UvA-DARE is a service provided by the library of the University of Amsterdam (https://dare.uva.nl)
UvA-DARE (Digital Academic Repository)
Prevalence morbidity and mortality among heroin users and methadone patients
Buster, M.C.A.
Publication date
2003
Link to publication
Citation for published version (APA):
Buster, M. C. A. (2003). Prevalence morbidity and mortality among heroin users and
methadone patients.
General rights
It is not permitted to download or to forward/distribute the text or part of it without the consent of the author(s) and/or copyright holder(s), other than for strictly personal, individual use, unless the work is under an open content license (like Creative Commons).
Disclaimer/Complaints regulations
If you believe that digital publication of certain material infringes any of your rights or (privacy) interests, please let the Library know, stating your reasons. In case of a legitimate complaint, the Library will make the material inaccessible and/or remove it from the website. Please Ask the Library: https://uba.uva.nl/en/contact, or a letter to: Library of the University of Amsterdam, Secretariat, Singel 425, 1012 WP Amsterdam, The Netherlands. You will be contacted as soon as possible.
0 0
M a r c e ll C . A . B u s t e r
DepartmentDepartment of Epidemiology, Documentation and Health Promotion £DC, MunicipalMunicipal Health Service, Amsterdam
G l e ll H . A . v a n B r u s s e l
DepartmentDepartment of Social and Mental Health, Municipal Health Service^ Amsterdam.
W i mm v a n d e n B r i n k
P M A M A H U A M P P
S
T
I
M
A
T
I
N
GG T H E
N U M B E RR O F O P I A T E
U S E R SS I N A M S T E R D A M
B YY C A P T U R E R E C A P T U R E ;
T H EE I M P O R T A N C E O F
C A S EE
D
E
F
I
N
I
T
I
O
N
EuropeanEuropean Journal Of Epidemiology 2001; 17: 935-94
A b s t r a c t t
Onee of the objectives of Amsterdam's methadone maintenance treat-mentt is maximising its coverage among problematic opiate users. In orderr to evaluate what proportion is reached, the capture-recapture methodd is conducted to estimate the prevalence of problematic opiate use.. Samples of opiate users in contact with police, hospital or treat-mentt are used. The treatment sample is limited to the low-threshold treatmentt sample treatment with minimal requirements to the clients-Basedd on differences of log likelihood ratio, Akaike's and Bayesian Informationn Criteria, log linear models are selected. The size of the populationn of problematic opiate users in 1997 is estimated to be 41300 -95% confidence interval CI: 3753-4566- Within three months 50% was registered:: 16% at the police, 2.5% at the hospital and 40% at treat-ment.. This study shows that the Amsterdam methadone treatment pro-grammess succeed in reaching a high proportion of problematic opiate users.. The estimation of the prevalence of problematic opiate users is consideredd to be valid. However, if, instead of the low-threshold treat-ment,, the total treatment sample had been used, the population of interestt and the sampled population would not match correctly, and prevalencee would have been overestimated.
I n t r o d u c t i o n n
Largee scale methadone programmes exist in Amsterdam since 1981. A treatmentt system w i t h three different treatment modalities is designedd in order to contact all heroin users; particularly those heroinn users who could not be reached with abstinence-oriented treat-mentt programmes."1213 Until now, no adequate estimation of the
num-P M A M A H U A M num-P num-P
3 44 P r e v a l e n t * o f p r o b l e m a t i c o p i a t e u s e r »
berr of opiate users is described in the scientific literature.14 The esti-mationn that is generally used is based on a two sample Capture Recapturee C/RC method in which the annual sample of opiate users arrestedd by the police and receiving methadone at the police station is comparedd with the annual sample of opiate users receiving methadone att the Municipal Health Service MHS.iS This way, the number of opiate userss was estimated as 5177 at 1997.16
However,, the C/RC method is vulnerable for biases as there are severall assumptions that can be violated. First, it is assumed that the capturee probabilities for different sources are not all dependent. Thus, iff only two sources are used they are assumed to be independent. In thiss study three samples -hospital, police and methadone
aree used and it is, therefore, possible to adjust for potential depen-denciess by using log-linear methods. Second, the population must be closed;; i.e. the same individuals must be present throughout the periodd in which samples are taken. In this study we try to minimise thee effect of violation of the closed population assumption by using a threee months sampling period. Third, all true matches and only true matchess are identified. In this study the linkage of the samples is facilitatedd because two of the three samples police and treatment-aree extracted from the same register with unique identification codes. Thee hospital sample is extracted from another register but sufficient informationn is available to match the individuals of different samples accurately.. The fourth assumption is that registration probabilities shouldd be homogeneous across all individuals in the population." This seemss less of a problem, in previous studies in Glasgow, Scotland,18 andd Barcelona, Spain,19 it was concluded that the observed hetero-geneityy in sex and age groups did not affect the results. The fifth assumptionn is of special importance if C/RC is used in epidemiological studies.. This assumption is related to case definition: there needs to bee correct matching between the population of interest and the popu-lationn appearing on the lists.110 In this study we focus on the latter assumption. .
