An augmented rehabilitation package incorporating immersive virtual reality during early recovery after ARDS: A randomized controlled trial

Highlights

• •Immersive VR was incorporated into rehabilitation during early recovery after ARDS.
• •The package improved short-term MoCA scores compared with usual rehabilitation.
• •The package improved short-term SF-36 scores compared with usual rehabilitation.
• •The package was associated with lower short-term HADS-A scores.
• •Findings do not establish a VR-specific effect because controls were not time- or attention-matched.

Abstract

Background

Survivors of acute respiratory distress syndrome (ARDS) frequently experience cognitive impairment, psychological symptoms, and reduced health-related quality of life after intensive care. Immersive virtual reality (VR) may provide an engaging, multimodal adjunct to conventional rehabilitation, but evidence in the ARDS recovery population remains limited.

Methods

In this single-center, parallel-group randomized controlled trial, 88 ARDS patients in the rehabilitation phase were randomized 1:1 to a 4-week augmented rehabilitation package incorporating immersive VR in addition to usual care (n = 44) or usual care alone (n = 44). One control participant was lost to follow-up, and 87 participants were included in the available-case analysis. The primary outcome was change in Montreal Cognitive Assessment (MoCA) score from baseline to 4 weeks. Secondary outcomes included the 36-Item Short Form Health Survey (SF-36) and Hospital Anxiety and Depression Scale (HADS) scores. The main between-group analysis used analysis of covariance (ANCOVA), with follow-up score as the dependent variable and baseline score as the covariate.

Results

Baseline imbalances were observed for age, MoCA, SF-36, and HADS-A scores. After adjustment for corresponding baseline scores, the VR group showed greater improvement in MoCA score than the control group (adjusted between-group effect: 1.42, 95% CI 0.86–1.98; p < 0.001). The VR group also showed higher SF-36 scores (adjusted effect: 13.20, 95% CI 11.85–14.54; p < 0.001) and lower HADS-A scores (adjusted effect: −2.55, 95% CI −3.23 to −1.88; p < 0.001). In the VR group, 29 of 44 participants (65.9%) reported being very satisfied or satisfied with the intervention.

Conclusions

An augmented rehabilitation package incorporating immersive VR was associated with short-term improvements in cognition, health-related quality of life, and anxiety symptoms during early ARDS recovery. Because the package also entailed additional structured rehabilitation exposure, staff attention, interaction, repeated activity, feedback, and treatment time, and because the control condition was not time- or attention-matched, the observed differences cannot be attributed specifically to VR. These single-center, short-term, hypothesis-generating findings require confirmation in larger studies with time- and attention-matched controls. Trial registration: Chinese Clinical Trial Registry, ChiCTR2600122078; retrospectively registered on 8 April 2026, after completion of participant recruitment.

Background

Acute respiratory distress syndrome (ARDS) is a common severe condition among critically ill patients, and its diagnosis and severity classification are commonly based on the Berlin definition [1]. With improvements in ventilation strategies and intensive care, more patients survive the acute phase and enter a prolonged recovery period. However, survival does not necessarily mean full recovery. Post-intensive care syndrome (PICS) encompasses physical, cognitive, and psychological impairments after critical illness, and these sequelae may affect long-term function and health-related quality of life [2]. Cognitive impairment after ARDS is common but heterogeneous across studies, and it may affect independence, return to work, and perceived recovery [3]. Cognitive impairment is also frequently observed among broader ICU survivor populations during early follow-up, supporting the need to address cognition during the post-ICU rehabilitation window [4].

Conventional critical care rehabilitation usually emphasizes respiratory training, exercise rehabilitation, early mobilization, and psychological support. These components remain essential, but adherence may be reduced by fatigue, dyspnea, anxiety, low motivation, and delayed feedback. Cognitive rehabilitation strategies already exist; however, they are not always easily integrated into routine post-ICU rehabilitation pathways. Immersive VR provides a potential complementary tool because it can combine respiratory pacing, simulated movement tasks, attention-demanding interaction, relaxation, and immediate multisensory feedback within a safe and engaging environment. Prior studies and reviews suggest that VR-based interventions are feasible in ICU-related populations and may reduce stress or anxiety, but the available protocols and populations remain heterogeneous [[5], [6], [7]]. Evidence from pulmonary rehabilitation also supports the feasibility of immersive VR as an adjunct to respiratory rehabilitation, although the COPD population differs from ARDS survivors [8].

