Tele-Intensive Care Unit-associated Early Recognition of In-hospital Hemodynamic Events and Clinical Outcomes: A Multicenter Observational Study

Abstract

Background and aims

In-hospital cardiac arrest (IHCA) remains a major cause of mortality among hospitalized patients globally. This prospective, multicenter, observational study assessed the associations of a tele-intensive care unit (Tele-ICU) hemodynamic surveillance program and clinical outcomes in adult inpatients monitored across three tertiary-care hospitals in India.

Patients and methods

From September 2024 to August 2025, 5,253 adult inpatients in ICUs, general wards, and Emergency Departments (EDs) were continuously monitored by a centralized Tele-ICU hub. Pre-specified thresholds for hemodynamic parameters generated real-time alerts, which were verified and adjudicated by Tele-ICU intensivists according to a standardized protocol. Outcomes, interventions, and mortality were analyzed using full multivariable logistic regression, with clearly defined denominators and clustering by site to account for within-site addressed.

Results

A total of 2,278 patients (43.3%) experienced clinically significant hemodynamic alerts. The system's alerting and verification protocol demonstrated a sensitivity of 79.2% and specificity of 80.1%, with consistent performance across sites and locations. Median acute physiology and chronic health evaluation (APACHE) II score was 16 [interquartile range (IQR) = 12–22]. Patients experiencing verified alerts had a mortality rate of 12% (n = 273), which is lower than the APACHE-predicted rate of 24.5% (risk-adjusted mortality ratio = 0.69; calibration plot provided). Multivariable logistic regression, including age, sex, APACHE II, diagnosis group, comorbidities, admission location, and time to intervention, showed that early intervention ≤15 minutes after alert was associated with lower odds of in-hospital mortality [adjusted odds ratio (aOR) 0.65, 95% CI: 0.52–0.81, p < 0.001; C-statistic 0.83].

Conclusion and clinical significance

The early recognition and verification of hemodynamic alerts in Tele-ICU were associated with improved clinical outcomes and a lower risk-adjusted mortality.

How to cite this article

Dharanindra M, Potineni RB, Rayana S, Thommandru S, Shaik J, Jampala K, et al. Tele-Intensive Care Unit-associated Early Recognition of In-hospital Hemodynamic Events and Clinical Outcomes: A Multicenter Observational Study. Indian J Crit Care Med 2026;30(1):40–49.

Highlights

Tele-intensive care unit (Tele-ICU) was associated with prompt detection of hemodynamic events and immediate management, which may be associated with lower rates of in-hospital cardiac events and mortality.

Introduction

Globally, In-hospital cardiac arrest (IHCA) remains a cause for a high mortality rate among inpatients. Contemporary guidance from the American Heart Association (AHA) and European Resuscitation Council (ERC) recommends an integrated IHCA chain of survival, which begins with prevention and early recognition, followed by early cardiopulmonary resuscitation (CPR), rapid defibrillation, advanced resuscitation, post-cardiac arrest care, and recovery/rehabilitation.^1–3 Despite advances in resuscitation science, IHCA survival to hospital discharge remains suboptimal at approximately 22–25% globally, underscoring the critical importance of the prevention link.^4,5

The pathophysiology of IHCA often involves a predictable deterioration pattern characterized by hemodynamic instability, respiratory compromise, and altered mental status preceding the arrest event.⁶ Studies have demonstrated that up to 80% of IHCAs are preceded by abnormal vital signs in the preceding 6–8 hours, presenting opportunities for early intervention.⁷ This recognition prompted widespread implementation of rapid response systems (RRSs) and early warning scores, which have shown variable success in reducing IHCA incidence and improving outcomes.⁸

The Tele-ICU system continuously monitors patients and alerts staff to hemodynamic events. Early identification of these events facilitates prompt therapy and may prevent complications from tissue hypoperfusion, helping to achieve patient-centric outcomes such as decreased ICU/hospital stays, shorter mechanical ventilation times, and a reduction in other acute illness complications.⁹

Tele-ICUs—remote telemedicine systems that combine continuous audio-visual connections, centralized monitoring, and decision-support algorithms—have become essential to improving critical care delivery in areas where intensivist availability is limited.^10,11

In-hospital cardiac arrest incidence is ~1–1.6 per 1,000 admissions in the UK (4) and up to 9–10 per 1,000 in US data.^12,13 Large registries and national audits report IHCA incidences ranging from ~1 to 10 per 1,000 admissions, with temporal improvements in risk-adjusted survival over the past two decades (from ~14 to >22% in the US registries), likely reflecting improvements across all chain links (early recognition, CPR quality, defibrillation, post-arrest care) rather than any single intervention such as Tele-ICU.^4,12,13 However, most IHCAs still present with non-shockable rhythms and have poor outcomes, underscoring the importance of preventing progression from deterioration to arrest.^5,14