Thee primary goal of this study is to estimate the size of the pop-ulationn of problematic opiate users and the extent to which methadonee treatment is reaching this population. Furthermore, to con-tributee to the discussion on the use of the C/RC method in epidemiol-ogy.. By repeating the C/RC method with different treatment samples wee describe the effect of incorrect matching of the population of interestt and the sampled population.
M e t h o d s s
Threee modalities of outpatient methadone treatment can be distin-guishedd in Amsterdam. The first modality is the 'tow-threshold* methadonee maintenance programme of the MHS. In this programme,
P r e v a l e n c ee o f p r o b l e m a t i c
P M A M A H U A M P P
o p i a t e e u s e r s s 3 5 5
{~J{~J All opiate users -non-problematic included-\\ ' Problematic opiate users
HiddenHidden population non-problematic opiate users
££ Hidden population problematic opiate users
(~J(~J Opiate users registered at Methadone Treatment HighHigh threshold treatment
^^ Low threshold treatment
OpiateOpiate users registered at Hospitals
^fcc Opiate users registered at Police Stations
FigureFigure J Opiate users in
A m s t e r d a m :: d i f f e r e n t s u b g r o u p s referredd to in this s t u d y
abstinencee of illegal drugs is not a p r e c o n d i t i o n to be t r e a t e d and demandss on p a r t i c i p a n t s are l i m i t e d to those c o n s i d e r e d essential for thee health status of the clients e.g. registration, X-thorax screening
andand a periodical medical check up- Several t i m e s a week, clients
receivee m e t h a d o n e a n d , if necessary, a d d i t i o n a l m e d i c a t i o n at the c l i n -icss or methadone bus. The second m o d a l i t y is a general p r a c t i t i o n e r s
CPsCPs methadone m a i n t e n a n c e p r o g r a m m e . Opiate users treated w i t h
methadonee at t h e i r GP receive methadone at the p h a r m a c i s t for 7- 14 days.133 They are e x p e c t e d to be able to c o n t r o l t h e i r a d d i c t i o n i.e. not toto sell their methadone at the black market, or swap it for
heroin-Thee MHS advises CPs to l i m i t methadone p r e s c r i p t i o n to r e l a t i v e l y sta-blee heroin users stable h o u s i n g , insurance, l i m i t e d i l l i c i t d r u g use etc. andd to refer 'difficult' p a t i e n t s to the MHS. " The t h i r d m o d a l i t y is an a b s t i n e n c e - o r i e n t e dd m e t h a d o n e r e d u c t i o n p r o g r a m m e and consists of opiatee users m o t i v a t e d to q u i t the use of o p i a t e s . Treatment at the CP orr the abstinence o r i e n t e d p r o g r a m m e is not possible for all opiate userss and is r e f e r r e d to as 'high-threshold methadone treatment'. In a d d i t i o n ,, specialised physicians and nurses of the MHS v i s i t opiate userss incarcerated at A m s t e r d a m police s t a t i o n s and opiate users a d m i t t e dd in h o s p i t a l s of A m s t e r d a m . This assistance is needed becausee c r a v i n g , w i t h d r a w a l and behavioural p r o b l e m s o f t e n c o m p l i
-P M A M A H U A M -P -P
I II P r e v a l e n t * o f p r o b l e m a t i c o p I * t u s e r s
catee hospital admissions and incarceration of opiate users.
Thee Amsterdam policy of discouragement limits treatment possi-bilitiess for heroin users without a residence permit. Opiate users born inn the Netherlands, Surinam or the Dutch Antilles further called 'Dutch
andand Surinam opiate users* generally have a residence permit and
max-imall access to treatment. Therefore this subgroup is described sepa-rately. .
Inn this study we attempt to estimate the total population of problematicc opiate users Figure / To estimate this number, the C/RC analysiss is limited to samples with problematic opiate users only. Opiatee users committing criminal offences and those with serious healthh problems those at high risk to he arrested hy the police or admittedadmitted in a hospital- are considered as problematic opiate users.
Furthermore,, opiate users who would participate in a low-threshold treatmentt programme if they demand for treatment are considered as problematicc opiate users since they would be referred to high-thres-holdd treatment programmes otherwise. The purple surface within the innerr dotted circle represents the hidden number of problematic opiatee users that we attempt to estimate.
Calculatingg the coverage of treatment, all clients of low-thres-holdd methadone treatment and an unknown proportion of high-thres-holdd methadone treatment participants are considered as problematic opiatee users. Outside the bold circle non-problematic opiate users are shown.. Part of them are stable high-threshold methadone treatment participants.. The C/RC analysis is repeated with the use of the total treatmentt sample and the high treatment sample only. In the discus-sionn section the question is raised whether the results obtained from thesee analyses represent the prevalence of the total population of opi-atee users or not.
** Data sources and matching
Threee lists are used to estimate the number of opiate users in the city off Amsterdam: a list of participants of the methadone programmes, a listt of arrested opiate users who received methadone at police sta-tions,, and a list of opiate users admitted to a hospital. The period of observationn is limited to the first three months of 1997.
