Predictors of suicidality in critical care ICU patients after discharge: A cross-sectional study

Abstract

Background:

Suicidality is a critical concern among patients recovering from intensive care unit (ICU) admissions due to their heightened vulnerability to psychological stressors and psychiatric illness. Quantitative studies specifically examining how these ICU-related stressors correlate with suicidal ideation remain limited, indicating a critical need for further research in this area.

Aim:

This study explores the direct and indirect factors associated with suicidality and predictors of suicidality among patients discharged from the ICU.

Materials and Methods:

This cross-sectional observational study was conducted at a tertiary care hospital in Uttar Pradesh, India. 315 patients were screened, and 250 adult participants were selected using convenience sampling. These participants were discharged from the medical ICU and recruited during follow-up visits to the psychiatry outpatient department between July 2021 and July 2022. Inclusion criteria encompassed individuals aged 18 years or above who had been discharged within 1 month of ICU stay and provided informed consent. Demographic and clinical variables were collected, including psychiatric diagnoses according to ICD-10 criteria. Suicidality was broadly defined to include suicidal ideation, planning, and attempts. Statistical analyses included Chi-square tests, Structural Equation Modeling (SEM) for direct and indirect associations, and machine learning-based Decision Tree Classification for prediction analysis.

Results:

The majority of participants were aged >30 years (83.9%), female (56.4%), and from urban areas (52.8%). Significant associations with suicidality were observed for family history of mental illness (P = 0.004), substance use (P < 0.001), medical comorbidities (P < 0.001), and co-occurring psychiatric illnesses along with depression (P < 0.001). SEM revealed that severe depression, co-occurring psychiatric illnesses, substance use, and extended ICU stays (>7 days) directly influenced suicidality, with past psychiatric history exerting an indirect effect through the severity of depression. Decision tree analysis ranked “more than one co-occurring psychiatric illness along with depression” as the most critical predictor, followed by “duration of ICU stay,” “severity of depression,” “past psychiatric history,” and “substance use history.”

Conclusion:

This study highlights the complex interplay of clinical and psychiatric factors associated with suicidality among post-ICU patients. The findings underscore the importance of comprehensive psychiatric screening and targeted interventions for high-risk individuals during their recovery phase to mitigate suicide risk.

Suicidal thoughts among patients who have survived critical illness in intensive care units (ICUs) represent a significant mental health issue that is gaining attention in the medical community.[1] With advancements in critical care medicine resulting in increased survival rates, a substantial number of patients discharged from ICUs confront enduring psychological challenges.[2] A growing body of evidence indicates that survivors of critical illness often suffer from post-intensive care syndrome (PICS), which includes cognitive, psychological, and physical impairments.[3] Among these psychological effects, conditions such as depression, anxiety, and post-traumatic stress disorder (PTSD) are commonly reported.[4] Suicidal ideation, a severe outcome of such mental health disturbances, remains poorly understood in the context of post-ICU patients, despite the profound impact it can have on long-term recovery and quality of life.[1,2] Psychological effects such as depression, anxiety, and PTSD are prevalent in ICU survivors, with approximately 30%, 20%, and 30%, respectively, reporting these symptoms within the first-month postdischarge.[5,6] These conditions, alongside persistent physical impairments and chronic pain, contribute significantly to the psychological burden and may lead to suicidal ideation in nearly 20% of survivors.[7]

Suicidal ideation, a serious consequence of these mental health disturbances, remains inadequately explored in post-ICU patients, despite its potential impacts on long-term recovery trajectories and overall quality of life.[1,2] Empirical evidence reveals that suicidal thoughts affect approximately 20% of ICU survivors within 1 month of discharge.[8] The interplay between psychological distress and suicidal ideation underscores the urgency of addressing mental health in post-ICU care. Factors such as prolonged sedation, mechanical ventilation, and the psychological toll of invasive medical procedures contribute significantly to the emotional challenges faced during recovery.[9,10,11] However, quantitative studies specifically examining how these ICU-related stressors correlate with suicidal ideation remain limited, indicating a critical need for further research in this area. This study aims to systematically identify the predictors of suicidal ideation in ICU survivors, addressing a gap in research and paving the way for targeted mental health interventions post-ICU discharge.

