Factors associated with post-traumatic growth in severe COVID-19 patients discharged from intensive care: the influence of near-death experiences Authors: Amaury C. Mengin, Julie Rolling, Walid Oulehri, Pierre Orselli, Adrien Gras, Fabrice Berna, Pierre Vidailhet Published: 2026 ABSTRACT Intensive Care Unit (ICU) hospitalizations confront individuals with imminent death and may cause post-traumatic stress disorder (PTSD). In contrast, traumatic events may lead to positive psychological outcomes such as post-traumatic growth. As consciousness is altered, near-death experiences (NDEs) may occur and be associated with post-traumatic growth. Our cross-sectional, monocentric study aims to determine the factors associated with post-traumatic growth in severe, ICU-discharged COVID-19 patients, including the influence of NDEs and memory characteristics. We recruited participants three months after their ICU hospitalization. Patients benefitted from a multidisciplinary medical assessment, including a clinical interview with a psychiatrist. Self-administered questionnaires evaluated PTSD symptoms, post-traumatic growth, and NDEs. Fifty-nine patients were included in the study. Mean post-traumatic growth levels were relatively high, while the occurrence of PTSD was low (3.4%). An NDE was identified among 20.3% of participants. Post-traumatic growth was relevantly associated with the presence of an NDE, intense and intrusive memories and higher PTSD symptoms levels. Patients discharged from ICU after severe COVID-19 report high levels of post-traumatic growth, which may be associated to the occurrence of a near-death experience and higher distress. Better identification and psychological support about NDEs could improve post-ICU psychological recovery and well-being. Introduction Post-traumatic growth (PTG) has been defined as “the positive legacy of trauma” by Tedeschi and Calhoun. It includes five dimensions (new possibilities, relationships with others, personal strength, appreciation of life, and spiritual changes)1,2. The occurrence of post-traumatic growth has been studied in various contexts involving distressing events, such as physical or sexual assault, terrorist attacks, and chronic illnesses like cancer and HIV3–6. Several factors that contribute to PTG have been identified, including cognitive processing, positive coping strategies, event-centrality, and personality traits7,8. However, studies on PTG in patients discharged from intensive care units (ICU) are scarce in the current literature and are primarily represented by case studies or qualitative studies9,10. ICU hospitalizations often involve direct confrontation to the risk of death. In this regard, an ICU hospitalization meets criterion A of the DSM-5 for Post-Traumatic Stress Disorder (PTSD), as it involves the exposure to a life-threatening illness, as considered by item 12 of the Life Events Checklist for DSM-5 (LEC-5)11,12. Item 12 has been identified as belonging to “Accidental/Injury Traumas”, one of the three clusters of potentially traumatic events measured by the LEC-5. The other two clusters are “Victimization Trauma” and “Predominant Death Threat Traumas”13. Previous studies have estimated that 19% of patients experience PTSD following an ICU stay, though there is significant variability across studies14,15. Overall, ICU hospitalizations increase the risk of PTSD and other psychiatric conditions such as depressive or anxiety disorders16. In contrast, post-traumatic growth is a positive outcome that can also emerge after potentially traumatic events, such as an ICU hospitalization. However, the importance of post-traumatic growth and the factors associated with its occurrence in such a specific context are still unknown. The context of the COVID-19 pandemic has provoked a sudden and massive increase of admissions in intensive care units due to severe COVID-19 worldwide, and particularly in the first regions of France affected by the pandemic like Alsace17.The international lockdown declared during this period weakened the protective factors from PTSD for the patients admitted in ICU (e.g., unavailability of social support from their families or ICU personal due to work overload). Sustained PTSD symptoms were already observed among patients during the SARS epidemic in China18. In contrast, the unexpected and collective confrontation to a life-threat provoked by the COVID-19 pandemic might have potentialized lifestyle changes and post-traumatic growth among those exposed. Younger age, social support and receiving mental health support might be associated with higher PTG among COVID-19 survivors19–21. Although ICU hospitalization involves an exposure to an imminent risk of death, the experience greatly differs from other traumatic or distressing events. First, these patients are at risk of having no memory of the ICU experience. This risk increases with age, the severity of the underlying medical condition, and the depth of sedation22. Administering corticosteroids or 50% glucose due to hypoglycemia may also increase that risk23. Nevertheless, studies have found that the majority of patients (between 68% and 82%) have vivid memories of their time in the ICU22,24. Following an ICU stay, the presence of distressing or delusional memories, the duration of sedation and the administration of benzodiazepines are associated with a higher risk of PTSD25,26. Conversely, the influence of memory characteristics on post-traumatic growth remain unexplored, though the presence of vivid memories may be necessary to draw positive lessons from a traumatic situation7. Memories of ICU stays often consist of a mix of factual memories (real events) and hallucinatory memories (intrapsychic events). In a recent qualitative review of the literature, Maartmann-Moe et al. identified three main themes of ICU memories: (1) memories of surreal dreams and confusion, (2) memories of care ranging from a sense of security to alienation and (3) memories of being