Evidence specifically addressing immersive VR rehabilitation in ARDS survivors remains limited. In particular, it is unclear whether adding VR to usual rehabilitation during the early ARDS recovery phase provides incremental benefits beyond conventional respiratory, motor, and psychological training. This study therefore evaluated a 4-week augmented rehabilitation package incorporating immersive VR and examined its association with cognitive function, health-related quality of life, anxiety and depressive symptoms, and participant-reported satisfaction in a single-center randomized controlled trial.

Methods

Study design and participants

This study was conducted as a single-center, parallel-group randomized controlled trial. Participants were patients with ARDS who were hospitalized in the critical care rehabilitation ward of The First Hospital of Jiaxing and had entered the rehabilitation phase between March 2024 and September 2024. ARDS was diagnosed according to the Berlin definition [1]. Patients were eligible whether ARDS was the primary reason for ICU admission or developed during the ICU stay, provided that they had been weaned from invasive mechanical ventilation and were clinically stable enough to participate in rehabilitation.

The inclusion criteria were defined as follows: (1) age ≥ 18 years; (2) a history of ICU treatment for ARDS, followed by successful weaning from invasive mechanical ventilation and transfer to the critical care rehabilitation phase; (3) relatively stable vital signs and ability to tolerate rehabilitation training; (4) current communication ability sufficient to understand instructions and complete scale assessments, judged with reference to the patient's pre-ICU communication status when available; and (5) written informed consent provided by the patient or legal guardian. The exclusion criteria were: (1) history of severe neurological disease; (2) severe visual or auditory impairment precluding VR participation; (3) active psychiatric disorder requiring antipsychotic medication or active delirium that prevented scale assessment or safe participation; (4) life-threatening comorbidity with an expected survival of less than 3 months; and (5) inability to complete follow-up assessments.

The study protocol was approved by the Medical Ethics Committee of The First Hospital of Jiaxing on March 18, 2021 (Approval No. LS2021-KY-275). All participants or their legal guardians provided written informed consent after receiving a comprehensive explanation of the study.

Sample size estimation

The planned sample size was set at 88 participants (44 per group), allowing for an anticipated 15% attrition rate. This target was determined with reference to the reported burden and heterogeneity of cognitive impairment after ARDS and to the feasible number of eligible patients during the recruitment period [3]. The absence of a formal power calculation based on a prespecified continuous between-group effect for the primary MoCA outcome means that this trial is exploratory, and the findings require confirmation in larger multicenter studies.

Randomization and blinding

Eligible patients were assigned to the control group or the VR group at a 1:1 ratio using a random number sequence method with a block size of 4. To ensure allocation concealment, the random sequence was generated by an independent staff member uninvolved in the study implementation, and the allocation results were sealed in consecutively numbered, opaque envelopes that were opened sequentially according to the order of patient enrollment.

Due to the nature of the intervention, blinding was not feasible for patients and intervention providers; however, outcome assessors and data analysts remained blinded to the grouping information to minimize measurement and analysis bias. Participant screening, exclusion, randomization, follow-up, and analysis are summarized in Fig. S1.

Intervention methods

Composition of the research team

The research team comprised critical care rehabilitation physicians, psychotherapists, rehabilitation technicians, rehabilitation specialist nurses, VR engineers, and nursing researchers. Specifically, the critical care rehabilitation physicians and psychotherapists were responsible for formulating the rehabilitation training protocols, while the VR engineers managed VR content development and technical support. The intervention was jointly implemented by the rehabilitation specialist nurses, rehabilitation technicians, and psychotherapists, with nursing researchers responsible for data collection, follow-up, and data organization.

Intervention methods in the VR group

A preliminary VR rehabilitation training protocol was developed based on a literature review and group discussions, with the final version established following two rounds of expert consultation (Table 1). This protocol comprised three core modules: respiratory training, exercise rehabilitation, and psychological intervention. Upon development by engineers, all VR rehabilitation content was evaluated by an associate chief physician specializing in critical care rehabilitation to ensure safety and operability prior to clinical application.

