ICU-specific virtual reality for mental health after critical illness: an international three-arm randomized clinical trial Authors: Denzel L. Q. Drop, Johan H. Vlake, Jasper van Bommel, Evert-Jan Wils, Denise E. Hilling, Tim I. M. Korevaar, Fabio S. Taccone, Marianne Devroey, Ralph K. L. So, Corstiaan A. den Uil, Anna F. C. Schut, Adrienne M. T. J. Raben, Jan H. Elderman, Sefanja Achterberg, Barbara Festen, Annemieke Dijkstra, Margo M. C. van Mol, Christian Jung, Diederik Gommers, Michel E. van Genderen Published: 2026 ABSTRACT BackgroundSurvivors of critical illness frequently experience symptoms of post-intensive care syndrome (PICS), including post-traumatic stress disorder (PTSD), anxiety, and depression, with downstream impact on health-related quality of life (HRQoL). We evaluated whether an intensive care unit-specific virtual reality (ICU-VR) video improves mental health outcomes and whether timing (early vs. late) matters.MethodsWe conducted an international, three-arm, multicentre randomized clinical trial across eleven hospitals. Adults (≥ 18 years) with ICU stay ≥ 72 h and mechanical ventilation ≥ 24 h were randomized 1:1:1 to (1) standard care, (2) early ICU-VR (within 7–15 days post-ICU), or (3) late ICU-VR (≈ 3 months, during aftercare). The primary endpoint was PTSD symptom severity at six months (T3) measured by the Impact of Event Scale-Revised (range 0–88). Secondary endpoints included anxiety/depression severity and prevalence, and HRQoL.ResultsWe randomized 344 participants (median age 61 years; 62% male); follow-up among survivors at six months (T3) was 80%. At six months, PTSD symptom severity did not differ between early ICU-VR and control (β = 0.32; 95% CI –3.38 to 4.02; p = 0.87) nor late ICU-VR and control (–0.82; 95% CI –4.62 to 2.97; p = 0.67). No between-group differences were observed for anxiety, depression, or HRQoL (all adjusted p > 0.05). Patients rated ICU-VR highly for improving understanding of the ICU experience and overall satisfaction (median 8/10).ConclusionsIn this multicentre randomized clinical trial, a single ICU-VR video did not improve PTSD, anxiety, depression, or HRQoL at six months versus standard care. Findings should not be generalized to multi-session, therapist-guided, or personalized VR interventions. Future trials should target high-risk patients, make multi-session mandatory, and personalize ICU-VR.Trial registrationThis study was prospectively registered Netherlands: Trial Register/International Clinical Trials Registry Platform, NL9812, October 21, 2021.Supplementary InformationThe online version contains supplementary material available at 10.1186/s13054-026-06186-4. Introduction Survival after critical illness has improved considerably over the past decades, yet many intensive care unit (ICU) survivors face persistent physical, cognitive, and psychological sequelae collectively referred to as post-intensive care syndrome (PICS) [1, 2]. Among these, post-traumatic stress disorder (PTSD), anxiety, and depression are common and strongly associated with reduced health-related quality of life (HRQoL) [3]. Several interventions, such as ICU diaries and follow-up clinics, aim to mitigate post-ICU psychological distress. However, meta-analyses have shown mixed or modest effects, highlighting a need for innovative and scalable approaches that address both mental health and survivors’ desire to understand what occurred during their ICU stay [4–6]. Additionally, previous studies have shown that ICU survivors often seek clarity about their ICU stay to make sense of their experience [6]. Given that exposure therapy is the gold standard for PTSD treatment, it is hypothesized that immersive exposure to the ICU environment, paired with treatment-related information, may help reduce delusional memories and mental health distress [6–10]. Virtual Reality (VR) offers a fully immersive, three-dimensional audiovisual experience capable of delivering realistic information and supporting psychological processing, while in PTSD treatment, VR exposure therapy can reduce symptoms through graded exposure and emotional processing [11–13]. Therefore, VR has shown promise in a wide range of patient populations [9, 11–15]. VR is believed to enhance psychological recovery by offering realistic, immersive exposure to ICU-related environments, potentially more effective than imaginal exposure and comparable to in vivo exposure [9, 12, 13, 16–18]. Notably, patients are often more willing to engage in VR exposure, which can be delivered in a gradual, safe, and controlled manner [13]. This enables repeated exposure to feared stimuli, fostering emotional processing and adaptive responses. However, such therapy requires multiple sessions, a trained therapist, and explicit activation of trauma-related fear structures [12, 13, 16, 19]. To meet the need of ICU survivors for more clarity about their ICU stay, we developed the Intensive Care Unit-specific Virtual Reality (ICU-VR) intervention [6]. ICU-VR