Trajectories of antidepressant medication use in individuals before and after being granted disability pension due to common mental disorders- a nationwide register-based study

All individuals aged 19–64 and living in Sweden who were granted DP due to CMD for the first time during 2009–2010 (N = 4642) comprised the study base. CMD was defined as ‘depressive episode’ (F32), ‘recurrent depressive disorder’ (F33), ‘phobic anxiety disorder’ (F40), ‘other anxiety disorder’ (F41), ‘obsessive-compulsive disorder’ (F42), and ‘reaction to severe stress and adjustment disorder’ (F43) [21‐24], based on the corresponding codes of the International Classification of Diseases (ICD) version 10 [25].

The study population was identified through register linkage at individual level using the unique ten-digit personal number of all residents in Sweden. Information was obtained from the following nationwide registers:

1) Longitudinal integration database for health insurance and labour market studies (LISA) (with information on: sex, age, education, country of birth, type of living area, family situation, and emigration) from Statistics Sweden.

2) (i) National patient register (date and main diagnosis of in- and specialized outpatient care); (ii) Drug register (information on prescribed dispensed drugs, including dates, dosages, amount, Anatomical Therapeutic Chemical (ATC) Classification System code [26], defined daily dose per dispensed pack; and (iii) Cause of death register (date of death) – all three from the National Board of Health and Welfare.

3) Micro-data for analyses of social insurance (MiDAS) (date and main disability pension diagnosis) from the National Social Insurance Agency.

Antidepressants were identified based on the respective ATC code (N06A) [26]. Levels of defined daily dose (DDD) of ADs per year during the 3 years before and 3 years after the date of being granted DP were assessed. The DDD, as defined by ‘the WHO collaborating centre for Drug Statistics Methodology’, is the tentative average maintenance daily dose for a drug, used as main indication in adults [27]. A time scale on an annual basis was introduced where t0 represented the first date of being on DP, t-1 to t-3 referred to the three respective years prior to DP, and t + 1 to t + 3 indicated the 3 years after. The total amount of DDDs for a given year was calculated by summing up the DDDs of any AD medication during that year. If more than one AD was purchased during a year, then annual DDDs of all AD were summed up together for that particular year.

There is an upper limit of how much an individual has to pay out of pocket for medication during a 12-months period. The limit was between 100 and 126 Euros during the study period [28, 29]. Individuals can buy for up to a maximum of 3 months of prescribed medication at a time.

Socio-demographics (sex, age, education, country of birth, type of living area, and family situation) were measured on 31-Dec of the year preceding DP granting and categorised as shown in Table 1. Among the medical factors, previous healthcare was measured during the 3 years before starting DP (from t-3 to t0) and categorized as ‘yes’ and ‘no’. Regarding underlying diagnoses of previous healthcare, F00-F99 (ICD-10) codes were categorized as mental and all other as somatic diagnoses. The main DP diagnosis was the main one given when granting the DP (t0). We did not consider alteration in the main DP diagnosis, if any, during the study period. Diagnoses were categorized as indicated in Table 1. Annual in- and specialized outpatient healthcare use was considered as a proxy of the severity of the underlying disorder.

All residents in Sweden aged 19–64 years, with a reduced work capacity due to disease or injury, can be granted DP from the Social Insurance Agency, for full- (100%) or part-time (75, 50, or 25% of ordinary working hours) [3]. People aged 19–29 years can be granted temporary DP if their work capacity is reduced due to disease or injury, not only for paid work but also to complete upper-secondary education, and individuals between 30 and 64 years of age can be granted permanent DP [3].

We calculated the individual annual DDDs for each specific AD for the six studied years, considering the date of granted DP; t0. Then we summed up DDDs of all specific ADs the individual bought during a one-year interval. Annual cumulative DDDs exceeding 1500 (around 4 DDDs) were deemed unusual (might be due to special cases, large purchases before traveling abroad, or error in data). So any such annual DDDs were truncated at a level of 1500 (equals 194 yearly purchases).

Thereafter, we used group-based trajectory modelling to estimate trajectories of AD among individuals with incident DP due to CMD during 2009–2010, for six time points (i.e. within a six-year window, starting from 3 years before and ending at 3 years after the date being granted DP). These models estimate changes in AD patterns over time in multiple subgroups within the cohort and estimate a regression model for each discrete group, and assess proportions of individuals in each group.

Additionally, this flexible model allows for different polynomials of the outcome [30]. We used the Bayesian information criterion (BIC) to test the best-fitted model related to the number of groups between 2 to 8, and in parallel, we also considered the proportions of the individuals in each group. While six and seven group models had better BIC values compared to the five-group model, there were very few individuals in some of the groups. Therefore, the model with five groups was considered most appropriate.

