Skip to main content

Foundational Trials

TELESCOPE

Image: Shutterstock / ChaNaWiT

Publication

  • Title: Effect of Tele-ICU on Clinical Outcomes of Critically Ill Patients: The TELESCOPE Randomized Clinical Trial
  • Acronym: TELESCOPE (TELE-critical care verSus usual Care On ICU PErformance)
  • Year: 2024
  • Journal published in: JAMA
  • Citation: Pereira AJ, Noritomi DT, dos Santos MC, et al; TELESCOPE Trial Investigators. Effect of Tele-ICU on Clinical Outcomes of Critically Ill Patients: The TELESCOPE Randomized Clinical Trial. JAMA. 2024;332(21):1798-1807.

Context & Rationale

  • Background
    • Tele-ICU (tele–critical care) is promoted as a strategy to extend intensivist expertise to ICUs with limited on-site specialist staffing, potentially improving quality, safety, and efficiency.
    • Prior evidence for tele-ICU benefit (mortality/length of stay) was largely observational and heterogeneous (models range from continuous remote monitoring with authority to advisory scheduled rounds), leaving uncertainty about causal effects and transportability.
    • Brazilian public hospital ICUs commonly face workforce and process constraints (including limited structured multidisciplinary rounds and audit/feedback), making this a plausible “high-yield” setting for tele–critical care evaluation.
  • Research Question/Hypothesis
    • Does an ICU-level tele–critical care programme (daily telemedicine multidisciplinary rounds plus monthly audit/feedback and protocol dissemination) reduce ICU length of stay compared with usual care in adult general ICUs in Brazilian public hospitals?
    • Hypothesis: tele–critical care would shorten ICU length of stay via improved processes of care and earlier readiness for ICU discharge.
  • Why This Matters
    • Tele-ICU programmes are resource-intensive (technology, specialist time, governance), and policy adoption often precedes definitive randomised evidence.
    • A large pragmatic trial in under-resourced public ICUs could clarify whether tele–critical care meaningfully improves patient-centred outcomes and ICU performance at scale, informing investment decisions in similar health systems.

Design & Methods

  • Research Question: In adult general ICUs in Brazilian public hospitals, does a tele–critical care programme reduce ICU length of stay compared with usual care?
  • Study Type: Pragmatic, multicentre, parallel-group, cluster-randomised clinical trial (ICU as the unit of randomisation), conducted in 30 adult general ICUs across Brazil; batched randomisation (5 blocks of 6 ICUs) with a 2-month baseline period per ICU, a 1-month transition/implementation period, and a 13-month intervention period.
  • Population:
    • Setting (cluster level): Adult general (mixed medical–surgical) ICUs in public hospitals (all Brazilian regions); ICUs had on-site physicians and nurses and ≥8 ICU beds.
    • Key ICU exclusions: Presence of structured daily multidisciplinary rounds led by an on-site board-certified intensivist; regular audit/feedback on performance indicators; fully implemented local clinical protocols (per trial definitions); specialised/step-down ICUs.
    • Patient inclusion: Adults (≥18 years) admitted to participating ICUs during the trial periods.
    • Patient exclusions: <18 years; legal custody (“justice-related custody”); admissions during the waiting/transition period between randomisation and evaluation start.
    • Analysis populations: Primary outcome analysed on first ICU admission only; secondary outcomes generally analysed including index admissions and readmissions (as pre-specified in the trial report).
  • Intervention:
    • Tele–critical care rounds: Daily (Monday–Friday) telemedicine multidisciplinary rounds led by a remote board-certified intensivist with an ICU nurse and ICU respiratory therapist, conducted with the local ICU team at the bedside using a structured daily care plan.
    • Audit/feedback: Monthly virtual meetings to discuss ICU performance indicators with local ICU leadership.
    • Protocols/training: Dissemination of 19 evidence-based clinical protocols and e-learning materials (text and video) for ICU physicians and multidisciplinary staff.
    • Technology: Mobile rack with all-in-one PC, microphone, and 4K 360° camera with remote control; software previously used in the tele–critical care network.
    • Pre-specification: Protocol and statistical analysis plan were published separately before/alongside trial reporting.12
  • Comparison:
    • Usual care: Standard bedside ICU care delivered by the local ICU team; no tele–critical care daily rounds, monthly audit/feedback, or protocol e-learning package during the trial evaluation period.
    • COVID-19 co-intervention: After the onset of the COVID-19 pandemic, COVID-19 protocols and training videos were shared with both groups to mitigate post-randomisation bias (potentially narrowing separation for pandemic-era care processes).
  • Blinding: Unblinded (cluster-level service intervention); objective outcomes (e.g., length of stay, mortality, device-associated infections) reduce detection bias risk, but performance bias and co-intervention effects remain plausible.
  • Statistics: Sample size of 15,000 patients planned to detect a 1.5-day absolute reduction in ICU length of stay (baseline mean 8 days, SD 10) with 80% power at a 2-sided 5% significance level, assuming an intraclass correlation coefficient of 0.018 and variable cluster sizes; primary analysis by intention-to-treat using mixed-effects modelling (log-transformed ICU length of stay with random ICU effects and adjustment for key covariates).12
  • Follow-Up Period: Outcomes assessed from ICU admission until ICU discharge or death; follow-up truncated at 90 days while in hospital from ICU admission (per trial report definitions).