was designed as an informational VR intervention explaining ICU procedures, devices, and the environment to promote sense-making and memory reconstruction, and volunteers in a pilot study preferred the video content to be delivered by VR rather than a 2D screen [18]. Similar to other VR interventions, ICU-VR was previously found to be safe in critical illness survivors [14, 20, 21]. In addition, a pilot randomized clinical trial in sepsis survivors suggested potential benefits when ICU-VR was delivered soon after ICU discharge, whereas a follow-up trial during post-ICU aftercare visits in COVID-19 ICU survivors did not [14, 20]. This discrepancy raised the possibility that timing may influence effectiveness. We therefore conducted an international, three-armed, multicentre randomized clinical trial to assess the efficacy and optimal timing of this ICU-VR video for reducing mental health distress (PTSD, anxiety, depression), and improving HRQoL up to six months after ICU discharge. We also explored patients’ perspectives on the ICU-VR video and ICU aftercare. Methods Study design This international, three-armed, multicentre randomized clinical trial was conducted across eleven hospitals in the Netherlands and Belgium between December 2021 and July 2024 (Table S1). Participant inclusion was evaluated each quarter and stopped after exceeding the predefined sample size in each treatment arm. The study protocol was approved by the Medical research Ethics Committees United (NL78555.100.21) and by the local review boards [22]. This study was prospectively registered (Netherlands Trial Register/International Clinical Trials Registry Platform, NL9812, October 21, 2021). The trial followed the CONSORT 2025 and RATE-XR guidelines and can be classified as a Technology Readiness Level (TRL) 8 study [23–25]. Study participants Eligible participants were adults (aged ≥ 18 years) admitted to the ICU for ≥ 72 h, receiving mechanical ventilation ≥ 24 h, proficient in Dutch or French. Exclusion criteria included active psychiatric disease (patients with a history of mental illness were not excluded), a history of or a primary neurological impairment necessitating ICU treatment, cognitive impairment at screening defined by a Telephone Interview for Cognitive Status (TICS) score < 27, active delirium, absence of a formal home address, or patients with a life expectancy < 48 h, including those receiving palliative care [26–28]. Written informed consent was obtained. Randomization and masking Participants were randomized 1:1:1 (control / early ICU-VR / late ICU-VR) using variable blocks stratified by site via a centralized, internet-based randomization system (Castor EDC, Amsterdam, the Netherlands). Blinding of the participant and researcher (i.e. those who assess the patient/participant and those who apply the ICU-VR intervention) was not feasible, given the nature of the study (i.e. the inability of blinding the VR intervention for the participant and the researchers and the inability to blind the data collector since the correct questionnaires must be provided). Although the analyst was unaware of the randomization allocation during data analysis. ICU-VR intervention The ICU-VR video was developed by an interdisciplinary team of intensivists, ICU nurses, a psychologist, and a psychiatrist, without commercial interest. Also, one former ICU patient provided input into the content of the ICU-VR video. It consisted of a 14-minute 360° informational video filmed within each participating ICU, featuring the local environment to enhance immersion. The video provided explanations of mechanical ventilation, monitoring, and typical ICU routines, avoiding graphic or distressing content. The VR was delivered using PICO G2 headsets with headphones. ICU nurses or research staff trained in device operation, patient briefing, and emotional safety conducted the sessions. Participants received standardized instructions, could pause or stop anytime, and were monitored for cybersickness or discomfort. The duration of the ICU-VR video was fourteen minutes to limit the risk of cybersickness. Further details on the development process, script, and a YouTube version of ICU-VR are available elsewhere and a translation of the script can be found in the published study protocol [18, 22]. An overview of the intervention is provided in the supplementary file (Fig. S1). Study procedures The study flowchart is provided in the supplementary file (Fig. S2) and a detailed overview of the questionnaires is provided in the published study protocol [22]. Patients were approached after ICU discharge, at the hospital ward, to receive study-related information and enrolled within seven days after ICU discharge and completed follow-up questionnaires one (T1), three (T2), and six (T3) months post-ICU discharge on paper or online. Non-responders were contacted by telephone as reminder. Standard care was applicable to all participants, comprising an invitation to a follow-up clinic