STRENGTHS AND LIMITATIONS OF THIS STUDY
-
This prospective cohort study examined the association between social support prior to intensive care unit (ICU) admission and mental health after ICU discharge.
-
The sample size was relatively small, and patients were recruited from a single institution, which may have influenced external validity.
-
Social support after ICU discharge was not measured; it was not possible to include a discussion on the impact of social support after ICU discharge.
Introduction
Although advances in intensive care have improved the survival rate of critically ill patients, the health-related quality of life (HRQoL) of patients discharged from the intensive care unit (ICU) has declined.1 2 Similarly, in patients with a COVID-19 critical illness, HRQoL has been shown to decline after discharge from the ICU.3 Post-ICU outcomes, including the short-term survival rate and long-term, multifaceted outcomes, must be improved. Patients discharged from the ICU experience some cognitive and physical decline,2 which negatively affects the HRQoL after ICU discharge. Furthermore, mental health problems, such as post-traumatic stress disorder (PTSD),3 anxiety and depression may contribute to decreased HRQoL.4 These cognitive, physical and psychological impairments that occur during and after critical illness are known as post-intensive care syndrome.5
Several studies have reported the occurrence of PTSD after ICU discharge. One meta-analysis demonstrated that the prevalence of PTSD‐related symptoms associated with ICU stay was 19.83% (95% CI: 16.72% to 23.13%).6 A previous study in the UK suggested that 45.7% of patients had anxiety and 41% had depression 3 months after discharge from the ICU.7 Symptoms, such as constant fear, panic attacks, crying without a reason and feelings of isolation,8 affect postdischarge rehabilitation and communication, not only hindering physical recovery but also negatively impacting return to work.9 The presence of mental health problems in patients after discharge from the ICU can have a serious impact on their daily life, which is manifested as a decrease in HRQoL.
Social support may result in better health and physical function. Studies involving patients with cancer indicate that social support reduced depressive symptoms and improved HRQoL.10 In patients with cardiac diseases, social support was found to decrease depressive symptoms.11 However, previous reports on the association between social support and mental health after ICU discharge are inconsistent. Deja et al
12 reported that social support was negatively associated with PTSD symptoms in ICU patients admitted for acute respiratory distress syndrome. In contrast, Kapfhammer et al
13 found no correlation between social support and PTSD. Moreover, there are few reports on the causal relationship between social support before ICU admission and mental health after ICU discharge.
This study aimed to investigate the association between social support before ICU admission and mental health after ICU discharge.
Methods
Study design and population
This prospective cohort study enrolled patients from an eight‐bed medical–surgical ICU of a hospital in Japan between December 2020 and June 2022. This work follows the Strengthening the Reporting of Observational Studies in Epidemiology checklist,14 which is a reporting guideline for observational studies.
Participants
The inclusion criteria were: (1) patients aged >18 years, (2) ICU stay of least 48 hours and (3) subsequent discharge from the hospital. The exclusion criteria were: (1) abnormal central nervous system function (determined based on diagnostic imaging), including stroke, traumatic brain injury and cerebral tumours, (2) severe cognitive impairment, (3) severe psychotic illnesses under treatment, (4) withdrawal and withholding during ICU admission, (5) admission to the ICU within the previous 3 months, (6) direct transfer to another hospital during their ICU stay, (7) inability to complete a self-administered questionnaire, (8) hospital admission postdischarge, (9) residence in a care home 3 months after discharge from the ICU, (10) no telephone communication and (11) refusal to participate in the survey.
Patients were further screened based on their medical records. Between 2 days after admission to the ICU and 2 weeks after discharge from the ICU, the researcher visited the patients and explained the study. After confirming that the effects of sedation were minimal, namely, that the patients’ level of consciousness was alert and the Japanese version of the Confusion Assessment Method for the ICU15 16 was negative, and after obtaining written consent, the patients were asked to complete a questionnaire regarding social support. A mental health questionnaire was sent via the mail and collected 3 months after ICU discharge. If no reply was received within 2 weeks, the patients were contacted by phone. If it was difficult for the patient to reply by mail, answers to the questionnaire were collected via telephone interviews.
Data collection
The survey items on the questionnaires were based on the Duke Social Support Index Japanese (DSSI-J),17 Impact of Event Scale-Revised (IES-R)18 and Hospital Anxiety and Depression Scale (HADS).19 Data on patient characteristics, delirium during ICU stay and hospital outcomes were retrospectively collected from the participants’ medical records.
