Association of Treatment-Resistant Depression With Patient Outcomes and Health Care Resource Utilization in a Population-Wide Study

This cohort study examines data for episodes of unipolar major depressive disorder that met criteria for treatment-resistant depression to assess its impact on individuals and society and consider potential prognostic factors.


eTable 1. Stockholm MDD Cohort
The Stockholm MDD Cohort (SMC) 1 comprises all patients with MDD in the region of Stockholm between 2010 and 2018 and includes data from the following data sources:

VAL
EHR SSIA The Stockholm regional healthcare data warehouse (called VAL) includes information on all contacts with healthcare financed by the Region Stockholm. Data for primary care, secondary care and hospitalizations are available from the 1980s (inclusion of diagnoses from primary care from 2003). The International Classification of Diseases Version 10 (ICD-10) has been used since 1997. VAL also contains demographic information on patient age, sex, migration status and death. Information on prescription drugs dispensed in the ambulatory setting have been included since July 2010. The drug dispensing data come from the same source as the Swedish Prescribed Drug Register, with a population coverage of over 99%. The Anatomical Therapeutic Chemical (ATC) classification system is used to code dispensed drugs. 2 Stockholm with its 2.4 million citizens accounts for 24% of the Swedish population. TakeCare is presently the most widely used EHR system in the healthcare region of Stockholm. The EHR installation in the Stockholm Healthcare Region has over 50,000 active users representing several groups of healthcare professionals. The installation handles more than 4 million patient records and covers more than 88% of inpatient care and 75% of outpatient care. 3 The  ICD10:  I20-I25, I50, I110, I42 (excl I42.1 and I42.2), I43, I48, I60-I64, G45, I70-I72, I73 This variable indicates in which MDD episode the patients currently are in (i.e., if the variable is equal to 3 this means that the patient has had two previous depressive episodes in their disease history (data available from 1 January 1997) Healthcare level This variable presents the healthcare level where the first MDD diagnosis within the current MDD episode was recorded. In order to make this variable as simple as possible we divided this variable into two levels only: (1) Non-psychiatric care and (2) psychiatry care. The level non-psychiatric mainly comprise patients diagnosed in primary healthcare settings.

Treated in psychiatric care
This binary variable (yes/no) presents if the patient within three years before index (start of current MDD episode) has been treated in psychiatric healthcare.

Treatment with antidepressants
This binary variable (yes/no) presents if the patient within three years before index (start of current MDD episode) has been treated with antidepressants (ATC: N06A) Treatment with add-on medication This binary variable (yes/no) presents if the patient within three years before index (start of current MDD episode) has been treated with add-on medication (olanzapine, quetiapine, risperidone, aripiprazole or lithium) Treatment with ECT/rTMS This binary variable (yes/no) presents if the patient within three years before index (start of current MDD episode) has been treated with either ECT or rTMS. Treatment with psychotherapy This binary variable (yes/no) presents if the patient within three years before index (start of current MDD episode) has been treated with psychotherapy.

Psychiatric comorbid conditions Sleep disorders
This binary variable indicates if the patient within the past three years before initiation of current MDD episode either has been diagnosed for sleep disorders (ICD10:G47 or F51) or at least one filled prescription of a sedative (ATC: N05C) Anxiety This binary variable indicates if the patient within the past three years before initiation of current MDD episode either has been diagnosed for anxiety (ICD10:F41) or at least one filled prescription of an anxiolytic (ATC: N05B) Substance use This binary variable indicates if the patient within the past three years before initiation of current MDD episode either has been diagnosed with mental and behavioral disorders due to psychoactive substance use (ICD10:F10-F19