Findings from multicenter trials, observational studies, systematic reviews, and complementary randomized studies show a consistent association between Tele-ICU implementation and lower ICU mortality (risk ratio ~0.80) and shorter ICU length of stay (LOS) (about 0.6 days) and demonstrate the feasibility of large-scale Tele-ICU implementation across multiple ICUs, supporting its integration into daily critical care practice.¹⁵

During COVID-19, Tele-ICU implementation was associated with best practices implementation and reduced inter-hospital transfers (~70%). This approach was also associated with enhanced local personnel capabilities and improved life-saving measures in under-resourced environments.¹⁶

In the Indian healthcare context, where the intensivist-to-population ratio remains critically low (1:1,00,000 compared to optimal ratios of 1:30,000), Tele-ICU implementation represents a scalable solution for extending specialized critical care expertise. However, limited data exist on the effectiveness of Tele-ICU systems for IHCA prevention in Indian healthcare settings.¹⁷

Primary research question: Is early intervention (≤15 minutes) following Tele-ICU hemodynamic alerts associated with reduced in-hospital mortality and lower IHCA incidence?

This study assesses the associations between a Tele-ICU-enabled hemodynamic surveillance program and clinical outcomes in adult inpatients monitored across three tertiary-care hospitals in India.

Patients and Methods

Study Design and Study Site

This was a prospective, multicenter, observational study conducted at a centralized Tele-ICU hub located in Aster Ramesh Hospital, Vijayawada, connected to three tertiary care hospitals in Andhra Pradesh, India: Aster Ramesh Hospitals, Vijayawada (200 beds), Guntur, and Ongole (400 beds). The Tele-ICU system was first implemented across the participating hospitals in 2020 and has been in continuous use since then. Ethical committee clearance has been taken at the centralized ethical committee of Aster Ramesh Group of hospitals (Cluster ethical committee) for conducting the study at all three centers (Approval ID: IECRH19072025_1) on July 19, 2024. The study was conducted in accordance with the Declaration of Helsinki and the guidelines of Good Clinical Practice. All patient-related images or photographs used in this study are fully de-identified and included in accordance with the journal's policy on patient privacy and confidentiality.

Study Objectives

Primary Objective

To quantify associations between Tele-ICU-enabled alert verification and hospital mortality [adjusted for acute physiology and chronic health evaluation (APACHE) II and relevant covariates].

Secondary Objectives

• To evaluate the diagnostic accuracy (sensitivity, specificity) and timeliness of intervention for Tele-ICU alerts.

• To compare pre- and post-site-level IHCA rates and estimate cardiac arrests potentially prevented.

Primary Outcome

In-hospital mortality (mortality during index hospitalization).

Secondary Outcomes

• Diagnostic performance of the alert system (sensitivity/specificity).

• Time to intervention.

• IHCA incidence rates.

• LOS and resource utilization.

Participating Centers

The three participating hospitals included two National Accreditation Board for Hospitals and Healthcare (NABH) Providers accredited (Vijayawada and Ongole) and one Joint Commission International (JCI) accredited (Guntur) multiple-specialty corporate hospitals with National Board Accreditation (NBE) for postgraduate and super specialty courses. Each hospital manages a mixed case load, including cardiology, pulmonology, neurology, nephrology, cardiothoracic surgery, neurosurgery, and general surgery. Intensive care unit nurse-to-patient ratios are maintained at 1:1 in the ICUs and Emergency Department (ED), and at 1:7 in wards. An intensivist is available round the clock, supported by continuous Tele-ICU monitoring.

Study Criteria and Recruitment Flow

From September 2024 to August 2025, a total of 5,667 adult subjects were admitted across the three participating hospitals and screened for eligibility. Of these, 414 patients were excluded for the following reasons: Those required monitoring <2 hours [Discharge against medical advice (DAMA) from ER, n = 182], do not attempt resuscitation (DNAR)/comfort care orders at admission (n = 129), patients transferred/discharged within 2 hours of admission (n = 57), pre-existing permanent cardiac pacemakers/implantable cardioverter-defibrillator (ICD) (n = 34), and pregnancy (n = 12). After these pre-specified exclusions, 5,253 adult inpatients were enrolled and continuously monitored via the Tele-ICU hub for ≥2 hours. All included patients met all eligibility criteria before monitoring began. There were no patients who lost to follow-up (all outcomes available). Figures 1 and 2 depict the entire recruitment and analysis process.