Methadonee prescriptions of treatment centres and police sta-tionss are alt centrally registered at the Central Methadone Register
CMR.CMR. Within this register each individual has its own unique
identifi-cationn code independent of location of prescription. Hospital admis-sionss are registered separately.
Bothh CMR and the register of hospital admissions provide suffi-cientt information to identify individual opiate users -surname,
fore-namename initials, date of birth, country of birth and gender- initially,
P r « v * l « n c aa o f p r « b I
F M A M A H U A M P P
o p l a t ** u * 3 7 7
PP - appearing in the Police sample HH - appearing in the Hospital sample TT - appearing in the Treatment sampie NN - total
TotalTotal population
low*thresholdd treatment sample only highh threshold treatment sample only totall treatment sample
Samples Samples HH yes TT yes PP yes 8 8 1 1 9 9 HSCrf f HH yes TT yes PP no 36 6 14 4 SO O HH yes TT no PP yes 6 6 13 3 5 5 HH no TT yes PP yes 203 3 42 2 24S S Hyes s TT no PP no 52 2 74 4 38 8 HH no TT yes PP no 1078 8 1047 7 2125 5 HH no TT no PP yes 431 1 592 2 389 9 N N 1814 4 1783 3 2861 1
DutchDutch an4 Surinam population only low-thresholdd treatment sample only
high-thresholdd treatment sample only totall treatment sample
4 4 1 1 5 5 23 3 13 3 36 6 3 3 6 6 2 2 157 7 36 6 193 3 40 0 50 0 27 7 782 2 739 9 1S21 1 285 5 406 6 249 9 1294 4 12S1 1 2033 3
Taal*Taal* 1 Description of the study populationn observed during the first threee months of 1997.
datee óf birth and the first four surname characters. Careful checking off both matched and unmatched records was conducted to minimise misclassification. .
Th*Th* slz* of thm study population
Duringg the first three months of 1997, a total of 2429 opiate users participatedd in the three outpatient methadone treatment modalities: 13255 at the MHS, 198 at an abstinence oriented programme and 981 at thee GP. Sixty-five patients attended both high-threshold programmes
ll number of high-threshold patients is 1179- Seventy-five high-threshold par-ticipantss also participated in a low-threshold programme number of exclusivelyy high-threshold patients is 1104- The mean and median percentage of dayss of treatment participation during these first three months of 19977 was 70% and 89%, respectively. Moreover, 102 opiate users had aa registered hospital admission and 648 opiate users received methadonee at the police office.
Ann opiate user can either be in methadone treatment or not, be arrestedd or not and be admitted in a hospital or not. if three samples aree used, eight different combinations are possible. These combina-tionss vary from appearing in all three samples to appearing in none of thee samples. The number of people appearing in none of the samples iss unknown. This is the hidden population that can be estimated using thee C/RC method. If all samples are independent, the expected fre-quenciess of the different cells are the product of the size of the esti-matedd total population and the chances to appear, or not to appear, in
P M A M A H U A M P P
3 11 P r e v a l e n c e o f p r o b l e m a t i c o p i a t e
thee different samples during the sampling period. The number of uniquee persons appearing in the seven known celts are described in Tablee 1. A distinction is made between low-threshold, high-threshold andd total treatment samples.
AppliedApplied formula*
Twoo sample Capture-Re capture: If a is the number of Individuals appearing in bothh samples and b and c are unique individuals within each sample. Totall population: N = {(a+b)(a+c))/a.
Variance:: VAR(N) - «a+b) x (a+c) x <b) x <c»/(a)3.
InformationInformation Criteria
Aikake'sAikake's Information Criterion: A1C = Cz - 2 X
(df>-BayeSianBayeSian Information Criterion: BIC — G - (In ^observed) x Wf).
Czz - likelihood ratio dff - degrees of freedom
EstimatingEstimating the population size
Generall log linear analysis SPSS-9.0- is used to perform a three-samplee C/RC analysis."0 This method analyses the frequency counts of observationn in each cross-classification category in a cross-tabulation •i.e.•i.e. Table / A structural zero for incomplete tables is defined by a celll structure variable."3 Différent models are compared in order to findd the best fitting model. The aim of model selection is to select a modell that adequately describes the data with as few parameters as possible.. Rival models were compared using the Log Likelihood ratio G22 for models with different degrees of freedom, the Aikaike's Informationn Criterion AIC and the Bayesian information criterion as proposedd by Schwarz BlC are used for models with the same degrees off freedom."4 The simplest model with the lowest AIC, BlC score was selectedd in the absence of a more complex model with a significantly lowerr Log likelihood ratio.
Beforee conducting a three sample C/RC analysis, two sample C/RC estimationss are presented to illustrate the relations between different samples.. Formulas to perform the two sample C/RC analysis and infor-mationn criteria are shown in the Box."1""
EstimatingEstimating the coverage among problematic opiate users.