visit approximately three months post-ICU discharge and the use of ICU diaries according to local practice guidelines. During the follow-up clinic appointment in all the participating centres, participants had an extensive consultation with an intensivist and ICU nurse to assess the complex need of these ICU survivors and coordinate follow-up care planning. If deemed necessary, participants could be referred to other healthcare professionals for further mental or physical support. Early ICU-VR participants additionally received one to three ICU-VR video sessions, depending on their preference, between seven and fifteen days after ICU discharge, if still admitted to the hospital ward. Late ICU-VR group participants received the ICU-VR video once during the scheduled follow-up clinic. No methodological differences were present between the early ICU-VR and late ICU-VR groups other than the frequency (1–3 versus 1) and timing (between seven and fifteen days after ICU discharge versus during the follow-up clinic) of the ICU-VR sessions. Frequency and timing of ICU-VR video sessions were based on pilot studies and routine scheduling of follow-up care [14, 18, 20]. Although a previous pilot study showed that patients preferred a median of one session, early ICU-VR participants were allowed up to three ICU-VR sessions, in line to our previous study when ICU-VR was offered multiple times, because we sought to assess this preference in a larger sample and because VR exposure therapy typically requires multiple sessions [12, 13, 20]. Late ICU-VR participants had one session owing to logistical feasibility, aligned with routine follow-up scheduling, and consistent with our previous study in COVID-19 ICU survivors [14]. Study outcomes The primary outcome was PTSD symptom severity at six months (T3) since this represents the differential timepoint between acute and delayed-onset PTSD [29]. Secondary outcomes included anxiety and depression, HRQoL, and participants’ satisfaction with ICU aftercare and ICU-VR. Additional exploratory outcomes comprised return to work, financial decline, and healthcare utilization. Finally, a post-hoc exploratory analysis examined whether the presence of probable ASD/PTSD-related symptoms, anxiety, or depression at baseline was associated with more severe psychological symptoms six months after ICU discharge. More detailed information on the study outcomes can be found in the statistical analysis plan (SAP) (supplementary notes 2). Measurements Recommended questionnaires with validated translations and cut-off values were used to assess mental health symptoms [30]. PTSD was assessed using the Impact of Event Scale-Revised (IES-R; range 0–88; higher = worse), with a prevalence cut-off value of ≥ 24 [31–33]. IES-R scores at baseline (T0) are considered as ASD/PTSD-related symptoms, since, officially, PTSD needs a duration of at least a month to be diagnosed [29]. Anxiety and depression were assessed using the Hospital Anxiety and Depression Scale (HADS; range 0–21; higher = worse), with a prevalence cut-off value of ≥ 8 [34–36]. HRQoL was assessed using the EuroQol-5dimensional-5level (EQ-5D-5 L) and 36-Item Short Form Health Survey (SF-36, version 1). The EQ-5D-5 L assess the quality of life in five dimensions after which an health state can be calculated, ranging from − 0.446 (worst quality of life) to 1.000 (best quality of life) [37–39]. The SF-36 assesses HRQoL across eight domains, which are directly transformed to a scale ranging from 0 (worst score) to 100 (best score) on the assumption that each question carries an equal weight [40, 41]. These scores are standardized so that the mean is 50 (SD 10) for the general population, which allows for a consistent comparison of individual scores to the general population’s average [42–44]. Patient satisfaction with ICU aftercare and ICU-VR was evaluated through questionnaires adapted from the Patient Satisfaction Questionnaire and Family Satisfaction with ICU Care tools, altered to the needs of this study by a multidisciplinary group of members of the RATE-XR steering group through extensive discussions (supplementary notes 1) [23, 45, 46]. Demographic information, medical and psychiatric history, ICU treatment details (including presence of delirium during ICU admission and before performing any study-related actions), as well as data on return to work, financial decline, and healthcare utilization were collected through participant questionnaires and review of electronic health records. Statistical analysis Using a Cohen’s d = 0.56, 2-sided α value of 0.05, a power of 0.80, and a 1:1:1 allocation ratio, we calculated a sample size per group of 60 participants. More detailed information on the power analysis can be found in the SAP (supplementary notes 2) and published study protocol [14]. Anticipating a loss to follow-up of 40%, we aimed to include at least 100 participants in each study arm. Statistical significant p-values were corrected for multiple