Figs 1A to C.

(A) Tele-ICU hub with central monitoring station. The intensivist monitoring has access to hemodynamic parameters. Audiovisual feed and EMR of the patient; (B) The monitors in the Tele-ICU have various settings to prevent visual fatigue and strain. Magnification of rhythm, reduction of rhythm; (C) Point-of-care alert system at the nursing station. Hemodynamic alerts are communicated to bedside nurses using a two-way speaker placed in the clinical area. EMR, electronic medical record; Tele-ICU, tele-intensive care unit

Fig. 2.

STROBE-compliant flow diagram showing total admissions screened (n = 5,667), exclusions (n = 414) with reasons, and the final monitored cohort (n = 5,253) included for Tele-ICU alert performance and outcomes analysis. DNAR, do not attempt resuscitation; ICD, implantable cardioverter-defibrillator; STROBE, strengthening the reporting of observational studies in epidemiology; Tele-ICU, tele-intensive care unit

Recruitment and Bias Minimization Statement

All admissions to ICUs, EDs, and monitored wards during the study window were systematically screened according to protocol. Exclusions were prospectively applied and fully documented, with reasons for exclusion tabulated. Although strict eligibility and universal monitoring reduce the risk of selection bias, potential residual bias may persist since non-monitored short-stay or DNAR patients could differ systematically in unmeasured ways from the included cohort. No participant was lost to follow-up, as electronic health records provided complete outcome ascertainment.

Selection Bias Considerations

The clear documentation of all screened, excluded, and monitored patients enables robust transparency of selection bias. Most exclusions were due to short monitoring duration (<2 hours) or palliative intent (DNAR). Recruitment and exclusion rates were similar across the three centers, as the patient populations and cluster medical services were similar across the three centers, reducing the risk of site-specific selection bias.

Sample Size and Power Considerations

This pragmatic, multicenter observational study enrolled all eligible adult inpatients (n = 5,253) continuously monitored across three tertiary care hospitals from September 2024 to August 2025. No formal priori sample size calculation was performed due to the real-world, universal monitoring design. Post-hoc power analysis, based on observed event rates (12% in-hospital mortality among 2,278 patients with verified hemodynamic alerts), site clustering, and multivariable adjustment, demonstrated approximately 80% power (α = 0.05, two-sided) to detect an odds ratio of 0.75 or smaller for the primary association between early intervention (≤15 minutes post-alert) and reduced in-hospital mortality.

Tele-ICU Infrastructure and Staffing

The Tele-ICU hub included:

• High-definition audiovisual communication systems.

• Real-time access to electronic medical records across all sites.

• An integrated multiparameter central monitoring system.

• Automated and human-verified real-time alerts.

• Secure data transmission protocols complying with patient privacy regulations.

Tele-ICU Staff Qualifications

The Tele-ICU staff included qualified intensivists and critical care nurses who worked around the clock.

• Intensivists: MD/DNB with DM/FNB critical care medicine.

• Nurses: Minimum of 3 years ICU experience with critical care certification.

• Technical support: 24/7 IT support for system maintenance.

Monitoring Protocol

Patients were continuously monitored through the centralized Tele-ICU hub. Physiological parameters, including heart rate (HR), blood pressure (BP), oxygen saturation (SpO₂), electrocardiogram (ECG), and respiratory rate (RR), were integrated into the Tele-ICU platform with automated alerts and audiovisual communication to bedside teams (Fig. 1).

Alert Thresholds (Pre-specified) and Standardization

Hemodynamic Alerts

• Tachycardia: HR >120 bpm (ICU/ED), >100 bpm (ward) for >60 seconds.

• Bradycardia: HR <50 bpm (all locations) for > 60 seconds.

• Hypotension: Mean arterial pressure (MAP) <65 mm Hg or systolic blood pressure (SBP) <90 mm Hg for >5 minutes.

• Hypertension: SBP >180 mm Hg or diastolic BP >110 mm Hg for >10 minutes.

Respiratory Alerts

• Hypoxemia: SpO₂ <90% for >60 seconds or abrupt decline by >5%

• Tachypnea: RR >24/min for ≥5 minutes.

• Bradypnea: RR <8/min for ≥2 minutes.

Temperature and Neurological Alerts

• Fever: Core temperature >38.5°C.

• Hypothermia: Core temperature <35.0°C.

