Suicide‐specific mortality among patients with treatment‐resistant major depressive disorder, major depressive disorder with prior suicidal ideation or suicide attempts, or major depressive disorder alone

Abstract Background The impact of treatment‐resistant depression (TRD) or prior suicidal ideation/suicide attempt (SI/SA) on mortality by suicide among patients with major depressive disorder (MDD) is not well known. This retrospective, observational, descriptive cohort study characterized real‐world rates of suicide‐specific mortality among patients with MDD with or without TRD or SI/SA. Methods Adult patients with MDD among commercially insured and Medicare enrollees in Optum Research Database were included and assigned to three cohorts: those with treatment‐resistant MDD (TRD), those with MDD and SI/SA (MDD+SI/SA), and those with MDD without TRD or SI/SA (MDD alone). Suicide‐specific mortality was obtained from the National Death Index. The effects of demographic characteristics and SI/SA in the year prior to the end of observation on suicide‐specific mortality were assessed. Results For the 139,753 TRD, 85,602 MDD+SI/SA, and 572,098 MDD alone cohort patients, mean age ranged from 55 to 59 years and the majority were female. At baseline, anxiety disorders were present in 53.92%, 44.11%, and 21.72% of patients with TRD, MDD+SI/SA, and MDD alone, respectively. Suicide‐mortality rates in the three cohorts were 0.14/100 person‐years for TRD, 0.27/100 person‐years for MDD+SI/SA, and 0.04/100 person‐years for MDD alone. SI/SA during the year prior to the end of observation, younger age, and male sex were associated with increased suicide risk. Conclusions Patients with TRD and MDD+SI/SA have a heightened risk of mortality by suicide compared with patients with MDD alone. Suicide rates were higher in patients with recent history versus older or no history of SI/SA, men versus women, and those of young age versus older age.

attempted suicide, major depressive disorder, mortality, suicidal ideation, suicide, treatmentresistant depression INTRODUCTION As one of the leading causes of disability, major depressive disorder (MDD) has been ranked as the third greatest cause of disease burden worldwide and is projected to rank first within the next decade (Bains & Abdijadid, 2021;Malhi & Mann, 2018;WHO, 2017). In the United States, the 12-month prevalence of MDD is 10.4%, and the lifetime prevalence is 20.6% (Hasin et al., 2018;WHO, 2017). MDD is often a chronic, recurrent illness, with a recurrence rate of approximately 50% after the first, 70% after the second, and 90% after the third depressive episode (Bains & Abdijadid, 2021).
While many treatment modalities exist, 10%−30% of patients with MDD do not respond to oral antidepressant medications (Al-Harbi, 2012;Kaur & Sanches, 2021;Khan & Brown, 2015;Cepeda et al., 2018;Zhdanava et al., 2021). Patients with MDD who fail to achieve adequate response to two or more antidepressant treatments within the current depressive episode are considered to have treatment-resistant depression (TRD) (Kaur & Sanches, 2021;Kubitz et al., 2013;McIntyre et al., 2014). Results from the Sequenced Treatment Alternatives to Relieve Depression (STAR*D) study, which was conducted to guide selection of subsequent treatments for patients with MDD who did not respond to or tolerate treatment, showed that patients requiring additional treatment steps (i.e., predefined treatment transitions following treatment intolerance or lack of remission) had increasingly lower remission rates (remission rates of 36.8%, 30.6%, 13.7%, and 13.0% for the first, second, third, and fourth treatment steps, respectively) and higher relapse rates, highlighting the challenges in treating patients with TRD (DiBernardo et al., 2018;Rush et al., 2006). Compared with patients with MDD alone, patients with TRD have an increased frequency of substance use disorders, anxiety, insomnia, pain, and other psychiatric conditions (Cepeda et al., 2018). Furthermore, patients with TRD appear to have more pronounced decreases in daily functioning and health-related quality of life (Cepeda et al., 2018;DiBernardo et al., 2018;Johnston et al., 2019). In addition, patients with TRD experience a more severe and protracted course of illness, have significant short-and long-term social impairment, and are more likely to attempt suicide compared with patients with treatmentresponsive MDD (Li & Zheng et al., 2019;Madsen et al., 2021;Vergunst et al., 2013).
Globally, more than 60% of individuals who have attempted suicide have MDD, and patients with this disorder have a 20-fold higher risk of suicide than the general population (Borentain et al., 2020;Nock et al., 2009;Xin et al., 2018). The prevalence of suicidality, including suicidal ideation (SI) and suicide attempts (SA), is increased in patients with MDD compared with matched controls without MDD (Cai et al., 2021; Center for Behavioral Health Statistics & Quality, 2018). Among people who die by suicide, MDD is the most prevalent mental health diagnosis recorded (Bertolote et al., 2002;Brådvik, 2018;Yeh et al., 2019).
In a comparative analysis of patients with TRD and managed depression, which used Medicare Standard Analytic File claims for a nationally representative 5% sample of fee-for-service beneficiaries from 2001 to 2009, the SA or self-inflicted injury rate was sevenfold higher in patients with TRD compared with those with treatment-responsive MDD (Feldman et al., 2013).
While all-cause mortality and suicidality among patients with MDD have been well researched, there is a paucity of data assessing rates of suicidality in patients with MDD that is treatment resistant or in patients with MDD who have a history of suicidal behavior. The current study was conducted to assess real-world rates of suicidespecific mortality among cohorts of adult patients with MDD that was treatment-resistant (TRD cohort), patients with MDD and a history of SI/SA (MDD+SI/SA cohort), and patients with MDD without SI/SA or TRD (MDD alone cohort).