MATERIALS AND METHODS

This cross-sectional observational study was conducted at a tertiary care hospital in Uttar Pradesh, India, focusing on patients who had recovered from various medical conditions after being discharged from the medical ICU. Data were collected during the patient’s follow-up visits to the outpatient department (OPD) as part of routine care or through liaison consultations with psychiatric services. We employed a convenience sampling technique, recruiting individuals from July 2021 to July 2022. The primary aim of this research was to investigate predictors of suicidality among patients discharged from the ICU. The objectives of this study were to identify the direct and indirect factors associated with suicidality among post-ICU patients during their recovery phase and to establish the key predictors of suicidality. To achieve this, we used Structural Equation Modeling (SEM) to analyze associations and a Decision Tree Classification (DCT) model to identify key predictors of suicidality.

Participants were recruited based on predefined inclusion criteria and provided informed consent before enrolment. The inclusion criteria required patients to be aged 18 years or older, have been admitted to the medical ICU, and have visited the psychiatry outpatient department (OPD) within 1 month of discharge. Patients were excluded if they were uncooperative for assessment, medically unstable, or declined to participate. Written informed consent was obtained after researchers thoroughly explained the study’s objectives and procedures, ensuring that participants fully understood the purpose of the research. Any patient meeting the inclusion and exclusion criteria and providing consent within the specified study period was included. A total of 315 patients were screened, and 250 participants were ultimately included using convenience sampling. Since a convenience sampling method was employed, no formal sample size calculation was conducted. A total of 315 patients were screened, out of which 250 participants were ultimately included using convenience sampling. A total of 65 patients were excluded due to various reasons: 30 patients were medically unstable, 20 patients were uncooperative for assessment, and 15 patients declined to participate in the study [Figure 1].

Figure 1.

Inclusion and exclusion of the participants for the study

Demographic information such as age, gender, and educational background was collected alongside clinical details, including the nature of their illness, utilization of oxygen support during hospitalization, the duration of their ICU stay, and any history of psychiatric illness. Psychiatric screening was conducted by a trained psychiatrist with 5 years of experience, and diagnoses were made according to ICD-10 criteria. In our study, suicidality was considered an umbrella term, encompassing all facets of suicidal behavior and thoughts, including ideation, planning, and attempts. The data collected were primarily categorical, so the Chi-square test was used to compare independent groups, and the threshold for statistical significance was set at P < 0.05.

To examine both the direct and indirect associations of various factors with suicidality, we employed SEM, which allows for the simultaneous analysis of multiple dependent and independent variables, accounts for latent constructs, and provides insights into causal pathways that traditional regression models may overlook. To establish the hierarchy of risk factors and predict suicidality, we utilized a DCT model with an 80:20 train–test split. DCT was selected due to its ability to handle imbalanced datasets and small sample sizes, making it more robust for predictive modeling in this study than traditional logistic regression. Additionally, DCT provides easily interpretable decision rules, which enhance its clinical applicability. Unlike conventional regression models, which assume linear relationships, SEM accommodates complex, multilayered associations, while DCT excels at handling nonlinear interactions and ranking risk factors based on importance. By combining these two approaches, we ensured a comprehensive understanding of both association and prediction. All statistical analyses were conducted using SPSS Version 27.0.0, while R Version 4.3.3 was used for SEM analysis and decision tree modeling.

RESULTS

The study analyzed 250 subjects with suicidality discharged from ICU and revealed that most were aged above 30 years (83.9%), were female (56.4%), belonged to urban areas (52.8%), and lived in nuclear families (62.4%) [Table 1]. Suicidality was found to have significant associations with family history of mental illness (67.6%, P = 0.004), substance use (78.4%, P < 0.001), medical comorbidities (74.8%, P < 0.001), and co-occurring psychiatric illnesses along with depression (87.2%, P < 0.001). Among the individuals discharged, a significant number (92.8%, P < 0.001) required oxygen support. Similarly, significant numbers of individuals with suicidality (92.8%, P < 0.001) required a stay of 7 days or more in ICU as cited in Table 2. Among the participants with suicidality, a significant number (87.6%, P < 0.001) perceived their physical condition as severe and severe depression was noted among a significant number, 26.8% (P < 0.001). The SEM analysis identified factors influencing suicidality among participants discharged from ICU [Figure 2]. Substance abuse in dependence patterns, co-occurring mental illness, and duration of stay in ICU were found to be associated directly with suicidality and past history of psychiatric illness was indirectly associated with suicidality, whereas the severity of depression was associated directly as well as indirectly with suicidality.