Table 1.

Content and implementation of the VR-based rehabilitation program.

Phase	Component	Intervention content	Frequency and duration	Provider
Phase I Inpatient rehabilitation	Respiratory training	Patients wore VR equipment and viewed relaxing natural scenery while performing paced deep breathing exercises synchronized with breathing rhythm, combined with effective cough training.	10 min per component; total 30 min per session; 5 sessions per week	Rehabilitation nurse
	Exercise rehabilitation	Simulated walking scenarios in VR (e.g., beach or park environments) combined with bed-based joint range-of-motion exercises of the upper and lower limbs and progressive muscle strengthening exercises (e.g., straight-leg raising).	10 min per component; total 30 min per session; 5 sessions per week	Rehabilitation nurse; rehabilitation therapist
	Psychological intervention	Patients viewed relaxing VR environments while performing breathing relaxation techniques and mindfulness exercises to promote emotional regulation and psychological relaxation.	10 min per component; total 30 min per session; 5 sessions per week	Psychotherapist
Phase II Outpatient follow-up rehabilitation	Respiratory training	Dynamic VR breathing scenarios (e.g., grassland or ocean environments) were used to enhance engagement in breathing exercises and reinforce respiratory rhythm training.	10 min per component; total 30 min per session; once weekly	Rehabilitation nurse
	Exercise rehabilitation	Progressive functional training in VR environments, including standing balance exercises, stepping training, and gait training within simulated walking scenarios.	10 min per component; total 30 min per session; once weekly	Rehabilitation therapist; rehabilitation nurse
	Psychological intervention	Patients viewed calming VR environments (e.g., starry sky or grassland scenes) while performing breathing relaxation techniques and mindfulness training to alleviate anxiety and promote relaxation.	10 min per component; total 30 min per session; once weekly	Psychotherapist

Note: Each intervention component lasted 10 min, with a total intervention duration of 30 min per session. During the inpatient rehabilitation phase, training was conducted five sessions per week; during the outpatient follow-up phase, training was conducted once weekly. VR scenarios and training intensity were dynamically adjusted according to patients’ heart rate, blood pressure, oxygen saturation, and subjective tolerance. Training was paused or terminated if patients developed obvious discomfort or abnormal vital-sign fluctuations.

The intervention lasted for 4 weeks and consisted of an in-hospital rehabilitation phase followed by a post-discharge outpatient phase (Table 1). During the in-hospital phase, VR sessions were delivered five times per week; after discharge, sessions were delivered once weekly during scheduled critical care rehabilitation outpatient visits. The reduced outpatient frequency reflected the transition from ward-based daily rehabilitation to follow-up-based rehabilitation and was applied according to the prespecified protocol. Throughout training, patient vital signs were monitored and recorded in real time using a Mindray VS900 mobile monitoring platform integrated with 5G technology. For the VR training, a head-mounted display (PICO neo3) was used to provide immersive visual and auditory scenarios, including natural landscapes, beach walks, forest exploration, and relaxation environments. The protocol followed a progressive principle, moving from static to dynamic tasks and from simple to more complex activities.

Training scenarios and difficulty levels were dynamically adjusted according to heart rate, blood pressure, oxygen saturation, subjective tolerance, and intervention-team assessment. Each session lasted 30 min, with respiratory training, exercise rehabilitation, and psychological relaxation each lasting approximately 10 min. Session completion, required breaks, interruptions, and reasons for discontinuation were recorded by the intervention team. Training was immediately paused or terminated if patients developed obvious discomfort, dizziness, nausea, abnormal heart-rate or blood-pressure fluctuations, or a decrease in oxygen saturation.

Intervention methods in the control group

The control group received a routine rehabilitation regimen developed by a multidisciplinary team and consisting of the same three core domains: respiratory training, exercise rehabilitation, and psychological intervention. Respiratory training included daily deep-breathing and effective cough exercises, with oxygen saturation maintained above 95% when clinically feasible. Exercise rehabilitation followed a stepwise strategy from bed-based activity to sitting balance, standing, and walking training according to functional status. Psychological intervention comprised routine relaxation training and mindfulness-based support. The key distinction was that the VR group received the same rehabilitation domains delivered through immersive scenario-based VR modules in addition to usual care, whereas the control group received routine rehabilitation without VR.

