Publication

• Title: Integrated Telehealth Rehabilitation and Quality of Life in Mechanically Ventilated Adults: A Randomized Clinical Trial
• Acronym: Tele-Rehab MV / Tele-Rehab
• Year: 2026
• Journal published in: JAMA
• Citation: Rosa RG, Moraes RB, Trott G, et al; Tele-Rehab Trial Group. Integrated telehealth rehabilitation and quality of life in mechanically ventilated adults: a randomized clinical trial. JAMA. Published online June 10, 2026.

Context & Rationale

• Background

  • Survivors of acute respiratory failure requiring invasive mechanical ventilation commonly experience persistent physical weakness, dysphagia, cognitive impairment, anxiety, depression, reduced independence, and impaired health-related quality of life.
  • Professional guidance increasingly emphasises analgesia optimisation, lighter sedation, delirium prevention, spontaneous breathing trials, early mobilisation, ventilator liberation, and post-ICU recovery planning. ¹²
  • Despite strong biological and face-valid rationale, randomised trials of ICU rehabilitation and post-ICU follow-up have produced inconsistent effects on long-term quality of life.
  • Most earlier trials tested one phase of care, usually ICU mobilisation alone or postdischarge rehabilitation alone, rather than a continuous recovery pathway spanning ICU, ward, and community.
  • Telehealth offered a plausible delivery mechanism in Brazil because public hospitals often face rehabilitation workforce constraints, geographical barriers, and variable access to specialist multidisciplinary follow-up.
  • The Tele-Rehab programme was originally developed from a quality-improvement initiative for critically ill patients with COVID-19, then broadened to acute hypoxaemic respiratory failure when COVID-19 ICU admissions declined. ³⁴
  • The trial’s central innovation was not simply “remote rehabilitation”; it was a telehealth-enabled implementation system integrating ventilator liberation, ward risk stratification, targeted rehabilitation, nurse navigation, and postdischarge support.
• Research Question/Hypothesis

  • The trial tested whether an integrated multicomponent telehealth-based rehabilitation strategy would improve EQ-5D-3L health-related quality of life at 90 days after hospital discharge.
  • The target population was adults with acute hypoxaemic respiratory failure requiring invasive mechanical ventilation in Brazilian public hospitals.
  • The hypothesis was that recovery-oriented care starting in ICU, continuing through the ward, and extending after discharge would improve patient-centred recovery more effectively than usual care delivered in isolated care phases.
  • The trial addressed an important gap: whether a system-level recovery pathway can improve outcomes when implemented remotely in hospitals with limited baseline ICU rehabilitation infrastructure.
• Why This Matters

  • ICU survivorship is now a major critical care outcome domain; survival alone is an incomplete endpoint when survivors face prolonged disability.
  • Health systems increasingly invest in telehealth, remote rehabilitation, remote monitoring, and navigator programmes, but the evidence base remains uneven.
  • A beneficial integrated strategy could provide a scalable way to implement ICU liberation and rehabilitation in resource-constrained settings.
  • A neutral or mortality-driven result would still be informative because it would distinguish ICU process improvement from true survivor functional recovery.
  • The ACCOMPLISH trial, published contemporaneously, sharpens the interpretation: remote monitoring after serious infection did not increase postdischarge home days, showing that telehealth surveillance alone should not be assumed to improve recovery. ⁵

