Electroconvulsive therapy and psychiatric readmission in major depressive disorder – A population‐based register study

The primary aim was to determine whether electroconvulsive therapy (ECT) is associated with reduced risk of psychiatric readmission in major depressive disorder (MDD).


| INTRODUCTION
Major depressive disorder (MDD) is a common mental illness with an estimated global point prevalence of 4.7%. 1 It is currently ranked as one of the top five leading causes of years lost because of disability by the World Health Organization. 2 MDD causes great societal costs, 3 where one study found 10% of total costs to stem from admissions and readmissions. 4 Studies have found several factors associated with increased risk of relapse and recurrence in patients with MDD, but the results have been inconsistent. To date, the most established predictors are childhood maltreatment, residual symptoms after treatment, and history of prior episodes. There is also some evidence, albeit weaker, that a comorbid anxiety disorder and younger age of onset could predict increased risk of relapse and recurrence. 5 Patients admitted for MDD are frequently readmitted within one year of discharge. 6 Some studies have found that antidepressant drugs 5,7 and lithium prevent recurrence and readmission. [8][9][10] Electroconvulsive therapy (ECT) is considered a highly effective treatment and is mainly recommended for severe cases of MDD according to the American Psychiatric Association. ECT should be considered for patients with psychotic, catatonic, or suicidal features, and in those who are insufficiently responsive to medication or psychotherapy. ECT could also be considered for patients who have responded well to ECT in the past and for those who prefer it over other treatments. 11,12 While ECT is mainly used as an acute treatment for depression and is much less commonly used as a continuation/maintenance treatment, it is uncertain whether it reduces the risk of psychiatric readmission in MDD.
Most previous studies investigating this subject have included study participants with mixed severe affective disorders. Slade et al. 13 included 162,169 patients with MDD, bipolar disorder, or schizoaffective disorder and found ECT to be associated with reduced 30-day readmission risk. Tor et al. 14 studied 121 patients with MDD or bipolar disorder and found a 35% reduction of readmissions within one year after ECT. The association was greater in patients with MDD than bipolar disorder. However, Lin et al. 15 included 1,568 patients with MDD or bipolar disorder insufficiently responsive to antidepressant drugs and found no statistically significant association on psychiatric readmission within one year after ECT. Similar results were found by Stoudemire et al. 16 who studied 94 older patients with MDD.
Thus, there is currently limited evidence supporting that ECT reduces the risk of psychiatric readmission in patients with MDD, although it is known that ECT is very efficient in achieving acute response and remission. 17

| Aims of the study
The primary aim of this study was to determine whether electroconvulsive therapy is associated with reduced risk of psychiatric readmission compared with nonelectroconvulsive therapy treatment in patients with major depressive disorder. The secondary aims were to investigate whether a potential association differed across subgroups of the study population and to identify risk factors for readmission.

| Study design
This was a registered study based on data from multiple Swedish population-based registries. Patients who had been admitted for moderate-to-severe MDD to any Swedish hospital from 1 January 2012 to 30 September 2018 were included. Study participants were divided into two groups depending on whether they received ECT or non-ECT treatment during inpatient care. Factors reported affecting the risk of psychiatric readmission were identified through a literature search. These potential confounders were included as covariates in both univariate and multivariate logistic regression analyses.

Significant Outcomes
• Electroconvulsive therapy (ECT) was associated with reduced risk of psychiatric readmission in certain subgroups of patients with major depressive disorder. • Subgroups with superior outcomes after ECT compared with non-ECT treatment included older patients, those with psychotic features, prior psychiatric hospitalizations, or family history of suicide. • Age-dependent effects of ECT on readmission risk need to be further investigated.

Limitations
• Substantial differences have been reported between patients treated with ECT and those receiving non-ECT treatment. • Some factors previously reported to affect the risk of psychiatric readmission were not accessible in register data.

| Participants
All adult patients admitted with a main diagnosis of the first episode or recurrent moderate-to-severe MDD to any Swedish hospital from 1 January 2012 to 30 September 2018 were identified using the Swedish National Patient Register. Inclusion criteria were discharged before 1 October 2018 and no prior admission for mania, schizoaffective disorder, or schizophrenia. All psychiatric hospitalizations that met the inclusion criteria within this period were counted as new primary admissions. This meant that study participants readmitted for MDD were included several times in some analyses. Follow-up was set at 30 and 90 days from discharge. Readmissions with a main diagnosis of any psychiatric disorder within this period were identified. Study participants who died within the 90-day follow-up period were excluded.

