Long‐term mortality associated with depression among South Korean survivors of extracorporeal membrane oxygenation

Abstract Introduction Depression is an important sequela in critically ill patients. However, its prevalence after extracorporeal membrane oxygenation (ECMO) therapy and its association with long‐term mortality remain controversial. Methods Data were extracted from the South Korean National Health Insurance Service database in this population‐based cohort study. Adults who received ECMO therapy from 2006 to 2014 were included. Survivors of ECMO were defined as patients who underwent ECMO and survived over 365 days after the initiation of ECMO therapy. Results A total of 3,055 survivors of ECMO were included in the final analysis. They were classified into the pre‐ECMO depression group (n = 275 [9.0%]), post‐ECMO depression group (n = 331 [10.8%]), and other ECMO survivor group. In the multivariable Cox regression model, a 1.52‐fold higher mortality was observed in the post‐ECMO depression group than in the other groups (hazard ratio, 1.52; 95% confidence interval, 1.17–1.96; p = .002). However, there was no statistically significant difference between the pre‐ECMO depression group and the other groups (p = .075). Conclusions The prevalence of pre‐ and post‐ECMO depression was 9.0% and 10.8%, respectively. Additionally, post‐ECMO depression was associated with an increased 5 year all‐cause mortality; however, pre‐ECMO depression was not.

used worldwide as a rescue therapy in many critically ill patients with coronavirus disease 2019 (Barbaro et al., 2020;Yang et al., 2020).
Thus, the clinical utility of ECMO will be more emphasized in the future.
As approximately 40%-60% of patients treated with ECMO survive to hospital discharge (Nasr et al., 2019), over 40% of patients undergoing ECMO may be expected to survive. A previous study reported that the 87% of ECMO patients who survived the first 90 days after ECMO therapy survived for 5 years (von Bahr et al., 2017), suggesting that quality of life among survivors of ECMO is an important public health issue. Depression may be an important psychological sequelae of ECMO therapy. It is commonly observed in survivors of critical illness who were admitted to the intensive care unit (ICU) (Battle et al., 2015), with a reported prevalence of 46% in a single-center retrospective study. (Battle et al., 2015) The prevalence of depression was 40% in survivors of critical illness from 26 ICUs in the United Kingdom (Hatch et al., 2018). Recently, singlecenter cohort studies have reported that the prevalence of depressive symptoms among survivors of ECMO ranged from 15%-42% (O'Brien et al., 2020;Sanfilippo et al., 2019;Sylvestre et al., 2019).
However, these studies had small sample sizes (O'Brien et al., 2020;Sanfilippo et al., 2019;Sylvestre et al., 2019) and the association of post-ECMO depression with long-term mortality was not identified.
Recently, the long-term quality of life following ECMO therapy has been emphasized as an important health issue (Wilcox et al., 2020), and the early identification and treatment of mental illnesses, such as depression, following ECMO therapy among survivors might improve long-term outcomes regarding quality of life (Knudson et al., 2019). Furthermore, depression following critical illness was associated with an increased relative long-term mortality among survivors of critical illness (Hatch et al., 2018). Therefore, depression following ECMO therapy might be associated with increased longterm mortality among survivors of ECMO; however, there have been no studies on this topic. Therefore, we aimed to investigate the prevalence of pre-and post-ECMO depression and examine its impact on the long-term mortality of survivors of ECMO, using data from the National Health Insurance Service (NHIS) database in South Korea.

| Study design and approval
This population-based cohort study was conducted according to the Strengthening the Reporting of Observational Studies in Epidemiology guidelines (von Elm et al., 2014). The study protocol was approved by the Institutional Review Board of Seoul National University Bundang Hospital (X-2001-586-902) and the Health Insurance Review and Assessment Service (NHIS-2020-1-125). The requirement for informed consent was waived because the data analyses were performed retrospectively using anonymized data derived from the South Korean NHIS database.

| NHIS database and study population
This study used data from the health records obtained from the South Korean NHIS database. In South Korea, information concerning disease diagnoses and prescription of drugs and/or procedures is registered in the NHIS database so that patients can receive financial support for treatment. Data were extracted by an unaffiliated, independent medical record technician at the NHIS center. All adult patients (age ≥18 years) who received ECMO therapy between 2006 and 2014 were included. Those who only received ECMO therapy during surgery were excluded. The prescription codes for ECMO in the database were O1901-1904, and the use of a Novalung (O1905) was not considered as ECMO in this study. We defined survivors of ECMO as those who had survived over 365 days after the initiation of ECMO therapy.