Design & Methods

• Research Question:
  • Among adults with acute hypoxaemic respiratory failure requiring invasive mechanical ventilation, does an integrated multicomponent telehealth-based rehabilitation intervention improve EQ-5D-3L health-related quality of life at 90 days after hospital discharge compared with usual care?
• Study Type:
  • Investigator-led, multicentre, stepped-wedge, cluster randomised clinical trial.
  • Conducted in adult ICUs of 20 Brazilian public hospitals.
  • Hospitals with multiple ICUs were treated as one cluster.
  • Clusters crossed from usual care to intervention at prespecified 2-month intervals after an initial control period, continuing until all clusters had crossed over.
  • Randomisation was stratified by ICU bed number, ≤10 versus >10 beds, and generated by a statistician not involved in trial conduct.
  • Allocation concealment used a secure password-protected system accessible only to the statistician.
  • Cluster assignments were disclosed to local teams 60 days before implementation to permit training.
  • The protocol and statistical analysis plan were published before the main trial report. ³⁴
• Population:
  • Eligible clusters were adult ICUs in Brazilian public hospitals with at least 8 beds.
  • Participating ICUs required capacity to admit patients with acute respiratory failure and suspected COVID-19, and internet access for remote care.
  • Adult patients aged ≥18 years were eligible if admitted to a participating ICU with acute hypoxaemic respiratory failure requiring invasive mechanical ventilation.
  • SARS-CoV-2 infection had to be part of the differential diagnosis; confirmed COVID-19 was not required, and patients remained eligible if an alternative diagnosis was later established.
  • Key exclusions were estimated life expectancy of 3 months or less, absence of a legally authorised representative in cases of communication barrier, lack of telephone contact, anticipated inability to complete longitudinal telephone follow-up for reasons other than lack of access to a telephone or smartphone, and previous enrolment.
  • Consent for in-hospital data collection was waived because the intervention targeted organisational processes of care and did not involve untested clinical procedures.
  • Written consent for post-ICU follow-up was obtained at hospital discharge from patients or legally authorised representatives.
• Intervention:
  • The intervention was a multicomponent telehealth-based rehabilitation programme spanning ICU, ward, and postdischarge phases.
  • ICU tele-bundle: local ICU staff delivered the bundle during daily multidisciplinary rounds, with remote review by a critical care teleconsultant during three 1-hour telehealth sessions per week.
  • ICU components targeted analgesia optimisation, sedation minimisation, systematic spontaneous breathing trials, delirium prevention, early mobilisation, and removal of unnecessary invasive devices.
  • Ward tele-bundle: centralised rehabilitation specialists used smartphone videoconferencing with local ward-team support.
  • Ward assessment screened for nutritional impairment, oropharyngeal dysphagia, reduced physical capacity, anxiety, and depression using standardised tools.
  • Ward rehabilitation plans included nutritional guidance, swallowing rehabilitation, physical therapy, respiratory therapy, and psychological follow-up.
  • Postdischarge tele-bundle: a 2-month personalised centralised telerehabilitation programme was delivered via smartphone videoconferencing.
  • Postdischarge components included tailored physical therapy, respiratory therapy, swallowing rehabilitation, psychological follow-up, and handoff to the primary care physician.
  • Critical care nurse telenavigators integrated all three phases, coordinated communication, addressed implementation barriers, and maintained continuity across ICU, ward, and postdischarge care.
  • Study-provided smartphones were used for ward sessions and were supplied to postdischarge survivors without compatible devices.
• Comparison:
  • During control periods, ICUs continued usual care according to local protocols.
  • Usual care did not include the structured telehealth implementation support, centralised ward rehabilitation assessment, nurse telenavigation, or 2-month postdischarge telerehabilitation programme.
  • Baseline usual care was variable across hospitals: 10/20 hospitals (50.0%) had daily multidisciplinary rounds, 13/20 (65.0%) used standardised sedation protocols, 14/20 (70.0%) used daily spontaneous awakening trials, and 11/20 (55.0%) had early mobilisation protocols.
• Blinding:
  • Participants, ICU teams, ward teams, and implementation teams were not blinded.
  • Blinding was not feasible because the intervention changed visible care processes, training, rounds, teleconsultations, videoconferencing, and postdischarge rehabilitation.
  • Centralised telephone outcome assessors remained blinded to cluster assignment.
  • Mortality and mechanical ventilation duration were comparatively objective, but extubation practices, ward discharge decisions, and self-reported recovery outcomes remained vulnerable to performance and response bias.
• Statistics:
  • Power calculation: A minimum of 18 clusters enrolling 100 patients each, total 1800 patients, was required to detect a 0.09 between-group difference in EQ-5D-3L utility score with 95% power at a 2-sided α of .05, assuming SD 0.28, intracluster correlation coefficient 0.05, and a 4-sequence, 5-period stepped-wedge design; 20 clusters were enrolled to allow for cluster loss.
  • The primary analysis was conducted at the individual participant level while accounting for stepped-wedge cluster randomisation.
  • The primary analysis set included all enrolled participants except those transferred to another hospital during ICU or ward stay and those without available follow-up consent/outcome data.
  • Participants were analysed according to randomised group assignment.
  • Missing outcome data were not imputed in the primary analysis.
  • The primary outcome was analysed using a generalised linear mixed model following the Hussey-Hughes approach, with random intercept for ICU, fixed effects for study period, and adjustment for age, sex, Charlson Comorbidity Index, and SAPS-3.
  • Prespecified sensitivity analyses included multiple imputation for transferred or lost-to-follow-up participants, zero-inflated Gaussian mixed-effects modelling, and analyses stratified by implementation fidelity.
  • Secondary and subgroup analyses were not adjusted for multiple comparisons and should be interpreted as exploratory.
• Follow-Up Period:
  • Patients were enrolled between June 2024 and May 2025.
  • Follow-up continued through September 2025.
  • The primary outcome was assessed 90 days after hospital discharge.
  • The postdischarge intervention lasted 2 months.
  • Centralised telephone follow-up occurred at prespecified postdischarge windows up to 90 days.

Key Results

This trial continued to completion. No cluster withdrew. Enrolment was extended because ICU admissions for acute hypoxaemic respiratory failure declined, slowing accrual, but the trial was not stopped early for efficacy, harm, or futility.