| Data sources
The Swedish National Patient Register provides information about demographic factors, diagnoses, and inpatient and specialized outpatient care. Reporting to the register is mandatory for healthcare providers. It has a high degree of completeness, where some of the most frequently used variables are 99% complete. 18 The International Statistical Classification of Diseases and Related Health Problems -Tenth Revision (ICD-10) is used to report diagnoses. 19 Study participants were identified through searches in the register for specific ICD-10 codes (F32.1 and F33.1 for moderate major depressive disorder, F32.2 and F33.2 for severe major depressive disorder without psychotic features, and F32.3 and F33.3 for severe major depressive disorder with psychotic features). Psychiatric readmission was defined as readmission with a main diagnosis of any psychiatric disorder (ICD-10 codes F01-F99) within the follow-up period. Further variables obtained from the register were gender, age, psychiatric comorbidity, number of prior psychiatric hospitalizations, and whether the admission was voluntary or involuntary. The Swedish National Quality Register for ECT (Q-ECT) contains information about indication for referral, ECT setting, response to treatment, and adverse effects. 20 All Swedish hospitals that provide ECT report to the Q-ECT, which is an opt-out register. Of all patients treated with ECT in 2018, 91% were included in the Q-ECT. 21 The longitudinal integrated database for health insurance and labour market, studies (LISA) provides information about demographic and socioeconomic factors. It includes all Swedish citizens and residents above 15 years of age. 22 Information about marital status, cohabitation status, level of education, and employment status was obtained from the LISA.
The Swedish Prescribed Drug Register contains information about all prescription drugs collected at Swedish pharmacies. It is classified according to the Anatomical Therapeutic Chemical (ATC) classification system. 23 ATC codes were used to identify relevant drugs each study participant had collected within 100 days before admission (antidepressants, antipsychotics, anxiolytics, benzodiazepines, and lithium). The Antidepressant Treatment History Form (ATHF) is used to rate the adequacy of a patient's antidepressant treatment. 24 A slightly modified version of the ATHF was used in this study where antidepressant treatment within one year before admission was evaluated for each study participant based on drug doses collected at pharmacies. This modified version has been previously described by Brus et al. 10 The Multi-Generation Register provides information about over 11 million index persons in Sweden. 25 It was used to identify first-degree relatives of the study participants. The register was linked to the Swedish National Patient Register to identify relatives with affective disorders (MDD or bipolar disorder).
The Swedish Cause of Death Register contains information about all deaths that have occurred in Sweden. 26 These data were used to exclude subjects who died during the 90-day follow-up period and to identify first-degree relatives who have committed suicide.

| Statistical analyses
Chi-square test and Mann-Whitney U-test were used to compare group characteristics. Logistic regression analyses were conducted using univariate and multivariate models. The covariates included in the analyses were identified through a literature search and were consequently included in the multivariate analyses as they could be potential confounders. These covariates were inpatient treatment (ECT or non-ECT treatment), gender, age, marital status, cohabitation status, level of education, employment status, psychiatric comorbidity (anxiety disorder, alcohol use disorder, substance use disorder, and personality disorder), depression severity, whether the admission was voluntary or involuntary, number of prior psychiatric hospitalizations, ATHF score, psychiatric prescription drugs collected within 100 days before admission (antidepressants, antipsychotics, anxiolytics, benzodiazepines, and lithium), and family history of affective disorders or suicide. Study participants with missing data for demographic variables were imputed into the largest category. A separate model included all covariates mentioned as well as continuation ECT within 14 days after discharge. Multivariate logistic regression analyses were also conducted after stratification to identify differences in readmission risk between subgroups of the study population treated with or without ECT. Thereafter, the same multivariate model was applied by adding interaction terms to determine whether these differences were statistically significant.
Two sensitivity analyses were performed. In the first sensitivity model, only the first admission for each study participant was included. Finally, matching (1:1) was conducted as a second sensitivity model using only the first admission for each patient. Variables matched exactly were depression severity, number of prior psychiatric hospitalizations, and collection of antidepressants within 100 days before admission. Propensity score matching was used for all remaining covariates mentioned above. The threshold for statistical significance (alpha) was set at 0.05. Data management and statistical analyses were performed using SAS software (

| Ethical considerations
This study was granted approval by the regional ethical review board in Uppsala, Sweden (2014/174). Data collected from Swedish population-based registries were pseudonymized, and study participants were not identifiable at any time. Therefore, they were not informed of the study and were not required to provide written consent.