Study design
This retrospective, observational, descriptive cohort study was designed to estimate real-world rates of suicide-specific mortality Similar requirements have been used previously (Solberg, 2006  A healthcare database validated algorithm was used to define TRD (Cepeda et al., 2018). Following the date of established MDD, patients had to have at least three distinct antidepressants (at the ingredient level) within a 365-day window or at least one antidepressant and at least one atypical antipsychotic within a 365-day window during the identification period for inclusion in the TRD group. The TRD index date was the date of service associated with the latest qualifying pharmacy fill. Patients with a history of dementia or psychosis in the 6 months prior to TRD index date were excluded from this cohort. Established MDD for the TRD cohort was defined as two or more medical claims with a diagnosis of MDD on separate dates of service within a 365-day window or an MDD-related inpatient stay (an inpatient stay with a diagnosis of MDD, in any position, at any time during the inpatient stay). The earliest date of service associated with the second MDD diagnosis or discharge date for a qualifying inpatient stay was denoted as the date of established MDD (see Table S1). Two algorithms that were created to identify SAs in claims databases and validated against medical-chart review were used. The algorithms followed a modified version of Simon and colleagues' algorithm, which determined self-harm or a probable suicide attempt from ICD-9 codes indicating intentional self-harm or undetermined intent (Simon et al., 2018). The algorithms in the current study did not include injuries of uncertain intent, however, because Barak-Corren and colleagues' algorithm (Barak-Corren et al., 2017) found these codes had low predictive values. The included codes were related to self-inflicted injury and drug poisoning. Equivalent ICD-10 codes were added to cover the recent data and concepts/codes for SI/thoughts (Cepeda et al., 2020). These algorithms reported positive predictive values ranging from 70% to 100%.

Patient cohorts
The MDD alone cohort comprised patients with at least two medical claims with a diagnosis of MDD, in any position, on separate dates of service during the identification period who did not meet TRD or SI/SA inclusion criteria.

Data sources
The study was conducted using the ORD and the NDI.  Table S2). Additionally, a sensitivity analysis was conducted using two additional suicide definitions: a probable definition that excluded poisonings and accidents of undetermined intent and a strict definition that excluded accidental poisoning (see Table S3).
NDI-sourced fact and cause-of-death information (suicide or not) was deterministically linked to patient records.

Statistical analysis
Three cohorts were captured. Univariate and bivariate descriptive statistics for the sample disposition, descriptive statistics about patient characteristics of each cohort (age, sex, comorbidities, and medication history), and suicide-specific mortality in each group were calculated.

