille DT, H, GP 1212 (67) LT 327 (18) CO, DS, TP 276 (15) Lyon DT 645 (47) GP 281 (21) CO, LT, DS, TC, TP 437 (32) Marseille DT 470 (61) LT, H 269 (29) CO 95 (10) Metz DT 140 (28) H 156 (31) CO, DS, TP, LT, TC 206 (41) Rennes DT 161 (39) LT 130 (32) CO, TP, GP 118 (29) Toulouse DT, LT 647 (56) GP 424 (37) CO 80 (7) Note: Specialised Drug addiction treatment centres (DT), Low-Threshold services (LT), General Practitioners (GP), Hospital units (H) such as infectious diseases, accident and emergency wards, etc. Drug Squads (DS), users Referred Treatment by the court (RT), Treatment units in Prison (TP), or data detained by the Central Office for the repression of drug-related offences in drug misuse (CO) and the Toulouse (south-east) urban area with 72 districts and 761 107 inhabitants. Data sources Data were collected over a 6-month period between 2005 and 2006: from these sources: drug treatment agencies, general practitioners, hospital units (such as infectious diseases, attendees at accident and emergency), low threshold agencies, social services and law enforcement sources, such as drug squads, users referred treatment by the court, treatment units in prisons (data collection in prison was delayed for 2 months compared with other data sources in order to allow problem drug users entering prison to have already been captured by other data sources over the last 2 months), and data detained by the OCRTIS (the Central Office for the Repression of drugrelated Offences in drug misuse). Various data sources were combined in three samples following both statistical criteria (odds ratio between two data sources higher than one, suggesting a positive dependency, leads to a combination of both data sources) and a field criterion (when it is locally known that two data sources are positively related). The combination of the data sources in three samples as well as the number of records collected from each sample is shown in table 1. Furthermore, Supplementary table 1 gives details on drugs detected in different cities and figures on polydrug uses. Inclusion criteria Subjects were included in the study if they were resident for >3 months within one of the six cities, if they mentioned illicit drug use over the last 30 days (cannabis excluded): opiates, cocaine/crack, other stimulants and/or hallucinogens (LSD, Ketamine, etc.), and if they were aged 15 64 years. Individual identifiers The individual identifiers used for matching people from each sample were the initial of the first name, the three first characters of the surname, the sex and date of birth. Due to the nature of the study, no consent of participant was collected. Nevertheless, the participation did not lead to any consequence for the individual, nor medical, social, neither legal. Log-linear models Matched data were analysed using log-linear models. All possible models were run from independent to all two-way interactions (we assumed that there were no three-way interactions). To select the best model, the absolute goodness of fit (G 2, the deviance between observed and expected values) of the model was approximated by the chi-squared distribution. A lower deviance implied that the observed and expected values were closer, indicating that the model fitted better. Models with different degrees of freedom were compared using a log-likelihood ratio test. For models with the same number of interactions, the Akaike Information Criterion (AIC) was used. 28 In each retained model, the expected number for the unknown cell and the associated confidence interval were obtained using the goodness-of-fit-based method. 29,30 Log-linear modelling was computed using the SAS V9.1 software package and the cell counts obtained by the Bishop formula 4 using Microsoft Excel. Ethical approval The study protocol was approved by the CNIL (French National Commission for Data protection and liberties), which ensures that participants freedom and rights were protected. Results Records and overlaps We obtained 1815 records corresponding to 1683 problem drug users in Lille; 7.5% of them were observed in several samples. In Lyon, 1363 records corresponded to 1267 subjects, 7.3% of them were in more than one sample. From 929 records in Marseille, 799 individuals were identified and only 4.2% in