vulnerable or near death27. Regarding memories of proximity to death, a specific type of experience has been identified among survivors: near-death experiences (NDEs). Greyson describes them as “profound psychological events with mystical and transcendental elements occurring in individuals who are close to death or in situations of psychological danger”28. These experiences appear to occur in 12% to 14% of patients resuscitated following cardiac arrest29,30. In patients affected by COVID-19, the severe acute respiratory syndrome might have provoked transient decreases in cerebral oxygen levels and hypercapnia, which could be at play in the neural mechanisms of NDE31. NDEs are generally perceived positively, contain cross-cultural features and stereotypical stages32. Nevertheless, distressing NDEs, culture-specific features, and non-specific NDEs were described33–35. Ring identifies five stages that constitute a “core experience” and tend to manifest in this order: (1) peace and well-being, (2) separation from the physical body, (3) entry into a dark region, (4) vision of a brilliant light, (5) entry through the light into another realm36. Greyson developed and validated a scale (the Near-Death Experience Scale) to identify NDEs and measure their intensity28. Moody includes changes in life conduct as one of the fifteen common traits of NDEs37. Furthermore, Storli et al. identify the search for meaning after an intensive care hospitalization as a crucial stage in regaining life balance following these experiences38. Finally, only one study showed that NDEs were associated with greater post-traumatic growth than close brushes with death in the absence of an NDE. However, near-death events in this population of healthy volunteers were heterogeneous and occurred an average of 34.5 years prior to the study39. Our main objective is to assess the factors associated with post-traumatic growth (PTG) in patients three months after being discharged from ICU due to severe COVID-19. We will focus specifically on the association with NDEs and memory characteristics. We hypothesize that the occurrence of NDEs will be associated with higher levels of post-traumatic growth. Furthermore, we hypothesize that the vividness, strength and centrality of memories during the ICU hospitalization will be associated with higher levels of post-traumatic growth. Materials & methods Study design and population This cross-sectional, monocentric study examined patients discharged from ICU due to severe COVID-19 infection at the University Hospital of Strasbourg in France from March 1, 2020, to May 15, 2020. Inclusion criteria were: (1) patients hospitalized in ICU with (2) diagnosis of COVID-19 confirmed with either detection of SARS-CoV-2 via reverse transcriptase-polymerase chain reaction (RT-PCR) assays on nasopharyngeal smear or bronchoalveolar lavage according to the WHO guidance, or with a chest computer tomography scan compatible with COVID-19; and (3) patients survival after hospital discharge. All patients were invited by mail to participate in a three-month follow up after ICU discharge, and a telephone confirmation for further evaluations in an ambulatory setting. Exclusion criteria for our study were insufficient French language level and suffering from severe cognitive or sensory disabilities impeding the interview. The study and experimental protocol were approved by the Research Ethics Committee of Strasbourg university hospital (RNI 2020 - HUS N° 7890). Written informed consent was obtained from all study participants. Assessments were performed in the anesthesia outpatient facility between June 15 and December 15, 2020. First, patients were examined by an intensivist. Then, all patients participated in several clinical and functional examinations to provide an overview of their status three month after severe COVID-19. A multidisciplinary team comprising neurologists, neuropsychologists, psychiatrists, pulmonologists, internists and intensivists then evaluated patients on the same day. Concerning psychiatric assessment, the patients first completed questionnaires about their mental health and the memory of their hospitalization. A general psychiatric interview was then conducted to identify past and current psychiatric diseases, psychotropic drugs, and any necessary medication adjustments. Finally, the examiner asked the patients to recall the memory of their hospitalization and rated the characteristics of this memory. Measures Sociodemographic data included age, gender, education level, family and professional situation. Post-traumatic growth was measured using the Post-traumatic Growth Inventory (PTGI) questionnaire adapted in French1,40. This questionnaire contains 21 items divided into 5 dimensions: relationship with others (7 items), new possibilities (5 items), personal strength (4 items), appreciation of life (3 items), and spiritual changes (2 items). The total score ranges from 0 to 105. A score of 63 or over is considered a moderate to high PTG. The reference event used for PCL-5, PTGI, NDE and event-centrality was the ICU stay. The internal consistency of the questionnaire was appropriate in our sample (Cronbach’s α = 0.95). Near-death experiences (NDE) were self-assessed with the Greyson NDE scale28. This phenomenological scale is used to identify the presence of a NDE memory by assessing the presence of specific affective, cognitive, transcendental, and paranormal features (e.g. “Did scenes from your past come back to you?”). This 16-item multiple-choice validated scale enables to quantify the richness of the experience (total score ranging from 0 to 32)41. Each item is scored on an ordinal scale ranging from 0 to 2 (0 = “not present”, 1 = “mildly or ambiguously present”, and 2 = “definitively present”). This scale allows a standardized identification of NDEs, where subjects getting a score ≥ 7 are considered as plausibly having experienced an NDE. The internal consistency of the questionnaire