• Altered mental status: Glasgow Coma Scale decrease ≥2 points.

The Tele-ICU system continuously monitored hemodynamic, respiratory, and neurological parameters. However, the present study analysis focuses specifically on hemodynamic events and oxygen desaturation.

Thresholds differed slightly by location due to expected baseline physiologic differences between ICU and ward patients (e.g., lower tachycardia threshold on wards to increase early detection). Thresholds were standardized across all sites before the study.

Hemodynamic Event Detection and Management

The Tele-ICU system automatically gives alerts when any physiological parameter deviates from its normal range. A critical care physician at the Tele-ICU hub immediately informed the bedside medical staff (physician and nurse). The management was initiated according to standard protocols, including fluid resuscitation (250–500 mL crystalloid bolus) or administration of vasopressor agents for hypotension, antihypertensive therapy, and neurological assessment for hypertension. For patients with tachycardia, including ventricular tachycardia (VT) and supraventricular tachycardia, an ECG was performed, and amiodarone was administered, along with other supportive measures. In a few patients, bradycardia spontaneously reverted, and in others with persistent bradycardia, they were treated with atropine and adrenaline. Inotropes were used for heart failure. Rapid response system or code blue activation was done when indicated (Fig. 1).

Data Collection

All Tele-ICU alerts, associated interventions, and patient outcomes were recorded in the Tele-ICU database. Data collected included patient demographics (age, gender), admission diagnosis, comorbidities, ICU and hospital LOS, mortality, APACHE II and APACHE IV scores (calculated within 24 hours), and details of hemodynamic alerts generated by the Tele-ICU, along with their management.

Verification/Adjudication Protocol for Alerts

All automated alerts were reviewed in real time by a Tele-ICU intensivist, who confirmed clinical significance using pre-specified threshold criteria and bedside information obtained through audiovisual monitoring and electronic medical record (EMR) review. A 10% random sample underwent secondary review within 24 hours by another intensivist. Disagreements were resolved by the head of the department, critical care medicine, if needed. For patients with multiple alerts of the same type (e.g., recurrent bradycardia), only the first verified event per patient was included in the analysis to avoid duplication.

Process Measures

• Frequency and type of first verified alert per patient by location.

• Time from alert generation to bedside intervention.

• False positive and false negative rates.

• Escalation of care requirements.

Clinical Outcomes

• ICU and hospital LOS.

• Requirement for mechanical ventilation and/or vasopressor support.

• Unplanned ICU transfers.

• In-hospital mortality.

• Estimated cardiac arrests prevented.

Statistical Analysis

Data were collected by using a pre-designed collection form. A thorough error check was done. Statistical analysis was performed using IBM SPSS Statistics version 29.0 and R version 4.3.0. Continuous variables were summarized as medians with interquartile ranges (IQRs), and categorical variables were presented as frequencies and percentages. Missing data were minimal (<5%) and were handled using complete case analysis as summarized in Supplementary Table S3. Associations between categorical variables were tested using the Chi-square test, and non-parametric continuous data were compared using the Mann–Whitney U test. Alert rates were expressed per 1,000 patient days.

Multivariable Modeling Strategy

A multivariable logistic regression model for in-hospital mortality forced inclusion of all clinically relevant covariates without automated selection: Age, sex, APACHE II score, primary diagnosis group (cardiovascular, respiratory, others), comorbidities (hypertension, diabetes mellitus, coronary artery disease, chronic kidney disease), admission location (ICU/ED/ward), and time to intervention (≤15 vs >15 minutes). APACHE IV was excluded due to high correlation with APACHE II (r = 0.87).

Missing Data in Regression Models

Complete case analysis was used (no multiple imputations). In the total monitored cohort (n = 5,253), missing data were minimal (<5%) across all variables. For the primary multivariable model: ~200 (3.8%) missing APACHE II/IV scores, <100 (1.9%) missing comorbidity data, and <50 (0.9%) intervention time (Supplementary Table S3). Age, sex, and outcome data had 0% missingness across all 5,253 monitored patients (Supplementary Table S3).