Table 1.

Sociodemographic details of the subjects

Characteristics of the subjects		Subjects with suicidality (n=250)
Age	18-30 yrs.	40 (16%)
	31 – 45 yrs.	71 (28.4%)
	46-60 yrs.	53 (21.2%)
	>60 years	86 (34.3%)
Sex	Female	141 (56.4%)
	Male	109 (43.6%)
Domicile	Urban	132 (52.8%)
	Rural	118 (47.2%)
Educational Status	Less than 5 years of education	72 (28.8%)
	Completed school	95 (38%)
	Graduate and above	83 (33.2%)
Family type	Nuclear family	156 (62.4%)
	Joint family	94 (37.6%)

Table 2.

Comparison of clinical variables among the study population

Clinical variables (Total subjects=250)	Yes n (%)	No n (%)	P
History of mental illness in the past	129 (51.6%)	121 (48.4%)	0.87
History of substance use in dependence pattern (in the past 1 year)	196 (78.4%)	54 (21.6%)	<0.001
Family history of mental illness	169 (67.6%)	81 (32.4%)	0.004
More than one co-occurring Psychiatric illness along with depression	218 (87.2%)	32 (12.8%)	<0.001
Medical comorbidity	187 (74.8%)	63 (25.2%)	<0.001
Requirement of oxygen support during ICU stay	232 (92.8%)	18 (7.2%)	<0.001
	Less than 7 days n (%)	7 days or more n (%)	P
Duration of ICU stay	18 (7.2%)	232 (92.8%)	<0.001
	Mild/Moderate n (%)	Severe n (%)	P
Perceived severity of physical condition during ICU stays	31 (12.4%)	219 (87.6%)	<0.001
Depression	67 (26.8%)	193 (77.2%)	<0.001

Figure 2.

SEM analysis of factors influencing suicidality

A past psychiatric history indirectly influences suicidality by influencing the severity of depression, which has a direct impact on suicide risk. More than one co-occurring psychiatric illness along with depression, the severity of depression, and longer hospital directly contribute to suicidality and substance use in dependence pattern in the past 12 months, and longer hospital stays were found to directly influence suicidality, suggesting that extended stays may lead to increased psychological distress. The perceived severity of the physical condition and the need for oxygen in the ICU did not show any effect on suicidality cited in Figure 2. Decision tree analysis [Figure 3] shows the hierarchy of risk factors in predicting suicidality. On comparison, “more than one co-occurring psychiatric illness along with depression” ranked as the most important predictor of suicidality. It was followed by factors such as the “duration of ICU stay”, “severity of depression”, “past psychiatric history”, and “substance use history” as the top five notable predictors in order of importance of prediction. The sensitivity of decision tree model was 85%, the specificity was 83%, and the area under the curve (AUC) was 87%.

Figure 3.

Decision tree analysis of predictors of suicidality

Figure 2 shows SEM analysis of factors influencing suicidality. Factors influencing suicidality directly are indicated by bold arrows, while factors influencing suicidality indirectly are indicated using dotted arrows. (Only the significant variables are mentioned in the figure).

Figure 3 shows the hierarchy of predictors of suicidality after decision tree analysis of risk factors.

DISCUSSION

The present study examined the factors influencing suicidality in 250 subjects, revealing key associations with age, gender, and family history of mental illness. A majority of the participants were 30 years and above (83.9%), females (56.4%), and participants from urban areas (52.8%). This demographic profile is consistent with previous studies which have shown that females and individuals from urban areas are at a higher risk for suicidality.[5] Higher educational status also played a significant role, with 38% of participants completing school and 33.2% graduating or above. Contrary to findings in our study, previous studies have highlighted lower educational levels being linked to higher suicidality risk. There is a higher education level and living in urban society. This finding may be attributed to the higher stress levels, social pressures, and increased awareness of mental health challenges often associated with urban living and higher education levels, potentially contributing to suicidality risk.[12,13] Suicidality was found to have significant associations with family history of mental illness substance use, medical comorbidities, and co-occurring psychiatric illnesses along with depression. A family history of mental illness was found to be a significant predictor of suicidality in our study. This is in congruence with the well-documented fact of role of a family history of mental illness in increasing suicidality risk, in the existing literature. Thus, family history of mental illness significantly increases the risk of suicidality through inherited vulnerabilities and learned maladaptive coping mechanisms.[14,15]