testing by using the Benjamini-Hochberg with a maximum of 5% expected false positive results. No imputation of missing data was necessary since follow-up rates were sufficient, according to the study protocol. The main analysis was an intention-to-treat (ITT), complemented with predefined per-protocol, subgroup, and complete case analysis (supplementary notes 2 and 3). All data was gathered using Castor EDC (Castor EDC, Amsterdam, the Netherlands). Analyses were performed using R statistics (R Foundation for Statistical Computing, Vienna, Austria, 2015). All p-values were tested two-sided and considered statistically significant if ≤ 0.05. Continuous variables were summarized as medians (95% range) and categorical variables as absolute and relative frequency. Outcomes between groups were analysed with mixed effects linear and logistic regression models. Regarding the primary outcome, the outcome at each timepoint served as dependent variable, randomisation group as independent variable, and a random intercept/slope for each study site was used. Outcomes of mixed effects linear and logistic regression models were presented as the coefficient of the model, which implies the estimated mean difference between groups, including its 95% CI, as the log of coefficient of the model, that is, the Odds Ratio (OR), including its 95% CI, respectively. More detailed information on the sample size calculation and statistical analysis for each outcome and group can be found in the SAP (supplementary notes 2). Any deviation from the SAP and an elaboration on the post-hoc analysis is described in the supplementary file (supplementary notes 3). Results Study population A total of 344 ICU survivors were randomized: 114 early ICU-VR, 114 late ICU-VR, and 116 control (Fig. 1). The median age was 61 years; 62% (n = 212) were male. Median APACHE IV score was 70, median ICU length of stay 11 days, and the median duration of mechanical ventilation 5 days. Baseline characteristics were balanced across groups (Table 1). Among survivors, follow-up response at six months was 80%. The number of included patients per study site are presented in Table S1. Fig. 1Flowchart of participant screening and follow-up. All completed questionnaires were included in the primary intention-to-treat analysis. Death of participants during follow-up could happen both before and after completing baseline questionnaires. ICU, Intensive Care Unit; VR, Virtual Reality; ICU-VR, Intensive Care Unit-specific Virtual Reality Flowchart of participant screening and follow-up. All completed questionnaires were included in the primary intention-to-treat analysis. Death of participants during follow-up could happen both before and after completing baseline questionnaires. ICU, Intensive Care Unit; VR, Virtual Reality; ICU-VR, Intensive Care Unit-specific Virtual Reality Table 1Baseline demographics and treatment-related characteristics of participantsParticipant characteristics1, 2Early ICU-VRn = 114Late ICU-VRn = 114Controln = 116Age, median (95% range)59 (23–78)62 (30–81)61 (28–77)Sex, male, n (%)72 (63)71 (62)69 (59)BMI prior to ICU admission, median (95% range)27.1 (20.0-43.6)26.3 (17.2–41.3)26.1 (19.4–37.1)Educational level, n (%) Academic education 13 (11)20 (18)15 (13) Higher vocational education 30 (26)18 (16)36 (31) Post-secondary vocational education 38 (33)40 (35)32 (28) Secondary education 19 (17)21 (18)12 (10) Primary education 7 (6)4 (4)8 (7)History of mental illness, yes, n (%)17 (15)15 (13)22 (19)Previous ICU admission in the past 5 years, yes, n (%)19 (17)22 (19)20 (17)Charlson Comorbidity Index (CCI) score, median (95% range)3 (0–7)3 (0–7)3 (0–9)Reason for ICU admission, n (%) Medical 59 (52)49 (43)60 (52) Surgical, planned32 (28)22 (19)33 (28) Surgical, acute23 (21)43 (38)23 (20)Sepsis during ICU admission, yes, n (%)36 (32)43 (38)42 (36)Delirium during ICU admission, yes, n (%)45 (40)48 (42)57 (50)Acute Physiology and Chronic Health Evaluation IV score (APACHE IV), median (95% range)71 (32–123)65 (20–122)71 (25–120)Simplified Acute Physiology Score (SAPS II), median (95% range)41 (22–74)39 (15–76)41 (20–72)Length of hospital stay, days, median (95% range)29 (10–89)35 (11–200)30 (10–119)Length of intensive care unit stay, days, median (95% range)9 (3–56)13 (3–91)11 (4–79)Duration of mechanical ventilation, days, median (95% range)4 (2–25)6 (2–33)4 (2–39)1 Percentages are taken from the total randomized sample size per group. Baseline data was retrieved from the baseline questionnaire and electronic healthcare records and was not available within the early, late, and control group for sex regarding 5 [4%], 8 [7%], and 7 [6%] participants, for educational level regarding 7 [6%], 11 [10%], and 13 [11%] participants, for the history of mental illness regarding 5 [4%], 5 [4%], and 9 [8%] participants, and for previous ICU admissions regarding 1 [1%], 2 [2%], and 2 [2%] patients2 Percentages are rounded, therefore a total sum may not be 100%. Continues variables are