Collinearity Assessment

Variance inflation factors (VIFs) were calculated; the maximum VIF was 2.3 (all VIFs < 5), confirming no multicollinearity concerns. Interaction terms: No interaction terms were pre-specified or tested, as the primary hypothesis examined the main effect of early intervention. Functional form assumptions for continuous predictors were assessed using restricted cubic splines, confirming approximate linearity for age and APACHE II score. Robust standard errors clustered by the three hospital sites accounted for within-site correlation. Model performance was evaluated using the C-statistic (0.834, 95% CI: 0.816–0.852) and Hosmer–Lemeshow test (p = 0.187). A sensitivity analysis excluded patients with DNAR orders. Diagnostic accuracy: Alerting algorithm performance used 2 × 2 verification tables with sensitivity [TP/(TP + FN)] = 79.2%, specificity [TN/(TN + FP)] = 80.1%, PPV [TP/(TP + FP)] = 68.7%, NPV = 87.1%, and false alarm rate = 31.3% (Supplementary Table S2). Cost-effectiveness analysis included implementation costs (equipment, staffing, training), operational costs (annual staffing, maintenance), and cost savings from prevented complications/reduced transfers, expressed as cost per QALY gained. Statistical significance was set at p-value < 0.05. This study adhered to STROBE guidelines (checklist provided as Supplementary Material).

Results

Patient Characteristics of the Total Monitored Cohort (n = 5,253)

A total of 5,253 patients were monitored during the 12-month study period. Of these, 64% (n = 3,362) were males and 36% (n = 1,891) were females. The median age was 58 years (IQR = 48–72). Hemodynamic events were detected in 2,278 (43.3% of 5,253) patients with verified alerts. Of these, 456 (20% of 2,278) required intervention following Tele-ICU alerts. The present study specifically evaluated patients who experienced hemodynamic events and oxygen desaturation.

Characteristics of Patients Who Experienced Hemodynamic Alerts (n = 2,278)

Among the 2,278 (43.3% of 5,253) patients with hemodynamic alerts, 60% were male and 40% were female. The median age in this subgroup was 64 years (IQR = 52–73). The most common admission diagnosis was cardiovascular diseases (30.0%). Significant comorbidities included hypertension (31.6%), diabetes mellitus (28.9%), and coronary artery disease (19.7%). The median APACHE II score was 16 (IQR = 12–22) and the APACHE IV score was 58 (IQR = 42–78), indicating moderate to high severity of illness (Table 1). Sites contributed proportional cases, with no significant recruitment variation.

Table 1.

Baseline characteristics of patients who experienced hemodynamic alerts (n = 2,278), stratified by intervention status

Characteristics	Total patients with alerts (n = 2,278)	Alert group with intervention (n = 456)	Alert group without intervention (n = 1,822)
Age (years), median (IQR)	64 (52–73)	67 (55–76)	63 (51–72)
Male, n (%)	1,366 (60.0)	274 (60)	1,092 (60)
Female, n (%)	912 (40.0)	182 (40)	730 (40)
Admission diagnosis, n (%)
Cardiovascular	683 (30.0)	137 (30.0)	546 (30.0)
Respiratory	501 (22.0)	100 (21.9)	401 (22.0)
Neurological	319 (14.0)	64 (14.0)	255 (14.0)
Gastrointestinal	273 (12.0)	55 (12.1)	218 (12.0)
Infectious diseases	228 (10.0)	46 (10.1)	182 (10.0)
Others	274 (12.0)	54 (11.9)	220 (12.1)
Comorbidities, n (%)
Diabetes mellitus	658 (28.9)	86 (18.9)	506 (27.7)
Hypertension	719 (31.6)	105 (23)	670 (37)
Coronary artery disease	450 (19.7)	114 (25)	306 (16.7)
Chronic kidney disease	287 (12.6)	78 (17.1)	159 (8.7)
COPD/asthma	96 (4.2)	59 (13)	137 (7.5)
Previous cardiac arrest	68 (3.0)	14 (3)	44 (2.4)
APACHE II score, median (IQR)	16 (12–22)	18 (14–25)	15 (11–21)
APACHE IV score, median (IQR)	58 (42–78)	68 (52–89)	55 (39–75)
LOS (days), median (IQR)
ICU	5 (3–8)	6 (4–9)	5 (3–7)
Hospital	8 (5–12)	9 (6–14)	8 (5–11)
Mortality, n (%)
ICU	202 (8.9)	68 (14.9)	134 (7.0)
ED	71 (3.1)	19 (4.2)	52 (2.8)
General ward	0 (0.0)	0 (0.0)	0 (0.0)
Overall hospital mortality	273 (12.0)	91 (20.0)	182 (10.0)

Percentages calculated using n = 2,278 as the denominator; APACHE, acute physiology and chronic health evaluation; COPD, chronic obstructive pulmonary disease; ED, emergency department; ICU, intensive care unit; IQR, interquartile range; LOS, length of stay

Length of Stay

Among 2,278 patients with alerts, ICU LOS median was 5 days (IQR = 3–8) and hospital LOS median was 8 days (IQR = 5–12). Among 456 intervened, ICU LOS was 6 days (IQR = 4–9) and hospital LOS was 9 days (IQR = 6–14) (Table 1).