This study found that substance abuse was a predictor of suicidality. This finding also aligns with existing literature, which suggests that substance use exacerbates impulsivity, emotional dysregulation, and social isolation, and neurobiological changes, thereby increasing suicidality risk.[16,17] Medical comorbidities also emerged as another significant predictor, further supporting the well-established link between physical health and suicidality. This finding is consistent with literature indicating that medical comorbidities contribute to suicidality by increasing psychological distress, reducing quality of life, and fostering feelings of hopelessness, particularly in chronic or debilitating conditions.[18,19]

The SEM analysis provided valuable insights into factors influencing suicidality directly or indirectly. Past psychiatric history was found to indirectly influence suicidality by exacerbating the severity of depression, which directly contributed to suicide risk. This finding is consistent with the literature, which highlights how past psychiatric disorders increase emotional vulnerability, impair coping mechanisms, and perpetuate cycles of distress, ultimately elevating the risk of suicidality. Additionally, the presence of multiple co-occurring psychiatric disorders alongside depression, the severity of depression itself, and prolonged hospital stays were identified as direct contributors to suicidality. These results align with evidence suggesting that complex psychiatric and medical conditions amplify emotional dysregulation, psychological burden, and hopelessness, thereby heightening the risk of suicidal behaviors.[16,17,18,19]

The decision tree analysis highlighted “more than one co-occurring psychiatric illness along with depression” as the most significant predictor of suicidality, followed by the duration of ICU stay, severity of depression, past psychiatric history, and substance use. These findings align with research indicating that individuals with multiple co-occurring psychiatric disorders experience heightened emotional distress and impaired coping mechanisms, significantly increasing their vulnerability to suicidality.[20,21] The impact of extended ICU stays on suicidality underscores the psychological toll of prolonged hospitalization, including factors such as social isolation, sleep disturbances, and the trauma of invasive procedures. This finding is consistent with studies suggesting that ICU experiences are associated with long-term psychological consequences, such as depression and PTSD, which further elevate suicide risk.[22,23]

Limittaions

The cross-sectional design limits causal inferences, and the reliance on self-reported data for substance use and mental health history may introduce recall bias. Additionally, suicidality was used as an umbrella term without stratification based on severity or lethality, which is a limitation. The lack of a standardized rating scale for suicidality assessment further limits the study. Moreover, while substance use in a dependence pattern emerged as a significant predictor, stratification based on the type of substance was not conducted. The study also did not account for the impact of ICU procedures such as CSF tapping, RT feeding, and catheterization, which may contribute to patient discomfort. Furthermore, economic factors, indirect costs of treatment, and other psychosocial aspects such as sleeping patterns, flashbacks, and cohabitant loss were not explored in detail. Last, since a convenience sampling method was employed, no formal sample size calculation was conducted, and the study was restricted to subjects with suicidality, limiting the generalizability of the findings to the broader population.

CONCLUSION

This study highlights the multifactorial nature of suicidality, with co-occurring psychiatric disorders, severity of depression, and ICU stay duration identified as key risk factors. Comprehensive suicide prevention efforts should focus on assessing both current mental health status and past psychiatric history, alongside managing physical and psychological distress during hospitalizations. Additionally, screening for a family history of psychiatric illness among ICU patients could serve as a vital strategy for early intervention and prevention of suicidality.

Future directions

Future research should focus on conducting longitudinal studies to further investigate the interplay of psychiatric and physical factors in suicidality. Such studies could help in developing targeted, evidence-based interventions tailored to ICU patients and high-risk populations. Additionally, exploring the role of family support and post-discharge care could provide deeper insights into suicide prevention strategies.

Authors’ contributions

Conceptualization: MKS; Data analysis: AJV.; Manuscript writing, editing, and results verification. AJV, AS.; SS: Manuscript editing, and finalizing the manuscript; BS, SS, STK.

Data availability statement

The data supporting the findings of this study are available from the corresponding author upon reasonable request.

Ethical approval

This study was approved by the Institutional Ethics Committee (IEC) of the Autonomous State Medical Council (ASMC/IEC/2020-21/06, dated 7^th May 2021).

Conflicts of interest

There are no conflicts of interest.

Informed consent

Written informed consent was obtained from all before their inclusion in the study.

Funding Statement

Nil.