Key Results

This trial was not stopped early. It completed enrolment across 30 ICUs, with 15,230 patients included during the intervention period (7,471 in tele–critical care ICUs and 7,759 in usual care ICUs).

Outcome Tele–critical care Usual care Effect p value / 95% CI Notes
Primary: ICU length of stay (days), first ICU admission Mean (SD) 8.1 (10.0)
Median (IQR) 4.6 (2.0–10.1)		 Mean (SD) 7.1 (9.0)
Median (IQR) 3.8 (1.8–8.7)		 Adjusted % change 8.2% 95% CI −5.4% to 23.8%; P=0.24 Adjusted median difference 0 days (95% CI −2 to 3)
Hospital mortality 3106/7471 (41.6%) 3119/7759 (40.2%) OR 0.93 95% CI 0.78 to 1.12; P not reported Adjusted absolute difference −1.2% (95% CI −4.2% to 1.9%)
Incidence of CLABSI (per 1,000 central venous catheter-days) 59/50,527 (1.2) 29/44,562 (0.7) IRR 1.15 95% CI 0.43 to 3.07; P not reported Directionally higher in tele–critical care ICUs; uncertainty large
Incidence of ventilator-associated events (per 1,000 ventilator-days) 656/40,231 (16.3) 596/34,856 (17.1) IRR 1.02 95% CI 0.82 to 1.27; P not reported Adjusted absolute difference 2.5 (95% CI −28.2 to 33.3)
28-day ventilator-free days Mean (SD) 8.2 (11.0) Mean (SD) 9.2 (11.0) IRR 1.00 95% CI 0.95 to 1.05; P not reported Adjusted absolute difference 0.0 (95% CI −0.4 to 0.4)
Receiving enteral or parenteral feeding (per 100 patient-days) 59,757/67,367 (88.7) 55,317/62,386 (88.7) IRR 0.99 95% CI 0.92 to 1.06; P not reported Process-of-care outcome; no meaningful separation in point estimate
Light sedation or alert/calm (per 100 ventilator-days) 14,531/40,231 (36.1) 12,114/34,856 (34.8) IRR 1.10 95% CI 0.92 to 1.32; P not reported Process-of-care outcome
Normoxaemia (SpO2 92–96%) (per 100 patient-days) 35,587/67,367 (52.8) 23,168/62,386 (37.1) IRR 1.10 95% CI 0.75 to 1.61; P not reported Large unadjusted separation; adjusted uncertainty wide
ICU-level standardised mortality ratio (SMR) Mean (SD) 1.06 (0.3) Mean (SD) 1.07 (0.2) Adjusted difference −0.03 95% CI −0.15 to 0.08; P not reported No signal of improvement in ICU-level risk-adjusted mortality
ICU-level standardised resource use (SRU) Mean (SD) 2.04 (1.1) Mean (SD) 1.78 (0.9) Adjusted difference −0.03 95% CI −0.38 to 0.31; P not reported No signal of improved resource efficiency
  • Primary outcome showed no reduction in ICU length of stay: 8.1 (10.0) vs 7.1 (9.0) days; adjusted % change 8.2% (95% CI −5.4% to 23.8%; P=0.24).
  • Hospital mortality was similar: 41.6% vs 40.2% (OR 0.93; 95% CI 0.78 to 1.12; P not reported).
  • Intervention delivery was moderate: tele-visits on 32,074/46,920 eligible patient-days (68%) with 111,901/151,723 recommendations accepted (74%); despite this, patient-level and ICU-level outcomes did not improve.