reported as medians (95% range), categorical variables as numbers and percentages. BMI, Body Mass Index; ICU, Intensive Care Unit Table 1Baseline demographics and treatment-related characteristics of participantsParticipant characteristics1, 2Early ICU-VRn = 114Late ICU-VRn = 114Controln = 116Age, median (95% range)59 (23–78)62 (30–81)61 (28–77)Sex, male, n (%)72 (63)71 (62)69 (59)BMI prior to ICU admission, median (95% range)27.1 (20.0-43.6)26.3 (17.2–41.3)26.1 (19.4–37.1)Educational level, n (%) Academic education 13 (11)20 (18)15 (13) Higher vocational education 30 (26)18 (16)36 (31) Post-secondary vocational education 38 (33)40 (35)32 (28) Secondary education 19 (17)21 (18)12 (10) Primary education 7 (6)4 (4)8 (7)History of mental illness, yes, n (%)17 (15)15 (13)22 (19)Previous ICU admission in the past 5 years, yes, n (%)19 (17)22 (19)20 (17)Charlson Comorbidity Index (CCI) score, median (95% range)3 (0–7)3 (0–7)3 (0–9)Reason for ICU admission, n (%) Medical 59 (52)49 (43)60 (52) Surgical, planned32 (28)22 (19)33 (28) Surgical, acute23 (21)43 (38)23 (20)Sepsis during ICU admission, yes, n (%)36 (32)43 (38)42 (36)Delirium during ICU admission, yes, n (%)45 (40)48 (42)57 (50)Acute Physiology and Chronic Health Evaluation IV score (APACHE IV), median (95% range)71 (32–123)65 (20–122)71 (25–120)Simplified Acute Physiology Score (SAPS II), median (95% range)41 (22–74)39 (15–76)41 (20–72)Length of hospital stay, days, median (95% range)29 (10–89)35 (11–200)30 (10–119)Length of intensive care unit stay, days, median (95% range)9 (3–56)13 (3–91)11 (4–79)Duration of mechanical ventilation, days, median (95% range)4 (2–25)6 (2–33)4 (2–39)1 Percentages are taken from the total randomized sample size per group. Baseline data was retrieved from the baseline questionnaire and electronic healthcare records and was not available within the early, late, and control group for sex regarding 5 [4%], 8 [7%], and 7 [6%] participants, for educational level regarding 7 [6%], 11 [10%], and 13 [11%] participants, for the history of mental illness regarding 5 [4%], 5 [4%], and 9 [8%] participants, and for previous ICU admissions regarding 1 [1%], 2 [2%], and 2 [2%] patients2 Percentages are rounded, therefore a total sum may not be 100%. Continues variables are reported as medians (95% range), categorical variables as numbers and percentages. BMI, Body Mass Index; ICU, Intensive Care Unit Baseline demographics and treatment-related characteristics of participants 1 Percentages are taken from the total randomized sample size per group. Baseline data was retrieved from the baseline questionnaire and electronic healthcare records and was not available within the early, late, and control group for sex regarding 5 [4%], 8 [7%], and 7 [6%] participants, for educational level regarding 7 [6%], 11 [10%], and 13 [11%] participants, for the history of mental illness regarding 5 [4%], 5 [4%], and 9 [8%] participants, and for previous ICU admissions regarding 1 [1%], 2 [2%], and 2 [2%] patients 2 Percentages are rounded, therefore a total sum may not be 100%. Continues variables are reported as medians (95% range), categorical variables as numbers and percentages. BMI, Body Mass Index; ICU, Intensive Care Unit Mental health distress and quality of life At 6 months (T3), there were no significant differences in PTSD symptom severity between early ICU-VR and control (β = 0.32; 95% CI –3.38 to 4.02; p = 0.87) or late ICU-VR and control (β –0.82; 95% CI –4.62 to 2.97; p = 0.67) (Fig. 2). Similarly, there were no significant differences between groups in terms of PTSD, anxiety, or depression symptom severity, mental HRQoL, or overall HRQoL (all p > 0.05; Fig. 2), nor in the prevalence of probable PTSD, anxiety, or depression between groups (all p > 0.05; Fig. 2). Furthermore, no group differences were observed at the earlier follow-up time points of one month (T1) and three months (T2) post-ICU discharge (all p > 0.05; Fig. 2).Fig. 2Mental health distress and health-related quality of life. Time points: T0 (at ICU discharge), T1 (1 month), T2 (3 months), T3 (6 months). ASD, acute stress disorder; PTSD, post-traumatic stress disorder; HRQoL, health-related quality of life; IES-R Impact of Event Scale-Revised; HADS-A, Hospital Anxiety and Depression Scale-Anxiety; HADS-D, Hospital Anxiety and Depression Scale-Depression; MCS-36, Mental Component Scale-36; EQ5D-TTO, EuroQol 5-dimensions time trade off; ICU-VR, Intensive Care Unit-specific Virtual Reality. IES-R scores at T0 are considered as ASD/PTSD-related symptoms, since, officially, PTSD needs a duration of at least a month to be diagnosed [29]. Late ICU-VR participants received the intervention during the T2 aftercare visit; consequently, they could only differ from controls from T3 onwards, and their data are therefore only presented separately at T3 Mental health distress and health-related quality of life. Time points: T0 (at ICU discharge), T1 (1 month), T2 (3 months), T3 (6 months). ASD, acute stress