Too calculate the coverage of health services among problematic opiate users,, the percentage of opiate users covered by low-threshold methadonee treatment, police and hospitals can be deduced directly fromm the estimation of the size of problematic opiate users and its 9596
oo f pp r o b ee m a t I c P M A M A H U A M P P o p l a t ** u s « r s a t t PP * Police sample HH - Hospital sample T ** treatment sample SDSD = Standard Deviation TotalTotal population Low-threshold d High-threshold d Totall treatment ïplesïples used H/T T Estimation n 3072 2 7507 7 4199 9 SD D 343 3 1778 8 351 1 P/T T Estimation n 4069 9 166 637 6197 7 SD D 462 2 4069 9 595 5 P/H H Estimation n 4721 1 4721 1 4721 1 SD D 1159 9 1159 9 1159 9
DutchDutch and Surinam population only Low-threshold d High-threshold d Totall Treatment 2504 4 3945 5 2996 6 372 2 935 5 298 8 2694 4 9575 5 3980 0 379 9 2393 3 438 8 4490 0 4490 0 4490 0 1597 7 1597 7 1597 7
TableTable 2 Size of the population estimatedd by a two-sample capture recapturee method
CI.. However, an unknown part of the participants of high-threshold programmess is considered as problematic opiate user as shown in FigureFigure J This part contributes to the coverage of the problematic opiatee users by health services as well. Because no specific client informationn is available, the proportion of problematic opiate users amongg high-threshold treatment can only be determined indirectly. To estimatee this proportion, the proportion problematic opiate users that iss admitted In a hospital or captured by the police is estimated first. Thee proportion of high-threshold treatment participants that is admit-tedd in a hospital or captured by the police is expected to be lower. The quotientt of the two proportions is used to indicate the proportion of problematicc opiate users within high-threshold programmes.
R e s u l t s s
TwoTwo sample capture recapture estimations
Tablee 2 shows the two sample C/RC estimations. The estimations vary widelyy -from 3072till 16 637- Highest estimations are calculated if a high-thresholdd treatment sample is used. This may be attributed to a nega-tivee dependency. High-threshold treatment clients probably are at lowerr risk to be arrested by the police or to be admitted in a hospital thann other opiate users. In this case the overlap is small and the esti-matedd size of thé population is probably too large. Lowest estimations aree calculated if treatment and hospital samples are compared. This mayy be attributed to a positive dependency. Foreign opiate users cann only receive methadone treatment if they are sex workers
«
P M A M A H U A M P P P r e v a l e n c ee o f r o b I * « t i c c c a r s s PP - Police sample HH - Hospital sample TT - Treatment sample Model Model HT.TP.HP P HTtt TP TP,, HP HT,, HP HT T TP P HP P --NNobsobs " number of registered opiate users CC11
- likelihood ratio
AICAIC — Akalke's Information Criterion
BICBIC - Bayeslan Information Criterion proposed by Sehworz NN « estimated size of the total population
Bold;Bold; best fitting model
SamplesSamples used DfDf Cz 00 0 . 0 0 11 0 . 1 7 11 3 . 6 4 11 1.28 22 1 . 6 2 22 5.42 22 6 . 0 0 33 6 . 5 5 Low-thresholdd treatment AIC AIC 0 . 0 0 0 - 1 . 8 3 3 1.64 4 - 0 . 7 2 2 - 2 . 3 8 8 1.42 2 2 . 0 0 0 0.55 5 Police, , Hospital l HHQbQbs-s- 1*1* BIC BIC 0.00 0 - 7 . 3 3 3 6 6 2 2 - 1 3 . 4 4 - 9 . 5 9 9 -9.01 1 0 0 N N 6115 5 5 5 4 9 9 3371 1 4 1 0 3 3 4 1 3 0 0 3 5 9 4 4 3 9 4 8 8 3 9 7 6 6 CC2 2 0.00 0 0.26 6 0.01 1 0 . 8 6 6 15.2 2 1 . 7 2 2 9.43 3 19.6 6 High-thresholdd treatment Police,, Hospital NNobsobs - 1783
AIC AIC BIC BIC N N
Totall treatment Police,, Hospital
NNobsobs=2861 =2861
AIC AIC BIC BIC N N
0.00 0 - 1 . 7 4 4 - 1 . 9 9 9 - 1 . 1 4 4 11.2 2 2 . 2 8 8 5.43 3 13.6 6 0.00 0 -7.23 3 - 7 . 4 8 8 - 6 . 6 2 2 0 . 2 3 3 - 1 3 . 3 3 - 5 . 5 4 4 - 2 . 8 8 8 100 120 5 1 5 3 3 7 3 1 7 7 166 5 4 0 133 781 6 2 2 6 6 144 2S6 122 511 0 . 0 0 0 1.33 3 3 . 6 7 7 0.28 8 1 . 3 9 9 3.71 1 13.4 4 14.1 1 0 . 0 0 0 - 0 . 6 7 7 1.67 7 - 1 . 7 2 2 - 2 . 6 1 1 9 9 9.42 2 8.13 3 0.00 0 -6.63 3 - 4 . 2 9 9 - 7 . 6 8 8 1 4 . 5 5 - 1 2 . 2 2 - 2 . 5 0 0 - 9 . 7 5 5 7302 2 5 8 1 7 7 4 4 7 6 6 6 2 3 5 5 6 1 8 S S 4 4 9 9 9 5881 1 5 8 5 0 0
TableTable 3 Using different treatment
sampless first three months of
1997-InIn a three sample capture recapture methodd leads to different estimations off the number of opiate users
publicpublic health reasons' or if they have severe illnesses humanitarian reasons-reasons- Therefore this positive dependency will be more outspoken iff the total population of opiate users 'including foreigners- is
esti-mated. .