• The primary EQ-5D-3L result was statistically significant but interpretively delicate: death was scored as 0, median EQ-5D-3L was 0 in both groups, and survivor-only EQ-5D-3L did not improve.
• The most compelling clinical separation was ICU-process related: mechanical ventilation duration was 9.9 ± 10.3 days versus 15.5 ± 15.9 days; adjusted mean difference −6.2 days; 95% CI −8.5 to −3.9; P<.001.
• Subgroup analyses did not show convincing heterogeneity: age ≤65 years adjusted mean difference 0.06 (95% CI 0.00 to 0.11) versus >65 years 0.05 (95% CI 0.00 to 0.11), interaction P=.91; COVID-19 diagnosis −0.09 (95% CI −0.31 to 0.13) versus non-COVID-19 0.06 (95% CI 0.01 to 0.11), interaction P=.17; SAPS-3 <64 0.06 (95% CI 0.00 to 0.11) versus ≥64 0.05 (95% CI −0.01 to 0.11), interaction P=.84; men 0.06 (95% CI 0.00 to 0.12) versus women 0.05 (95% CI 0.00 to 0.11), interaction P=.80.

Internal Validity

• Randomisation and Allocation:
  • The stepped-wedge cluster design was appropriate for a hospital-level intervention that changed ICU workflow, ward transitions, telehealth infrastructure, and postdischarge rehabilitation.
  • Randomisation was computer-generated, stratified by ICU bed number, and concealed using a secure password-protected system accessible only to the statistician.
  • All 20 clusters completed the trial, avoiding cluster attrition.
  • Cluster assignments were disclosed 60 days before implementation, which was operationally necessary for training but created a theoretical risk of anticipatory practice change before formal crossover.
  • The analysis accounted for cluster and period effects, which is essential in stepped-wedge trials.
• Dropout and Post-randomisation Exclusions:
  • Of 4692 adults screened, 1916 were enrolled: 1063 in intervention periods and 853 in usual-care periods.
  • Primary outcome analysis included 1687 patients: 941 in the intervention group and 746 in usual care.
  • A total of 229 enrolled patients were excluded from the primary analysis: 122/1063 (11.5%) in the intervention group and 107/853 (12.5%) in usual care.
  • Reasons for exclusion from the primary analysis were balanced: transfer to another hospital during ICU or ward stay, 82 intervention and 69 usual care; and loss to follow-up, 40 intervention and 38 usual care.
  • Excluding transferred patients limits a pure all-enrolled interpretation, especially because transfers may reflect severity, bed pressure, geography, or health-system factors.
  • The main sensitivity analysis including transferred and lost-to-follow-up participants with multiple imputation gave a similar point estimate but wider uncertainty: EQ-5D-3L 0.159 versus 0.124; adjusted difference 0.045; 95% CI −0.005 to 0.096.
• Performance and Detection Bias:
  • Bedside blinding was not feasible.
  • Unblinded teams could change sedation practice, extubation readiness, mobilisation thresholds, ward transfer decisions, discharge planning, or goals-of-care discussions.
  • Centralised telephone outcome assessors were blinded, reducing detection bias for postdischarge assessments.
  • Mortality is objective, but mechanical ventilation duration is partly behaviour-sensitive because sedation, spontaneous breathing-trial use, extubation thresholds, reintubation thresholds, and tracheostomy practice may vary by team and period.
  • Patient-reported and survivor-only outcomes were vulnerable to missingness, survivor selection, and response bias.
• Protocol Adherence:
  • Implementation fidelity was meaningful but incomplete.
  • Mean overall implementation fidelity was 69.6% (95% CI 66.0% to 73.2%).
  • ICU tele-bundle adherence was 80.4% (95% CI 75.2% to 85.6%).
  • Ward tele-bundle participation was 65.0% (95% CI 59.5% to 70.5%).
  • Postdischarge tele-bundle participation was 61.8% (95% CI 55.8% to 67.0%).
  • Among 300 intervention-group ICU survivors who met rehabilitation criteria, 269 (89.7%) attended at least 1 ward session and 222 (74.0%) attended at least 1 postdischarge session.
  • Median videoconference encounters per patient were 5.0 (IQR 2.0 to 11.0) during the ward phase and 11.5 (IQR 3.0 to 21.8) during the postdischarge phase.
  • Fidelity assessment used semistructured weekly interviews and encounter participation rather than universal direct observation, which is pragmatic but imperfect.
• Baseline Characteristics:
  • Baseline characteristics were broadly comparable across groups.
  • Mean age was 61.3 ± 17.2 years in the intervention group and 59.9 ± 17.5 years in usual care.
  • Age ≥65 years was present in 518/1055 (49.1%) versus 392/851 (46.1%).
  • Male sex was 603/1057 (57.0%) versus 473/852 (55.5%).
  • Race categories reflected Brazilian census definitions: Brown race was reported by 712/1056 (67.4%) versus 556/851 (65.5%), White race by 276/1056 (26.2%) versus 249/851 (29.2%), and Black race by 57/1056 (5.4%) versus 36/851 (4.2%).
  • Years of formal education differed numerically: median 1.0 year (IQR 0 to 8) versus 4.0 years (IQR 0 to 8).
  • Charlson Comorbidity Index was 1 (IQR 1 to 3) versus 2 (IQR 1 to 3); Charlson ≥2 occurred in 514/1062 (48.3%) versus 445/851 (52.2%).
  • SAPS-3 at ICU admission was 64 (IQR 54 to 75) versus 64 (IQR 55 to 74).
  • Vasopressor use at ICU admission was 547/1061 (51.5%) versus 447/846 (52.8%).
  • The cohort was extremely high risk: 90-day mortality was 71.8% in the intervention group and 78.3% in usual care, making all death-weighted outcomes highly mortality-sensitive.
• Heterogeneity:
  • Clinical heterogeneity was substantial, reflecting broad acute hypoxaemic respiratory failure rather than a narrow ARDS phenotype.
  • Primary causes of acute hypoxaemic respiratory failure included bacterial pneumonia, extrapulmonary sepsis, COPD exacerbation, aspiration pneumonitis, acute cardiogenic pulmonary oedema, COVID-19, pulmonary embolism, asthma, and other causes.
  • Cluster heterogeneity was also substantial because hospitals differed in baseline staffing and processes of care.
  • Before intervention implementation, only 2/20 hospitals (10.0%) reported 24/7 intensivist coverage; median ICU beds were 37 (IQR 19 to 55), median nurse-to-patient ratio was 5 (IQR 5 to 10), and median physiotherapist-to-patient ratio was 10 (IQR 9 to 11).
  • The stepped-wedge analysis adjusted for cluster and period, but residual confounding from unmeasured organisational change remains plausible.