several samples. In Metz, 467 individuals were identified from 502 records and 6.8% were present in several samples. In Rennes, 409 records corresponded to 351 persons; 15.4% of them were observed in several samples. From 1151 records in Toulouse, 1090 persons were reported and 5.5% of them were observed in several samples. Overlaps among samples are detailed in Supplementary table 2. Demographic characteristics More than 75% of the subjects were male and half were over 33 years old (table 2). Women were more present in Marseille, Rennes and Toulouse. Mean age ranges from 29.7 in Rennes and 35.2 in Marseille. Most of the individuals lived in stable accommodation: from 59% in Rennes to 74% in Lyon. Nevertheless, unstable accommodation is quite common (30% in Marseille), even being homeless (10% in Rennes). Selected models Table 3 presents the best-fitting log-linear models showing interactions, AIC, G 2 and P-value and their estimates. Models selected in Lille and Rennes contained one significant interaction term between treatment data and harm reduction data. The positive interaction terms (0.16 in Lille and 0.36 in Rennes) suggested that individuals attending the treatment agencies are more likely to also frequent low threshold services. These results seem consistent because social and health services
34 European Journal of Public Health Table 2 Age/sex distribution of individuals in each of the six data sets Number in data set Female (%) Age Living status Mean (SD) Stable accommodation (%) Unstable accommodation (%) Homeless (%) Prison (%) Lille 1683 17 32.6 (6.48) 72.7 14.6 6.5 6.2 Lyon 1267 18 33.5 (7.75) 73.9 13.9 3.9 8.3 Marseille 799 21 35.2 (7.37) 61.2 29.9 0.4 8.5 Metz 467 20 30.5 (7.66) 63.2 24.1 5.3 7.4 Rennes 351 23 29.7 (7.71) 58.7 12.6 9.9 18.7 Toulouse 1090 25 34.4 (7.85) 67.4 15.9 1.8 14.9 Table 3 Best-fitting log-linear models: interactions, DF, G2 and AIC, estimates and prevalence Interactions DF G 2 AIC Population aged 15 64 No. of obs Estimate of unobs Overall 95% CI a Prevalence (ø) 95% CI estimate a Lille Positive: A B 2 3.20 0.80 728 173 1683 6207 7900 6300 10 200 10.8 8.6 14.0 Lyon Positive: A C 2 1.54 2.46 788 893 1267 7163 8400 6300 11 800 10.6 8.0 15.0 Marseille Independent 3 3.98 2.02 543 206 799 4757 5600 4200 7700 10.3 7.7 14.2 Metz Independent 3 4.48 1.52 212 632 467 1844 2300 1700 3200 10.8 8.0 15.0 Rennes Positive: A B 2 0.76 3.24 196 389 351 1142 1500 1100 2300 7.6 5.6 11.7 Toulouse Negative: A C 2 1.92 2.08 534 132 1090 4288 5400 4300 6900 10.1 8.0 12.9 a: Rounded to nearest 100 work in close collaboration in Lille and Rennes, as reported by local experts. Law enforcement data was independent from other samples. Both models fitted the data well, as demonstrated by the relatively small deviance values. In Metz and Marseille, best models were the independent models. This was consistent with the independent work of the different agencies in these cities. In Lyon and Toulouse, the best-fitting model supposed the statistical independency of the GPs data. This revealed a local trait: general practitioners who met some drug users seem to have very few relationships with other health, social or police services, as reported during the study. The positive interaction in Lyon between social, health and law enforcement workers was quite plausible because some drug users observed by repressive sources are also treated by hospital units. In Toulouse, a negative interaction between police data and treatment data is supported by local perception. Estimates The estimates (table 3) ranged from 1500 problem drug users in Rennes to 8400 in Lyon and 15- to 64-year-old population prevalence ranging from 7.6ø in Rennes to 10.8ø in Lille and Metz. Discussion To estimate the size of the population of problem drug users in six cities, capture recapture analyses were performed using three samples of individuals. We found very similar and credible estimates of the prevalence of problem drug use in the 15- to 64-year-old population in Lille, 10.8ø (range 8.6 14.0), in Lyon with 10.6ø (range 8.0 15.0), in Marseille, 10.3ø (range 7.7 14.2), in Metz, 10.8ø (range 8.0 15.0) and in Toulouse, 10.1ø (range 8.0 12.9). Nevertheless, the upper limit of confidence interval was somewhat lower in Toulouse and this was