Internal Validity

  • Randomisation and allocation: Cluster randomisation (ICU-level) in 5 batched blocks with a restricted algorithm to minimise baseline imbalance (region, ICU size, baseline case-mix and performance metrics); allocation concealment at the cluster level was not feasible after assignment given the service-delivery nature of the intervention.
  • Dropout/exclusions: No reported cluster dropouts; patients admitted during the transition/waiting period were excluded by design; primary outcome used first ICU admission only (readmissions excluded for primary analysis).
  • Performance/detection bias: Unblinded trial introduces risk of co-intervention and behavioural change in both arms; key outcomes are objective (length of stay, mortality, device-associated infections), limiting detection bias.
  • Protocol adherence and fidelity:
    • Tele-visits occurred on 32,074/46,920 eligible patient-days (68%).
    • Median proportion of eligible patient-days visited per ICU was 72% (IQR 62%–87%); 6 ICUs exceeded 70% of eligible days visited.
    • Total recommendations: 151,723; accepted: 111,901 (74%); median accepted proportion per ICU 76% (IQR 66%–83%).
    • Monthly performance indicator meetings completed: 149 (62%).
  • Baseline characteristics: During the intervention period, groups were broadly comparable in age and illness severity (e.g., SAPS 3 mean 55.6 vs 54.8; SOFA median 6 vs 5), with some case-mix differences (e.g., invasive mechanical ventilation at ICU admission 48.7% vs 42.2%); analyses were adjusted for key prognostic covariates.
  • Heterogeneity: ICU-level heterogeneity (organisation, baseline performance, pandemic timing) is intrinsic to the setting; modelling accounted for clustering and allowed secular trends to vary across clusters, but residual heterogeneity and effect-modification by context remain plausible.
  • Timing: Implementation followed a defined transition period; the intervention overlapped substantially with the COVID-19 pandemic (11,192/15,230 patients [73.5%] admitted after the pandemic began), potentially diluting effects or shifting outcome determinants (capacity strain, discharge constraints).
  • Dose (intensity) of intervention: The tested model was scheduled tele-rounds and performance feedback (not continuous remote monitoring); tele-round delivery was below the aspirational threshold of covering ≥70% of eligible days in many ICUs (median 72% but with variability), potentially limiting effect size.
  • Separation of the variable of interest:
    • Eligible tele-round days: 46,920 patient-days (after excluding weekends/holidays) with tele-visits on 32,074 (68%).
    • Recommendation acceptance: 111,901/151,723 (74%).
    • Monthly indicator meetings: 149 (62%).
  • Crossover/contamination: Formal crossover not reported; COVID-19 protocols/training materials were shared with both arms after pandemic onset, which could reduce between-group separation for pandemic-era care processes.
  • Outcome assessment: Primary outcome (ICU length of stay) is objective but can be influenced by system-level discharge barriers; secondary outcomes include objective measures (mortality, device-associated infections) and process metrics derived from routinely collected data.
  • Statistical rigour: A priori protocol and statistical analysis plan support analytical transparency and reduce selective reporting risk.12

Conclusion on Internal Validity: Moderate: the pragmatic cluster design, large sample, objective outcomes, and published protocol/SAP strengthen causal inference, but unblinded delivery, variable fidelity, pandemic-era co-interventions, and system-determined length-of-stay constraints plausibly limited detectable effects.

External Validity

  • Population representativeness: Adult general ICUs in Brazilian public hospitals without structured daily intensivist-led multidisciplinary rounds or mature audit/feedback infrastructure; this matches many resource-constrained ICUs but not high-intensity, intensivist-staffed academic ICUs.
  • Intervention applicability: The evaluated tele–critical care model was primarily scheduled weekday rounds plus audit/feedback and protocol dissemination; findings are most applicable to similar “advisory/rounds-based” tele-ICU programmes rather than continuous monitoring models with direct decision-making authority.
  • Health-system translation: Generalisability may be strongest to low- and middle-income settings with workforce gaps and public-sector constraints; applicability to settings with different discharge pathways, ICU bed pressures, and governance (which strongly influence ICU length of stay) may be limited.

Conclusion on External Validity: Good for public-sector, under-resourced general ICUs considering an advisory rounds-based tele–critical care model; limited for high-intensity ICUs or tele-ICU systems built around continuous surveillance and stronger remote authority.

Strengths & Limitations

  • Strengths: Large pragmatic cluster-randomised evaluation at scale (30 ICUs; >15,000 patients); clinically and policy-relevant setting (public Brazilian ICUs); objective primary and key secondary outcomes; detailed implementation metrics (tele-round coverage, recommendation volume/acceptance); published protocol and SAP supporting transparency.12
  • Limitations: Unblinded service-delivery intervention (performance bias); heterogeneous context across ICUs and substantial COVID-19 overlap (including shared pandemic protocols); intervention intensity constrained to weekday rounds with moderate coverage (68% of eligible patient-days); primary outcome (ICU length of stay) is system-sensitive and may be less responsive to clinical decision support in strained public systems; secondary outcomes often reported without p values (interpretation relies on confidence intervals).