was appropriate in our sample (Cronbach’s α = 0.86). Characteristics of hospitalization memories were measured through twelve questions rated on a 1 to 5 Likert scale by the interviewer and one specific question. The questions were inspired by the work of Johnson et al.42. The memories characteristics were chosen as they represented features that are typically associated with the recollective qualities of autobiographical memories and were used in previous studies assessing memories of potentially traumatic events43,44. The twelve questions assessed: 1/Precision of the memory (setting and context of the event), 2/Vividness, 3/Strength of the impression of reliving the event, 4/Continuity and coherence, 5/First-person or third-person visualization, 6/Valence of associated emotions 7/Intensity of the physical reaction while recalling the memory, 8/Flashback manifestations, 9/Frequency of intrusive manifestations of the memory, 10/Uncontrollability (feeling that the memory is stronger than oneself), 11/Sense of ownership of the memory, 12/Level of certainty concerning the event. Participants finally specified whether recalling the memory affected their mood (negatively, positively) or not. Event-centrality was measured by the Centrality of Event Scale (CES) – 7 items version45. This scale assesses the subjective centrality of an event is for one’s identity. It consists of 7 items rated on a Likert scale ranging from 0 to 5 (total score ranging from 0 to 35) (e.g., “This event tells a lot about who I am”). This measure has a good reliability (Cronbach’s α = 0.88). Psychopathology Post-traumatic stress disorder (PTSD) was assessed with Post-traumatic Checklist for DSM-5 (PCL-5)46. It consists of a 20-item scale, with each item scored on a scale ranging from 0 (not at all) to 5 (extremely). This scale was used as a continuous scale and as a diagnostic measure based on DSM-5. PTSD was defined following the DSM-5 diagnostic rule, which requires at least 1 B item (questions 1–5), 1 C item (questions 6–7), 2 D items (questions 8–14) and 2 E items (questions 15–20) scored as 2 (i.e., “moderately”) or higher to diagnose PTSD. This calculation method was validated in French47. Partial PTSD was defined when two or three of the DSM-5 Criteria B, C, D or E were met (McLaughlin et al., 2015). The internal consistency of the questionnaire was appropriate in our sample (Cronbach’s α = 0.91). Anxiety and depression were measured with the Hospital Anxiety and Depression (HAD) Scale48. This 14-item scale encompasses 2 dimensions, with 7 items dedicated to anxiety and 7 items dedicated to depression and was validated in French49. Each item is scored on a scale ranging from 0 to 3. A score over 7 for each dimension respectively is considered as probable anxious or depressive disorder. The internal consistency of the questionnaire was appropriate in our sample (Cronbach’s α = 0.84 for anxiety subscale and α = 0.82 for depression subscale). Suicidal risk was assessed with the suicidal scale of the Mini-International Neuropsychiatric Interview (M.I.N.I.)50. The MINI is a short diagnostic structured interview exploring 17 disorders according to the Diagnostic and Statistical Manual, fifth version (DSM-5) diagnostic criteria and was validated in French51. The suicidal scale includes 6 “yes/no” questions, and defines 4 risk levels: null, low, medium, or high. A low suicide risk was established when the participants answered ‘yes’ to either Q1, Q2 or Q6. A medium suicide risk was established when the participants answered ‘yes’ to Q1 or Q2 + Q6. A high suicide risk was established when the participants answered ‘yes’ to Q4, Q5 or Q3 + Q652,53. ICU parameters We measured the total doses received for the main psychotropic ICU drugs received by each patient in milligrams (hypnotics [propofol, midazolam], opioids [sufentanil, remifentanil], sedatives [clonidine], neuromuscular blockers [curare] and neuroleptics [in haloperidol equivalency]); and the total sedation time (in hours). Statistical analysis Descriptive statistics Sociodemographic characteristics, psychopathology, ICU parameters and memories characteristics were described as mean (SD) for continuous variables and as frequency (%) for categorical data. Inferential statistics The data were analyzed using Bayesian methods. From a probabilistic point of view, the goal of the study was to compute the probability that post-traumatic growth would be greater in the near-death experience group than in the no near-death experience group. For the main outcome analyses, we used weakly informative priors. The main outcome was modeled using a beta regression. The best model was selected Deviance Index Criterion (DIC); the lowest DIC indicates the best model. For each analysis on the main outcome, we present the posterior estimates and posterior probabilities. For readers less familiar with Bayesian methods, posterior estimates can be interpreted as effect sizes with associated uncertainty. Posterior estimates were presented as the posterior mean (M) of each parameter (expressed in odds-ratio for the binary variables), with 95% credible intervals (2.5%–97.5%) reflecting uncertainty. Posterior probabilities (Pr) were used to quantify the strength of evidence for an effect. Values greater than 0.975 or lower than 0.025 were interpreted as strong evidence, although these thresholds were not treated as strict decision rules. The result section will report statistically relevant effects, namely those associated with a probability higher than 0.975 or lower than 0.025. For instance, the difference of post-traumatic growth between NDE and non-NDE participants will be considered statistically relevant if Pr (NDE > no NDE) or Pr (NDE < no NDE) are either > 0.975 or < 0.025 (reminding also that Pr (NDE > no NDE) = 1 – Pr (NDE < no NDE)). In addition, regression coefficients were considered statistically relevant when their probability to be higher than 0 was either > 0.975 or < 