Outcome measures

All outcome measures were assessed at baseline and upon completion of the 4-week intervention by uniformly trained evaluators.

Cognitive function was evaluated using the Chinese version of the MoCA. The MoCA assesses multiple domains, including visuospatial ability, executive function, naming, attention, language, abstraction, delayed recall, and orientation, with a total score of 30 points; higher scores indicate better cognitive function. The MoCA is widely used as a screening tool for cognitive impairment [9].

The Chinese version of the SF-36 was used to evaluate health-related quality of life. The SF-36 includes eight domains: physical functioning, role-physical, bodily pain, general health, vitality, social functioning, role-emotional, and mental health. Each domain is scored from 0 to 100, with higher scores indicating better health-related quality of life. The SF-36 is a widely used generic measure of health-related quality of life [10].

Psychological status was assessed using the HADS, which consists of the HADS-Anxiety (HADS-A) and HADS-Depression (HADS-D) subscales. Each subscale contains seven items scored from 0 to 3; a subscale score of ≥8 indicates possible anxiety or depressive symptoms, and a score of ≥11 suggests more marked symptoms. The HADS is widely used to assess anxiety and depressive symptoms in medical populations [11].

Statistical analysis

Data analysis was performed using Statistical Package for the Social Sciences (SPSS) 26.0 software. Continuous variables were initially subjected to normality and homogeneity of variance tests. Measurement data conforming to a normal distribution were expressed as mean ± standard deviation (SD), whereas those not following a normal distribution were presented as median (interquartile range) [M (P25, P75)]. Categorical data were described as frequencies and percentages [n (%)].

Regarding the comparison of baseline data, continuous variables exhibiting a normal distribution and homogeneity of variance were analyzed using the independent samples t-test, while those with non-normal distributions were examined using the Mann–Whitney U test Categorical variables were analyzed using the Pearson χ² test; however, Fisher's exact test was utilized when the theoretical frequency requirements were not met.

This study employed the MoCA score as the primary outcome measure, while the SF-36 and HADS scores served as secondary outcome measures. Descriptive statistics were calculated for all scores in both groups before and after the intervention. Within-group comparisons between pre- and post-intervention scores were conducted using paired t-tests; however, if the data did not follow a normal distribution, the Wilcoxon signed-rank test was utilized.

To evaluate the intervention effect while controlling for baseline outcome differences, between-group efficacy was assessed using ANCOVA, with the post-intervention score as the dependent variable, study group as a fixed factor, and the corresponding baseline score as the covariate. The homogeneity of regression slopes was checked before interpreting ANCOVA results. Change values for each outcome (Δ value = post-intervention score - baseline score) were also calculated and compared between groups using independent-samples t tests for normally distributed variables or Mann-Whitney U tests otherwise. For HADS scores, a decrease indicated symptom improvement, whereas for MoCA and SF-36 scores, an increase indicated improved outcomes.

To verify the robustness of the primary analytical results, a sensitivity analysis was further performed using non-parametric tests on the change values.

The available-case analysis set included all randomized participants with observed follow-up outcome data. Dropout cases were included in the participant-flow description and baseline characterization but were not included in outcome analyses because follow-up outcome data were unavailable. Missing data were not imputed, and analyses were based on observed values. All tests were two-sided, with statistical significance defined as p < 0.05. In addition to p values, between-group mean differences and 95% confidence intervals (95% CIs) were reported to improve interpretability. Because multiple secondary outcomes were tested without multiplicity adjustment, the secondary results are interpreted as exploratory.

Results

Comparison of baseline characteristics between the two groups

A total of 126 patients were assessed for eligibility. Thirty-eight were excluded, including 26 who did not meet the inclusion criteria, 9 who declined to participate, and 3 who were excluded for other reasons. Eighty-eight participants were randomized, with 44 allocated to the control group and 44 allocated to the VR group. One participant in the control group was lost to follow-up, resulting in 43 control participants and 44 VR participants in the final available-case analysis (Fig. S1). Baseline characteristics were generally similar between groups for sex, BMI, education, smoking and drinking history, comorbidities, and ARDS severity. However, the VR group was younger than the control group, and baseline MoCA, SF-36, and HADS-A scores differed significantly between groups, whereas baseline HADS-D scores did not differ significantly. These baseline characteristics are summarized in Table 2.