Calculation of time-at-risk
The mortality rates due to suicide were estimated as the number of observed suicides divided by person-years-at-risk. Patients were considered at risk starting at the index date and contributed time-at-risk

Protection of human subjects
The study was a secondary analysis of de-identified data, which was reviewed by the New England Institutional Review Board (IRB) and determined to be exempt because it does not meet the definition of human subject research (NEIRB# 17-1298974-1).

Patients
A total of 807,148 patients had MDD and met the inclusion criteria. Patient attrition for the three cohorts of interest is described in

Suicide rates
Rates of suicide were 0.14 per 100 person-years for patients with TRD, 0.27 per 100 person-years for patients with MDD+SI/SA, and 0.04 per 100 person-years for patients with MDD alone ( Table 2). Having SI or SA during the year prior to the end of observation (first of death, end of enrollment, or end of study period) markedly increased the risk of suicide by more than 10 times.
Stratifying by age and sex found that rates of suicide were roughly three times higher in men than women across each of the three cohorts.
Suicide was less frequent in patients aged 65 years and older, with rates roughly three to four times higher in those aged 18−44 and 45−64 years ( Table 3).

DISCUSSION
The current study using data from the ORD and the NDI found that the rates of death by suicide was 0.14, 0.27, and 0.04 per 100 person-years for adult patients with TRD, MDD+SI/SA, and MDD alone, respectively. Suicide rates were highest for younger patients and men. Rates of suicide increased substantially with recent SI/SA.
MDD, especially when complicated by treatment resistance, is a common psychiatric disorder associated with high risk of suicide; approximately 15% of patients with MDD are at high risk for suicide, and 30% of patients with TRD are estimated to have at least one SA (Bachmann, 2018;Bergfeld et al., 2018;Orsolini et al., 2020;WHO, 2017). The current study found that rates of suicide were high in patients with TRD. Previous studies evaluating mortality and suicide risk in patients with MDD and TRD have shown greater rates of attempted suicide, all-cause mortality, and suicide-related mortality for patients with TRD Dold et al., 2018;Gronemann et al., 2021;Li et al., 2019;Nelsen & Dunner, 1995;Reutfors et al., 2018). Compared with patients with MDD, an increased suiciderelated mortality has been observed in patients with TRD even at mild levels of depression symptoms (Dold et al., 2018;Li et al., 2019). In a recent study, patients with TRD had a higher percentage of deaths due to traceable external causes (i.e., suicides and accidents) than patients with MDD (50% vs. 37%; Brenner et al., 2021). An age-sex matched study demonstrated that patients with TRD were more likely to have made a SA than those with MDD (Nelsen & Dunner, 1995).
Rehospitalization for SI/SA within a year is common among patients with MDD previously hospitalized for SI/SA (Cepeda et al., 2020).
Although SAs are up to 30 times more common than suicides, they are the most important predictors of repeated SAs and suicide (Bachmann, 2018;Ribeiro et al., 2016;WHO, 2021). In line with these previous reports, the current study found high suicide rates in patients with MDD+SI/SA, which increased substantially in patients with recent SI/SA (1 year before death). Other studies showed increased rates of suicide in patients with MDD+SI/SA, especially in those patients with recent SI/SA (Li et al., 2017;Reutfors et al., 2018Reutfors et al., , 2021Simon et al., 2018;Whittier et al., 2016). Although recent SI and SA before death TA B L E 2 Incidence rates of suicide overall and in patients with SI/SA during the last year of observation in the three cohorts . Note: Suicide was defined using ICD-10 codes for accidental poisoning (X40-X44), intentional self-poisoning (X60-X69), intentional self-harm (X70-X84), poisonings of undetermined intent (Y10-Y19), accidents of undetermined intent (Y20-Y34), sequelae of intentional self-harm (Y87.0), sequelae of events of undetermined intent (Y87.2), and sequelae of unspecified external cause (Y89.9). ICD-10, International Classification of Diseases-10; MDD, major depressive disorder; SA, suicide attempt; SI, suicidal ideation; TRD, treatment-resistant depression.