0.025. Probabilities lower than 0.05 or higher than 0.95 will be reported as marginally relevant. Multivariable regression analyses were then conducted with post-traumatic growth as a dependent variable to address our main hypothesis. Given the number of univariate analyses and with aim to reduce multivariable model complexity and the risk of overfitting4,54,55, we selected variables showing a tendency of association at the univariate level (Pr > 0.90) or identified as influencing PTG in the literature (gender, age). Multivariable models were then built progressively by sequentially adding selected covariates to examine whether the observed associations remained stable after adjustment. Missing data were handled using joint model-based imputation in JAGS. Under a Missing Completely at Random (MCAR) assumption, missing values in the binary predictors were imputed using a Bernoulli model, missing values in continuous covariates (PCL-5, centrality) were imputed using Student-t model (given their non-normal distribution), and missing values in discrete ordinal covariates (memories characteristics) were handled using a multinomial distribution with Dirichlet prior. Missing values in the continuous outcome (PTG) were estimated directly within the beta regression model as latent variables. Group differences were examined using Bayesian models, estimating posterior distributions of mean differences (age) and proportions (gender). A burn-in of 5000 iterations, followed by 105,000 iterations was used for each of the 3 McMC chains, yielding a final 110,000 iterations-sample for retrieving posterior distribution characteristics. The convergence and autocorrelation of the McMC sample chains were checked with the Gelman-Rubin factor (Rhat) and graphically. Convergence was observed in each case. All computations were done with R version 4.2.3 through R-Studio 2024.04.1 + 748 and JAGS software(s)56. Results Sociodemographic characteristics of the participants One hundred and sixty-four patients were hospitalized in ICU for severe COVID-19 in Strasbourg University Hospital between March and May 2020. The 131 patients who survived were invited to participate in a multidisciplinary follow-up 3 months after ICU discharge. Among them, 79 completed a psychiatric assessment 3-months after their hospitalization. Finally, 59 patients were included in our study (the flow chart of our study is presented in Fig. 1). Fig. 1Flow chart. Flow chart. Among the participants included, 47 (80%) were males. The median [IQR] age was 63 [29 – 80] years, and 18 (37%) patients had a bachelor level or higher. Fifty-three patients (91%) had children, 47 (80%) were married or in a relationship, 30 (61%) were retired. Mean ICU stay duration was 12.2 ± 8.4 days. The relation between post-traumatic growth and near-death experiences The mean post-traumatic growth score was 44.5 (SD = 29.1), with 17 patients (28.8%) presenting a moderate to high PTG. The mean NDE score was 3.6 (SD = 4.6) on a scale ranged from 0 to 32. An NDE (score ≥ 7) was identified among 12 (20.3%) participants. Figure 2 presents the relevant association between post-traumatic growth and the presence of an NDE (Pr (NDE > no NDE) = 0.999). All PTG dimensions were relevantly associated with the presence of an NDE (Pr (NDE > no NDE) > 0.992), except “appreciation of life” which was positively associated with NDE but remained under the relevance threshold (Pr (NDE > no NDE) = 0.912). Fig. 2Comparison of post-traumatic growth scores between patients reporting a near-death experience and those who do not. NDE, Near-Death Experience; PTGI, Post-Traumatic Growth Inventory. ***Pr (NDE > No NDE) > 0.999. Comparison of post-traumatic growth scores between patients reporting a near-death experience and those who do not. NDE, Near-Death Experience; PTGI, Post-Traumatic Growth Inventory. ***Pr (NDE > No NDE) > 0.999. Other factors associated with post-traumatic growth Sociodemographic characteristics Age, gender (male), retirement and having children were positively associated with higher PTG, under the relevance threshold (Prs < 0.928). Being married was not associated with PTG (Pr = 0.554), while higher education level was negatively associated with PTG, under the relevance threshold (Pr = 0.108). Memories characteristics Fifty-nine hospitalization memories were recorded and assessed by the interviewers. The sense of ownership and certainty of the memory were rated high (i.e. M between 4.30 and 4.58/5), emotional valence was rather positive (M = 3.00). Precision, vividness and continuity of the memories were rather high (i.e. M between 3.06 and 3.57/5), while the strength of the memory, intrusions, flashbacks, uncontrollability and the presence of a physical reaction while recalling the memory were low (i.e. M between 1.25 and 1.51/5). We compared the characteristics of these memories between participants reporting an NDE and those who did not in Table 1. Table 1Characteristics of hospitalization memories and their comparison between patients with and without NDE.No NDE(n = 47)NDE(n = 12)m (± sd)m (± sd)Precision3.51 (± 1.66)3.82 (± 1.54)Vividness3.12 (± 1.62)2.82 (± 1.60)Strength1.45 (± 1.13)1.36 (± 0.92)Continuity & Coherence3.38 (± 1.82)2.36 (± 1.75)$First-person to Third-person1.85 (± 1.46)2.10 (± 1.52)Emotional valence3.00 (± 1.00)3.00 (± 1.26)Physical reaction1.45 (± 0.88)1.73 (± 1.27)Flashbacks frequency1.95 (± 1.54)2.18 (± 1.47)Intrusions1.44 (± 1.02)1.73 (± 1.35)Uncontrollability1.27 (± 0.78)1.18 (± 0.60)Sense of ownership4.62 (± 1.00)4.40 (± 1.35)Certainty4.40 (± 1.17)3.90 (± 1.20)Centrality21.30 (± 8.81)29.40 (± 5.87)*This table present memories characteristics among the participants and compares these characteristics between participants reporting a near-death experience (NDE) and those who do not (no NDE). The memories were rated by the interviewers on a Likert-scale ranging from 0 to 5. Event-centrality was self-rated by the patients, the total score ranges from 0 to 35 .