Inn case the number of people appearing in the overlap cell is low,, standard deviations will be large. As a consequence, largest stan-dardd deviations are found if the hospital and police samples are com-paredd and if the high-threshold treatment sample is compared to the hospitall or police sample.
Three-sampleThree-sample estimation of the population of opiate uxors.
Tablee 3 shows the results of the three sample C/RC analyses of the totall population. If the low-threshold treatment sample is used, the estimatedd size of problematic opiate users is 4130 -95% CI:
3753-4566-PP r o ff p r e b P M A M A H U A M f * * t i c o p i a t e « g o r s 4 1 1 PP = Pof/ce sample HH - Hospital sample TT <- Treatment sample Model Model HT.TP.HP P HT,, TP TP,, HP HTtt HP HT T TP P HP P
--NobsNobs " number of registered opiate users CC11 -= likelihood ratio
AICAIC - Akafke's Information Criterion
etcetc - Bayesian information Criterion proposed by Schwarz NN = estimated Size o, 'the'the total population
Bold;Bold; best fitting model
SamplesSamples used DfDf C2 00 0.00 11 0.17 11 1 .33 11 1.28 22 1.61 22 Ï.SS 22 1.37 33 1 . 7 3 Low-thresholdd t r e a t m e n t AIC AIC 0.00 0 -1.83 3 -0.67 7 -0.72 2 -2.39 9 -2.4S S -2.63 3 7 7 Police, , Hospital l Nobs*Nobs* 1814 BIC BIC 0.00 0 -7.00 0 -5.84 4 -5.88 8 -12.7 7 -12.8 8 -13.0 0 - 1 9 . 8 8 N N 3 7 6 4 4 4 1 4 0 0 2851 1 2 8 0 7 7 2 8 2 5 5 2 9 2 3 3 2 7 8 9 9 2 8 0 7 7 C2 2 0.00 0 0.17 7 0.29 9 0.17 7 3.25 5 0 . 3 4 4 8.15 5 10.15 5 High-thresholdd treatment Police,, Hospital NNotot,,ss-- 1783
AIC AIC BIC BIC N N AIC AIC
Totall treatment Police,, Hospital NN0bs0bs = 286J BIC BIC N N 0.00 0 -1.83 3 -1.71 1 -1.83 3 -0.75 5 3 . 6 6 6 4.15 5 4.15 5 0.00 0 -6.96 6 -6.84 4 -6.96 6 -11.01 1 - 1 3 . 9 3 3 -6.11 1 -11.24 4 7206 6 4 6 3 4 4 4 0 9 3 3 9585 5 8 8 2 4 4 4 2 0 3 3 8Q79 9 7 6 7 9 9 0 0 0.34 4 2.17 7 0.55 5 0 . 6 6 6 3.18 8 5.95 5 6.15 5 0 0 -1.66 6 0.17 7 -1.45 5 - 3 . 3 4 4 -0.82 2 1.95 5 0.15 5 0.00 0 -7.28 8 -5.45 5 -7.07 7 - 1 4 . 5 7 7 -12.05 5 -9.28 8 -16.70 0 5285 5 5394 4 3 1 7 4 4 3 9 9 5 5 4 0 0 5 5 3 2 6 6 6 3 8 4 0 0 3 8 5 0 0
TableTable 4 Estimation of the population off opiate users limited to the Dutch and Surinamm population by a three sample capturee recapture method with various treatmentt samples -first three months
ofof
1997-Iff the high-threshold treatment sample is applied, the model with adjustmentt for the negative dependency between the treatment and policee sample is selected lower part Table 3- This leads to an esti-matedd number of 6226 -95% CI 4647-8688- Similar to the two sample estimationss the CI is large because of small overlaps between samples. Duee to a sparse sensitive cell of individuals appearing within
high-thresholdthreshold treatment and police and hospital sample- both the estima-tionn as confidence intervals are inflated if the saturated model is used. Iff the total treatment sample is used, the selected model is similar to thee model with the low-threshold treatment sample. The estimated numberr of opiate users, however, increases to 6185 -95% ci: 5697-6766-Thee size of the population is comparable to that of the high-threshold treatment.. If the sample population is limited to the Dutch and Surinamm and the low-threshold treatment sample is used, the crude
4 2 2
P M A M A H U A M P P
P r e v a l e n t ** « f PP r e b ee m a t oo p I a t e u s e r s s
'Number:'Number: proportion N3
Percentage:Percentage: (N1/N2) / (N3/N4). TotalTotal population
Numberr %
Sample Sample
Nll total no. problematic opiate users N22 hospital or police
N33 high-threshold treatment N44 W2 - N3
Problematicc opiate users within N3*
DutchDutch anil Surinam only
Numberr % 4130 0 736 6 1104 4 57 7 320 0 100% % 1 8 %% 16%-20% 100% % 5% % 2 9 %% 26%-32% 2807 7 512 2 789 9 50 0 274 4 100% % 1 8 %% 17%-20» 100% % 6% % 3 5 %% 32% -38%
TableTable S Estimation of problematic opiatee users among the
high-thresholdd treatment sample
•first•first three months of