Hemodynamic Event Detection and Distribution

Hemodynamic events were detected in 2,278 (43.3% of the monitored cohort) patients. The most frequent event was tachycardia (39%) (Table 2). Among 2,278 patients, a significant majority (83%, n = 1,891) of subjects experienced at least one hemodynamic event during their hospital stay, 13% (n = 296) experienced two hemodynamic events, and 4.0% (n = 91) experienced more than two events. Most hemodynamic events were observed in ICUs (n = 1,551, 68%), followed by ED (n = 709, 31%), and a small proportion in wards (n = 18, 0.8%) (Table 2). The higher event rate in ICUs reflects the higher acuity of patients. The location-wise distribution of verified hemodynamic alerts, associated interventions, and mortality is provided in Supplementary Table S1.

Table 2.

Distribution of hemodynamic events by hospital location among monitored patients with verified alerts (n = 2,278)

Alert type	ICU, n (%)	Ward, n (%)	Emergency department, n (%)	Total, n (%)
Tachycardia (HR >120 bpm)	604 (39)	7 (39)	278 (39)	889 (39)
Bradycardia (HR <50 bpm)	403 (26)	5 (28)	185 (26)	593 (26)
Hypotension (SBP <90 mm Hg)	186 (12)	3 (17)	85 (12)	274(12)
Hypertension (SBP >180 mm Hg)	16 (1)	1 (5)	7 (1)	24 (1)
Hypoxia (SpO2 <88%)	342 (22)	2 (11)	154 (22)	498 (22)
Total events	1,551 (68)	18 (0.8)	709 (31.2)	2,278 (100)

Percentages represent the proportion of patients with verified alerts in that location (denominator = 2,278); HR, heart rate; ICU, intensive care unit; SBP, systolic blood pressure; SpO₂, peripheral capillary oxygen saturation

Tele-ICU System Performance

The Tele-ICU alert system demonstrated strong diagnostic performance characteristics. The system achieved a sensitivity of 79.2% (95% CI: 77.1–81.2%) and specificity of 80.1% (95% CI: 78.9–81.3%), indicating that most true hemodynamic events were correctly identified while minimizing false alarms. Based on these operating characteristics, the diagnostic performance confusion matrix included 1,804 true positives, 592 false positives, 474 false negatives, and 2,383 true negatives (Supplementary Table S2). The positive predictive value was 68.7% (95% CI: 66.4–71.0%), and the negative predictive value was 87.1% (95% CI: 85.8–88.3%) with a false alarm rate 31.3%. Performance remained consistent across hospital areas: ICUs (sensitivity 82.5%, specificity 76.8%), wards (sensitivity 75.3%, specificity 83.2%), and EDs (sensitivity 78.9%, specificity 79.4%). The median time from alert generation to bedside intervention was 12 minutes (IQR = 8–18 minutes), and 89.4% of interventions were implemented within 30 minutes of alert generation, demonstrating prompt clinical response to alerts (Table 3).

Table 3.

Clinical interventions, response times, and outcomes among patients with verified Tele-ICU alerts (n = 456)

Outcome/intervention	Alert group with intervention (n = 456)
Immediate interventions, n (%)	456 (100)
Medication adjustment	278 (61.0)
Fluid resuscitation	182 (40.0)
Oxygen therapy	164 (36.0)
Mechanical ventilation	41 (9.0)
Vasopressor support	68 (15.0)
Central line placement	32 (7.0)
Chest physiotherapy	59 (13.0)
Other interventions	91 (20.0)
Escalation of care, n (%)
Transfer to ICU	91 (20.0)
Transfer to a higher acuity unit	46 (10.0)
Increased monitoring	228 (50.0)
Rapid response team activation	32 (7.0)
Code blue activation	14 (3.1)
No escalation required	45 (9.8)
Time metrics, median (IQR)
Alert to intervention (minutes)	12 (8–18)
Alert to physician assessment (minutes)	15 (10–22)
Patient outcomes within 24 hours, n (%)
Clinical improvement	319 (70.0)
Remained stable	91 (20.0)
Clinical deterioration	32 (7.0)
Cardiac arrest prevented	14 (3.1)
Unplanned ICU transfer	23 (5.0)

Percentages calculated using n = 456 as the denominator; ICU, intensive care unit; IQR, interquartile range; Tele-ICU, tele-intensive care unit

Interventions and Escalation of Care

Following Tele-ICU alerts, immediate clinical interventions were implemented in 456 (20% of 2,278) of the patients with verified events. The most common interventions were medication adjustments in 278 patients (61.0% of 456), fluid resuscitation in 182 patients (40.0% of 456), oxygen therapy modifications in 164 patients (36.0% of 456), and vasopressor support in 68 patients (15.0% of 456). Mechanical ventilation was required in 41 patients (9.0% of 456). Many patients required multiple or a combination of interventions (Table 3).