$Pr (NDE > No NDE) < 0.050; *Pr (NDE > No NDE) > 0.975. Here, the probability that continuity & coherence of the memory is superior in the ‘NDE’ group than in the ‘No NDE’ group is low (marginally relevant). The probability that centrality of the memory is superior in the ‘NDE’ group than in the ‘No NDE’ group is high (relevant). Table 1Characteristics of hospitalization memories and their comparison between patients with and without NDE.No NDE(n = 47)NDE(n = 12)m (± sd)m (± sd)Precision3.51 (± 1.66)3.82 (± 1.54)Vividness3.12 (± 1.62)2.82 (± 1.60)Strength1.45 (± 1.13)1.36 (± 0.92)Continuity & Coherence3.38 (± 1.82)2.36 (± 1.75)$First-person to Third-person1.85 (± 1.46)2.10 (± 1.52)Emotional valence3.00 (± 1.00)3.00 (± 1.26)Physical reaction1.45 (± 0.88)1.73 (± 1.27)Flashbacks frequency1.95 (± 1.54)2.18 (± 1.47)Intrusions1.44 (± 1.02)1.73 (± 1.35)Uncontrollability1.27 (± 0.78)1.18 (± 0.60)Sense of ownership4.62 (± 1.00)4.40 (± 1.35)Certainty4.40 (± 1.17)3.90 (± 1.20)Centrality21.30 (± 8.81)29.40 (± 5.87)*This table present memories characteristics among the participants and compares these characteristics between participants reporting a near-death experience (NDE) and those who do not (no NDE). The memories were rated by the interviewers on a Likert-scale ranging from 0 to 5. Event-centrality was self-rated by the patients, the total score ranges from 0 to 35 .$Pr (NDE > No NDE) < 0.050; *Pr (NDE > No NDE) > 0.975. Here, the probability that continuity & coherence of the memory is superior in the ‘NDE’ group than in the ‘No NDE’ group is low (marginally relevant). The probability that centrality of the memory is superior in the ‘NDE’ group than in the ‘No NDE’ group is high (relevant). Characteristics of hospitalization memories and their comparison between patients with and without NDE. This table present memories characteristics among the participants and compares these characteristics between participants reporting a near-death experience (NDE) and those who do not (no NDE). The memories were rated by the interviewers on a Likert-scale ranging from 0 to 5. Event-centrality was self-rated by the patients, the total score ranges from 0 to 35 . $Pr (NDE > No NDE) < 0.050; *Pr (NDE > No NDE) > 0.975. Here, the probability that continuity & coherence of the memory is superior in the ‘NDE’ group than in the ‘No NDE’ group is low (marginally relevant). The probability that centrality of the memory is superior in the ‘NDE’ group than in the ‘No NDE’ group is high (relevant). PTG was relevantly associated with two characteristics, namely, the strength of the impression of reliving the event while recalling it (Pr (β > 0) = 0.975) and the frequency of intrusive memories of the event (Pr (β > 0) = 0.988). PTG was marginally associated with the presence of flashbacks (Pr (β > 0) = 0.969). PTG and the occurrence of an NDE were both relevantly associated with event-centrality (Prs (β > 0) > 0.998). Psychopathology Table 2 recapitulates the scores for psychopathological variables in our sample. Only two patients (3.4%) met the criteria for PTSD. Ten patients (16.9%) met the criteria for partial PTSD. PTG was relevantly associated with higher PTSD scores (Pr (β > 0) = 0.997). PTG was positively associated with higher depression or anxiety scores, but these associations remained under relevance levels (Prs (β > 0) < 0.882). To note, experiencing an NDE was associated with higher PTSD scores - under the relevance level (Pr (β > 0) = 0.906), but was not associated with a higher risk of presenting partial or full PTSD (Pr (PTSD > no PTSD) = 0.641). Table 2Psychiatric disorders at 3 months post-ICU.m (± sd)n (%)PTSD (PCL-5 score)9.72 (± 10.30) Partial PTSD10 (16.9) PTSD2 (3.4)Anxiety (HAD-A score)5.08 (± 4.19) Possible anxious disorder (score > 7)4 (6.8) Probable anxious disorder (score > 10)8 (13.6)Depression (HAD-D score)4.21 (± 3.64) Possible depressive disorder (score > 7)6 (10.2) Probable depressive disorder (score > 10)3 (5.1)Suicidal risk (MINI) Mild4 (6.8) Severe1 (1.7)HAD-A, Hospital Anxiety and Depression – Anxiety subscale; HAD-D, Hospital Anxiety and Depression – Depression subscale; MINI, Mini Internation Neuropsychiatric Interview; PCL-5, Post-traumatic CheckList for DSM-5. Table 2Psychiatric disorders at 3 months post-ICU.m (± sd)n (%)PTSD (PCL-5 score)9.72 (± 10.30) Partial PTSD10 (16.9) PTSD2 (3.4)Anxiety (HAD-A score)5.08 (± 4.19) Possible anxious disorder (score > 7)4 (6.8) Probable anxious disorder (score > 10)8 (13.6)Depression (HAD-D score)4.21 (± 3.64) Possible depressive disorder (score > 7)6 (10.2) Probable depressive disorder (score > 10)3 (5.1)Suicidal risk (MINI) Mild4 (6.8) Severe1 (1.7)HAD-A, Hospital Anxiety and Depression – Anxiety subscale; HAD-D, Hospital Anxiety and Depression – Depression subscale; MINI, Mini Internation Neuropsychiatric Interview; PCL-5, Post-traumatic CheckList for DSM-5. Psychiatric disorders at 3 months post-ICU. HAD-A, Hospital Anxiety and Depression – Anxiety subscale; HAD-D, Hospital Anxiety and Depression – Depression subscale; MINI, Mini Internation Neuropsychiatric Interview; PCL-5, Post-traumatic CheckList for DSM-5. ICU parameters Table 3 describes the main psychotropic drugs administered during ICU stay and the total sedation time for each patient. No ICU parameter was associated with the occurrence of post-traumatic growth. However, higher doses of sufentanil or neuroleptics were associated with a lesser probability of presenting an NDE (Prs (NDE > no NDE) = 0.015 and Pr (NDE > no NDE) = 0.010) respectively). Higher doses of sufentanil were associated with lower PTSD scores (Pr (β > 0) = 0.020). Table 3Total sedation time and main psychotropic drugs administered in ICU.ParameterMedian [Q1-Q3]Mean (SD)Total sedation time (in hours)11 [5–30]13.0 (10.0)Drug doses administered (in mg) Midazolam4 [0–8]5.7 (6.2) Propofol4 [0–11]6.6 (7.5) Sufentanil1 [0–8]5.1 (6.9) Remifentanil0 [0–6]4.5 (7.5) Clonidine4 [0–8]4.9 (5.7) Curare3 [2–8]5.3 (5.0) Neuroleptics (Haloperidol equivalence)0 [0–2]1.7 (3.0)Six patients did not receive any sedative drug. Among them, two reported a near-death experience. Table 3Total sedation time and main psychotropic drugs administered in ICU.ParameterMedian [Q1-Q3]Mean (SD)Total sedation time (in hours)11 [5–30]13.0 (10.0)Drug doses administered (in mg) Midazolam4 [0–8]5.7 (6.2) Propofol4 [0–11]6.6 (7.5) Sufentanil1 [0–8]5.1 (6.9) Remifentanil0 [0–6]4.5 (7.5) Clonidine4 [0–8]4.9 (5.7) Curare3 [2–8]5.3 (5.0) Neuroleptics (Haloperidol equivalence)0 [0–2]1.7 (3.0)Six patients did not receive any sedative drug. Among them, two reported a near-death experience. Total sedation time and main psychotropic drugs administered in ICU. Six patients did not receive any sedative drug. Among them, two reported a near-death experience. Multivariable analyses To investigate to which extent the presence of an NDE contributed to post-traumatic growth after considering adjusting variables, we performed five successive multivariable regression analyses (presented in Table 4). The presence of an NDE remained relevantly associated with post-traumatic growth in every model (Prs (NDE > no NDE) ≥ 0.987), while higher PTSD scores were relevantly associated with a higher PTG in two models (Pr (β > 0) = 0.988) and remained marginally relevantly associated with PTG when taking memories characteristics into account (Pr (β > 0) = 0.971). Table 4Multivariable analysis of factors associated with post-traumatic growth.Model 1Model 2Model 3Model 4Model 5M2.5%97.5%PrM2.5%97.5%PrM2.5%97.5%PrM2.5%97.5%PrM2.5%97.5%PrIntercept− 0.29− 0.610.03− 0.61− 0.97− 0.26− 0.56− 0.90− 0.20− 0.50− 0.86− 0.13− 0.91-1.580.21Sex1.780.773.520.9041.950.863.830.9411.550.673.020.8301.400.602.780.7541.260.532.560.654Age0.16− 0.120.450.8710.15− 0.120.430.8620.19− 0.080.470.9170.14− 0.140.430.8340.17− 0.120.470.876NDE3.451.646.40 0.999* 3.111.465.73 0.998* 2.571.185.03 0.987* 2.681.135.38 0.987 * PCL-50.340.040.64 0.986* 0.350.040.66 0.988* 0.31− 0.010.63.971$Centrality0.22− 0.120.660.9010.15− 0.220.520.789Strength0.18− 0.130.490.881Intrusions0.01− 0.290.290.537Flashbacks0.09− 0.310.290.661DIC167.1210.8366.3520.9813.0This table presents the results of the separate multivariable Bayesian beta regression analyses performed with post-traumatic growth as a dependent variable. Each line presents the result for a predictive variable, in five successive models. NDE is considered as a binary (presence of an NDE, i.e. Greyson NDE scale ≥ 7. PCL-5 and centrality are treated as continuous variables.M: mean of the posterior distribution of regression coefficient β (and its 95% credible interval) for each linear variable, expressed in odds-ratio for the binary variables (e.g., sex, NDE); Pr: probability that the regression coefficient β > 0 (Pr (β > 0)). NDE, Near-Death Experience; PTSD, Post-traumatic Stress Disorder. DIC: Deviance Index Criterion (DIC) assesses the goodness of fit and complexity of each model; a lower DIC indicates a better model. $Pr (β > 0) > .950 (marginally relevant); *Pr (β > 0) > .975 (relevant). Probabilities over the relevance threshold (Pr > 0.975) are in bold. Table 4Multivariable analysis of factors associated with post-traumatic growth.Model 1Model 2Model 3Model 4Model 5M2.5%97.5%PrM2.5%97.5%PrM2.5%97.5%PrM2.5%97.5%PrM2.5%97.5%PrIntercept− 0.29− 0.610.03− 0.61− 0.97− 0.26− 0.56− 0.90− 0.20− 0.50− 0.86− 0.13− 0.91-1.580.21Sex1.780.773.520.9041.950.863.830.9411.550.673.020.8301.400.602.780.7541.260.532.560.654Age0.16− 0.120.450.8710.15− 0.120.430.8620.19− 0.080.470.9170.14− 0.140.430.8340.17− 0.120.470.876NDE3.451.646.40 0.999* 3.111.465.73 0.998* 2.571.185.03 0.987* 2.681.135.38 0.987 * PCL-50.340.040.64 0.986* 0.350.040.66 0.988* 0.31− 0.010.63.971$Centrality0.22− 0.120.660.9010.15− 0.220.520.789Strength0.18− 0.130.490.881Intrusions0.01− 0.290.290.537Flashbacks0.09− 0.310.290.661DIC167.1210.8366.3520.9813.0This table presents the results of the separate multivariable Bayesian beta regression analyses performed with post-traumatic growth as a dependent variable. Each line presents the result for a predictive variable, in five successive models. NDE is considered as a binary (presence of an NDE, i.e. Greyson NDE scale ≥ 7. PCL-5 and centrality are treated as continuous variables.M: mean of the posterior distribution of regression coefficient β (and its 95% credible interval) for each linear variable, expressed in odds-ratio for the binary variables (e.g., sex, NDE); Pr: probability that the regression coefficient β > 0 (Pr (β > 0)). NDE, Near-Death Experience; PTSD, Post-traumatic Stress Disorder. DIC: Deviance Index Criterion (DIC) assesses the goodness of fit and complexity of each model; a lower DIC indicates a better model. $Pr (β > 0) > .950 (marginally relevant); *Pr (β > 0) > .975 (relevant). Probabilities over the relevance threshold (Pr > 0.975) are in bold. Multivariable analysis of factors associated with post-traumatic growth. This table presents the results of the separate multivariable Bayesian beta regression analyses performed with post-traumatic growth as a dependent variable. Each line presents the result for a predictive variable, in five successive models. NDE is considered as a binary (presence of an NDE, i.e. Greyson NDE scale ≥ 7. PCL-5 and centrality are treated as continuous variables. M: mean of the posterior distribution of regression coefficient β (and its 95% credible interval) for each linear variable, expressed in odds-ratio for the binary variables (e.g., sex, NDE); Pr: probability that the regression coefficient β > 0 (Pr (β > 0)). NDE, Near-Death Experience; PTSD, Post-traumatic Stress Disorder. DIC: Deviance Index Criterion (DIC) assesses the goodness of fit and complexity of each model; a lower DIC indicates a better model. $Pr (β > 0) > .950 (marginally relevant); *Pr (β > 0) > .975 (relevant). Probabilities over the relevance threshold (Pr > 0.975) are in bold. Discussion Our study suggests that higher post-traumatic growth after