Table 2.

Baseline Characteristics of the Study Population by Group.

Characteristic	Control (n = 43)	VR (n = 44)	Statistic	p value
Age, years	76.19 ± 11.46	70.89 ± 13.03	t = 2.016	0.047
BMI, kg/m²	23.30 ± 4.84	23.26 ± 4.04	t = 0.045	0.964
MoCA at baseline	13.98 ± 6.49	17.11 ± 7.29	t = −2.121	0.037
SF-36 at baseline	54.00 ± 16.24	63.95 ± 14.35	t = −3.027	0.003
HADS-A at baseline	12.21 ± 5.42	9.50 ± 6.99	t = 2.023	0.046
HADS-D at baseline	9.91 ± 3.03	9.64 ± 2.99	t = 0.419	0.676
Sex: Male	23 (53.5%)	26 (59.1%)	χ² = 0.096	0.756
Female	20 (46.5%)	18 (40.9%)
Education: Primary school or below	31 (72.1%)	30 (68.2%)	χ² = 4.910	0.178
Junior middle school	10 (23.3%)	6 (13.6%)
High school/technical school	1 (2.3%)	6 (13.6%)
College or above	1 (2.3%)	2 (4.5%)
ARDS severity: Mild	36 (83.7%)	30 (68.2%)	χ² = 4.423	0.110
Moderate	5 (11.6%)	13 (29.5%)
Severe	2 (4.7%)	1 (2.3%)

Note. Continuous variables are presented as mean ± SD and categorical variables as n (%). p values for continuous variables were derived from independent-samples t tests, and p values for categorical variables from chi-square tests. Abbreviations: ARDS, acute respiratory distress syndrome; BMI, body mass index; HADS-A, Hospital Anxiety and Depression Scale-Anxiety subscale; HADS-D, Hospital Anxiety and Depression Scale-Depression subscale; MoCA, Montreal Cognitive Assessment; VR, virtual reality.

Comparison of cognitive function between the two groups

MoCA scores increased from baseline in both groups, with a greater improvement in the VR group. In the control group, MoCA scores increased from 13.98 ± 6.49 to 16.26 ± 4.91, with a change of 2.28 ± 1.68. In the VR group, MoCA scores increased from 17.11 ± 7.29 to 20.43 ± 7.24, with a change of 3.32 ± 1.34. The change-score difference favored the VR group by 1.04 points (95% CI 0.39–1.69; p = 0.002). ANCOVA using post-intervention MoCA score as the dependent variable and baseline MoCA score as the covariate also showed a significant adjusted between-group effect (1.42, 95% CI 0.86–1.98; p < 0.001) (Table 3 and Fig. 1).

Table 3.

Primary and secondary outcome measures at baseline, follow-up, and change from baseline.

Outcome	Control baseline	Control follow-up	Control change	VR baseline	VR follow-up	VR change	Adjusted group effect	95% CI	p value
MoCA	13.98 ± 6.49	16.26 ± 4.91	2.28 ± 1.68	17.11 ± 7.29	20.43 ± 7.24	3.32 ± 1.34	1.419	0.861 to 1.977	<0.001
SF-36	54.00 ± 16.24	53.23 ± 13.27	−0.77 ± 3.64	63.95 ± 14.35	75.19 ± 14.29	11.23 ± 3.36	13.198	11.853 to 14.544	<0.001
HADS-A	12.21 ± 5.42	11.26 ± 4.60	−0.95 ± 1.13	9.50 ± 6.99	6.52 ± 5.84	−2.98 ± 2.52	−2.555	−3.226 to −1.883	<0.001
HADS-D	9.91 ± 3.03	8.23 ± 1.46	−1.67 ± 1.63	9.64 ± 2.99	7.18 ± 0.97	−2.45 ± 2.13	−0.946	−1.131 to −0.762	<0.001

Note. Values are presented as mean ± SD. Change values are calculated as follow-up minus baseline. Adjusted group effects represent the VR group relative to the control group and were estimated using analysis of covariance with follow-up score as the dependent variable, study group as the fixed factor, and the corresponding baseline score as the covariate. CI, confidence interval.