corroborated by an examination of other regional indicators which suggested that the prevalence of problem drug use in Toulouse could be slightly less than that in Lyon, Lille, Marseille and Metz. In Rennes, the prevalence appeared lower with a prevalence of 7.6ø (range 5.6 11.7). According to local experts, six estimates generated by our models are credible. Comparison with old estimates These last capture recapture survey was conducted in France in 1999 in five cities (Lens, Lille, Marseille Nice and Toulouse) 9 for problem opiates and cocaine users. By restricting our present analysis to the same target group, it is possible to compare our results with those obtained in 1999 for Lille, Marseille and Toulouse. Thus, in 2006 we found a prevalence of problem opiate/cocaine use among the whole population of 6.4ø (4.9 7.9) in Lille, 6.1ø (4.2 7.9) in Marseille and 6.7ø (5.2 8.3) in Toulouse. The prevalence of problem opiate/cocaine users was estimated as follows in 1999: 6.2ø (5.2 7.2) in Lille, 6.4ø (5.8 7.6) in Marseille and 4.3ø (4.0 4.7) in Toulouse. Because the confidence intervals do not overlap each other, the prevalence of problem opiate/ cocaine users in Toulouse seems to be higher in 2006. The evolution of the prevalence of the problem opiate/cocaine use between 1999 and 2006 are not significant in Marseille and Lille. These comparisons should be interpreted cautiously since the participating sources and geographical areas are slightly different between 1999 and 2006, due to possible changes in the social, health and/or legal framework of the cities. Comparison with other data A survey among the general 15- to 64-year-old population allows the departmental prevalence of people who experimented with heroine, cocaine, crack, amphetamines, LSD or ecstasy over the previous year to be extracted. 31 Direct prevalence estimates of drug experimentation in six French departments (administrative divisions) are: 16.1ø in Bouche-du-Rhoˆne (Marseille department), 14.1ø in Rhoˆne (Lyon department), 9.3ø in Nord (Lille department), 7.7ø in Haute-Garonne (Toulouse department), 4.9ø in Ille-et-Vilaine (Rennes department) and 0.0ø in Moselle (Metz department). The ranking of the six departments according to these prevalence differed from the ranking of the six cities according to problem drug use
Capture recapture drug users in France 35 prevalence. It was not surprising to note that the prevalence of problem drug use in Marseille and Lyon was lower than the drug experimentation prevalence in the general population of Bouche-du-Rhoˆne and Rhoˆne. In Lille, Metz, Rennes and Toulouse, the prevalence of problem drug use is slightly higher than the drug experimentation prevalence in the respective departments. This could be due to a possible and expected general under-registering of drug experimentation in the general population survey as well as to its general methodology: the use of telephone excludes the homeless and the poorest people, and maybe a great part of the problem drug users, and the sampling design does not ensure that data is representative at the department level. When compared with departmental health and social drugrelated indicators in 2004 (available at: http://www.ofdt.fr/ ofdtdev/live/donneesloc/indic.html), such as the rate of drug users seeking treatment, injecting material sales, substitution substance sales and the individuals arrested by the police for drug use, our prevalence estimates with capture recapture are consistent. The department of Rennes had the lowest drugrelated indicators in 2004 and was ranked at the bottom of the French departments. On the contrary, the department of Metz is at the top of the French department list in terms of these four drug-related indicators. This could corroborate the relatively high prevalence we found in Metz with capture recapture. The other prevalence of problem drug use in Lille, Lyon, Marseille and Toulouse estimated with capture recapture was relatively closed. The prevalence in Marseille could be under-estimated because a number of data sources declined to participate in the study referring to organizational difficulties in collecting data. Consider the definition of problem drug use The definition of problem drug use per se is a major concern. The substances