We used three administrative databases of the Province of Quebec: la Régie de l’Assurance Maladie du Québec (RAMQ), Med-Écho, and Le fichier des événements démographiques du Québec (birth and death registries) of l’Institut de la Statistique du Québec (ISQ). The RAMQ database contains information on medical services (diagnoses and procedures) received by all Quebec residents. All diagnoses are classified according to the International Classification of Diseases, Ninth Revision (ICD-9)
Although RAMQ covers all Quebec residents for the cost of physician visits, hospitalisations, and procedures, it only covers a portion of residents for the cost of medications. The RAMQ drug plan covers individuals 65 years and older, recipients of social assistance (welfare recipients), and workers and their families (adherents) who do not have access to a private drug insurance programme, accounting for approximately 43% of the overall Quebec population.15
It is also estimated that 30% of women between 15 and 45 years of age in Quebec are covered by the RAMQ drug plan for their medication (RAMQ data). The Med-Écho database is a provincial database, which records acute care hospitalisation data for all Quebec residents; it also records gestational age for planned abortions, miscarriages, and deliveries. ISQ provides demographic information on the mother, father, and baby, as well as birthweight and gestational age for live births and stillbirths.
The RAMQ and Med-Écho databases have often been used in the past for epidemiological research.16–18
The ISQ database has also been used in epidemiological studies.19
Data recorded in the medication database of the RAMQ have been suitably evaluated and found to be comprehensive and valid.20
The same was found for medical diagnoses recorded in the Med-Écho database.21
The RAMQ, Med-Écho, and ISQ databases were linked together to create the ‘Medications and Pregnancy’ cohort which contains data on all pregnancies that occurred in Quebec between 1 January 1998 and 31 December 2002. The linkage between the three databases was performed using women's ‘Numero d’Assurance Maladie’, which is the unique identifier for all Quebec residents (RAMQ and Med-Écho), and using mothers and babies dates of birth, first names, and family names (RAMQ and ISQ). This cohort is composed of women with a diagnosis or procedure code related to pregnancy.
Within the ‘Medications and Pregnancy’ cohort, women meeting the following eligibility criteria were included in this study: they had to be (1) between 15 and 45 years of age on the date of entry in the cohort defined as the first day of gestation and (2) continuously insured by the RAMQ drug plan for at least 12 months before the first day of gestation, during the pregnancy, and for at least 12 months after the end of pregnancy. The end of pregnancy was defined as the calendar date of a planned abortion, miscarriage, or delivery. If a woman had more than one pregnancy between 1998 and 2002, the first pregnancy meeting eligibility criteria was included for analysis.
The prevalence of antidepressant use during the ‘12 months before pregnancy’ was calculated by dividing the number of women receiving at least one antidepressant in this 12-month period by the total number of women in the cohort. The same calculations were performed for the time period ‘during pregnancy’ and for the time period ‘12 months after the end of pregnancy’. In addition, the prevalence of antidepressant use in the first trimester (≤14 weeks of gestational age), second trimester (>14 to ≤26 weeks of gestational age), and third trimester (>26 weeks of gestational age) of pregnancy was calculated by dividing the number of women filling at least one antidepressant prescription in the respective trimesters by the number of women in the study at that time (depending on the outcome of the pregnancy, some women were counted in the denominator only in the first or second trimester). When the duration of a prescription overlapped between trimesters, women were defined as exposed in both time periods. The prevalence of antidepressant use according to pharmacological class and type stratified by trimester was calculated in the same manner.
For each dispensed prescription of an antidepressant, the daily dosage was calculated. This daily dosage was then compared with the optimal range recommended by The Public Health Agency of Canada
Since lower dosages are often prescribed at the initiation of a treatment and thus could decrease the overall daily dosage, we used the lower range of what is considered the minimum threshold of pharmacological efficacy.22
The daily dosage for each prescription could fall into one of the following three categories: optimal dosage, underdosage, or overdosage. For a given antidepressant, the percentage of prescriptions with optimal dosage was calculated by dividing all prescriptions that were optimally prescribed by the total number of prescriptions. The same calculation was performed to determine the percentage of prescriptions with sub-dosage and overdosage. We restricted these calculations to women with at least one diagnosis of major depressive disorder (ICD-9 codes: 296.2 and 296.3)14
either during pregnancy or during the 12 months before pregnancy. This was necessary since these dosage recommendations are intended for women with major depressive disorders.
We determined the percentage of subjects who had at least one switch from one class or type of antidepressant to another in the first trimester.
Women were considered exposed to antidepressants on the first day of gestation and at the end of pregnancy if they filled a prescription or if the duration of a prescription overlapped on these days. In addition, we allowed a 7-day grace period between consecutive prescriptions of antidepressants, and thus, women were considered exposed if the first day of gestation or end of pregnancy fell during this grace period.
The following variables were considered as potential predictors of receiving at least one antidepressant on the first day of gestation and were measured on this day: maternal age, maternal place of residence (urban versus rural), maternal RAMQ drug plan status (adherent versus welfare recipient), and calendar year. The following variables were also considered as potential predictors of receiving at least one antidepressant on the first day of gestation and were measured in the year before pregnancy: number of different types of medications used other than antidepressants, number of different prescribers for all medications, planned abortions or miscarriages, number of visits to the physician, visits to the emergency department and/or hospitalisations, and diagnosis of depression (ICD-9 codes: 296.x; 300.4; 309; 311).23
Predictors of antidepressant use at the end of pregnancy were determined only for women who had delivered a baby (live birth and stillbirth) and thus excluded for women whose pregnancies resulted in planned abortions or miscarriages.
As our cohort spanned a 5-year period, we controlled for any time trend in prescribing practices by adjusting the estimates for the calendar year on the first day of gestation.
This study was approved by the Sainte-Justine Hospital Ethics Committee, and by the Commission d’Accès à l’Information du Québec, the agency granting ethics clearance for the use of linked administrative data.
Descriptive statistics were used to summarise the characteristics of the study population. McNemar's test was used to compare the prevalence of antidepressant use before versus during the first trimester of pregnancy and before versus after pregnancy. Unconditional logistic regression models were performed to identify and quantify predictors of antidepressant use on the first day and on the last day of gestation, separately. All analyses were two-tailed, and P≤ 0.05 was considered significant. SAS version 8.2 (SAS Institute Inc., Cary, NC, USA) was used to perform the analyses.