Patient demographics, diagnoses and the Charlson Comorbidity Index20 were extracted from the participants’ medical records. The Acute Physiology and Chronic Health Evaluation II score21 was calculated using data obtained within 24 hours of ICU admission. The severity of illness during ICU stay was assessed using the Sequential Organ Failure Assessment Score.22 23 Daily sedatives and/or opioid medication doses were recorded throughout the ICU stay. Delirium was routinely assessed by nurses at the bedside at least twice a day using the Intensive Care Delirium Screening Checklist.24 Similarly, the Richmond Agitation‒Sedation Scale (RASS)25 was used for routine assessment. A positive result obtained at least once in the screenings was considered to indicate delirium, and the duration of delirium (total number of days) was recorded. We defined the coma RASS score as <−4.
Social support
We used the DSSI-J,17 a short version of the Duke Social Support Index,26 to evaluate social support. It is a self-administered questionnaire consisting of 15 items with three factors: emotional support, instrumental support and cognitive evaluation support. The reliability and validity of the Japanese version have been examined.17
Mental status
We used the Japanese versions of the IES-R18 and HADS19 without modifications to assess mental status. The IES-R is widely used to assess PTSD-related symptoms. It is a self-administered questionnaire consisting of 22 items: 8 items for intrusive symptoms, 8 items for avoidance symptoms and 6 items for hyperarousal symptoms. Each item was rated on a 5-point scale from 0 to 4. A cut-off score of 24/25 was used.18 The reliability and validity of the Japanese version have been established, and it is widely used in Japan.
The HADS has been widely used to evaluate the degree of anxiety and depressive symptoms. It is a self-administered questionnaire that consists of 14 items. Each item is scored on a 4-point scale from 0 to 3, with a total score ranging from 0 to 42. Half of the items are related to anxiety symptoms, and the other half are related to depressive symptoms. A score ≥8 in each domain indicates substantial anxiety or depressive symptoms in the Japanese version of the HADS.19 The reliability and validity of the Japanese version were examined.
Statistical analysis
Sample size
The pwr package in the statistical analysis software R was used to calculate the sample size in the general linear model. Based on the primary endpoint, the objective variable, such as the IES-R score, an effect size of 0.13, a power of 0.8, a significance level of 0.05 and four explanatory variables (DSSI-J, age, sex, and years of education), the minimum sample size was 104.
Data analysis
Descriptive statistics were used for analysis. Continuous variables are represented as medians with IQRs unless otherwise specified. The normality of the data was verified by visual inspection and using the Shapiro-Wilk test. Categorical variables are represented as numbers, percentages and 95% CIs. The Fisher’s exact test was used to compare two or more categorical variables.
The RASS Score was calculated as the median of the daily measurements throughout the sedation period. The doses of sedatives and/or opioid medications were calculated as the median dose per day throughout the treatment period.
The missing items on the HADS and IES-R were imputed using the ‘half rule’ defined as follows: If half of the items on the subscale had a response, the mean of the items with a response was imputed.27 Multiple imputations by chained equation (MICE)28 were also attempted for items that were not imputed by the ‘half rule’ owing to the large number of missing items in order to avoid selection bias. MICE was used for the total IES-R and HADS subscales. The imputed data were only used for multivariate analysis. The IES-R and HADS data were not normally distributed and therefore transformed to natural logarithms. When each score was 0, the natural logarithm transformation resulted in infinity, and +1 was added to the score, which was transformed into the natural logarithm for analysis.
Covariates affecting social support and PTSD-related symptoms were predefined based on previous studies. Age, sex and years of education were selected as covariates affecting these variables, as well as anxiety and depression symptoms. To adjust for these covariates, the relationship between social support before admission and mental health after ICU discharge was evaluated using linear regression. To examine in detail the association between mental health problems and social support, linear regression analyses were conducted separately for men and women.
All statistical analyses were two-sided at a 5% significance level, except for interaction analyses. Due to the underpowered nature of the interaction analysis, a two-sided significance level of 20% was used for all interactions.29 All statistical analyses were performed with R (V.4.2.1.) software using the rms package.
Sensitivity analysis
A sensitivity analysis was used to test the robustness of the model assumptions. The results of the linear regression analysis, with the objective variable treated as a continuous variable, were compared with the results of the logistic regression analysis, in which the objective variable was treated as a binary variable with a cut-off value, to confirm the consistency of results when using different statistical analysis methods.
Missing values were compared with the results of the multiple imputation method and those of the complete case analysis.
Since exposure to accidents and disasters might affect PTSD, anxiety and depressive symptoms, a similar analysis excluding trauma patients was performed, and the results were compared.
Patient and public involvement
There was no patient or public involvement in the design, conduct, reporting or dissemination plans of this research.