  • No prespecified subgroup showed statistically convincing effect modification.
• Timing:
  • The ICU tele-bundle was delivered during the biologically plausible window in which sedation minimisation, daily spontaneous breathing trials, mobilisation, delirium prevention, and device removal can shorten mechanical ventilation.
  • The ward and postdischarge bundles were delivered only to patients who survived long enough to receive them, making later-phase exposure conditional on post-randomisation survival.
  • The dominant outcome separation occurred during the index hospitalisation, particularly in ICU mortality and mechanical ventilation duration.
  • This timing pattern supports an ICU-process mechanism more strongly than a direct postdischarge telerehabilitation effect.
• Dose:
  • The ICU intervention dose was substantial: three 1-hour telehealth sessions per week embedded into local multidisciplinary rounds, supported by training, site visits, Plan-Do-Study-Act cycles, and peer-learning meetings.
  • The ward and postdischarge rehabilitation dose was personalised rather than fixed.
  • Postdischarge participation was incomplete: cluster-level participation was 61.8%, and 222/300 eligible intervention patients attended at least 1 postdischarge session.
  • Among 284 survivors eligible for postdischarge follow-up, 63 (22.2%) required study-provided smartphones.
  • The dose was adequate to test a pragmatic service model but insufficient to isolate the effect of any single rehabilitation discipline, session frequency, or phase of care.
• Separation of the Variable of Interest:
  • Process separation was shown by implementation fidelity: 80.4% ICU tele-bundle adherence, 65.0% ward tele-bundle participation, and 61.8% postdischarge tele-bundle participation.
  • Clinical separation was largest for mechanical ventilation duration: 9.9 ± 10.3 days versus 15.5 ± 15.9 days; adjusted mean difference −6.2 days; 95% CI −8.5 to −3.9.
  • Primary outcome separation was small in absolute utility terms: EQ-5D-3L 0.16 ± 0.31 versus 0.12 ± 0.28; adjusted mean difference +0.049.
  • Survivor-only quality-of-life separation was absent: 0.60 ± 0.32 versus 0.59 ± 0.32; adjusted mean difference −0.045.
  • Mortality separation was clinically important: 71.8% versus 78.3%; adjusted difference −7.6 percentage points; 95% CI −14.7 to −0.6.
  • The trial therefore separated ICU liberation and mortality outcomes more convincingly than survivor functional recovery outcomes.
• Key Delivery Aspects:
  • The intervention relied on existing local multidisciplinary teams plus centralised specialists, making it pragmatic and scalable in concept.
  • The centralised workforce included critical care teleconsultants, nurse telenavigators, physiotherapists, speech-language pathologists, psychologists, and nursing coordination.
  • The intervention was complex and bundled, making causal attribution difficult.
  • The comparator was usual care in Brazilian public ICUs with variable baseline bundle adoption, not a highly protocolised, high-resource control environment.
• Crossover:
  • Classic individual crossover was not relevant because this was a cluster stepped-wedge trial.
  • Contamination was possible because clusters knew their future transition date 60 days before implementation.
  • Peer-learning meetings were restricted to hospitals within the same randomisation sequence to reduce cross-sequence contamination.
  • Any anticipatory uptake of intervention practices during usual-care periods would probably dilute the observed treatment effect, but it cannot be quantified from the published data.
• Adjunctive Therapy Use:
  • The trial did not report granular patient-level use of all ICU co-interventions beyond the intervention components.
  • Usual care already included several evidence-informed practices in some hospitals: sedation protocols in 65.0%, daily spontaneous awakening trials in 70.0%, and early mobilisation protocols in 55.0%.
  • The trial therefore tested an implementation-and-integration strategy rather than the simple presence versus absence of evidence-based ICU liberation practices.
• Outcome Assessment:
  • The primary outcome, EQ-5D-3L utility at 90 days after hospital discharge, is patient-centred and relevant.
  • Assigning death a utility score of 0 integrates mortality and quality of life into a single distribution.
  • In this high-mortality cohort, that approach made the primary outcome strongly death-weighted.
  • The observed intraclass correlation coefficient for the primary outcome was 0.02, lower than the assumed 0.05.
  • The 9-level clinical status ordinal scale captured death and transitions in care but lacks a well-established minimal clinically important difference.
  • Survivor-only questionnaires had small denominators and substantial missingness.
• Statistical Rigor:
  • The trial exceeded the planned 1800-patient target, enrolling 1916 patients.
  • The primary model matched the stepped-wedge design with cluster and period adjustment.
  • Adjustment for age, sex, Charlson Comorbidity Index, and SAPS-3 was prespecified and clinically appropriate.
  • The primary result was statistically fragile: adjusted mean difference 0.049; 95% CI 0.0002 to 0.098; P=.04.
  • The multiple-imputation sensitivity analysis including transferred and lost-to-follow-up participants had a similar point estimate but crossed zero: adjusted difference 0.045; 95% CI −0.005 to 0.096.
  • The zero-inflated Gaussian mixed-effects model gave a similar estimate: adjusted difference 0.049; 95% CI 0.000 to 0.099.
  • Secondary outcomes and subgroup analyses were exploratory because they were not adjusted for multiplicity.
  • The protocol history documents late extension of enrolment and clarification that transferred patients would be excluded from the primary analysis set; these changes were transparent but relevant to interpretation. ⁴
• Conclusion on Internal Validity:
  • Overall, internal validity is moderate to strong for testing a pragmatic hospital-level telehealth implementation strategy.
  • The findings that the intervention improved a death-weighted EQ-5D-3L outcome and shortened mechanical ventilation are credible, but attribution to postdischarge telerehabilitation is limited by the unblinded, multicomponent, stepped-wedge design and by the mortality-driven nature of the primary outcome.