)997-modell is selected and the estimated number of opiate users is 2807 %% Cl: 2568-3048- Table 4- If high-threshold and total treatment sampless are used, selected models are equal to those of the total pop-ulationn presented in Table 3. Again, the estimated sizes are compara-blee to each other but significantly higher than the estimation with the low-thresholdd treatment sample. A C/RC analysis with the total treat-mentt sample results in an estimated number of 4005 -95% Ci:
3653-4425-ProblemattcProblemattc opiate users In high- threshold samples Ass shown in Figure 1, part of the high threshold treatment sample belongss to the problematic opiate users. Table 5 shows that the per-centagee of high-threshold treatment participants has been deduced indirectly.. The 18% 95% CI: 16-20% opiate users appearing in the hospi-tall or police sample among the estimated population of problematic opiatee users is compared to the 596 observed among the high-thres-holdd treatment sample. This results in an estimated percentage of problematicc opiate users among the high-threshold sample of 5/188 x 10096 = 29% 95% CI: 26-32% A similar calculation among Dutch andd Surinam opiate users only, leads to an estimated percentage of problematicc opiate users among high-threshold treatment participants off 35% 95% CI:
32-38%-CoverageCoverage of the population
Dataa of the coverage of the problematic opiate users are presented in Tablee 6. Among problematic opiate users, the percentage that appearedd in the police sample within three months is 16%, the per-centagee that appeared in the hospital sample was 2.5% and the
PP r * v a o ff p r o It I « t i c c
P M A M A H U A M P P
o p i a t ee u s e r s 4 3 3
a.a. 1814 table 1 + 320 table 5 - 57 (already captured atat police, hospital)
b:b: 1294 + (274 -SO).
Location Location
Polite e
Hospital l Treatment t
Coveredd at any location Hiddenn population
Totall number problematic opiate users
Number r 648 8 102 2 1645 5 2077* * 2054 4 4 1 3 0 0 TotalTotal population Coveredd % 1 6 ** 14-17% 2.5%% 2.2-2.7% 40366 36-44% 50366 45-55% 50%% 45-55% 3753-4566 3753-4566
DutchDutch and Surinam only
Number r 4 4 9 9 70 0 1240 0 151Sb b 1288 8 2807 7 2568 2568 Coveredd % 16366 15-17% 2.5%% 2.3-2.7% 44%% 40-48% S4%% 50-58% 46%% 42-50% -3048 -3048
TableTable 6 Coverage of the population off problematic opiate users by health servicess first three months óf 1997 •
percentagee appearing in methadone treatment is estimated to be 40%. Withinn three months, 50% of the problematic opiate users comes into contactt with health services. If the population is limited to Dutch and Surinamm opiate users, coverage by methadone treatment is higher 44% andd total coverage increases to 54%.
D i s c u s s i o n n
Too estimate the size of the population of problematic opiate users in Amsterdamm a C/RC analysis was performed using three samples off opiate users, all registered during the first three months of 1997: opiatee users participating in low-threshold methadone treatment, arrestedd opiate users who received methadone at a police office and opiatee users admitted in a hospital. The prevalence was estimated as 41300 -95% Cl:3753-4566- A repeated analysts using the total methadone treatmentt sample both high and low-threshold' resulted in an esti-matedd number of 6185 -95% Cl: 5697-6766- opiate users. Restriction of thee treatment sample to high-threshold methadone treatment, result-edd in a comparable estimation but larger confidence interval.
Howw should this difference be interpreted? If we limit the sam-pless to the low-threshold treatment, police and hospital sample, all sampledd opiate users match correctly with the population of problem-aticc opiate users that we intend to estimate i.e. the fifth assumption
mentionedmentioned in the introduction- Hospitalised opiate users will have medicall problems, arrested opiate users judicial problems, Opiate
P M A M A H U A M P P
44 'S P r e v a l e n c e a f p r o b l e m a t i c o p i a t e u s e r s
userss treated at the low-threshold methadone treatment of the MHS are generallyy unable to control their addiction. Otherwise, they would be referredd to high-threshold methadone treatment and receive treat-mentt at their CP or abstinence oriented treatment centre.