Results
Participant characteristics
Figure 1 shows a flowchart of the study enrollment process. Of the 650 patients admitted to the ICU, 153 were enrolled in this study and completed a questionnaire on social support prior to ICU admission. The mental health questionnaires were sent 3 months after discharge from the ICU. The survey forms were sent via the mail 3 months after discharge from the ICU. Patients who had died, resided in a care home, were admitted to a hospital, or were readmitted to the ICU within 3 months were excluded. Three patients refused to provide consent and 116 returned the completed survey forms (response rate=90.6%). Of the 116 respondents who returned the survey form, 1 patient did not answer the survey form. Additionally, 12 non-responders and one patient who provided incomplete data were treated as non-responders. The IES-R had missing values in 6 of 115 (5.2%) complete questionnaires, and the ‘half rule’ was used to impute missing values. All IES-R scores were imputed using the ‘half rule’. HADS did not have missing values, eliminating the need for multiple imputations.
Flowchart of the patient enrollment process. CNS, central nervous system; ICU, intensive care unit.
The median age of the patients was 74.0 years (IQR=64.5–80.0). Regarding the type of admission, approximately 70% were admitted for elective surgery, with cardiac surgery occurring most frequently. Overall, 54% had a high school education, 31.1% were an active worker at the time of the study, 68.7% were unemployed and 23.5% were living alone (table 1). The clinical characteristics of the non-responders are presented in online supplemental table 1.
Supplemental material
Baseline patient demographics (N=115)
Mental status 3 months after ICU discharge
PTSD-related symptoms
The median IES-R total score was 3 (IQR=1–12); the prevalence of suspected PTSD (IES-R total >24) was 11.3%.
Anxiety and depressive symptoms
The median HADS anxiety score was 2.5 (IQR=1.00–5.75), and the prevalence of anxiety symptoms was 14.0%. The median HADS depression score was 5 (IQR=2.25–7.00), and the prevalence of depressive symptoms was 24.6%.
Relationship between social support and mental health
The results of the univariate and multivariate analyses using linear regression models of the association between PTSD-related symptom severity and social support variables are presented in online supplemental table 2. The multivariate linear model adjusted for predefined covariates revealed no independent factors associated with PTSD severity (online supplemental table 2).
Supplemental material
The results of the univariate and multivariate analyses using linear regression models for the association between anxiety symptom severity and social support variables are presented in online supplemental table 3. After adjusting for the predefined covariates, no independent factors associated with anxiety severity (online supplemental table 3) were observed. The results of univariate and multivariate analyses using linear regression models for the association between the severity of depressive symptoms and social support variables are presented in table 2. In the linear regression models adjusted for the predefined covariates, female sex (β=0.268, 95% CI: 0.005 to 0.531, p=0.046) and social support (β=−0.018, 95% CI: −0.029 to −0.006, p=0.002) were independent factors associated with depression severity (table 2). Additionally, figure 2 shows the relationship between mental health symptoms (PTSD, anxiety and depressive symptoms) and social support in men and women after adjusting for the predefined covariates. Sex differences were observed in the association between depressive symptoms and social support (p for interaction=0.056).
Supplemental material
Risk factors for depressive symptoms in the univariate and multivariate analyses using linear regression
Relationship between social support and mental health symptoms in men and women, after adjusting for age and education years as covariates, using a generalised linear model. (A) PTSD symptoms, (B) depressive symptoms and (C) anxiety symptoms. Grey areas show 95% CIs. DSSI-J, Duke Social Support Index Japanese; HADS, Hospital Anxiety and Depression Scale; IES-R, Impact of Event Scale-Revised; PTSD, post-traumatic stress disorder.
Sensitivity analysis
The results of the multivariate logistic regression and linear regression analyses were compared. PTSD-related symptoms were not significantly associated with social support (OR: 1.00, 95% CI: 0.954 to 1.06). Similarly, anxiety symptoms were not significantly associated with social support (OR: 0.987, 95% CI: 0.942 to 1.03). However, depressive symptoms were significantly associated with social support (OR: 0.959, 95% CI: 0.923 to 0.998), consistent with the results of the linear regression analysis.
Next, a complete case analysis was performed, excluding the data of the six patients with missing IES-R data. In the multivariate linear models adjusted for the predefined covariates, no independent factors were found to be associated with PTSD severity.
Four trauma patients were recruited, although they were excluded 3 months after ICU discharge, eliminating the need to exclude them from the analysis.
Discussion
This prospective cohort study examined the association between social support prior to ICU admission and PTSD, anxiety and depressive symptoms at 3 months after ICU discharge in patients admitted to the ICU for more than 48 hours. In a multivariate analysis adjusted for the confounding variables of age, sex, and years of education, social support prior to ICU admission was not associated with PTSD-related symptoms or anxiety symptoms after ICU discharge. However, social support prior to ICU admission (β=−0.018, 95% CI: −0.029 to −0.006, p=0.002) and the female sex (β=0.268, 95% CI: 0.005 to 0.531, p=0.046) were independent factors for depressive symptoms. Furthermore, sex differences were observed in the associations between depressive symptoms and social support (p for interaction=0.056).