disorder; PTSD, post-traumatic stress disorder; HRQoL, health-related quality of life; IES-R Impact of Event Scale-Revised; HADS-A, Hospital Anxiety and Depression Scale-Anxiety; HADS-D, Hospital Anxiety and Depression Scale-Depression; MCS-36, Mental Component Scale-36; EQ5D-TTO, EuroQol 5-dimensions time trade off; ICU-VR, Intensive Care Unit-specific Virtual Reality. IES-R scores at T0 are considered as ASD/PTSD-related symptoms, since, officially, PTSD needs a duration of at least a month to be diagnosed [29]. Late ICU-VR participants received the intervention during the T2 aftercare visit; consequently, they could only differ from controls from T3 onwards, and their data are therefore only presented separately at T3 Six months post-ICU discharge (T3), differences in the trajectories of both the overall HRQoL score (β = -0.06; 95% CI: -0.09, -0.03; adjusted for multiple comparisons p < 0.01) and the PCS-36 score (β=-4.88; 95% CI: -7.51, -2.24; adjusted for multiple comparisons p < 0.01) were observed in favour of the early versus late ICU-VR group. No other significant differences were found in the progression over time of mental health or HRQoL (all p > 0.05; Fig. 2). Post-hoc analyses showed no between-group differences at six months (T3) when stratified by baseline ASD/PTSD-related symptoms, anxiety, or depression (Table S2). Probable ASD/PTSD-related symptoms or anxiety at baseline was associated with higher PTSD (β = 13.89; 95% CI 8.27–19.51; adjusted p < 0.01) and higher anxiety (β = 3.72; 95% CI 0.96–1.85; adjusted p < 0.01) at six months (T3). Baseline depression was initially associated with higher depression at six months (T3) (β = 2.85; 95% CI 0.92–1.04; p < 0.01), though not after multiplicity correction (adjusted p = 0.06). At six months (T3) in the overall cohort, 14% (n = 36) met criteria for probable PTSD, 19% (n = 51) for anxiety, and 22% (n = 57) for depression (Table S3). Median symptom scores were: PTSD 7 (95% range 0–46), anxiety 3 (0–14), and depression 3 (0–14). Median mental HRQoL was 51 (95% range 18–64) and overall HRQoL 0.778 (95% range 0.111–1) (Tables S4–S6). Detailed time-point data are provided in Tables S3–S6. The median number of ICU-VR sessions received in the early ICU-VR group was one (95% range: 0–3). As such, all participants who received at least one ICU-VR session were included in the per-protocol analysis. No significant differences were identified in the predefined per-protocol, complete case, or subgroup analyses (all p > 0.05). Study intervention and follow-up care related outcomes In the early ICU-VR group, 97% (n = 111) of participants received ≥ 1 session (16% two; 4% three). In the late ICU-VR arm, 65% (n = 74) completed the session, primarily due to missed aftercare visits. Attendance at the follow-up clinic was 42% (control), 52% (early), and 65% (late). There were no between-group differences in work resumption, financial decline, or healthcare consultations after multiplicity adjustment (Tables S7-S11). No side effects in the intervention groups were reported resulting in termination of the ICU-VR video usage in participants. Satisfaction with ICU aftercare and the ICU-VR intervention Satisfaction with ICU aftercare in the overall cohort was 8 out of 10 (95% range: 5–10). Participants in the ICU-VR groups rated the intervention’s contribution to ICU aftercare with a median of 8 out of 10 (95% range: 1–10). Participants in both ICU-VR groups reported high appreciation for the intervention with improving ICU understanding (median 8 out of 10, 95% range: 1–10), the degree of recommendation to others to watch the ICU-VR intervention (8, 95% range: 4–10), happiness about having received ICU-VR (8, 95% range: 1–10), and general rating of ICU-VR (8, 95% range: 3–10). No significant differences regarding the satisfaction with ICU aftercare and the ICU-VR intervention were observed between groups, after adjusting for multiple testing (all p or adjusted p > 0.05; Fig. 3). Fig. 3Perspectives ICU-VR and ICU aftercare. ICU-VR, Intensive Care Unit-specific Virtual Reality; ICU, Intensive Care Unit. Scoring ranges from 1–10, with higher scores indicating a higher satisfaction Perspectives ICU-VR and ICU aftercare. ICU-VR, Intensive Care Unit-specific Virtual Reality; ICU, Intensive Care Unit. Scoring ranges from 1–10, with higher scores indicating a higher satisfaction Discussion This international, three-armed, multicentre randomized clinical trial found that a single general ICU-VR video session did not improve PTSD, anxiety, depression, or HRQoL up to six months after ICU discharge compared with standard care, when the ICU-VR video was delivered either shortly after ICU discharge or during aftercare. Satisfaction with ICU aftercare was high and comparable across groups. Participants in both intervention arms highly valued the ICU-VR video, particularly for improving their understanding of ICU treatment and the ICU environment. Three explanations may underlie the absence of