| Depression
The diagnosis of depression was evaluated using the following

| Study outcomes
The primary endpoint of this study was the 5 year all-cause mortality among survivors of ECMO, which was defined as death within 5 years of undergoing ECMO therapy. The patients' dates of death were extracted until 1 April 2020, and survival times were calculated from the date that ECMO therapy was started to the date of death or 1 April 2020, whichever occurred first. The secondary endpoint of this study was the development of post-ECMO depression, which was evaluated within 365 days from the date of starting ECMO therapy.

| Measurements of confounders
The data on confounders were extracted using ICD-10 codes registered within 1 year of starting the ECMO therapy and included demographic characteristics (age, sex, and place of residence at the time of ECMO therapy [Seoul, metropolitan city, and other areas]); annual income; underlying disability; other underlying psychiatric illness, such as post-traumatic stress disorder (PTSD, ICD-10: F43.1) and anxiety disorder (F41); and Charlson comorbidity index (CCI). The CCI was used to reflect the comorbid status of ECMO survivors at ECMO therapy because the CCI can predict in-hospital mortality among hospitalized patients (Sundararajan et al., 2004). The CCI was calculated as the sum of the 17 individual underlying diseases listed in Table S1 (1 point for 10 diseases, 2 points for 4 diseases, 3 points for 1 disease, and 6 points for 2 diseases); therefore, the range of CCI was from 0 (best score) to 27 (worst score) in this study. In addition, information on the other treatments administered, length of hospital stay, duration of ECMO therapy, annual case volume of ECMO therapy, and main diagnosis at the time of ECMO therapy were collected. The annual case volume of ECMO therapy was calculated as the total ECMO cases/9 years to reflect the capability of the ECMO centers (Barbaro et al., 2015), and it was divided into four groups (Q1 <14, Q2: 15-29, Q3: 30-64, and Q4 > 64). In South Korea, all individuals with disabilities are registered in the NHIS database to receive various benefits from the social welfare system. Disabilities were divided into 15 types: physical disabilities, brain lesion disabilities, visual disturbances, hearing disturbance, speech disabilities, autism, intellectual disorder, mental disorder, renal disorder, heart diseases, respiratory disorders, hepatopathy, intestinal fistulae, urinary fistulae, and epilepsy. Additionally, each disability was assigned one of six grades based on severity; we grouped the severity grades into two (1-3, severe disability; and 4-6, mild to moderate disability). The main diagnoses at the time of ECMO therapy were classified into four groups: (1) potential venoarterial (VA) ECMO group (cardiovascular disease or shock), (2) potential venovenous (VV) ECMO group (ARDS or respiratory failure), (3) potential extracorporeal cardiopulmonary resuscitation (ECPR) group (post-cardiac arrest), and (4) others.