Fig. 1.

Distribution and paired changes of cognitive, quality-of-life, and psychological outcomes.

Note: (A) MoCA score; (B) SF-36 score; (C) HADS-A score; (D) HADS-D score. Upper panels show individual score distributions at baseline and follow-up, and lower panels show mean scores with error bars. The p values below each baseline–follow-up pair indicate within-group paired comparisons. Between-group intervention effects were primarily evaluated using ANCOVA adjusted for corresponding baseline scores, as reported in Table 3.

Comparison of QoL between the two groups

The two groups showed different SF-36 patterns after intervention. In the control group, SF-36 scores changed from 54.00 ± 16.24 to 53.23 ± 13.27, with a change of −0.77 ± 3.64. In the VR group, SF-36 scores increased from 63.95 ± 14.35 to 75.19 ± 14.29, with a change of 11.23 ± 3.36. The change-score difference favored the VR group by 12.00 points (95% CI 10.51–13.50; p < 0.001). After adjusting for baseline SF-36 score, ANCOVA showed a significant adjusted between-group effect (13.20, 95% CI 11.85–14.54; p < 0.001) (Table 3 and Fig. 1).

Comparison of anxiety status between the two groups

HADS-A scores decreased in both groups, with a larger reduction in the VR group. In the control group, HADS-A scores decreased from 12.21 ± 5.42 to 11.26 ± 4.60, with a change of −0.95 ± 1.13. In the VR group, HADS-A scores decreased from 9.50 ± 6.99 to 6.52 ± 5.84, with a change of −2.98 ± 2.52. The change-score difference favored the VR group by −2.02 points (95% CI −2.86 to −1.19; p < 0.001). After adjustment for baseline HADS-A score, ANCOVA showed a significant adjusted between-group effect (−2.55, 95% CI −3.23 to −1.88; p < 0.001) (Table 3 and Fig. 1).

Comparison of depression status between the two groups

HADS-D scores decreased from 9.91 ± 3.03 to 8.23 ± 1.46 in the control group (change: −1.67 ± 1.63) and from 9.64 ± 2.99 to 7.18 ± 0.97 in the VR group (change: −2.45 ± 2.13). The between-group difference in change values did not reach statistical significance (VR minus control: −0.78, 95% CI −1.59 to 0.03; p = 0.058). However, ANCOVA adjusting for baseline HADS-D showed a significant adjusted between-group effect (−0.95, 95% CI −1.13 to −0.76; p < 0.001). The HADS-D result was therefore exploratory (Table 3 and Fig. 1).

Sensitivity analysis

Sensitivity analyses compared ANCOVA results with change-score comparisons and Mann-Whitney U tests. The direction and statistical interpretation of MoCA, SF-36, and HADS-A were consistent across analyses, supporting the robustness of these findings. For HADS-D, the ANCOVA result was statistically significant, whereas the change-score and nonparametric comparisons were not, indicating only partial consistency. The sensitivity analysis is presented in Table S1, and adjusted between-group effects with 95% CIs are shown in Fig. S2.

Acceptance of the VR intervention

The VR intervention showed acceptable participant-reported satisfaction. Among the 44 VR participants, 12 (27.3%) were very satisfied, 17 (38.6%) were satisfied, 10 (22.7%) were neutral, and 5 (11.4%) were dissatisfied. The combined number of very satisfied and satisfied participants was 29, corresponding to a total satisfaction rate of 65.9% (Fig. 2).

Fig. 2.

Distribution of participant-reported satisfaction with the VR intervention.

Note: Satisfaction was assessed only in the VR group (n = 44) and categorized as very satisfied, satisfied, neutral, or dissatisfied. Percentages were calculated using the VR group as the denominator.