Probabilities for an individual to be assigned to a specific trajectory group were calculated. The highest estimated probability was used to decide each individual’s group belonging. Côté et al. recommend that the average probability for individuals of a trajectory group should be ≥0.70 [31]. Such average probability for individuals of our cohort was 0.89, indicating a very good fit. It should be noted that all the time points represented an interval of 1 year calculated on the basis of DP granting date (t0), but for the sake of graphical presentation, t0 was calculated as the average of t-1 and t + 1, though t0 represented a date and not an interval (Fig. 1).

We calculated descriptive statistics of the socio-demographics and medical factors prior to DP, after assessing potential sex differences in these factors by chi²-test. We then estimated their associations in each AD trajectory group by chi²-test and multinomial logistic regression. The likelihood ratio tests were used to evaluate whether socio-demographic and medical factors were associated with type of trajectory group in the full model. Moreover, Nagelkerke pseudo R² values were estimated to evaluate the strength of these associations. By consecutively excluding and re-including each factor from the full model, we calculated differences in R² for each factor in order to examine the contribution of a given factor to the full model.

Analyses were carried out for the entire study population with DP due to CMD regardless of grade (part-time/full-time) as sensitivity analyses revealed no differences in the trajectories between individuals with full-time and part-time DP.

In case of death or emigration during the study period, due to differences in the exposure time, the data on DDDs for an individual was considered as missing for the whole year of the event and onward.

Data processing was performed using statistical software SAS for Windows version 9.4 (SAS-based procedure “Traj” [32]) and SPSS for Windows version 22.0 (chi²-test and multinomial logistic regression).

In this explorative study, we identified five different trajectories of DDDs of ADs over a 6-year period among all 4642 individuals granted DP due to CMDs during 2009–2010. For the vast majority of individuals (89%), DDDs of ADs – though on different levels – varied only slightly before and after granting of a DP. Out of them about a third of the individuals, who more often had stress-related DP diagnoses and less psychiatric care, received very low levels or no ADs during the years around the time of DP grant. Two smaller groups (6% each) registersshowed increases of DDDs up to granting of the DP, in one group DDDs levelled off afterwards and in the other group they kept increasing. Individuals in this latter group tended to be younger and were more likely to have an anxiety disorder as a DP diagnosis.

To the best of our knowledge, this is the first study attempting to disentangle different groups of trajectories related to the amount of ADs during the years before and after being granted DP due to CMD. Previous studies have not discriminated between specific DP mental diagnoses and have used other statistical methods, which are not capable to identify groups with varying trajectories [18, 19].

Main strengths of this study include the use of high quality population-based nationwide registers [33‐37] with longitudinal data linked at individual level. Further strengths result from the use of a large study group which comprised all individuals aged 18–64 from whole Sweden who were granted DP due to CMD during the studied exposure years (2009–2010). This means the study is not merely based on a sample. The register data also means that the study was not affected by recall bias regarding exposure and outcome measures. Moreover, there was no loss to follow up and DP diagnoses were set by the treating and certifying physician through thorough assessments of the patients’ disease, functioning, and work incapacity. An additional advantage is that a wide range of socio-demographic and medical factors could be included as potential confounders. Another important strength of our study is that a potential social gradient for AD purchase can be considered negligible. First, drug purchase is considerably subsidised in Sweden and the level of these subsidies did not change much during study period. Moreover, analyses were controlled for socio-economic factors.

Regarding limitations, the validity of sick-leave and DP diagnoses is often discussed; however, few related studies are carried out so far. A Swedish study from 1991 concluded that sick-leave diagnoses have high validity when compared to the diagnoses from medical records [38]. Additionally, granting of DP is preceded by a long process of medical evaluation and work-capacity assessments [3]. Moreover, due to the stigma around mental diagnoses [39, 40], we assume that a mental diagnosis is given as a main DP diagnosis only when the patient actually has a mental disorder and when the main reason of work disability cannot be attributed to a somatic diagnosis [41]. On the other hand, this implies that some people with CMD might not have been given CMD as the main DP diagnosis on the DP certificate. Thus, they would not be included in this study. This can also be seen as a strength, as our cohort of individuals with DP due to CMD is more strictly defined. Another limitation is that, as in all studies using drug registers, we have no information on whether the individual actually used the dispensed AD. However, as the AD cannot be prescribed for more than 3 months at a time, it is plausible that the patient would not repeatedly have bought medication not used.