measurable benefit in this largest clinical trial investigating the effect of a VR-based intervention in ICU aftercare, resulting in recommendations for future research. First, baseline mental health symptoms in this cohort (25% ASD/PTSD-related symptoms and 40% depression) was lower than in a prior single-centre feasibility study (50% ASD/PTSD and depression), reducing the capacity to detect improvement [20]. Consistent with earlier work, higher baseline symptoms predicted worse outcomes regardless of intervention [3]. Future studies should target high-risk individuals, though identifying them remains a key challenge since current evidence is insufficient to conclude which patient characteristics or ICU-related variables reliably predict PICS, largely due to methodological limitations in available studies. No validated PICS-specific risk stratification tool exists. This underscores the need for well-designed prospective studies that not only identify but also externally validate predictive models. Second, dose and format likely limited effect. In this study, the ICU-VR video was not used as a way of extensive exposure therapy, since exposure-based therapies rely on repeated, progressive engagement and emotional processing [13]. In this trial participants favoured a single ICU-VR video session, primarily addressing cognitive understanding rather than emotional reprocessing. Though this is in contrast to other trials that are extensively summarized in reviews and meta-analysis that examine the effect of VR on mental health symptoms in trauma care [12, 13, 16, 47]. In these studies, standard VR exposure therapy for non-ICU related PTSD consists of typically 8–14 progressive sessions enabling emotional processing and psychological theory suggests that repeated, progressive exposure is required to effectively reduce trauma-related distress [12, 13, 16, 19, 47]. Notably, even the maximum of three sessions offered in the early ICU-VR group falls well below the 8–14 progressive sessions typical of VR exposure therapy, which together with the predominantly single-session uptake likely explains the absence of a measurable effect and supports a dose-related interpretation. Additionally, in the VR mental health and trauma care reviews and meta-analysis, participants are also more intensively guided by psychiatric therapists than in our study [12, 13, 16, 47]. These differences may also explain the lack of measurable effect in mental health improvement despite high satisfaction with the ICU-VR video. Table S12 situates our informational ICU-VR intervention alongside prior ICU-VR studies, therapist-guided VR exposure therapy, and recent systematic reviews and meta-analysis of VR in critical illness, allowing to compare design features that may underlie the differing outcomes. Studies were selected purposively to illustrate the contrast between informational ICU-VR and therapist-guided VR exposure therapy, and to situate both within recent reviews and meta-analysis; this table is therefore illustrative and not intended as a systematic review. Methodologically pooled evidence is available in the cited reviews and meta-analysis. Future studies should investigate multi-session, therapist guided approaches, incorporating emotional processing, to study the effect of the ICU-VR video as a medium of exposure therapy. Third, timing may have been suboptimal. PTSD symptoms can emerge both early and late after discharge; a single informational session may miss the therapeutic window for those developing delayed-onset distress [29]. Tailored, timing-adaptive protocols warrant exploration. Various interventions to improve mental health post-ICU are often well accepted and widely implemented, but have reported conflicting or limited effectiveness and are often based on small sample sizes [48, 49]. ICU diaries, which aim to help patients make sense of their ICU stay, were not mandated in our study protocol and were applied according to local clinical practice, since they can be assumed that they had no differential influence on outcomes between groups, based on the current literature and randomization process with stratification per site [4, 50]. Similarly, early psychological assessments, while often appreciated by patients, have not shown significant benefits in reducing mental health distress or improving HRQoL [51]. Other RCTs evaluating coping skills training and primary care-based interventions have likewise failed to demonstrate improvements in mental health [52, 53]. In addition, ICU follow-up clinics, though widely implemented, appear to have limited or no effect on long-term mental health or HRQoL [3, 5, 54]. Furthermore, our findings align with this broader evidence that, while most single post-ICU interventions are often well appreciated by patients, measurable improvements in mental health outcomes remain difficult to achieve. In contrast, previous studies investigating VR-based interventions, including our pilot study, have shown promising