External Validity

• Population Representativeness:
  • The trial enrolled adults with acute hypoxaemic respiratory failure requiring invasive mechanical ventilation in Brazilian public hospitals.
  • Participants were severely ill, with median SAPS-3 of 64 in both groups and vasopressor use at ICU admission in approximately 52%.
  • Mortality was very high: 71.8% in the intervention group and 78.3% in usual care.
  • Most participants self-identified as Brown or Black according to Brazilian census categories, and formal education was low, improving relevance to populations often underrepresented in high-income critical care rehabilitation trials.
  • The trial excluded patients with estimated life expectancy less than 3 months and those unable to complete telephone follow-up for reasons other than lack of device access.
• Applicability:
  • The results are most applicable to resource-constrained public ICU systems with inconsistent implementation of evidence-based ICU liberation and rehabilitation processes.
  • Generalisability is uncertain in high-resource ICUs with 24/7 intensivist coverage, mature ABCDEF/PADIS bundle implementation, established early mobilisation, robust ward rehabilitation, and structured post-ICU follow-up.
  • Generalisability is also uncertain in lower-mortality cohorts because the primary outcome was strongly influenced by death.
  • The intervention requires telehealth infrastructure, centralised specialist rehabilitation personnel, nurse telenavigation, smartphone access or device provision, and local staff willing to engage in repeated implementation cycles.
  • The findings do not directly apply to children, non-invasively ventilated patients, patients ventilated for non-hypoxaemic indications, elective postoperative cohorts, or systems without capacity for remote rehabilitation and nurse navigation.
  • The ACCOMPLISH data reinforce that results from one telehealth model cannot be extrapolated to another: symptom-questionnaire remote monitoring after serious infection did not increase 90-day home days and worsened home days in adults aged ≥65 years. ⁵
• Conclusion on External Validity:
  • External validity is strongest for high-mortality, resource-constrained ICU systems where evidence-based liberation and rehabilitation processes are inconsistently implemented.
  • External validity is limited for lower-risk, high-resource systems already delivering mature ICU liberation bundles and structured post-ICU rehabilitation; in those settings, the incremental benefit of Tele-Rehab is uncertain.