Discussion

This single-center randomized controlled trial evaluated an augmented rehabilitation package incorporating immersive VR during the ARDS recovery phase. The VR group showed greater short-term improvement in MoCA and SF-36 scores and a larger reduction in HADS-A scores after adjustment for baseline values. HADS-D findings were inconsistent across analytical methods: ANCOVA showed a significant adjusted effect, whereas change-score and nonparametric sensitivity analyses did not reach statistical significance. These findings are clinically relevant because PICS commonly includes cognitive and psychological sequelae that influence long-term quality of life after critical illness [2]. The results also align with evidence that cognitive impairment after ARDS and ICU discharge is common and heterogeneous, supporting the need for structured rehabilitation approaches during early recovery [3,4].

The observed MoCA difference may reflect the combined effects of the augmented rehabilitation package, including structured activity, staff interaction, additional rehabilitation exposure, and immersive VR. Components of the package may have increased attention, task engagement, and multisensory feedback; however, the design does not isolate the contribution of immersive VR. However, the adjusted between-group MoCA effect was modest, and MoCA is a screening instrument rather than a detailed neuropsychological battery [9]. The cognitive result therefore provides supportive short-term evidence but does not establish durable cognitive recovery. Future trials should include longer follow-up and more detailed cognitive testing to determine whether this difference translates into sustained functional benefit.

The SF-36 findings indicate a short-term between-group difference associated with the augmented rehabilitation package, but they do not establish a VR-specific effect. One possible explanation is that immersive scenes and immediate feedback may improve training experience, self-efficacy, and willingness to participate. The slight decrease in SF-36 score in the control group may reflect short-term fluctuation during early post-ICU recovery and the influence of baseline imbalance rather than true deterioration attributable to routine rehabilitation. Nevertheless, the VR group had a higher baseline SF-36 score, and baseline perceptions of recovery may have influenced receptiveness or adherence in ways that statistical adjustment cannot fully capture. The SF-36 is a broad patient-reported measure, and future studies should analyze its physical and mental component domains separately to identify which quality-of-life dimensions are most responsive to VR [10].

The most consistent psychological result was the short-term between-group difference in HADS-A scores associated with the augmented rehabilitation package. This is compatible with previous ICU-related VR evidence suggesting that VR is feasible and may reduce stress or anxiety [5]. The multicenter feasibility trial of ICU-specific VR also supports the acceptability of this approach in ICU survivors, although that work focused primarily on feasibility and exploratory psychological outcomes [6]. The HADS-D result was sensitive to the analytical approach and remains exploratory, as statistical significance was observed only in the outcome-specific ANCOVA, whereas the change-score and nonparametric sensitivity analyses were not statistically significant. Depression-related improvement may require longer intervention periods, longer follow-up, or more targeted psychological content than the short 4-week protocol used here. The HADS is a symptom scale rather than a diagnostic psychiatric assessment [11].

The intervention was designed as an adjunct rather than a replacement for routine rehabilitation. Accordingly, the observed between-group differences may reflect the combined effects of immersive delivery, additional structured rehabilitation exposure, attention from intervention staff, repeated activity, feedback, and engagement rather than a VR-specific effect. The control intervention was not matched for time or attention, and actual rehabilitation exposure was not quantified in sufficient detail for a dose-response comparison. Although scenarios were adjusted according to vital signs and subjective tolerance, the study did not use a validated engagement scale. Participant satisfaction was acceptable, with 65.9% of VR participants reporting that they were very satisfied or satisfied, but reasons for dissatisfaction and minor adverse experiences such as dizziness or nausea were not systematically graded. Standardized treatment-dose, adverse-event, and acceptability reporting will be needed before broader implementation [5].

This study has several limitations. First, this was a single-center study with a modest sample size, and baseline imbalances in age, MoCA, SF-36, and HADS-A reduce the strength of causal inference. The outcome-specific ANCOVA models adjusted for the corresponding baseline outcome value but did not account for imbalance in age or other potentially prognostic characteristics. Given the modest sample size, residual confounding remains possible, and statistical adjustment may not fully resolve the baseline imbalances. Second, blinding of patients and intervention providers was not feasible. Third, the VR group received an adjunctive package that entailed additional structured rehabilitation exposure, staff attention, interaction, and activity, whereas the control group was not matched for time and attention. The observed effects therefore cannot be attributed specifically to VR.