Escalation of care occurred in 411 (90.0% of 456) patients, including 91 patients (20.0% of 456) transferred to the ICU, 46 (10.0% of 456) to a higher acuity unit, 228 patients (50.0% of 456) required increased monitoring, 32 patients (7.0% of 456) triggered rapid response team activation, and 14 patients (3.0% of 456) required code blue activation (Table 3). These findings demonstrated that Tele-ICU alerts were frequently followed by timely bedside interventions and escalation to appropriate care levels when needed.

Hemodynamic Events Associated with Mortality

Mortality occurred in 273 (12% of 2,278) patients who experienced hemodynamic alerts (Table 1). Patients experiencing tachycardia had a 1.8-fold increased risk of mortality (χ² = 33.2, p < 0.001), while bradycardia was associated with a 3.2-fold higher risk (χ² = 153.7, p < 0.001). Hypoxia (9%) and VT (6%) were also associated with increased mortality, though with a lower effect size compared with bradycardia (12%) and tachycardia (21%) (Supplementary Fig. S1).

APACHE and Risk-adjusted Outcomes

Acute physiology and chronic health evaluation II and IV scores were calculated for ICU admissions and eligible patients in the ED/wards monitored for more than 2 hours (n = 5,055; missing <5% by site). Based on the APACHE II predicted number (671 deaths), the RAMR ratio was 0.69 (95% CI: 0.62–0.77), indicating lower mortality with the Tele-ICU monitored cohort than predicted.

Multivariable Logistic Regression Analysis

The multivariable logistic regression model for in-hospital mortality included age, sex, APACHE II score, comorbidities, admission location, and time-to-intervention. Functional form assumptions were assessed using restricted cubic splines, which confirmed linearity for age and APACHE II score. Robust standard errors clustered by hospital site were applied to account for within-site correlation.

After adjustment, APACHE II score ≥20 (OR 3.24, 95% CI: 2.67–3.94, p < 0.001), age >75 years (OR 1.89, 95% CI: 1.52–2.35, p < 0.001), chronic kidney disease (OR 1.67, 95% CI: 1.29–2.16, p < 0.001), and previous cardiac arrest (OR 2.45, 95% CI: 1.58–3.78, p < 0.001) were independently associated with increased mortality. Early intervention within 15 minutes of alert generation was independently associated with lower odds of in-hospital mortality (OR 0.65, 95% CI: 0.52–0.81, p < 0.001), suggesting a protective association between timely Tele-ICU-facilitated intervention and outcomes. The model demonstrated excellent discrimination (C-statistic 0.834, 95% CI: 0.816–0.852) and good calibration using the Hosmer–Lemeshow goodness-of-fit test (p = 0.187) (Table 4; Supplementary Fig. S2).

Table 4.

Multivariable logistic regression for in-hospital mortality (n = 2,278)

Variables	Adjusted OR (95% CI)	p-value
Age > 75 years	1.89 (1.52–2.35)	<0.001
APACHE II score >20	3.24 (2.67–3.94)	<0.001
Chronic kidney disease	1.67 (1.29–2.16)	<0.001
Previous cardiac arrest	2.45 (1.58–3.78)	<0.001
Hypertension	1.05 (0.87–1.28)	0.61
Diabetes mellitus	1.16 (0.94–1.42)	0.18
Coronary artery disease	1.23 (0.96–1.58)	0.09
Location: ICU (ref = ward)	1.34 (1.02–1.76)	0.04
Location: ED (ref = ward)	0.92 (0.66–1.28)	0.62
Early intervention ≤15 min	0.65 (0.52–0.81)	<0.001

Multivariable logistic regression using clustered standard errors by site; continuous variables checked for linearity; missing data were minimal (<5%); APACHE, acute physiology and chronic health evaluation; CI, confidence interval; ED, emergency department; ICU, intensive care unit; OR, odds ratio

IHCAs and Estimated Prevention

Baseline IHCA rates showed similar trends across sites (range 7.4–9.1 per 1,000 admissions), with no obvious differences in case mix or secular trends that would influence interpretation. Pre-Tele-ICU: Site 1 (8.2), site 2 (9.1), and site 3 (7.4) per 1,000 admissions. Post-Tele-ICU: Site 1 (4.3), site 2 (4.7), and site 3 (3.8) per 1,000 admissions, representing proportional reductions of 48, 48, and 49%, respectively (Supplementary Table S4).

