Longitudinal pharmacoepidemiological and health services research for substance users in treatment: protocol of the Belgian TDI-IMA linkage

Cases were selected based on patients’ first registration in the TDI-database between 01/01/2011 and 31/12/2014, following the case definition used by Antoine and colleagues for the Belgian TDI registration (protocol 2.0) [11]: in TDI information is collected (a) on every treatment episode (b) started by a person (c) in a treatment center (d) for his or her alcohol or illicit drug use.

(a) An episode is defined as the period between the start of the treatment, which is the first face-to-face contact between a professional and the patient, and the end of activities in the context of the program prescribed. In outpatient settings this end of the episode occurs when the patient stops attending treatment for a period longer than 6 months, whereas in inpatient settings, it is defined as the moment when the patient leaves the center and no further admission is foreseen. However, data only show an image at the start of the episode and no information is available about the duration of the episode. Patients also have the right to refuse registration. If in the period 2011 to 2014 patients were admitted to treatment more than once, data from the start of the first episode were used for the current linkage. Since some patients were in treatment before 2011, this means that this first episode does not necessarily correspond to the first treatment ever, nor does it mean that there have not been any other treatment episodes afterwards. Treatment is defined as any activity targeting a person with substance use problems directly in order to obtain results in terms of reducing or eliminating these problems.

(b) The registration concerns all individuals without any restriction, with the only condition that the patient should have had a face-to-face contact with a care giver for his or her substance use problem.

(c) Activities have to take place in a treatment center, which is defined as a facility or practitioner providing treatment for drug or alcohol addiction. It can be an outpatient or inpatient service, either specialized in addiction treatment or included in larger scale facilities targeting different groups of people, and sometimes but not always recognized within a convention of authorities such as the National Institute for Health and Disability Insurance (NIHDI) [11].

(d) Drug types that are registered are opioids, cocaine/crack, stimulants other than cocaine, hypnotics and sedatives, hallucinogens, volatile inhalants, cannabis and alcohol, and their subcategories.

The TDI-database was linked to the IMA-database for the period between 01/01/2008 and 31/12/2017, based on the national identification number (NIN) of the patient. This number is unique for every Belgian citizen and for other people living in Belgium with social security rights. It is used in TDI to avoid double counting over several treatment episodes and over different treatment centers. In TDI, the NIN was encrypted for the linkage by a trusted third party, eHealth. In IMA, the NIN was encrypted twice by a trusted third party, the Crossroads Bank for Social Security (CBSS). The linkage procedure has been established in line with all relevant national privacy rules [13].

Through this linkage, pharmacoepidemiological and health service data became available for at least 3 years before the first registered episode in the TDI-database till at least 3 years after the first episode. This time span not only allows analysis of short- and long-term effects of a specific treatment approach, but also the conducting of longitudinal analyses on pathways in health seeking behavior before and after specialized alcohol and drug treatment. The chosen reference period started in 2008 as the structure of the Belgian social security system changed fundamentally at that time. For instance, since 2008 all beneficiaries of the official health insurance agencies (including people who are working as independent contractor) have access to the same package of health care, including insurance for small risks.

An added asset of the linkage is the inclusion of a group of people who were not attending specialized alcohol and drug treatment between 2008 and 2014. This group was selected from the IMA-database and allows for comparative analyses. For every patient included in the TDI-database four comparators were drawn at random from the clients of the seven Belgian health insurance companies [14]. Cases and comparators were matched on sex, age and place of residence [15]. Sex and gender were used as basic matching variables. The matching on municipality was related to both the underlying regional differences in health care regulation and health care availability (with for instance geographically different access to specialized medical health care for substance use disorders) as well as socio-economical differences between patients in the different regions of Belgium. As a result, this matching procedure allows adjusting for confounders and at the same time it created a “comparative” group that is similar to the patients in TDI who are mainly men (71.8%) with an average age of 39.8 years [16].

Five exclusion criteria were applied in the matching procedure: (1) comparators passed away in the year in which the case had a first episode in treatment, (2) comparators used medication for opiate dependence, defined by the corresponding Anatomical Therapeutic Chemical Classification System Code (ATC-code: N07BC), between 2008 and 2014, (3) comparators used medication for alcohol dependence (ATC-code: N07BB) between 2008 and 2014, (4) comparators used opioid substitution medication (based on national nomenclature codes referring to extemporaneous preparations based on methadone) between 2008 and 2014 and (5) comparators attended a health service for alcohol or drug treatment (based on national nomenclature codes referring to treatment of substance use disorders) between 2008 and 2014. If it was not possible to select four comparators, the matching criteria were expanded to a 5 year age group instead of the same age (year) and to the arrondissement (an administrative level, grouping municipalities) instead of the place of residence (municipality) of the case.

Participation of treatment centers to the TDI registration is constantly evolving over time due to obligations by the different Belgian authorities [12, 17]. During the period 2011–2014 mainly specialized treatment centers with a convention with the NIHDI, centers with an agreement from the Walloon region, centers of mental health in Flanders and a group of voluntary hospitals participated in the data collection. The participating TDI data providers can choose between two registration systems: an online registration module that allows for a case-by-case registration and a repository module that functions as a secured mailbox through which structured files can be submitted, containing a complete dataset for a given registration year [11]. At the start of every treatment episode a form has to be completed by a trained care giver of the participating treatment center during a face-to-face interview with the patient. Data are registered (1) about the treatment center (identification on the level of the center itself, on the level of a unit or program or satellite within the center, the type of program and its geographical location), (2) about the patient (socio-demographic data (sex, age and nationality) and socio-economic information (type of living accommodation, type of household, educational level, work and income situation)), and (3) about the treatment episode (the start date, history of previous treatment for substance use disorders, the main source of referral, his/her substitution treatment situation, substances used, patterns of use for the main primary substance and injecting status) [11]. After quality checks, performed in the registration module or after reception of the file from the repository module, data is stored in a database where each record corresponds to a treatment episode.