causing problems, patterns of use, types of problem triggered by drug use and substance legal status may also vary across areas and over time. This is yet another reason for the cautions interpretation of comparisons. The definition of the target group we used is an operational one: in this context, problem drug use refers to all different forms of physical, psychological, social and legal problems caused by the use of opiates, cocaine, stimulants and hallucinogens. As a consequence, individuals who had never encountered any health or social problems due to their drug use but who were in contact with the police on at least one occasion because of their addiction were therefore considered as problem drug users and included in our study. The European definition of problem drug use is as follows: Injecting drug use or long duration/regular use of opiates, cocaine and/or amphetamines. Because France should meet the objectives of the EMCDDA, recent drug users of opiates, stimulants (cocaine, crack, amphetamines) were included in our study, as recommended by the EMCDDA. 32 We also added hallucinogen (LSD, ketamine... but cannabis excluded) users because some fieldworkers at the local level (in particular in Rennes) mentioned that using ketamine is a relevant problem that should be considered as a problem drug use. Nevertheless drug users who consumed only hallucinogens are a very small group and were they excluded of the samples, estimates of problem drug use using capture recapture would not have been very different. Limitations The large confidence intervals we obtained in six studies were probably due to small overlaps between samples. This means that our local estimates should be considered as rough estimates. We took care of the main four underlying assumptions when applying the method. 33 35 The closure assumption means that the true population is unaffected by migration, births and deaths during the study period. Data were examined over a 6-month period in each city and one of the selection criteria had to be present for more than 3 months there: it is therefore to be hoped that the population size was not affected too much by migrations. Moreover, the time span of 6 months is relatively short in drug user s career so that the closed population assumption involving new users (births) entering to the population and others leaving it (i.e. ceasing use, death) can be considered as covered by the data. The second assumption demands that identification of the overlap between different sources is not subject to error. In our protocol, the concise recording of individual identifiers and overlap identification was carefully carried out in order to minimize errors. The heterogeneity assumption may perhaps be the most problematic assumption when applying capture recapture methods. Drug users are diverse and may not have the same probability of being captured depending on gender, on age or on type of substance used. This variability could be controlled by stratifying the population by gender, age group and into three types of substance used (opiates, stimulants, hallucinogens). Nevertheless, the amount of data would be too poor in order to apply capture recapture, particularly for stimulant users and hallucinogen users. Moreover, a significant percentage of poly dug users are present. The Supplementary table 3 shows the distribution of drug detected by different sources. Figures underline that heterogeneity probably exists in Marseille and Toulouse where opiates users are more likely to enter treatment centres than being in the data from the Central office for the repression of drug-related offences in drug misuse. The contrary occurs for stimulant users who are in both cities more likely to be detected by the law enforcement services. In other cities (Lille, Lyon, Metz and Rennes), heterogeneity caused by the type of substance used would not seem to be a major problem. Furthermore, previous studies showed that this does not greatly affect the results 7,10,11,26 and the bias introduced by heterogeneity would appear generally less than that generated by dependencies between samples. 