Strengths & Limitations

• Strengths:
  • Large pragmatic cluster trial enrolling 1916 mechanically ventilated adults across 20 public hospitals.
  • Appropriate stepped-wedge design for a system-level intervention.
  • Integrated intervention spanning ICU, ward, and postdischarge phases.
  • Focus on a patient-centred primary outcome rather than only process metrics.
  • Centralised blinded outcome assessment after discharge.
  • Published protocol and statistical analysis plan.
  • High relevance to public-hospital and resource-constrained settings.
  • Detailed reporting of implementation fidelity and participation.
  • Large reduction in mechanical ventilation duration, supporting a plausible ICU liberation mechanism.
  • Study-provided smartphones reduced digital access barriers among eligible survivors.
• Limitations:
  • Unblinded clinical intervention with risk of performance bias.
  • Stepped-wedge design susceptible to secular trends, seasonal variation, and period effects.
  • Cluster assignments disclosed 60 days before implementation, creating potential for anticipatory behaviour.
  • Primary analysis excluded transferred and lost-to-follow-up participants.
  • Primary outcome was heavily influenced by death; survivor-only EQ-5D-3L did not improve.
  • Very high mortality limits generalisability and reduced survivor denominators for functional outcomes.
  • Multicomponent intervention prevents identification of the active ingredient.
  • Implementation fidelity was incomplete and did not show a simple dose-response relationship.
  • Secondary and subgroup analyses were exploratory and not adjusted for multiplicity.
  • Postdischarge rehabilitation exposure occurred only in survivors, making later-phase effects difficult to separate from survival selection.

Interpretation & Why It Matters

• Clinical practice

  Tele-Rehab supports telehealth-enabled implementation of integrated ICU liberation and rehabilitation pathways in resource-constrained settings where sedation minimisation, spontaneous breathing trials, mobilisation, and multidisciplinary coordination are inconsistently delivered.
• What improved

  The intervention improved death-weighted EQ-5D-3L, reduced 90-day mortality, shortened mechanical ventilation by 6.2 adjusted days, and increased days alive and free of hospital by 4.9 adjusted days.
• What did not improve

  Among survivors, measured EQ-5D-3L utility, anxiety, depression, cognitive impairment, instrumental activities of daily living, physical dependence, rehospitalisation, and return to work did not significantly improve.
• Mechanistic interpretation

  The most plausible active mechanism is improved ICU liberation and early care-process implementation, rather than a direct effect of postdischarge telerehabilitation on survivor function.
• How it changes thinking

  The trial shifts rehabilitation thinking from “post-ICU clinic” towards “continuous recovery system”, but it also shows that mortality-weighted quality-of-life outcomes can obscure whether survivors themselves function better.
• Role of ACCOMPLISH

  ACCOMPLISH provides an important contrast: remote symptom monitoring after serious infection did not increase home days, so the lesson from Tele-Rehab is not “telehealth works”; it is that telehealth may work when coupled to actionable, high-leverage ICU processes and structured rehabilitation pathways. ⁵