an intensive care hospitalization due to severe COVID-19 is associated with the occurrence of a near-death experience (NDE). Additionally, post-traumatic growth was higher in participants reporting intense memories and higher centrality of the event. Consistent with our main assumption, post-traumatic growth was positively associated with the occurrence of NDEs. Indeed, Moody specifies that “changes in life conduct” are a constitutional part of NDEs37. In a previous study, Khanna & Greyson (2015) showed a similar result among 251 survivors of a close brush with death. However, our study brings new insights on an unexplored population. Indeed, the sample explored by Khanna & Greyson was composed of healthy volunteers, mostly women (69% vs. 20% in our study). NDEs were frequent (90% vs. 23% in our study) and happened at a young age (mean age = 28.8 vs. 63.0 years in our study). Our study is the first to show an association between PTG and NDE in a clinical sample discharged from ICU. Moreover, we demonstrated that post-traumatic growth remained associated with the occurrence of an NDE when taking confounding factors into account. While cognitive and subjective factors associated with PTG were mainly identified in the literature (e.g., the presence of cognitive processing, positive coping, personality traits)7, its association with unusual intrapsychic experiences such as NDEs has yet to be clarified. Various phenomenological characteristics of NDEs could explain these results. First, NDEs are often reported as positive experiences, though the literature indicates that they can, in about 14% of cases, be negative and frightening experiences34. Surprisingly, the emotional valence of hospitalization memories was not superior in patients reporting an NDE. Although not entirely positive, the contrasting nature of the emotions experienced during an NDE could bring more lessons than a fundamentally negative resuscitation experience. Second, the increase of spirituality after NDE is documented, which may explain their association with higher post-traumatic growth57. Nevertheless, in our study, NDEs were positively associated with each dimension of post-traumatic growth and not only spiritual changes. Additionally, the common characteristics of NDEs, particularly as measured by Greyson’s scale, are currently subject to criticism regarding their relevance. The variety of NDE experiences might not be sufficiently covered by this type of scale, which could be caricatured in some of its transcendental elements35. In our study, the most frequent NDE features encountered were cognitive: modification of time perception (39% of patients) and reviewing past events (36% of patients). In contrast, literature showed that the most frequent NDE features were out-of-body experiences and a feeling of peacefulness58. Cognitive features may instead be more non-specific and result from the experience of the ICU stay. During an ICU stay, many hallucinatory experiences could also be classified as NDEs59. Other types of stress-related experiences specific to intensive care have not been measured in our study and have recently been compiled in a specific scale (the Stressful Memory Assessment Checklist for the Intensive Care Unit (SMAC-ICU))60. A recent study presented a qualitative analysis of the hallucinations reported by eight ICU-discharged COVID-19 patients. The authors stated that seven patients reported hallucinations consistent with peduncular hallucinosis, i.e. vivid visual hallucinations occurring alongside an acute lesion in the midbrain or another subcortical region61. Therefore, ICU-discharged patients with a history of hallucinations that are consistent with an NDE may have experienced other hallucinatory phenomena. Using specific scales or qualitative analyses in future studies could help disentangle these experiences from NDEs and compare their association with post-traumatic growth. Moreover, it is worth noting that NDEs can occur even when an individual is not actually close to death but fears its sudden onset (e.g., a near-road accident that almost happens but does not). NDEs may even occur outside of dangerous situations (e.g., meditative states or syncope) and are then called “NDE-like experiences”62. Therefore, reporting an NDE might also result from a certain level of subjectivity. Personality traits such as “openness” and a tendency toward fantasy could be related to reporting an NDE63. However, NDE experience itself may also impact personality and promote the trait of “openness”64. Finally, the influence of subjectivity in the emergence of post-traumatic growth is supported through its association with event-centrality. Event-centrality was also higher in patients reporting an NDE. The connection between PTG, PTSD and centrality has already been observed in previous studies, highlighting the difference between growth and resilience, the latter being considered as a form of “return to baseline” after a potentially traumatic event and not requiring that the event be emotionally impactful or central to the individual7,65–67. The patients included in our study were relatively less affected by PTSD compared to typical post-ICU populations (3.4% of PTSD compared to 22–44% in previous studies)14,68 or to other cohorts of COVID-19 patients discharged from ICU (13–28% of PTSD reported)69,70. However, our study was very similar to these cohorts in terms of gender and age (80% male, mean age 63, versus 72–74% male, mean age 62). In our study, social support may have constituted an important protective factor, also implicated in post-traumatic growth10. Indeed, most patients were in a relationship, and many expressed during interviews that they felt particularly supported by the healthcare providers and their relatives. Contrary to our expectations, social support might have been reinforced by the context of the COVID-19 pandemic, as the whole society, healthcare workers and relatives were committed to save