eAppendix. Statistics: prognostic model
Candidate prognostic variables for TRD in the MDD-cohort of the secondary study sample were defined and used as input in a risk prediction algorithm. The outcome of interest was TRD within one year after start of MDD. Variables were selected based on previous reports and availability in the data (eTable 4). Missingness patterns in each variable was investigated and Montgomery Åsberg Depression Rating Scale (MADRS-S) included 95.2% missing information, likely explained by the fact that: only MADRS-S recorded in a structured format within 14 days prior the start of the MDD-episode were included, structured assessment of clinical rating scales was not fully implemented in the region during the time period, and that we were only able to collect MADRS-S data covering approximately 50% of the population in Region Stockholm. We compared patients with and without records of MADRS-S and found no evidence of violations from the assumption that individuals with missing values on MADRS-S were missing at random (MAR); i.e., when conditioning on variables that explain why MADRS-S is missing, the pattern of missingness was random 21 . Thus, missing values of MADRS-S were imputed in 100 datasets using multiple imputation with chained equations, implemented in the R-package "mice". Lastly, we fitted a Cox's proportional hazard model including all candidate prognostic variables. To create a clinically useful model, we approximated the full model by using a fast backward algorithm on an ordinary least squares model in which the estimated linear predictor from the full Cox model was the outcome and all candidate variables were entered in exactly the same manner as in the full Cox model. Thus, in the first step R 2 = 1.0 by design, and by removing variables in a stepwise manner the full model could be approximated to an arbitrary level. The model was internally validated using 300 bootstrap samples. Discrimination was assessed by Harrell's c-index and calibration was assessed graphically by comparing observed event rates with the predicted risk at one and two years after index. The final model was presented as a nomogram (eFigures 3 and 4). In a sensitivity analysis, we used the same model but only including patients with MADRS-S (i.e., complete case analysis). We performed this sensitivity analysis to evaluate if the prognostic value of MADRS-S was similar in the two fitted models.

eFigure 1. Definition of MDD episode
Since we do not have clinical depression ratings scales recorded at regular timepoints for all patients, we established a definition in order to get a proxy for the length of an MDD episode (or rather, the duration of healthcare contacts related to MDD). This was operationally defined based on recorded activities related to depression: (1) For each patient we selected the patient's first recorded MDD-diagnosis (ICD10: F32-Depressive episode and F33-Recurrent depressive episode), initiating the first documented depressive episode (i.e., the index date) (2) For each patient we further analyzed the time from first recorded MDD-diagnosis to subsequently recorded events related to depression. If the time interval between the recording of depressive events was ≤365 days, the episode was categorized as ongoing while if the time interval was >365 days, the episode was categorized as closed at the date of the last recorded depressive event. As depressive events, we allowed (1) recording of depression diagnoses, (2) filled prescriptions of antidepressants (AD; ATC: N06A) and add-on medication for depression (lithium, risperidone, olanzapine, aripiprazole, and quetiapine (>100 mg)), electroconvulsive therapy (ECT), repetitive transcranial magnetic stimulation (rTMS) or treatment with psychotherapy. If the last recorded depressive event was a dispensation of either AD or add-on medication, we extended the episode with the number of dispensed tablets (i.e., a maximum of 100 days was added) (3) When an episode is categorized as closed, a subsequent recording of a depression diagnosis code would initiate the start of a new episode). eFigure 2. Matching TRD episodes to non-TRD episodes Each MDD episode that fulfill TRD criteria will be matched to a non-TRD depression episode on age, sex, socioeconomic status, and length of depressive episode. This means that a TRD episode can only be matched to a non-TRD episode having a duration that corresponds to the time when the matched TRD episode fulfills TRD criteria (i.e., the non-TRD episode must be ongoing at the time when the matched TRD-episode fulfill the TRD criteria). The matched non-TRD episode will be given the same index as the matched TRD episode. This means that if a patient enters into TRD 200 days after initiation of the depressive episode, the TRD index will be start of depressive episode + 200 days. The five matched non-TRD episodes will be given the same index i.e., start of depressive episode + 200 days. eFigure 4. Nomogram example A woman with an MDD episode initiated in psychiatric care (16 points), a MADRS-S score of 25 (41 points), previous use of sedatives (8 points) and twenty outpatient physician visits during the past 12 months (12 points) will have a total score of 16+41+8+12=77 points and a 1-year risk of fulfilling TRD criteria of approximately 3.3%. eFigure 7. Psychiatric comorbid conditions among TRD and matched non-TRD episodes eFigure 9. Treatments (on a substance level) when patients fulfill the TRD criteria ( N=12 793) This bar chart presents the antidepressant therapy at time of fulfilment of TRD criteria (i.e. initiation of third antidepressant treatment trial within the depressive episode). The bars are colored according to the following therapeutic groups: SSRI -selective serotonin reuptake inhibitors NaSSA -noradrenergic and specific serotonergic antidepressant SNRI -serotonin and norepinephrine reuptake inhibitors NDRI -norepinephrine-dopamine reuptake inhibitor ADD-ON -Add-on medication TCA -tricyclic antidepressants ECT/rTMS -electroconvulsive therapy or repetitive transcranial magnetic stimulation Agomelatine Vortioxetine Other For ATC codes (and clinical procedure codes) included in each group, see eTable 2.