| Statistical analysis
The clinicoepidemiological characteristics of the survivors of ECMO are presented as means with standard deviations for continuous variables and numbers with percentages for categorical variables. First, we fit the multivariable logistic regression model for development of post-ECMO depression within 365 days following ECMO therapy among survivors. All covariates were included in the multivariable logistic regression model for adjustment, except for pre-ECMO depression, which was not included because the pre-ECMO group had been treated for depression both before and after ECMO therapy.
Next, we fit the multivariable Cox regression model for the 5 year all-cause mortality among the survivors of ECMO to investigate whether pre-and post-ECMO depression were associated with 5 year all-cause mortality. All covariates were included in the model for adjustment. Additionally, we performed sensitivity analyses according to pre-and post-ECMO depression with or without other psychiatric illness (PTSD or anxiety disorder) using multivariable Cox regression modeling. Finally, we performed the subgroup analyses according to main diagnosis at ECMO therapy to examine the impact of potential type of ECMO and ECPR on associations of pre-and post-ECMO depression with 5 year all-cause mortality among survivors. We confirmed that there was no multicollinearity in all the multivariable models involving the entire cohort, with a variance inflation factor of <2.0. The results of the Cox regression are presented as hazard ratios (HRs) with 95% confidence intervals (CIs), and log-log plots were used to confirm that the central assumption of multivariable Cox model was satisfied. The results of the logistic regression models are presented as odds ratios (ORs) with 95% CIs.
The Hosmer-Lemeshow test was used to confirm the goodness of fit of the multivariable models at p > .05. R software (version 4.0.3; R Foundation for Statistical Computing) was used for all analyses, and a p-value < .05 was considered statistically significant. F I G U R E 1 Flowchart depicting the patient. selection process ECMO, extracorporeal membrane oxygenation From 2006From to 2014 patients received ECMO therapy at 72 hospitals in South Korea. After excluding 981 patients aged <18 years, 10,215 patients were screened. Of these patients, 3,055 (29.9%) survived for more than 365 days after the initiation of ECMO therapy and were included in this study. Among them, 275 (9.0%) and 331 (10.8%) were assigned to the pre-and post-ECMO depression groups, respectively. Five-year all-cause mortality occurred in 556 (18.2%) survivors of ECMO therapy. Figure 1 presents a flowchart depicting patient selection. The clinicoepidemiological characteristics of the patients are presented in Table 1. Table 2 shows the results of the multivariable logistic regression model for the development of depression within 1 year after ECMO therapy. An increase in age by 1 year was associated with a 1% increase in the development of post-ECMO depression (OR, 1.01; 95% CI, 1.00-1.02; p = .032), and an increase in the length of hospital stay by 1 day was associated with a 2% increase in the development of post-ECMO depression. Furthermore, the ECMO survivors with PTSD and anxiety disorder showed a threefold (OR, 3.00; 95% CI, 1.83-4.90; p < .001) and 1.37-fold (OR, 1.37; 95% CI, 1.06-1.77; p = .016) increased risk for the development of post-ECMO depression, respectively. Table 3 shows the results of the multivariable Cox regression analysis for the 5 year all-cause mortality among the survivors of ECMO. A 1.52-fold higher mortality was observed in the post-ECMO depression group than in the other groups (HR, 1.52; 95% CI, 1.17-1.96; p = .002). The survival plot derived from the multivariable model showed a similar trend as that shown in Figure S1.

| Five-year all-cause mortality
However, there was no statistically significant difference for the 5 year all-cause mortality between the pre-ECMO depression group and the other groups (HR, 0.74; 95% CI, 0.63-1.13; p = .075). In addition, an increase in age by 1 year was associated with a 4% increase in the 5 year all-cause mortality (HR,  derived from the multivariable model also showed a similar trend as shown in Figure S2.