COVID-19 patients’ lives. Social support is known to be a protective factor against PTSD and has been related to a better mental health-related quality of life after ICU discharge71–73. Although this hypothesis is highly expectative, several local mental health prevention measures dedicated to healthcare workers might have indirectly benefited to severe COVID-19 patients in terms of mental health17,74. Moreover, our patients benefitted from post-ICU follow-up, which was shown to decrease PTSD symptoms75. Finally, PTSD and PTG scores were positively associated. However, PTSD scores were low to moderate, while PTG was relatively high in our population. Our results are consistent with previous findings showing that individuals with few PTSD symptoms had low PTG, while individuals with moderate PTSD had high PTG76. Our study also suggests that a moderate traumatic impact of an ICU stay may be related to high levels of post-traumatic growth. Longitudinal post-ICU studies may reveal whether post-traumatic growth is mostly illusory (perceived but not observable changes) or constructive (associated with functional adjustment) in this context67,77. No association between post-traumatic growth and sedative drugs were found. However, we found a negative association between sufentanil doses and the probability of presenting an NDE or PTSD. Concerning PTSD, previous studies found a positive association between benzodiazepine levels and the risk of PTSD but not with opioid analgesics14,25. Though, the protective effect of morphine use against PTSD after traumatic injuries is documented78,79. Concerning near-death experiences, our study suggested a negative association between opioid analgesics levels and the probability of presenting an NDE. The main neurobiological mechanisms of NDEs evoked until now were ketamine-like effects (NMDA receptor blockade), disrupted serotoninergic activity or a massive release of endogenous endorphins80. As the administration of endogenous opioids may hinder the release of endogenous endorphins, our observations may align with these theoretical considerations81. Nevertheless, these assumptions require careful support by further fundamental and clinical research. Our study has some limitations. First, it involves a snapshot cross-sectional assessment of post-traumatic growth, as is the case with many studies. This type of assessment requires significant cognitive effort from participants and may lead to notable memory biases65. Moreover, this design prevents us from drawing conclusions about the causal relationship between NDEs and post-traumatic growth. Second, our study remained monocentric, and our sample was relatively small (n = 59). Unfortunately, 72 patients eligible for an outpatient clinic could not be included for several reasons (see Fig. 1). This could have impacted the representativeness of our sample, especially by underestimating PTSD prevalence. Therefore, we performed additional comparison analyses to verify the representativeness of our sample. These analyses indicated that age tended to be lower in the participants included in our study than in other discharged from ICU (Mean diff = − 3.19 years, 95% CI [− 7.12, 0.70], Pr = 0.945), although uncertainty remained, while no meaningful difference was observed in gender distribution between groups (Mean diff = 0.003, 95% CI [− 0.039, 0.049], Pr = 0.55). These results, although uniquely conducted on available data (age, sex) suggest a good representativeness of our sample. Moreover, two patients (among 28) who declined the multidisciplinary follow-up declared that they felt well, which supposes that the latter did not suffer from PTSD or other psychological sequalae. Eight patients benefited from another multidisciplinary follow-up and six came to the next follow-up appointment, which reinforces the hypothesis of a good representativeness of our sample. Thirdly, our study was conducted only 3 months after hospitalization. On the one hand, the feeling of having survived death and the resulting life changes (post-traumatic growth) may still be very strong at this time; on the other hand, PTSD may manifest later. The use of a self-report measure of PTSD might have underestimated PTSD prevalence. However, this tool was used in similar populations discharged from intensive care after severe COVID-19, and rather presents a tendency to overestimate PTSD prevalence70,82. Moreover, each patient was clinically assessed by a senior psychiatrist. Only one discrepancy was shown: a patient scored positive for PTSD on PCL-5 (score = 43) while the clinical evaluation was not in favor of PTSD. Fourthly, memories characteristics were assessed by the interviewers, which may have introduced observer bias. Therefore, the results concerning memories characteristics remain exploratory and should be interpreted with caution. The use of a self-report validated autobiographical memories characteristics questionnaire would be more relevant for future studies83. Finally, phenomenological studies of testimonies could enrich our description of NDEs84. While the narrow and caricatured definition of NDEs is currently criticized35, these phenomenological studies may allow to better understand the complexity of NDEs and their psychological consequences. Conclusion The occurrence of post-traumatic growth after an ICU hospitalization had remained a largely unexplored area in research. Nevertheless, post-traumatic growth could be a critical component of recovery in the aftermath of a sudden exposure to death. In a three-month follow up cohort of severe COVID-19 patients, post-traumatic growth is real, particularly associated with the occurrence of near-death experiences and the centrality of the hospitalization in patients’ lives. Although the literature on NDEs is extensive, their causes and impact on the lives of those who experience them have yet to be clarified, so that clinicians can better identify them, and support patients discharged from ICU.