However,, if the treatment sample is enlarged with opiate users participatingg in high-threshold treatment, the sample will partly con-sistt of non-problematic methadone cfients. We could argue that the populationn we intend to estimate is extended to all opiate users and thereforee meet the fifth assumption too see Figure 1 However, the populationn of non-problematic opiate users consists of two different populations;; problematic heroin users outside treatment and non-problematicc methadone clients inside treatment, in case non-problem-aticc methadone clients have always used heroin without additional problemss this assumption may have been met. However, many of thesee clients will not use heroin anymore and may be socially sta-bilisedd due to methadone treatment. Probably, they gradually turned fromm problematic heroin users into non-problematic methadone clients.. If methadone treatment is successfut, the proportion of non-problematicc methadone clients will be high. As a consequence, if a C/RC-methodd would be applied, the number of non-problematic heroin userss outside treatment appears to be higher. Most probably, this is an artefact.. Although all sampled individuals are opiate users, the sam-pless do not match correctly with the population of interest. Therefore, byy using the C/RC method with the total treatment or high-threshold treatmentt samples, it is impossible to estimate the size of the total
•problematic•problematic and non-problematic- population of opiate users and we havee to limit our pronouncements to the size of the population of
problematicc opiate users only.
Ann earlier estimation of the size of the population resulted in a higherr estimation than this study 5177. This estimation was based on thee annual police and the annual low-threshold treatment samples of
19977 and can be considered as an overestimation. This overestimation isis mainly due to the violation of the closed-population assumption; a similarr two sample C/RC estimation with three months samples result-edd in an estimated size of 4069 opiate users Table 2- The estimation givenn in this article reflects the size of the Amsterdam population of problematicc opiate users at risk to be captured by the police, admit-tedd in a hospital or low-threshold• methadone treatment. An unknownn proportion of problematic opiate users residing in prison duringg the sample period has not been taken into account.
Thee validity of the estimations depends on the success on meet-ingg the other assumptions. We were able to adjust for some depend-enciess by performing a three sample C/RC. The fact that all registers aree constructed at the MHS facilitated matching, but may imply the risk off a positive dependency between all samples. However, hospitals and
P M A M A H U A M P P
P r e v a l e n c ee o f p r o b l e m a t i c o p l a t » e s e r s 4 1
policee stations take the initiative to ask the MHS for assistance when ann opiate user is admitted or arrested. The short sampling period pre-ventedd a serious violation of the closed population assumption. This isis of special importance to estimate the population of opiate users includingg the foreign opiate users that show a higher mobility. Moreover,, the small sampling period prevents a cumulating number of wronglyy registered persons that can not be matched. The fourth assumption,, 'registration probabilities of the samples should be homo'
geneousgeneous for all individuals' appears to be violated. Male problematic opiatee users for example» are at higher risk of being arrested, older opiatee users are at higher risk of being admitted in a hospital. However,, as mentioned earlier, this kind of heterogeneity does not seemm to lead to biased results.1819 In epidemiology, homogeneity refers too the correct matching of the samples and the population of interest
mentioned earlier as the fifth assumption In this study all sampled personss are 'drawn' from the referent population of problematic opiate users.. If, for example, people who only used cannabis had been includedd in the hospital sample, the samples would be heterogeneous inn the sense that some individuals belong to a different referent popu-lation.. Because the other samples are limited to those who receive methadone,, the cannabis users will neither appear in the treatment samplee -even if they demand for treatment- nor in the police sample
eveneven if they are arrested- As a result of this kind of heterogeneity, prevalencee estimations will be severely biased.
Similarly,, estimated prevalence increases if a C/RC analysis is performedd with the total treatment sample. This increase is consider-edd to be an artefact and not to refer to a broader target population. Thiss is of major importance when a C/RC analysts is conducted in regionss with one treatment centre or high-threshold treatment only. In thatt case additional information is needed to distinguish problematic opiatee users and stabilised non-problematic opiate users in order to havee a correct match between the sampled individuals and the popu-lationn of interest.
Thee C/RC analysis is an attractive method because existing reg-isterss can be used. We had the opportunity to perform and compare variouss analyses. We conclude that with the registers we used it is not possiblee to estimate the total population of opiate users and that we havee to limit our prevalence estimation to the problematic opiate users.. This population, however, is the target population of a public healthh related measures concerning opiate users. To answer the ques-tionn to what extent health services are reaching problematic opiate userss we have to limit our pronouncements to the problematic opiate userss as well. Among this population, 2.5% appeared in the hospital sample,, 16% in the police sample and 32% is treated at the low-thres-holdd methadone treatment of the MHS within three months. Besides, an
P M A M A H U A M P P
4 11 P r e v a l e n c e o f p r o b l e m a t i c o p i a t e u s a r
unknownn percentage of high-threshold treatment clients is assumed to bee a problematic opiate user. Based on the prevalence of hospital admissionss and police arrests among these clients we estimated that 29%29% of them are problematic opiate users. Including these clients, f o r t yy percent of the problematic opiate users participates in methadonee treatment, within three months. Coverage among the Surinamm and Dutch population was higher: 4496 -95% C!: 40-48%- partici-patess in methadone treatment. This may be explained by the policy of discouragementt of Amsterdam; in order to prevent a cumulating num-berr of foreign heroin users in Amsterdam, treatment possibilities for heroinn users without a residence permit are limited. Although the esti-matedd coverage is based on a three months period, it is high com-paredd to other European or American cities.I»!»"4!17 Therefore, we can concludee that the Amsterdam treatment programmes have succeeded inn reaching a large proportion of the population of problematic opiate userss of Amsterdam.