Controversies & Subsequent Evidence

• Mortality-weighted quality of life:
  • The primary endpoint is clinically defensible because it prevents death from being ignored in a quality-of-life trial.
  • The primary endpoint is also difficult to communicate because the observed treatment effect was driven mainly by fewer deaths rather than higher survivor quality of life.
  • The primary adjusted mean difference of 0.049 lies within the estimated EQ-5D-3L minimal clinically important difference range used by the trialists, 0.03 to 0.06.
  • The median EQ-5D-3L utility score was 0 in both groups.
  • Survivor-only EQ-5D-3L did not improve: 0.60 ± 0.32 versus 0.59 ± 0.32; adjusted mean difference −0.045; 95% CI −0.138 to 0.045; P=.34.
  • The multiple-imputation sensitivity analysis including transferred and lost-to-follow-up patients crossed zero: adjusted difference 0.045; 95% CI −0.005 to 0.096.
  • The appropriate interpretation is therefore “improved death-weighted health-related quality of life”, not “better quality of life among survivors”.
• ICU bundle versus postdischarge telerehabilitation:
  • The mortality and mechanical ventilation signals are more consistent with ICU liberation processes than with postdischarge rehabilitation.
  • The intervention targeted analgesia optimisation, light sedation, spontaneous breathing trials, delirium prevention, early mobilisation, and device removal, all of which plausibly reduce ventilation duration and ICU complications.
  • Postdischarge rehabilitation was only available to survivors, so its effect is conditioned on post-randomisation survival.
  • The trial design cannot separate which component mattered most: sedation practice, spontaneous breathing trials, mobilisation, nurse navigation, ward risk stratification, postdischarge rehabilitation, or centralised specialist support.
  • Implementation trials of ICU liberation bundles have previously associated greater bundle delivery with improved ICU outcomes, although they do not isolate the Tele-Rehab intervention itself. ⁶⁷
• ACCOMPLISH changes the telehealth interpretation:
  • ACCOMPLISH tested a different but highly relevant telehealth model: symptom-questionnaire remote monitoring after hospitalisation for sepsis, COVID-19, or lower respiratory tract infection. ⁵
  • In ACCOMPLISH, no remote-monitoring strategy increased postdischarge home days at 90 days compared with structured telephone support: COR 0.96 (95% CrI 0.70 to 1.32) for RPM-low standard response, 0.86 (95% CrI 0.60 to 1.23) for RPM-high standard response, 1.01 (95% CrI 0.76 to 1.33) for RPM-low enhanced response, and 0.96 (95% CrI 0.69 to 1.36) for RPM-high enhanced response.
  • ACCOMPLISH also found a concerning subgroup signal in patients aged ≥65 years: standard response COR 0.56 (95% CrI 0.36 to 0.85) and enhanced response COR 0.67 (95% CrI 0.45 to 0.98), indicating fewer home days with remote monitoring.
  • That result argues against broad claims that telehealth contact or surveillance alone improves post-acute recovery.
  • Tele-Rehab’s effect should therefore be attributed to a structured intervention that changed bedside ICU processes and transitions of care, not to remote technology as a standalone therapeutic mechanism.
• Post-ICU rehabilitation evidence remains mixed:
  • iRehab tested a 6-week remote multicomponent rehabilitation programme in 429 UK ICU survivors who had received ≥48 hours of invasive mechanical ventilation and were recruited within 12 weeks of hospital discharge. ⁸
  • iRehab did not significantly improve the primary 8-week EQ-5D-5L outcome: 0.69 ± 0.26 versus 0.67 ± 0.27; adjusted mean difference 0.04; 95% CI −0.001 to 0.09; P=.05. ⁸
  • The accompanying JAMA editorial for iRehab emphasised the need for better phenotyping, targeting, dose selection, and survivorship models rather than uniform rehabilitation for all ICU survivors. ⁹
  • SUIVI-REA, a multidisciplinary post-ICU consultation trial, did not improve 1-year mortality or quality of life. ¹⁰
  • Tele-Rehab differs from iRehab and SUIVI-REA because it began in the ICU and targeted acute-care processes, not just postdischarge survivor rehabilitation.
• Telehealth-enabled transition programmes after sepsis remain difficult:
  • The STAR/ENCOMPASS sepsis transition trial tested proactive telehealth-based sepsis transition and recovery support in 3548 patients and did not reduce the primary composite of 90-day readmission or mortality: 48.2% versus 48.0%; adjusted OR 1.05; 95% CI 0.90 to 1.24; P=.53. ¹¹
  • The ENCOMPASS result also illustrates that readmission and survival can move in different directions, making composite outcomes in post-acute care difficult to interpret.
  • The JAMA Internal Medicine commentary accompanying STAR highlighted that readmission can sometimes represent rescue rather than failure, particularly in high-risk postsepsis populations. ¹²
  • Tele-Rehab avoided readmission as its primary endpoint, but its own death-weighted EQ-5D-3L endpoint introduces a different composite-outcome challenge.
• Tele-ICU and telehealth quality-improvement evidence:
  • The ERIC stepped-wedge trial showed that intensive care telehealth can improve process quality, supporting the plausibility of a telehealth-enabled implementation mechanism. ¹³
  • The TELESCOPE trial in Brazilian ICUs showed that tele-ICU infrastructure alone does not guarantee broad clinical outcome benefit. ¹⁴
  • The implication is that remote expertise must be attached to a specific, actionable, high-yield clinical process; technology alone is not the intervention.
• Guideline context:
  • The 2025 SCCM PADIS focused update issued conditional recommendations relevant to Tele-Rehab, including enhanced mobilisation/rehabilitation over usual mobilisation/rehabilitation in adult ICU patients. ¹
  • The 2018 PADIS guideline remains important background for analgesia, sedation, delirium, immobility, and sleep disruption management in ICU. ²
  • Future guidance should distinguish ICU liberation implementation, ward rehabilitation, postdischarge telerehabilitation, and symptom-based remote monitoring as separate interventions with different evidential bases.

Summary

• Tele-Rehab was a 20-cluster Brazilian stepped-wedge trial of an integrated ICU, ward, and postdischarge telehealth rehabilitation strategy in 1916 adults with acute hypoxaemic respiratory failure requiring invasive mechanical ventilation.
• The primary EQ-5D-3L utility score at 90 days after hospital discharge was higher with the intervention: 0.16 ± 0.31 versus 0.12 ± 0.28; adjusted mean difference 0.049; 95% CI 0.0002 to 0.098; P=.04.
• The primary result was largely mortality driven: 90-day mortality was 71.8% versus 78.3%; adjusted difference −7.6 percentage points; 95% CI −14.7 to −0.6; P=.03, while survivor-only EQ-5D-3L did not differ.
• Mechanical ventilation duration was substantially shorter with the intervention: 9.9 ± 10.3 versus 15.5 ± 15.9 days; adjusted mean difference −6.2 days; 95% CI −8.5 to −3.9; P<.001.
• In light of ACCOMPLISH and iRehab, Tele-Rehab is best interpreted as evidence for telehealth-enabled ICU liberation and integrated recovery implementation in resource-constrained systems, not as proof that remote monitoring or postdischarge telerehabilitation alone improves survivor quality of life.