| D ISCUSS I ON
This population-based cohort study using data from the Korean NHIS database showed that 9.0% of survivors of ECMO were already diagnosed with depression before ECMO therapy, and 10.8% were newly diagnosed with depression within 365 days after ECMO therapy. While post-ECMO depression was associated with a 1.52fold higher 5 year all-cause mortality, pre-ECMO depression was not. Additionally, this association was more evident in the post-ECMO depression group with an underlying anxiety disorder or PTSD in the sensitivity analyses as well as in the post-ECMO depression group that received ECMO therapy for cardiovascular disease or shock (potential VA ECMO group). Moreover, an older age, prolonged length of hospital stay, and underlying PTSD and anxiety disorder were associated with a higher incidence of post-ECMO depression in our cohort. This is the first study to report that post-ECMO depression might be a predictive factor for poor long-term prognosis among survivors of ECMO. Our results suggest that survivors of ECMO who acquired and were diagnosed with depression  (Kim et al., 2020), which is relatively lower than prevalence of pre-ECMO depression (9.0%) in our study.
Our study focused on the critically ill and survivors of ECMO. Thus, the prevalence of depression in the participants of our study might be higher than that of the general adult population in South Korea.
Our study seems to report a lower prevalence ( (Sanfilippo et al., 2019). However, our study might have suffered from detection bias (Sjolander et al., 2008) because we used the registered ICD-10 codes for depression. For example, survivors of ECMO living in areas with better hospital or outpatient clinic access would be more likely to have their depression diagnosed by a medical professional than those living in areas with poor access to medical facilities. Considering these biases, more research is needed in the future regarding the prevalence of depression among survivors of ECMO.
The relationship between post-ECMO depression and poor long-term survival is the most important finding of this study.
Some assumptions might be applied to explain this phenomenon.
Many complications have been reported among patients following ECMO therapy, with central nervous system (CNS) complications having an incidence of 8% (Zangrillo et al., 2013). These complications include ischemic and hemorrhagic brain injury that may result in increased mortality (Le Guennec et al., 2018). Ischemic damage is known to increase the development of depression in the poststroke period through a neuroinflammatory mechanism (Robinson & Spalletta, 2010). This systemic and neuroinflammation may lead to a breakdown of the blood-brain barrier, which results in an increased susceptibility of the CNS to certain drugs, such as sedatives and opioids. Thus, depression can develop as a consequence of systemic inflammation and neurotransmitter imbalances (Battle et al., 2015). Inflammatory responses due to the widespread activation of the innate immune system by ECMO therapy have been reported (Millar et al., 2016). Therefore, post-ECMO depression might reflect the sequelae of CNS complications among survivors of ECMO. Neurologic complications among critically ill patients were risk factors for higher long-term mortality (Sonneville et al., 2011), Based on this, the relationship between CNS complications following ECMO, post-ECMO depression, and long-term mortality might be understood.
In addition, the impact of stress on the development of depression among survivors of ECMO should be considered. A previous study reported that two-fifths of survivors of an admission to the ICU experienced unemployment of up to 12 months following discharge (Kamdar et al., 2020). Furthermore, 71% of American survivors of ARDS experienced lost earnings (Kamdar et al., 2017).
Unemployment is known to be closely related to depression (Dooley et al., 1994). Thus, in addition to the physiologic sequelae of ECMO therapy, the development of post-ECMO depression needs to be understood from a social perspective. Since depression is also a known risk factor for increased mortality in the general population (Gilman et al., 2017), the prevention and treatment of post-ECMO depression will be important for survivors of ECMO.
Another finding in the sensitivity analysis also showed that the post-ECMO depression group with other psychiatric illness (anxiety disorder and PTSD) had increased 5 year all-cause mortality among survivors of ECMO. The co-existence of other such psychiatric illnesses suggests that the ECMO patients had more severe depression than those without psychiatric illness (Zimmerman et al., 2018), and this might have influenced the increased 5 year all-cause mortality among survivors of ECMO. Thus, the patients with post-ECMO depression with other psychiatric illnesses, such as anxiety disorder and PTSD, might be at a higher risk for poorer long-term survival.
This study has several limitations. First, certain important variables, including body mass index, alcohol consumption, and smoking history, were not included in the analysis because the necessary information was not available in the NHIS database.
Second, we defined comorbidities using their ICD-10 codes to calculate the CCI; however, the ICD-10 codes might not accurately reflect the actual diseases in the study population. Third, we did not distinguish between VA and VV ECMO in this study due to the limited prescription codes for ECMO in South Korea. Therefore, information on the difference in the prevalence of depression between VA and VV ECMO is missing from our study. Fourth, as mentioned above, our analysis of the prevalence of depression might have suffered from detection bias due to its use of ICD-10 codes in the NHIS database. Lastly, we did not assess the cause of 5 year mortality among survivors of ECMO due to the limitations of our data source.
In conclusion, our study showed that 9.0% of survivors of ECMO were already diagnosed with depression before undergoing ECMO therapy, while 10.8% were newly diagnosed with depression within 365 days after ECMO therapy. Post-ECMO depression was associated with an increased 5 year all-cause mortality, but pre-ECMO depression was not. Our results suggest that survivors of ECMO who acquired and were diagnosed with depression within 365 days of ECMO therapy may be a high-risk group. Future studies are needed to investigate strategies to decrease post-ECMO depression to improve both quality of life and long-term survival among survivors of ECMO.

ACK N OWLED EG M ENT
None.

CO N FLI C T S O F I NTE R E S T
None declared.

PE E R R E V I E W
The peer review history for this article is available at https://publo ns.com/publo n/10.1002/brb3.2218.

DATA AVA I L A B I L I T Y S TAT E M E N T
The data that support the findings of this study are available on request from the corresponding author. The data are not publicly available due to privacy or ethical restrictions.