A point to be noted is, that we did not have the possibility to measure formally the severity of CMD, we have rather used information on in- or specialized outpatient healthcare use as a proxy of severity. Concerning the possibility of returning to work after DP granting, it should be noted that DP is a permanent measure for individuals aging from 30 to 64 years (around 75% of the study population). Moreover, as we have considered only the individuals who were on DP during the whole follow-up, it is unlikely for even the younger individuals (with temporal DP) to have returned to work. However, we do not have the information if some have returned to work after the follow-up period. Also, looking at the role of employers would have given further insight, but unfortunately, we do not have data on different aspects of employers.

The study revealed heterogeneity between the five groups identified based on the amounts of ADs. A third of the individuals in our study population had either no or very low annual DDDs of ADs during the 6 years of observation. This finding is surprising and may suggest that there is a possibility that these individuals might not have received optimal pharmacological treatment by ADs before being granted DP, which is in line with previous reports [16]. On the other hand, the individuals, if treated without pharmacotherapy, e.g., psychotherapy only, or other occupational therapy during some years of observation and later with ADs, might show low DDDs during the study period. Moreover, some in the cohort were prescribed other psychotropic drugs than ADs. Around 8–11% (depending upon the trajectory group) of the individuals in this cohort were prescribed anxiolytics or sedatives, separately or along with AD (data not shown). There is also a possibility of poor compliance to ADs due to their potential side effects, e.g., weight gain, decreased libido, diarrhoea, agitation etc. [42‐44], which might have led to low annual DDDs. Finally, one should also consider that individuals might have improved clinically through AD treatment, but not regarding their work capacity [45‐47].

Our analyses showed that most trajectory groups had relatively stable annual DDDs of AD around the time when granted DP. This is in contrast to findings from a Finnish project, where use of AD substantially increased before and decreased immediately after granted DP, especially among those with a mental DP diagnosis [18, 19]. These discrepancies in findings might arise from the differences in DP diagnoses, i.e. in our study the focus was on CMDs, while other previous studies included all mental disorders. One group differed considerably from the patterns of the other trajectory groups, i.e., the ‘low increasing’ group which showed a steep increase in DDDs of AD since the beginning of the study period (t-3) up to DP granting and continued to increase after that.

The study findings also indicate heterogeneity regarding socio-demographics and medical factors between the trajectory groups. Among the socio-demographic variables, age had the strongest association with trajectory groups in the full model. There were also observable differences in educational level and family situation between the trajectory groups. Differences in educational level or in the family status in the ‘low increasing’ trajectory group might have been due to the younger age distribution in that group. Regarding the level of education, individuals with higher educational level are likely to have more employment security regarding a current job, or even a greater choice of alternative jobs if they are not able to remain at the present one due to health reasons [48]. It is further not unlikely that these individuals may remain longer time at work with a later pensionary age despite a higher clinical severity and higher AD doses than their low educated peers.

Depressive disorders were the most common DP diagnosis in all the trajectory groups. However, anxiety disorders contributed equally much in the ‘low increasing’ group. This may be because many of them were young, and anxiety disorders usually have an earlier onset and they are detected earlier in life [49‐51]. Previous mental inpatient or specialized outpatient healthcare was proportionate to the amount of AD, i.e., the highest AD purchases were found in the constant ‘high group’ and the lowest in the ‘low constant’ group. It might be necessary to pay special attention to the ‘low increasing’ group because they also had a high use of previous mental healthcare. As this includes the youngest among all trajectory groups, we expected that they would less frequently be using healthcare [52, 53]. Thus, our results suggest that individuals belonging to this group might have suffered from severe mental disorders several years before granted DP, and that the disorders worsened with time leading to DP. Notably, this group also had the highest level of previous somatic inpatient care. Such high use of previous healthcare suggests that further research should also focus on somatic comorbidity. On the other hand, 20–60% of the individuals, depending on the trajectory groups, did not receive any specialized mental healthcare during 3 years prior to DP. This is not in accordance with the Swedish sick-leave guidelines [54], stating that all patients on sickness absence due to depression for more than 6 months, should be referred to a specialist. This may indicate that these individuals might have not received optimal treatment during the pre-DP years, leading to further worsening of work capacity and later resulting in exclusion from the labour market by being granted DP [16].

A large proportion of the study population (41%), who were more often granted DP due to stress-related mental disorders, did not receive psychiatric in- or specialized outpatient care during the 3 years before being granted DP. This is consistent with previous research suggesting that individuals with work disability due to stress-related mental disorders have a lower proportion of specialized health care than individuals with other common mental disorders e.g., depressive disorders [55].