Figure 1. Key methodological features of the TDI and SGP surveillance studies
Settings and study participants
The TDI and the SGP surveillance systems of provider-reported, care-based data are managed within Sciensano, the Belgian Institute for Health.
The Belgian Network of SGP consists of a sample of GPs who report the occurrence and characteristics of well-defined health-related events in their daily practice [7]. Data are reported weekly on standard forms for a period of at least one year. Across the study period, the network comprised 125 practices. Annual statistics showed that sentinel GPs are comparable to non-sentinel GPs for age and gender. The network covers about 1,5% of the Belgian population in most Belgian districts [8]. As Belgian GPs do not serve a defined practice population, the size of the SGP patient population - the denominator - is estimated by applying the ratio of patient contacts across the entire Belgian population to the sum of weekly patient contacts in the network.
The Belgian TDI study protocol has recently been described in detail [2]. The TDI register collects socio-demographic, treatment- and substance-related data about patients who started treatment for substance use disorders in a wide range of settings. Patients are interviewed by health professionals by means of a structured questionnaire. In order to detect multiple treatment episodes in the same patients, national identification numbers are used in accordance with the European General Data Protection Regulation. Data are gathered by Sciensano using a reporting module or a repository tool allowing batch transfer of data.
Participation in the TDI surveillance is only mandatory for particular types of treatment centres for substance use problems. Throughout the study period, 221 treatment centres reported cases. TDI data from 2014 show that the best participation rates were reached by centres that are specialised in substance use problems (56% to 100% of eligible centres) [2]. Participation rates were lower among hospitals and centres that also offer treatment for other mental health problems (17% to 52% of eligible centres).
In contrast to the SGP surveillance, the TDI surveillance is unable to estimate the incidence of treatment episodes for substance use problems in the total population for lack of data about the size of the denominator, that is the population covered by the register.
Patient population samples
To address our first research question, we selected data about the TDI patients that were referred or encouraged to seek specialist treatment by caregivers. To address our second research question, we included data from all the SGP patients.
Case definitions
The SGP instructions were based on the Belgian TDI protocol [2]. A new treatment episode starts with the first face-to-face contact with a GP/other caregiver for substance use problems. When the patient shows up with a similar treatment demand six months after the previous face-to-face contact, a new treatment episode starts. Treatment was defined as any activity directly targeting patients with substance use problems in order to ameliorate their mental, medical or social status. We explicitly described possible GP interventions aimed at reducing substance-related harm in active users, detoxification or abstinence, medical and non-medical problems, informal advice, counselling and support (e.g. a brief intervention). Excluded were interventions only targeting the physical consequences of substance use (e.g. infections or overdoses) or focusing mainly on problems other than substance use.
Variables
The variables in this paper (all described in Table 1) are (derived from) the items 1-6, 9-12, 14-15 and 17 of the TDI protocol 3.0 [1]. In the context of employment, accommodation and use of primary/only substance, ‘recent’ was understood as the last 30 days before the start of the new treatment episode. Patients who had recently been living at different places (friends’ home, street, shelters, etc.) or moved from one place to another, were considered as residing in unstable accommodation. Four variables were not recorded by the SGP, respectively treatment centre type, source of referral, highest educational level completed and recent accommodation. The variable ‘primary drug’ was reported in less detail by the SGP, e.g. the groups ‘cocaine or crack’ and ‘cannabis’ comprise three subcategories in the TDI. One additional variable was reported by the SGP, i.e. whether or not the patient was concurrently receiving specialist treatment for substance use problems.