| Participants
The inclusion criteria were met by 42,207 admissions. Subject death occurred within the 90-day follow-up period subsequent to 291 admissions (0.6% of the ECT group and 0.7% of the non-ECT group; p = 0.088), which were excluded. Thus, the final number arrived at 27,851 unique study participants contributing to a total of 41,916 admissions. ECT was used in 11,227 admissions (26.8%), whereas non-ECT treatment was used in the remainder. Compared with the non-ECT group, study participants treated with ECT were significantly older and more likely to be female, married, cohabiting, and highly educated. Additionally, patients in the ECT group were more likely to have a higher number of prior psychiatric hospitalizations, greater severity of illness, family history of affective disorders or suicide, and being involuntarily admitted. Full characteristics are presented in Table 1.

| Subgroups with differential associations between electroconvulsive therapy and readmission risk
Stratified multivariate logistic regression analyses demonstrated differences in readmission risk when subgroups of the ECT group were compared with the same subgroups of the non-ECT group ( Table 2). The factors associated with decreased risk of both 30-and 90-day readmission in study participants treated with ECT compared with those receiving non-ECT treatment were older age, being married, having psychotic features, prior psychiatric hospitalizations, higher ATHF score, having collected antidepressants, antipsychotics, or benzodiazepines within 100 days before admission, and having a family history of suicide. Younger age was associated with increased readmission risk within both 30 and 90 days in patients treated with ECT compared with those receiving non-ECT treatment. According to these estimates, the risk of readmission within 30 days is reduced by about 8 percentage points in patients aged 55 years and above with psychotic symptoms when treated with ECT compared with non-ECT treatment. Thus, the number needed to treat to prevent one readmission in this group is about 13 patients. The estimations indicate that the number needed to treat could be reduced to about 6 patients if continuation ECT is also provided.

| First sensitivity analysis
In the first sensitivity model, only the first admission for each study participant that met the inclusion criteria was included. There were no statistically significant differences in 30-or 90-day readmission risk in the multivariate analysis (OR: 0.93; 95% CI: 0.85-1.02; p = 0.112 and OR: 0.99; 95% CI: 0.92-1.07; p = 0.824, respectively). Full characteristics and logistic regression analyses for this model are presented in Appendix A.

| Second sensitivity analysis
Matching was conducted as a second sensitivity model.

| Risk factors for psychiatric readmission
Full logistic regression analyses of risk factors for psychiatric readmission in the study population as a whole are presented in Appendix C.

| DISCUSSION
The results of this study support that ECT reduces the risk of psychiatric readmission in certain subgroups of patients with MDD. In the main analysis, ECT was associated with reduced overall risk of both 30-and 90-day readmission in a multivariate model when compared with non-ECT treatment. The association of ECT on readmission risk, however, was not as pronounced as in previous studies with positive results. 13,14 Furthermore, although the results of both sensitivity analyses pointed in the same direction as the main analysis for readmission risk within 30 days, statistical significance was not reached in these analyses. This could be explained by insufficient power, as fewer study participants were included in these models, and that the association was unevenly distributed across subgroups with a modest overall effect. Moreover, there is risk of residual confounding. Therefore, there is still some uncertainty regarding to what degree ECT reduces the risk of psychiatric readmission in MDD. Having residual symptoms after treatment is a well-established risk factor for relapse and recurrence in MDD. 5 The likely explanation of the superior outcomes of certain subgroups treated with ECT compared with those receiving non-ECT treatment is that the ECT-treated patients had fewer and less severe symptoms at discharge (i.e., were more effectively treated). This is in line with other studies which have shown ECT to be more effective than pharmacotherapy for the treatment of inpatients with depressive illness. 17 Study participants treated with ECT for severe MDD with psychotic features were less likely to be readmitted than patients with the same diagnosis receiving non-ECT treatment in multivariate stratification and interaction analyses. By contrast, there was no statistically significant difference in readmission risk between study participants treated with ECT and those receiving non-ECT treatment in moderate-to-severe MDD without psychotic features. There is evidence supporting ECT as especially effective in preventing relapse and readmission in MDD with psychotic features. 10,27,28 Depressed patients in ECT trials have previously been described to be more severely ill than those participating in pharmacotherapy trials. 29 At the time of admission, the ECT group had more severe symptoms than the non-ECT group. Greater severity of illness has previously been associated with increased risk of relapse and readmission in MDD 30,31 and was also found to predict increased risk of both 30-and 90-day readmission in the present study. Although we did adjust for depression severity as diagnosed with ICD-10, it is likely that within each diagnostic category, the ECT group tended to have more severe symptoms than the non-ECT group. If this was the case, the true effects of ECT on readmission risk are in fact likely to be underestimated in our models. Note: Statistically significant results (p-value < 0.05) are presented using bold font. All models are adjusted for sex, age, marital status, cohabitation status, level of education, employment status, psychiatric comorbidity (anxiety disorder, alcohol use disorder, substance use disorder, and personality disorder), depression severity, whether the admission was voluntary or involuntary, number of prior psychiatric hospitalizations, ATHF score, psychiatric prescription drugs collected within 100 days before admission (antidepressants, antipsychotics, anxiolytics, benzodiazepines, and lithium), and family history of affective disorders or suicide.