As said, the IMA database includes (1) administrative socio-demographic data, (2) information on reimbursed health services and (3) information on reimbursed medication sold in public pharmacies. For the IMA socio-demographic database the seven Belgian health insurance agencies collect data twice a year with specific demographic and socio-economic indicators [14]. This data is transferred to IMA, where one aggregated file of the seven health insurance agencies is created. It results in one file with individual socio-demographic indicators for 99% of the people living in Belgium during the reference year. In this database information is registered about age, sex, vital status (alive or passed away in the year of registration), patient’s status concerning specific social categories (widowhood, invalidity category, chronic illness, incapacity benefits, allowance for people with invalidity…), information about rights concerning maximum costs for medical care of the patient, work status and province, arrondissement and degree of urbanization of the place where the patient is living [14].

For the IMA health services database, indicators are collected for all health care related expenses which are reimbursed according to the compulsory insurance [18]. Data are collected either at the counter of the health insurance agency, either through third parties such as hospitals, groups of nurses, homes for elderly, or together with pharmaceutical data. It consists of information about the unique code of the prescriber, the unique code of the service provider, professional qualification of the prescriber/service provider, medical service by nomenclature code, subcategories of medical services, day, month and year of service delivery, type and qualification of the health institution, amount of costs that are reimbursed by the health insurance agency, amount of costs paid by the patient and supplements linked to medical services, and ATC-code for medication that is provided through hospital pharmacies [18].

The IMA pharmaceutical database consists of all the expenditures for medication which are reimbursed according to the compulsory insurance and which are not distributed through hospital pharmacies [19]. The collection is done through the accounts department of the pharmacies and delivered electronically. Information is registered about the unique code of the prescriber, the unique code of the service provider, qualification of the prescriber/service provider, day, month and year of delivery of the product, a unique national code number for every product (CNK-code), ATC-code, Defined Daily Doses (DDDs), Quantities per package (QPP), Quantities per Unit (QPU), medical service by nomenclature code, amount of costs that are reimbursed by the health insurance agency, amount of costs paid by the patient and supplements linked to medical services [19].

Descriptive statistical analysis was performed using SAS software version 9.3 (SAS Institute Inc., Cary, NC).

The data can only be used for epidemiological research, based on clearly defined study protocols, with results published on an aggregated level, in a way that patients and care givers cannot be identified. Administrative use of data is neither allowed nor possible. Careful use of terminology is required, for example, data do not reflect the situation of people who use drugs (PWUD) in Belgium, but only of people who have entered treatment for substance use disorders.

The result of the linkage is one TDI-table with data for people who were in specialized treatment between 2011 and 2014 and tables for every of the three IMA-databases covering the period 2008 till 2017 with socio-demographic data, data about the use of reimbursed health services and data about reimbursed medication sold in public pharmacies, as well as a similar set of 10 tables for the three IMA-databases with corresponding information about the comparators. Moreover, there are two sets of 4 separate tables (one for every year between 2011 and 2014) with reference numbers linking cases with comparators: one set ordered on the NIN of cases, the other set ordered on the NIN of comparators. Table 1 gives an overview of the number of records for each table. IMA’s socio-demographic data were not yet available for 2016–2017 at the time of publication, as well as the health service data and the pharmaceutical data for 2017. For 2015 and 2016 socio-demographic data, the health service data and the pharmaceutical data were not complete because insurance agencies can provide updates and corrections up to 2 years after the end of the registration period.

Data in Table 1 give an indication to what degree people in alcohol and drug treatment are using health services in general. Indeed, not every case in TDI was recorded every year in the socio-demographic tables, for example out of the 30,905 people in treatment between 2011 and 2014 only 29,956 were registered in the socio-demographic table in 2008, meaning that some people were not registered in 2008 in one of the health insurance agencies. Moreover, some patients were recorded in the socio-demographic tables, but without records in the corresponding health services or pharmaceutical tables, meaning that they were registered in the general health system but that they did not make use of it. As shown in Table 2, from which data for 2016 were omitted because of lack of completeness, 94.6% of the records in the TDI-database have corresponding records for every year in the IMA socio-demographic tables. For 1.6%, only 1 year is missing. On the other hand, 35 records (0.1%) have only 1 year for which corresponding data in the IMA socio-demographic tables is available.

71.9% of the records in the TDI-database have corresponding records for every year in the IMA health services tables and almost half of them (45.5%) have corresponding records for every year in the IMA pharmaceutical tables. This means that more than half of the patients in the TDI-database did not use prescribed medication sold in public pharmacies at least 1 year during the reference period. On the other hand, cases purchased more medication and they addressed health services more often than comparators.

By applying the aforementioned matching criteria, to each person in TDI four comparators were matched. However, the number of comparators is not exactly four times the number of cases. Indeed, as shown in Table 2 there are 30,905 people in treatment for substance use disorders, but only 122,142 unique comparators, meaning that as a result of the stringent matching criteria 1478 comparators have been linked to more than one case.