36 The independence assumption, i.e. the absence of a threeway interaction between the three samples in each city, seems reasonable when considering both the chosen models with at least one sample likely to be independent of all the others, and the local situation on the field. The treatment units in prisons collected only few questionnaires, as only a small number of inpatients were included in the sample. Then, the possible bias may only very small if any and we did not detect any tendency to a negative interaction due to the 4 months of overlap with other data sources. Recommendations These approximate local estimates in Lille, Lyon, Marseille, Metz, Rennes and Toulouse seem to be valid, credible estimates. They carry clear implications for professionals and decision-makers at local level and will provide an opportunity to produce a new national estimate of problem drug use in the future. This survey should be repeated in order to produce a follow-up of the size of the problem drug users population. Specific attention should be focussed on the quality of the data-collection stage, especially the professionals involvement in the exhaustive data collection.
36 European Journal of Public Health Acknowledgements The authors thank all the participants who collected the data. Supplementary Data Supplementary Data are available at Eurpub online. Funding European and French monitoring centre for drug and drug addiction. Conflicts of interest: The authors are state employees and do declare that they have no financial interest in the funding nor the results. Key points Despite the clear advantages of the capture recapture method compared with general population surveys for problem drug use measurement, no survey had been conducted in France since 1999. A multicentric capture recapture survey was conducted in 2006 in six big French cities located in various geographical areas in order to provide French drug use prevalence indicators for France for the European monitoring centre for drug and drug addiction. This article shows that the prevalence of problem drug use was close to 0.1% among the 15 64 years old in five cities and lower in one, in the west on France (0.76%), despite relatively wide confidence intervals. Methodological concerns are discussed. Further surveys should be conducted to monitor the problem drug use and the development of local drug policies. References 1 European Monitoring Centre for Drugs and Drug Addiction. Estimating the prevalence of problem drug use in Europe. In: Monographs, Vol. 1. Lisbon: EMCDDA, 1997. 2 United Nations Office on Drugs and Crime. GAP Toolkit Module 2 estimating prevalence: indirect methods for estimating the size of the drug problem. Vienna: United Nations Publication, 2003, 91. 3 Wittes JT, Sidel VW. A generalization of the simple capture recapture model with applications to epidemiological research. J Chron Dis 1968;21:287 301. 4 Bishop YMM, Fienberg SE, Holland PW. Estimating the size of a closed population. discrete multivariate analysis: theory and practice. Cambridge, MA: MIT Press, 1975, 227 56. 5 Fienberg SE. The multiple recapture census for closed populations and incomplete 2k contingency tables. Biometrics 1972;59:591 603. 6 European Monitoring Centre for Drugs and Drug Addiction. Methodological guidelines to estimate the prevalence of problem drug use on the local level. Lisbon: EMCDDA, 1999, 72. 7 Domingo-Salvany A, Hartnoll RL, Maguire A, et al. Use of capture recapture to estimate the prevalence of opiate addiction in Barcelona, Spain, 1989. Am J Epidemiol 1995;141:567 74. 8 Bello PY, Chêne G. A capture-recapture study to estimate the size of the addict population in toulouse, France. In: Stimson GV, Hickman M, Quirk A, et al., editors. Estimating the prevalence of problem drug use in Europe. Council of Europe; Luxembourg: EMCDDA, 1997, 93 102. 9 Chevallier E. Estimations locales de la prévalence de l usage d opiacés et cocaïne en France. Une étude multicentrique à Lens, Lille, Marseille, Nice et Toulouse [Local estimates of the prevalence of opiates and cocaine use in France. A study carried out in Lens, Lille, Marseilles, Nice and Toulouse.]. Paris: Observatoire français des drogues et des toxicomanies, 2001, 112. http://www.ofdt.fr/ BDD/publications/docs/capture.pdf (accessed October 2008). 10 Frischer M. Estimated prevalence of injecting drug use in Glasgow. Br J Addict 1992;87:235 43. 