T A B L E 2 (Continued)
There was a strong correlation between a higher number of prior psychiatric hospitalizations and increased readmission risk in all analyses. To date, there is stronger evidence supporting a history of prior depressive episodes being associated with increased risk of relapse or recurrence than the specific number of prior depressive episodes or hospitalizations. 5 However, some studies have found a higher number of prior psychiatric hospitalizations to correlate with increased readmission risk in patients with MDD. 10,32,33 Additionally, some other studies have found a higher number of prior episodes to correlate with increased risk of recurrence. 34-36 Study participants with a history of prior psychiatric hospitalizations had decreased risk of readmission in multivariate stratification and interaction analyses, especially within 30 days, when treated with ECT compared with those receiving non-ECT treatment. Patients without prior admissions did not have a significantly reduced readmission risk when treated with ECT. These results indicate that ECT is an especially suitable choice for patients with severe and recurrent depressive episodes.
Age was negatively associated with readmission risk in all analyses. In multivariate stratification and interaction analyses, older age was consistently associated with reduced risk of both 30-and 90-day readmission in study participants treated with ECT compared with the same subgroups receiving non-ECT treatment. Furthermore, younger age was associated with increased readmission risk in patients treated with ECT compared with non-ECT treatment. ECT has been found to be especially effective in older patients. 37 Psychotic symptoms and psychomotor retardation are more common in older individuals, and patients with such symptoms tend to respond better to ECT, which could explain the results of the present study. 38,39 To date, there is sparse evidence supporting that older patients have decreased risk of readmission after ECT 40 and that younger individuals are at increased risk when compared with older patients treated with ECT. 10 The increased readmission risk in younger individuals treated with ECT in the present study could perhaps be explained by residual confounding. Nevertheless, a meta-analysis of 372 randomized placebo-controlled antidepressant drug trials found the risk of suicidal ideation and suicidal behavior (preparation, attempted suicide, or completed suicide) associated with antidepressant use to be strongly age-dependent. Increasing age was associated with a protective effect. 41 Age-dependent associations of ECT on readmission risk and of antidepressant drugs on suicidality could possibly be explained by differences in underlying disease mechanisms.
Most factors associated with readmission risk in the present study have been previously described to predict relapse or recurrence, including having a comorbid personality disorder, 42 anxiety disorder, 5 or higher ATHF score. 10 To the best of our knowledge, no previous study has found having a higher education level or being voluntarily admitted to correlate with decreased risk of readmission, or having collected antipsychotics or benzodiazepines within 100 days before admission to be associated with increased readmission risk. However, some studies have found a similar association for such drugs collected after discharge. 10,43 Study participants prescribed drugs other than antidepressants or lithium perhaps represent a subset of patients difficult to treat and, therefore, more likely to be readmitted. Nevertheless, when these study participants were treated with ECT, they were less likely to be readmitted than similar patients receiving non-ECT treatment. These results suggest that ECT is a valid option for such patients.
There were several other subgroups with decreased risk of readmission when treated with ECT compared with non-ECT treatment. Patients who were married, widowed, or unemployed were less likely to be readmitted after ECT. It has been suggested that family history of affective disorders or suicide should be considered when selecting appropriate candidates for ECT, 44 and in this cohort family history of affective disorders was more common in the ECT group than in the non-ECT group. This study could neither confirm nor refute differential effects of ECT on readmission risk among patients with or without family history of affective disorders. However, family history of suicide was associated with decreased risk of readmission in study participants treated with ECT compared with those receiving non-ECT treatment. Therefore, it could be appropriate to ask about family history when considering a patient for ECT. These patients could perhaps represent a subset of patients with more severe family history of psychiatric disorders. To the best of our knowledge, no previous study has identified family history of suicide to be predictive of response, remission, or readmission in MDD. Further studies are needed to investigate potential differential effects of ECT in patients with family history of suicide, psychiatric disorders, or specific genetic factors.
This study shows considerable risk of readmission after inpatient treatment for MDD in patients treated with and without ECT. In addition to treatment with antidepressants, which most patients had been prescribed, lithium 8,45 and continuation ECT have been shown to reduce the risk of relapse in MDD. 46,47 There was an association between reduced readmission risk and continuation ECT after discharge in this study. A small minority of study participants had these treatments; outcomes could likely be improved by more systematic use of lithium and continuation ECT.
To date, this is one of the largest studies to have investigated whether ECT reduces the risk of psychiatric readmission in MDD compared with non-ECT treatment. The large sample is facilitated by the high rate of ECT use in Sweden and the existence of national registries. 20 Patients treated at all Swedish hospitals were included. Nonetheless, there were several limitations to this study. Substantial differences have been reported in previous studies between patients treated with ECT and those receiving non-ECT treatment. 48 Similar differences were also found in the present study, limiting the comparability between the groups. Matching was conducted to minimize these differences but made each group considerably smaller and reduced the statistical power. Potential confounders were identified through a literature search. However, some of these factors were not accessible in register data and could not be included in the analyses, such as childhood maltreatment, age of onset, and medication compliance. Even though ECT can be delivered in an outpatient setting, most patients in Sweden receive their index series in an inpatient setting. This could have increased the readmission risk among ECT-treated patients.
A few individual patients received a very high number of ECT sessions in their treatment series, suggesting unusually severe and persistent symptoms. Such patients are so unusual that statistical adjustment becomes uncertain. However, as they are so few their overall impact on the results is small. Only a minor proportion of patients received continuation ECT after their index series, and these patients are likely to differ from other patients in regard to previous response to pharmacotherapy or ECT and perceived risk of relapse as assessed by the treating psychiatrist. Further studies are required to adequately assess the effects of continuation ECT on readmission risk. Finally, the diagnoses were made in clinical practice, which means that there may be some uncertainty in the classifications. For instance, it is probable that a small subset of patients with bipolar disorder were included in this study despite the fact that patients with prior admissions for mania were excluded.
This study suggests that ECT reduces the risk of psychiatric readmission in certain subgroups of patients with MDD. ECT-treated subgroups with superior outcomes on readmission risk compared with non-ECT treatment were older, unemployed, married, or widowed patients, those treated with antipsychotics or benzodiazepines before admission, with psychotic features, prior psychiatric hospitalizations, or family history of suicide. However, in patients below 35 years of age, ECT was found to be associated with increased readmission risk. This study is limited by the risk of residual confounding, especially since patients treated by ECT tend to have more severe symptoms and are more difficult to treat.