results when the intervention was applied during or after ICU admission [20, 21, 55–59]. Though, there is methodological heterogeneity between studies of different VR interventions, as is extensively summarized in recent systematic reviews and meta-analysis [21, 55–58]. Therefore further exploration of ICU-VR is justified and could include a more targeted and personalized ICU-VR approach, as we and other experts in the field previously advocated for [60, 61]. Such an approach may consist of preparatory application of ICU-VR interventions to reduce anticipatory anxiety. Moreover, although not therapeutically effective in this form, this ICU-VR video may have pragmatic value as an informational tool, since ICU survivors are in need for more treatment related information [6]. By standardizing the explanation of ICU care, it might potentially complement aftercare consultations and support patient education with minimal increase of staff workload. Several limitations must however be acknowledged. First, the ICU-VR video was only available in Dutch and French, which likely introduced selection bias by excluding non-native speakers. 181 potential participants were excluded based on a language barrier (1.3% of all screened patients), of which 125 (69%) were excluded at a participating hospital in Rotterdam, The Netherlands - a diverse and multicultural city. Since patients with such language barriers understand even less of what happened during their ICU admission, this group might benefit most from such interventions [62]. Additionally, like most trials in this field, we enrolled patients with an ICU length of stay ≥72 h, as some previous studies reported a longer ICU length of stay as a risk factor for PICS-related psychological symptoms [3, 63]. However, recent evidence suggests that psychological symptoms may also develop in this group [64]. Future studies should consider including these types of patients or focus on those at higher risk of mental health distress. Second, the study did not assess whether the ICU-VR video affected communication dynamics between staff and patients. While the intervention may partially replace informational aspects of aftercare, it may lack the empathy and warmth of direct human interaction. Still, satisfaction with aftercare did not differ between groups, suggesting a limited influence on communication dynamics. Future studies should formally assess the impact of the ICU-VR video on communication quality and patient-clinician relationships, using mixed-methods approaches (e.g. validated communication quality scales combined with qualitative interviews). Third, the ICU-VR video in the early group was delivered prior to discharge, whereas late ICU-VR video participants received the intervention during the aftercare clinic. Consequently, non-attendance or insufficient time of patients during the outpatient visit likely contributed to lower uptake in the late ICU-VR group, introducing selection bias. On the other hand, per protocol analysis showed no differences between groups in those who actually received the ICU-VR video intervention. Finally, the content of the ICU-VR video was similar for all participants, while individuals received different treatments and interventions during ICU admission. Patients noted that certain ICU procedures (e.g. restrictive mobility interventions), were missing, limiting their ability to relate the ICU-VR video to their own ICU experience. A future ICU-VR intervention should be developed with more input from former ICU-patients. Additionally, a more individual and flexible, modular ICU-VR video design, allowing patients to select content based on their own ICU experience, may improve engagement and effectiveness. Conclusions In this international, multicentre, randomized clinical trial, a single and non-personalized, ICU-VR video session did not improve PTSD, anxiety, depression, or HRQoL at six months among ICU survivors. Participants valued the intervention for enhancing understanding of their ICU stay, supporting its use as an informational adjunct to aftercare. Future studies should focus on personalized and modular VR content, with a multi-session, therapist guided, timing-adaptive VR design, ultimately in high risk patients, and explore pre-ICU applications to reduce anticipatory anxiety before elective procedures. Take-home message A single and non-personalized, ICU-VR video after ICU treatment is insufficient in improving mental health and quality of life in ICU survivors, but nonetheless highly valued for enhancing understanding of ICU stay, supporting its use as an informational adjunct to aftercare. Future studies should focus on personalized and modular VR content, with a multi-session, therapist guided, timing-adaptive VR design, ultimately in high risk patients. Supplementary Information Below is the link to the electronic supplementary material. Supplementary Material 1. Supplementary Material 1. Supplementary Material 2. Supplementary Material 2.