Interpretation & Why It Matters

  • Clinical meaning
    • In this pragmatic public-ICU setting, a weekday tele-rounds-based tele–critical care programme did not shorten ICU length of stay (8.1 vs 7.1 days; adjusted % change 8.2%; 95% CI −5.4% to 23.8%; P=0.24) and did not reduce hospital mortality (41.6% vs 40.2%; OR 0.93; 95% CI 0.78 to 1.12).
    • These results argue against assuming that tele-rounds alone, even with frequent recommendations (151,723 total; 74% accepted), necessarily translates into improved patient-centred outcomes or ICU-level efficiency.
  • Mechanistic inference
    • Null effects may reflect limited intervention “dose” (tele-visits on 68% of eligible days), constrained authority (advisory model), and/or dominant system constraints on ICU discharge and downstream bed availability.
    • The substantial pandemic-era overlap and shared COVID-19 materials across groups plausibly reduced between-group separation for key processes most sensitive to protocolised guidance.

Controversies & Subsequent Evidence

  • Tele-ICU is not a single intervention: Prior reviews synthesise heterogeneous tele-ICU models (continuous monitoring vs scheduled rounds; advisory vs decision authority), which complicates causal interpretation and helps explain variable effects across studies and settings.456
  • Outcome selection debate (ICU length of stay): ICU length of stay is objective but strongly shaped by organisational factors (step-down capacity, discharge pathways, bed strain), potentially reducing responsiveness to decision-support interventions; this can bias trials towards null even with improved clinical processes.
  • Implementation fidelity as a causal component: TELESCOPE achieved moderate tele-round coverage (68% of eligible patient-days) and incomplete monthly performance meetings (62%), raising the possibility that the tested “real-world dose” was insufficient to move outcomes despite high recommendation volume (151,723) and acceptance (74%).
  • Technology evaluation in complex systems: The accompanying editorial emphasised that complex technological innovations in critical care require rigorous evaluation and careful attention to context, implementation, and mechanism—frames that align with TELESCOPE’s null patient-centred results despite substantial intervention activity.3
  • Where TELESCOPE sits in the evidence base: Meta-analyses have reported associations between tele-ICU exposure and improved outcomes in some settings, but are limited by heterogeneity and (often) non-randomised primary studies; TELESCOPE adds large randomised evidence suggesting that benefits are not guaranteed for advisory rounds-based tele-ICU models in under-resourced public ICUs.456

Summary

  • Large pragmatic cluster-randomised trial (30 ICUs; 15,230 patients) evaluating a weekday tele-rounds-based tele–critical care programme in Brazilian public ICUs.
  • No reduction in ICU length of stay: 8.1 vs 7.1 days; adjusted % change 8.2% (95% CI −5.4% to 23.8%; P=0.24).
  • No reduction in hospital mortality: 41.6% vs 40.2% (OR 0.93; 95% CI 0.78 to 1.12; P not reported).
  • Moderate implementation fidelity: tele-rounds occurred on 68% of eligible patient-days; 151,723 recommendations issued with 74% acceptance; monthly performance meetings completed 62% of the time.
  • ICU-level performance metrics (SMR and SRU) did not improve, suggesting limited system-level impact of the evaluated tele-ICU model in this context.

Overall Takeaway

TELESCOPE is a landmark, policy-relevant cluster-randomised evaluation showing that a pragmatic tele–critical care model built around weekday tele-rounds, audit/feedback, and protocol dissemination did not improve ICU length of stay, mortality, or ICU-level efficiency in Brazilian public ICUs. The trial highlights that tele-ICU effectiveness is contingent on implementation fidelity, authority, and local system constraints—and that “tele-ICU” should not be assumed to deliver uniform outcome benefits across settings.

Overall Summary

  • 30-ICU pragmatic cluster RCT in Brazilian public hospitals; 15,230 patients during the intervention period.
  • No reduction in ICU length of stay (8.1 vs 7.1 days; adjusted % change 8.2%; 95% CI −5.4% to 23.8%).
  • No reduction in hospital mortality (41.6% vs 40.2%; OR 0.93; 95% CI 0.78 to 1.12).
  • Moderate fidelity (tele-visits on 68% of eligible patient-days; 74% of recommendations accepted) yet no measurable improvement in patient-centred outcomes.

Bibliography