ACKNOWLEDGMENTS
We would like to thank the patients, nurses and doctors who provided data to the Swedish National Quality Registry for ECT.

CONFLICT OF INTEREST
The authors report no conflicts of interest regarding this research.

PEER REVIEW
The peer review history for this article is available at https://publo ns.com/publo n/10.1111/acps.13373.

DATA AVAILABILITY STATEMENT
The data that support the findings of this study are available from the corresponding author, Axel Nordenskjöld, upon reasonable request.   Note: Statistically significant results (p-value < 0.05) are presented using bold font. All models are adjusted for gender, age, marital status, cohabitation status, level of education, employment status, psychiatric comorbidity (anxiety disorder, alcohol use disorder, substance use disorder, and personality disorder), depression severity, whether the admission was voluntary or involuntary, number of prior psychiatric hospitalizations, ATHF score, psychiatric prescription drugs collected within 100 days before admission (antidepressants, antipsychotics, anxiolytics, benzodiazepines, and lithium), and family history of affective disorders or suicide.

APPENDIX C RISK FACTORS FOR PSYCHIATRIC READMISSION
Several factors associated with increased or decreased risk of psychiatric readmission in the study population as a whole were identified (Table A5 and Table A6). Factors associated with increased risk of both 30-and 90-day readmission in the multivariate analyses were having a comorbid anxiety or personality disorder, severe MDD without psychotic features, higher number of prior psychiatric hospitalizations, higher ATHF score, and having collected antipsychotics, anxiolytics, or benzodiazepines within 100 days before admission. Older age was associated with decreased risk of both 30-and 90-day readmission. Family history of affective disorders was associated with increased readmission risk only within 30 days, whereas being voluntarily admitted was associated with decreased risk of readmission only within 90 days. Additionally, these analyses were also conducted including only the first admission for each patient and the matched sample. Full logistic regression analyses are presented in Table A7-A10.     T A B L E C 6 (Continued)