TRD
TA B L E 3 Incidence rates of suicide among patients stratified by age and sex. were observed and were confirmed as risk factors for suicide across cohorts in the current study, it is noteworthy that suicide rates in this study were 24 times higher (from 0.14 per 100 patient-years to 3.35

TRD
per 100 patient-years) in patients with TRD and recent SI/SA compared with the overall TRD cohort. This study examined groups of all patients with TRD and all patients with MDD+SI/SA, but not their mutually exclusive subgroups (e.g., patients with TRD with SI/SA and those without SI/SA). Exploring these sub-cohorts and the interaction between treatment resistance and history of SI/SA within patients with MDD and the rates of suicide in each sub-cohort could be a topic for future research.
Previous reports have highlighted the importance of including undetermined intent and accidental poisoning codes in suicide definitions because 90% and 80% of cases coded as undetermined and accidental drug deaths, respectively, are suicide (Rockett, 2020;Rockett, 2016 It is well established that depression is strongly related to SI/SA, but limited knowledge exists regarding the patient characteristics that increase suicide risk among patients with MDD and TRD, possibly due to differences that arise between regions and countries with respect to age, sex, and other variables (Brådvik, 2018;Bachmann, 2018). The current study found higher suicide rates in male patients compared with female patients. This aligns with the recent medical literature reporting that completed suicides are up to three times more common in men than women, even though SAs are significantly more common in female patients (Bachmann, 2018;Brådvik, 2018;Handley et al., 2018;WHO, 2017WHO, , 2021Finkelstein, 2015). However, these trends may be changing, as suicide rates increased by 55% in women from 2006 to 2018 compared with an overall increase of 28% during that entire reporting period of 1999−2018 (Hedegaard et al., 2020). Although it has been reported that suicides tend to occur more often in elderly than younger patients with depression (WHO, 2021;Bachmann, 2018;Hedegaard et al., 2020), the current study found the opposite, as rates of suicide in patients ≥65 years of age were three to four times lower than rates of suicide in patients aged 18-44 years. This supports a recent report that suicide rates among adults between 25 and 44 years of age have surpassed rates seen in older adults (aged ≥65 years) (Centers for Disease Control & Prevention, 2021b).
The current study used a robust healthcare database, the ORD, that included numerous patients. In addition, it allowed for linking to the NDI, the gold standard to study suicide rates, for fact and cause of death with the large population size. In terms of limitations, however, this study was restricted to enrollees of a commercial health plan or Medicare Advantage. The study was able to assess adults of all ages, and although patients 65 years of age or older are present, patient records required employment or Medicare Advantage coverage information to be included; therefore, the study excluded many of the elderly as well as those not eligible for employer-based insurance.
It is possible that the rates of completed suicides could differ from the broader population among employed individuals and their dependents. Our chosen observation window was relatively short (2.5 years).
Although the report of death is likely to be accurate, suicide as the cause of death was likely underreported (WHO, 2021). To mitigate the possibility of misclassification, a combination of suicides by any method and "accidents" of undetermined cause were considered death owing to suicide based on published data indicating that 80% of "accidental" drug intoxication deaths and 90% of undetermined drug intoxication deaths could be considered suicide (Rockett, 2020;Rockett, 2016). An additional limitation is that SI may be under-captured in medical claims data, as reported previously (Pilon et al., 2022).

CONCLUSION
This descriptive study using the ORD and the NDI databases found that incidences of death due to suicide were 0.14, 0.27, and 0.04 per 100 person-years for the TRD, MDD+SI/SA, and MDD alone cohorts, respectively. Suicide rates were substantially higher in patients with recent SI/SA. This study also found that suicide rates are higher in younger adults and were more prevalent in men than women. The high prevalence of suicidality (SI/SA) and suicide among patients with MDD and TRD seen in the current study suggests the need for a careful assessment and determination of suicide risk in this patient population.

AUTHOR CONTRIBUTIONS
All authors and the funder of this study participated in study design.
Due to the sensitive nature of the study, only Optum employees had access to and analyzed the full dataset. All authors reviewed and interpreted the final datasets for the MDD, TRD, and MDD+SI/SA cohorts, participated in the development and critical review of the manuscript, approved submission of the manuscript for publication, and are accountable for the accuracy and integrity of the work.

DATA AVAILBILITY STATEMENT
The data that support the findings of this study are available from Optum. Restrictions apply to the availability of these data, which were used under license for this study. Data are available from the authors with the permission of Optum.