11 Hay G, McKegany N. Estimating the prevalence of drug misuse in Dundee, Scotland: an application of capture recapture methods. J Epidemiol Commun Health 1996;50:469 72. 12 Domingo-Salvany A, Hartnoll RL, Maguire A, et al. Use of capture recapture to estimate the prevalence of opiate addiction in Barcelona, Spain, 1989. Am J Epidemiol 1995;141:567 74. 13 Korf DJ, Reijneveld MD, Toet J. Estimating the number of heroin users: a review of methods and empirical findings from the Netherlands. Int J Addict 1994;29:1393 417. 14 Gfroerer JC, Larson SL, Colliver JD. Drug use patterns and trends in rural communities. J Rural Health 2007;23:10 15. 15 Mastro TD, Kitayaporn D, Weniger BG, et al. Estimating the number of HIV-infected injection drug users in Bangkok: a capture recapture method. Am J Public Health 1994;84:1094 9. 16 Hope VD, Hickman MK, Tilling K. Capturing crack cocaine use: estimating the prevalence of crack cocaine use in London using capture recapture with covariates. Addiction 2005;100:1701 8. 17 Comiskey MC, Barry JM. A capture recapture study of the prevalence and implications of opiate use in Dublin. Eur J Public Health 2001;11:198 200. 18 Buster MCA, Van Brussel GHA, Van Brink W. Estimating the number of opiate users in Amsterdam by capture recapture: the importance of case definition. Eur J Epidemiol 2001;17:935 42. 19 Böhning D, Suppawattanabodee B, Kusolvisitkul W, et al. Estimating the number of drug users in Bangkok 2001: a capture recapture approach using repeated entries in one list. Eur J Epidemiol 2004;19:1075 83. 20 Hay G. Capture recapture estimates of drug misuse in urban and non-urban settings in the north east of Scotland. Addiction 2000;95: 1795 803. 21 Gemmel I, Millar T, Hay G. Capture recapture estimates of problem drug use and the use of simulation based confidence intervals in a stratified analysis. J Epidemiol Commun Health 2004;58:758 65. 22 Hickman M, Higgins V, Hope VD, et al. Injecting drug use in Brighton, Liverpool, and London: best estimates of prevalence and coverage of public health indicators. J Epidemiol Commun Health 2004;58:766 771. 23 Platt L, Hickman M, Rhodes T, et al. The prevalence of injecting drug use in a Russian city: implications for harm reduction and coverage. Addiction 2004;99:1430 8. 24 Brugha RF, Swan AV, Hayhurst GK, et al. A drug misuser prevalence study in a rural English district. Eur J Public Health 1998;8:34 6. 25 Hook EB, Regal RR. Internal validity analysis: a method for adjusting capture recapture estimates of prevalence. Am J Epidemiol 1995; 142(9, Suppl.):S48 52. 26 Hook EB, Regal RR. Capture recapture methods in epidemiology methods and limitations. Epidemiol Rev 1995;17:243 64. 27 Hook EB, Regal RR. The value of capture-recapture methods even for apparent exhaustive surveys. Am J Epidemiol 1992;135:1060 7. 28 Hook EB, Regal RR. Validity of methods for model selection, weighting for model uncertainty, and small sample adjustment in capture recapture estimation. Am J Epidemiol 1997;145:1138 44. 29 Regal RR, Hook EB. Goodness-of-fit based confidence intervals for estimates of the size of a closed population. Stat Med 1984;3:287 91. 30 Cormack RM. Interval estimation for mark-recapture studies of closed populations. Biometrics 1992;48:567 76. 31 Beck F, Legleye S, Spilka S. Drogues. In: Guilbert P, Guautier A, editors. Baromètre santé [Health Barometer]. INPES: Saint-Denis, 2007. 32 EMCDDA. Guidelines for the Evaluation of Treatment in the Field of Problem Drug Use A Manual for Researchers and Professionals. Office for Official Publications of the European Communities: Luxembourg, 2007, 76. 33 Hay G, Gannon M. Capture recapture estimates of the local and national prevalence of problem drug use in Scotland. Int J Drug Policy 2006;17:203 10.
Capture recapture drug users in France 37 34 Calkins RF, Aktan GB. Estimation of heroin prevalence in Michigan using capture-recapture and heroin problem index methods. J Drug Issues 2000;30:187 204. 35 Tilling K. Capture-recapture methods-useful or misleading? Int J Epidemiol 2001;30:12 4. 36 Hook EB, Regal RR. Effect of variation of probability of ascertainment by sources ( variable catchability ) and upon capture-recapture estimates of prevalence. Am J Epidemiol 1993;137:1148 66. Received 30 May 2008, accepted 21 November 2008