Overall Takeaway

Tele-Rehab is an important systems-level critical care rehabilitation trial because it tested recovery as a continuum beginning in the ICU, rather than as a postdischarge add-on. Its most practice-relevant signal is reduced mechanical ventilation duration and lower mortality in a high-risk public-hospital population; its main caution is that survivor quality of life did not measurably improve, and the benefit should not be generalised to remote monitoring alone.

Bibliography

• 1.Lewis K, Balas MC, Stollings JL, et al. A focused update to the clinical practice guidelines for the prevention and management of pain, anxiety, agitation/sedation, delirium, immobility, and sleep disruption in adult patients in the ICU. Crit Care Med. 2025;53(3):e711-e727.
• 2.Devlin JW, Skrobik Y, Gélinas C, et al. Clinical practice guidelines for the prevention and management of pain, agitation/sedation, delirium, immobility, and sleep disruption in adult patients in the ICU. Crit Care Med. 2018;46(9):e825-e873.
• 3.Pereira AJ, Moraes RB, Trott G, Rosa RG, dos Santos MC, Miozzo AP, et al. The impact of a multicomponent telemedicine-based intervention on quality of life in adults with respiratory failure requiring mechanical ventilation: protocol for a cluster stepped-wedge randomized clinical trial (Tele-Rehab MV Trial). Crit Care Sci. 2025;37:e20250136.
• 4.Cavaliere YF, Moraes RB, Trott G, Rosa RG, Zampieri FG, Pereira AJ, et al. Statistical analysis plan for a cluster stepped-wedge randomized clinical trial assessing the effects of a multicomponent telemedicine-based intervention on quality of life in adults with respiratory failure requiring mechanical ventilation (Tele-Rehab MV Trial). Crit Care Sci. 2026;38:e20260308.
• 5.Yende S, Talisa VB, Mayes K, Williams K, Malito A, Mayr FB, et al. Remote monitoring approaches to reduce readmissions after infection and sepsis: a randomized clinical trial. JAMA Netw Open. 2026;9(6):e2616641.
• 6.Pun BT, Balas MC, Barnes-Daly MA, Thompson JL, Aldrich JM, Barr J, et al. Caring for critically ill patients with the ABCDEF bundle: results of the ICU Liberation Collaborative in over 15 000 adults. Crit Care Med. 2019;47(1):3-14.
• 7.Barnes-Daly MA, Phillips G, Ely EW. Improving hospital survival and reducing brain dysfunction at seven California community hospitals: implementing PAD guidelines via the ABCDEF bundle in 6064 patients. Crit Care Med. 2017;45(2):171-178.
• 8.O’Neill B, Bradley JM, Connolly B, Bruce J, Ratna M, Lall R, et al; iRehab Trial Investigators. Remote multicomponent rehabilitation in intensive care unit survivors: a randomized clinical trial. JAMA. Published online May 18, 2026.
• 9.Moncion K, Smith-Turchyn J, Kho ME. The future of survivorship in critical illness. JAMA. Published online May 18, 2026.
• 10.Sharshar T, Grimaldi-Bensouda L, Siami S, et al; Suivi-Rea Investigators. A randomized clinical trial to evaluate the effect of post-intensive care multidisciplinary consultations on mortality and the quality of life at 1 year. Intensive Care Med. 2024;50(5):665-677.
• 11.Taylor SP, Eaton T, Rios A, Holcomb JB, Ferraro JP, Marquez-Martin E, et al. Proactive telehealth-based sepsis transition and recovery support, hospital readmission, and mortality: a randomized clinical trial. JAMA Intern Med. 2025;185(10):1238-1246.
• 12.Grady D, Durant RW, Anderson TS. What if readmission prevents death? The challenge of composite outcomes. JAMA Intern Med. 2025;185(10):1246-1247.
• 13.Spies CD, Paul N, Adrion C, et al; ERIC Study Group. Effectiveness of an intensive care telehealth programme to improve process quality (ERIC): a multicentre stepped wedge cluster randomised controlled trial. Intensive Care Med. 2023;49(2):191-204.
• 14.Pereira AJ, Noritomi DT, dos Santos MC, et al. Effect of tele-ICU on clinical outcomes of critically ill patients: the TELESCOPE randomized clinical trial. JAMA. 2024;332(21):1798-1807.
• 15.Prescott HC, Antonelli M, Alhazzani W, Møller MH, Alshamsi F, Azevedo LCP, et al. Surviving Sepsis Campaign: international guidelines for management of sepsis and septic shock 2026. Crit Care Med. 2026;54(4):725-812.