Table 1 Characteristics of new treatment episodes of substance use problems by data source: the TDI subpopulation referred/motivated by caregivers and the SGP population, Belgium 2016-7
|
TDI subpopulation of patients referred by caregivers (N=16,543)
|
SGP population (N=314)
|
|
n/valid N
|
%
|
n/valid N
|
%
|
Sex
|
|
|
|
|
Man
|
11,476/16,543
|
69.5
|
220/314
|
70.1
|
Age
|
|
|
|
|
<20
|
674/16,543
|
4.1
|
14/312
|
4.5
|
20-29
|
2,982/16,543
|
18.1
|
35/312
|
11.2
|
30-39
|
4,518/16,543
|
27.4
|
67/312
|
21.5
|
40+
|
8,327/16,543
|
50.5
|
196/312
|
62.8
|
Highest educational level completed
|
|
|
|
|
None or primary
|
3,865/13,877
|
27.9
|
|
|
Secondary
|
7,519/13,877
|
54.2
|
|
|
Tertiary
|
2,493/13,877
|
18.0
|
|
|
Recent stable accommodation
|
13,075/16,219
|
80.6
|
|
|
Recently employed
|
3,394/14,989
|
22.6
|
135/292
|
46.2
|
Region
|
|
|
|
|
Flanders
|
9,936/16,543
|
60.0
|
198/314
|
63.1
|
Wallonia
|
4,495/16,543
|
27.1
|
82/314
|
26.1
|
Brussels
|
2,112/16,543
|
12.8
|
34/314
|
10.8
|
Previous treatment
|
10,952/16,039
|
68.3
|
178/282
|
63.1
|
Type of substance use
|
|
|
|
|
Alcohol only (I)
|
7,354/16,543
|
44.5
|
176/314
|
56.1
|
Pharmaceuticals (and alcohol) (II)
|
1,047/16,543
|
6.3
|
46/314
|
14.7
|
Cannabis only/primarily (III-a)
|
2,077/16,543
|
12.6
|
30/314
|
9.6
|
Street drugs minus cannabis primarily (III-b)
|
6,065/16,543
|
36.7
|
62/314
|
19.8
|
Mono-substance use
|
10,427/16,543
|
63.0
|
254/314
|
80.9
|
Recent use of primary/only substance
|
|
|
|
|
No use in last 30 days
|
2,193/15,576
|
14.1
|
12/258
|
4.7
|
≤ 1 day a week
|
923/15,576
|
5.9
|
8/258
|
3.1
|
2-3 days a week
|
1,386/15,576
|
8.9
|
16/258
|
6.2
|
4-6 days/week
|
1,775/15,576
|
11.4
|
22/258
|
8.5
|
Daily
|
9,299/15,576
|
59.7
|
200/258
|
77.5
|
Type of treatment
|
|
|
|
Outpatient treatment
|
5,254/16,543
|
31.8
|
Inpatient treatment:
|
|
|
Inpatient, non-hospital
|
2,064/16,543
|
12.5
|
Psychiatric hospital
|
3,897/16,543
|
23.6
|
General hospital (psychiatric service)
|
5,220/16,543
|
31.6
|
Treatment for criminal law offenders
|
108/16,543
|
0.7
|
Source of referral
|
|
|
GP
|
4,515/16,543
|
27.3
|
Care services for substance use problems
|
2,349/16,543
|
14.2
|
Hospital
|
5,277/16,543
|
31.9
|
Medical-psycho-social services
|
4,402/16,543
|
26.6
|
We summarized the type of substance use into mutually exclusive groups to fit the observed use across settings. Group I covers the use of alcohol only. Group II spans the use of pharmaceuticals, i.e. hypnotics, sedatives or pharmaceutical opioids, i.e. mainly opioid analgesics. Group III encompasses the use of street drugs, i.e. opiates, cocaine, stimulants other than cocaine, cannabis, hallucinogens and volatile inhalants. Group III was divided into two groups with group III-a spanning the use of cannabis only or primarily. Group III-b contains any other use of street drugs but no primary cannabis use and is further described as ‘street drugs minus cannabis primarily’. The classification of the three groups is hierarchical in the sense that the use of pharmaceuticals (group II) may be combined with alcohol (group I), while the use of street drugs (group III) may be combined with alcohol (group I) and pharmaceuticals (group II). Methadone, buprenorphine and fentanyl were classified in group III.
Analysis
All data are episode-based. We used 95% proportion confidence intervals (CI) to describe patient population characteristics and bivariate associations. We used stepwise backward multiple logistic regression analysis to examine the research questions. Patient population characteristics that were significantly (p<0.05) associated with the dependent variables were included in the full models. We accounted for clustering of patients within practices or treatment centres by using robust standard errors. Interaction effects between independent variables were tested only in the multivariable logistic model examining the second research question. Data were analysed with Stata 15.