Acknowledgement Acknowledgement
TheThe work described in this publication has benefited from thethe European Network to Develop Policy Relevant Models andand Socio-Economic Analyses of Drug Use, Consequences andand Interventions, funded by TSER/DG12 -project ERB 41414141 PL 980030- and co-ordinated by the EMCDDA, Lisbon. Lisbon.
P M A M A H U A M P P t i cc o p i a t e
R E F E R E N C E S S
|11 Plomp HN, Van der Hek H, & Ader HJ. TheThe Amsterdam Methadone
DispensingDispensing Circuit: genesis and effectivenesseffectiveness of a public health modelmodel for local drug policy.
Addictionn 1996; 91: 71 1-21.
|22 Langendam MW, van Haastrecht HJA, vann Brussel CHA, Reurs H, van den Hoekk JAR, et al. Differentiation in thethe Amsterdam methadone
dispensingdispensing circuit: determinants ofof methadone dosage and site of
methadonemethadone prescription. Addiction 1998;; 93: 61-72.
|33 van Brussel GHA. Methadone
treatmenttreatment in Amsterdam: the criticalcritical role of general
practitioners.practitioners. Addiction Research 1996;; 3: 363-68.
|44 European Monitoring Centre for Drugs andd Drug Addiction. Report
CT.97.EP.05.CT.97.EP.05. Scientific Review of literatureliterature on estimating the prevalenceprevalence of Drug Misuse on the locallocal level. EMCDDA, Lisbon 1999.
|88 Frischer M, Leyland A, Cormack R, Goldbergg DJ, et al. Estimating the
populationpopulation prevalence of injecting drugdrug use and infection with
humanhuman immunodeficiency virus amongamong injection drug users in Glasgow,Glasgow, Scotland. American Journall of Epidemiology 1993;
138:138: 170-81.
|99 Domingo Salvany A, Hartnoll R, Maguiree A, Brugal T, Albertin P, Cayla JA,, Cassabona J, Suelves JM.
AnayticalAnaytical considerations in the useuse of capture-recapture to
estimateestimate prevalence: case studies ofof the estimation of opiate use in thethe metropolitan area of
Barcelona.Barcelona. Spain. American Journal off Epidemiology 1998; 148: 732-40.
100 Larson A, Bammer C. Why? Who?
How?How? Estimating numbers of illicit drugdrug users: lessons from a case studystudy from the Australian Capital Territory.Territory. Australian and New Zealandd Journal of Public Health
1996;; 20: 493-99.
|ll 5 Woodward JA, Retka RL. Ng L.
ConstructConstruct validity of heroin abuse estimators.estimators. International Journal of thee Addictions 1984; 10:93-111.
| I 66 Hook EB, Regal RR. Validity of methodsmethods for model selection, weightingweighting for model uncertainty andand small sample adjustment in capturecapture recapture estimation. Americann Journal of Epidemiology
1997;; 145: 1138-44.
| I 77 National Consensus Development Panell on effective Medical Treatment off Opiate Addiction. Effective MedicalMedical Treatment of Opiate Addiction.Addiction. JAMA 1998; 280:280: 1936-43.
|55 Korf DJ, Reijneveld SA, Toet J.
EstimatingEstimating the number of heroin users,users, a review of methods and empirical-findingsempirical-findings from The Netherlands.Netherlands. International Journal of thee Addictions 1994; 29: 1393-417.
|66 van Brussel CHA, Buster MCA. Care forfor the future, opiate addicts in Amsterdam,Amsterdam, trends and figures 1996-1998.1996-1998. GG&CD Amsterdam, Amsterdamm 1999.
\7\7 International Working Group for
Diseasee Monitoring and Forecasting.
Capture-recaptureCapture-recapture and Multiple-RecordRecord Systems estimation I:
HistoryHistory and theoretical
development.development. American Journal of Epidemiologyy 1995; 142: 1047-58.
111 1 Buning E, van Brussel GHA.
TheThe effects of harm reduction in Amsterdam.Amsterdam. European Addiction Researchh 1995; / : 92 98.
|122 European Monitoring Centre for Drugs andd Drug Addiction. Methodological GuidelinesGuidelines to Estimate the PrevalencePrevalence of Problem Drug Use onon the Local Level. EMCDDA. Lisbon
111 3 SPSS-9.0, advanced statistics.
Hickmann M, Cox S, Harvey J, Howes S, FarellFarell M, Frischer M, Stimson G, Taylorr C, Tilling K. Estimating the
prevalenceprevalence of problem drug use in innerinner London: a discussion of
threethree capture-, recapture studies. Addictionn 1999; / / : 1653-62.