Fourth, information on ICU and hospital course, concomitant treatment, actual session completion, and total rehabilitation exposure was not collected consistently enough for retrospective between-group comparison. The exact duration spent in the inpatient and post-discharge outpatient rehabilitation phases also varied with clinical discharge timing and was not captured in sufficient detail for phase-specific analysis. These gaps limit assessment of whether differences beyond the VR intervention contributed to the outcomes. Fifth, the primary analysis used an available-case approach without imputation; because one randomized control participant lacked follow-up outcome data, a small risk of attrition bias remains. Sixth, follow-up was limited to 4 weeks. The findings therefore provide evidence only of short-term differences in measured outcomes during early recovery and cannot establish durable cognitive, psychological, functional, or health-related quality-of-life benefits. Finally, the study relied on questionnaire and screening-scale outcomes and did not include objective neurobiological or long-term functional endpoints.

Overall, the findings support further study of an adjunctive immersive VR-based rehabilitation package but do not establish a VR-specific treatment effect. The results are preliminary and hypothesis-generating. Future multicenter trials should use stronger allocation balance, prospectively powered continuous endpoints, longer follow-up, standardized treatment exposure, detailed ICU and hospital-course data, and a time- and attention-matched control intervention. Objective process indicators such as adherence, engagement, training dose, adverse events, and functional recovery should also be incorporated to clarify whether VR provides benefit beyond additional rehabilitation exposure and structured attention. Longer follow-up is required to determine whether the observed short-term differences persist and translate into durable cognitive, psychological, functional, or health-related quality-of-life benefits.

Conclusion

An augmented rehabilitation package incorporating immersive VR was associated with greater short-term improvement in cognitive function, health-related quality of life, and anxiety symptoms among ARDS survivors during early recovery compared with usual rehabilitation. However, because the intervention also involved additional structured rehabilitation exposure, attention, and activity, the observed differences cannot be attributed specifically to VR. The depression-related findings were sensitive to the analytical approach and remain exploratory. These preliminary, hypothesis-generating results require confirmation in larger multicenter studies with standardized treatment exposure, a time- and attention-matched control intervention, more complete characterization of ICU and hospital course, and longer follow-up.

Authors' contributions

Maoxian Yang and Mengsun Li conceived and designed the study. Xuwen Wu, Liyan Yang, and Ya Shen performed the experiments. Qianqian Wang and Han Sheng analyzed the data. Mengsun Li and Zewei Jiang wrote the manuscript. All authors reviewed and approved the final version of the manuscript.

Ethics approval and consent to participate

This study was conducted in accordance with the Declaration of Helsinki. The study protocol was approved by the Medical Ethics Committee of The First Hospital of Jiaxing (Approval No. LS2021-KY-275). The record filing number is MR-33-26-026711. The trial was retrospectively registered with the Chinese Clinical Trial Registry on 8 April 2026, after participant recruitment had been completed (registration number: ChiCTR2600122078). Written informed consent was obtained from all participants or their legal guardians prior to enrollment.

Consent for publication

Not applicable.

Funding

This work was supported by 2021 Joint Research Fund of The First Hospital of Jiaxing/Affiliated Hospital of Jiaxing University (No. 2021YJKY038), the 2021 Academic Leaders Project (No. 2021LHJJ003), and the Construction Project of Provincial Clinical Key Specialty (No. 2024-ZJZK-002).

Data availability

All data generated or analyzed during this study are included in this article and/or its supplementary material files. Further enquiries can be directed to the corresponding author.

Declaration of competing interest

The authors declare no conflict of interest.

Appendix A. Supplementary data

The following are Supplementary data to this article:

Fig. S1

CONSORT flowchart of participant screening, randomization, follow-up, and analysis. Note: The flowchart displays the number of participants assessed for eligibility, reasons for exclusion, randomized allocation, follow-up, and final analysis.

Fig. S2

Forest plot of adjusted between-group effects for primary and secondary outcome measures. Note: Points represent adjusted between-group effects for the VR group relative to the control group from ANCOVA; horizontal lines represent 95% CIs; the dashed vertical line indicates an effect of 0.