Social support prior to ICU admission may not be associated with PTSD-related symptoms after ICU discharge. This result was contrary to the results of many studies showing an association between higher social support and a lower incidence of PTSD-related symptoms.30 31 However, some studies investigating the bidirectional association between social support and PTSD-related symptoms, not only among ICU patients, have reported that social support has no effect on subsequent PTSD-related symptoms.32 33 Contributing factors for the lack of an association between social support before ICU admission and PTSD-related symptoms after ICU discharge may include the type of trauma experienced.
Trauma type has been shown to be a significant predictor of PTSD.31 Studies of individuals exposed to combat/war and interpersonal violence showed effectively reduced PTSD-related symptoms with social support compared with studies of individuals exposed to natural disasters.31 Thus, the trauma requires a specific coping response for the support to be effective.34 35 ICU patients experience physical pain12 36 37 and delusional memories,38–40 including being killed by aliens, among a variety of other experiences. Depending on the type of trauma experienced by ICU patients, social support may have different effects on PTSD-related symptoms.
In addition, the country of origin and sociocultural characteristics may influence the occurrence of PTSD-related symptoms. Compared with European Americans, Asians and Asian Americans are less willing to seek social support when coping with stress,41 and they find social support less helpful in coping with stress.42 Moreover, the prevalence of PTSD may be relatively low in South and East Asian countries.43 44 Japan has the lowest prevalence of PTSD in the world.45 Further research is required to account for ethnocultural factors.
Although the association between social support and anxiety disorders is unclear, the negative effects of anxiety may be primarily caused by comorbid depression.46 Regardless of age, social support has a protective impact on depressive symptoms over the lifetime,47 and studies of cancer patients10 48 and older adults49 50 have shown similar results. The results of this study also suggest that social support before ICU admission may be associated with depressive symptoms after ICU discharge based on stress-buffering effects.
Several studies have shown that the relationship between depressive symptoms and social support differs between men and women.51 52 Women are more likely to be depressed than men, but the association between depression and social support was stronger for men than for women.51 52 Similarly, our study showed that women were independently associated with depressive symptoms, while men with low social support were more vulnerable to depressive symptoms than women. Some reports indicated that no stress-buffering effect was present in women,53 which might indicate that sex differences in stress-buffering effects may be present, which suggests that not only women but also men with low social support must be especially aware of depressive symptoms. In hospital settings, ICU nurses play an essential role in assessing a patient’s need for social support, including their social background and family, from the time of admission to providing care for life after discharge. Furthermore, current depressive symptoms could be more strongly associated with perceptions of social support because depression affects these perceptions. Therefore, it is important to pay attention to the changes in patient perceptions and consider providing the necessary social support when needed.
This study has some limitations. First, the sample size was relatively small, and patients were recruited from a single institution. The characteristics of the ICU patients (ie, a high percentage was admitted for elective surgery) reflected the findings. Thus, different populations of patients admitted to the ICU may have led to different results; thus, this may have affected the external validity. Second, as avoidance is a PTSD-related symptom, patients who refused to participate or were non-responders may have had PTSD-related symptoms. However, the number of non-responders was 13; hence, it was unlikely that this influenced the results. Third, social support after ICU discharge was not measured; therefore, it was not possible to include a discussion on the impact of social support after ICU discharge. Finally, recall bias may have occurred in responses to social support prior to ICU admission, which was unlikely to have affected the results, as the survey was conducted within a short time period of 2 weeks after discharge from the ICU.
Despite these limitations, the present findings are novel, as this is the first study to characterise the association between preadmission social support for ICU patients and their mental health after discharge from the ICU. The results of this study may help detect mental health problems after ICU discharge, especially in patients at a high risk for depressive symptoms.
Conclusions
Perceived social support prior to ICU admission is associated with depressive symptoms, but not with PTSD-related symptoms, after ICU discharge. Therefore, attention must be given to patients with low social support before ICU admission regarding depressive symptoms after ICU discharge.
Data availability statement
All data relevant to the study are included in the article or uploaded as supplementary information.
Ethics statements
Patient consent for publication
Ethics approval
This study involves human participants and was approved by Sapporo City University School of Nursing Ethics Review Committee (No.12, 10 June 2020). Participants gave informed consent to participate in the study before taking part.
Acknowledgments
We thank the patients who participated in this study and the hospital administrators and staff for their cooperation. We also thank Dr Jun Makino for his kind attention and cooperation in conducting this study.
This post was originally published on https://bmjopen.bmj.com