A total of 350 IHCAs occurred across the participating hospitals during the study period. Of these, 250 (71% of 350) occurred in ICUs, 90 (26% of 350) in ED, and 10 (3% of 350) in the general ward. Based on historical data and risk stratification, an estimated 38 cardiac arrests may have been prevented during the study period. Of these, 18 occurred in ICUs, 12 in wards, and 8 in ED. This represents a 47% lower overall cardiac arrest incidence compared to pre-Tele-ICU implementation data (8.2 vs 4.3 per 1,000 admissions, p < 0.001), which coincided with early detection and intervention facilitated by the Tele-ICU system. A pooled Poisson regression analysis indicates a statistically significant reduction in IHCA rates post-Tele-ICU implementation. An interrupted time series (ITS) analysis illustrating these trends is included in the Supplementary Figure S3 and Supplementary Table S4.

The expected IHCA burden was calculated by applying the pre-Tele-ICU incidence rate (8.2 per 1,000 admissions) to the monitored cohort (5,253). Observed IHCA incidence decreased to 4.3 per 1,000 admissions. Prevented cardiac arrests = Expected – observed (38 cardiac arrests), adjusted for APACHE II distribution.

Cost-effectiveness Analysis

Economic analysis demonstrated that the Tele-ICU program was cost-effective relative to accepted international thresholds. The initial setup cost was approximately 20 lakhs per 100 beds, with annual operational expenses (including staffing and maintenance) of 5 lakhs per 100 beds, and training and certification costs of 1 lakh per 100 beds. Based on the 5,253 patients monitored during the study period, the estimated cost per patient was ~1,000 rupees for 48 hours of monitoring. The program was associated with estimated cost savings through lower complication rates and avoided transfers.

Discussion

In this multicenter prospective observational study, more than 80% of patients experienced at least one hemodynamic event, and early recognition through Tele-ICU was linked to lower odds of mortality. The overall in-hospital mortality among patients with hemodynamic alerts (12%) was lower than rates reported by Sandroni et al.'s study (15–20%) and by Bowman et al. (25%).^6,18 Although direct comparisons are limited by differences in populations and inclusion criteria. These findings suggest that timely identification via Tele-ICU surveillance coincided with favorable clinical outcomes. In our study, 9% of patients required mechanical ventilation following hemodynamic deterioration. Although this indicates the Tele-ICU's capacity to identify patients at risk for respiratory compromise, the duration of mechanical ventilation was not recorded. Nonetheless, previous research studies have shown that associations exist between Tele-ICU-guided interventions, shorter ventilator days, and fewer complications.^19–21

Our risk-adjusted mortality ratio of (0.69, 95% CI: 0.62–0.77) indicates a favorable association between Tele-ICU monitoring and patient outcomes compared to APACHE II-predicted mortality. This finding aligns with meta-analyses showing that Tele-ICU implementation has been associated with 10–30% lower odds of ICU mortality.^9,20,21 In the present study, early intervention within 15 minutes of an alert was independently associated with lower odds of mortality (OR 0.65, 95% CI: 0.52–0.81), supporting a protective association between timely Tele-ICU-facilitated intervention and favorable outcomes.

In Western cohorts, Bergum et al. reported that the predominant causes of IHCA were pulseless electrical activity, asystole, and ventricular fibrillation (VF)/VT.²² Sandroni et al. identified VF/VT as a major cause of IHCA.⁶ In the present study, IHCA was primarily associated with tachycardia and bradycardia. This difference may reflect the focus of the Tele-ICU on early detection of hemodynamic instability rather than terminal arrhythmic events. Our findings support the concept that IHCA is often preceded by progressive hemodynamic deterioration rather than sudden arrhythmic death.²³

Tele-ICU monitoring coincided with early detection of pre-arrest conditions, alerting the bedside team to provide appropriate care before deterioration progresses to cardiac arrest. This may align with the first link in the AHA and ERC IHCA chain of survival. In resource-limited settings, where there is limited intensivist availability, Tele-ICU programs were associated with standardized care practices and reduced unnecessary interhospital transfers.

The present study provides important data on Tele-ICU implementation in a middle-income country healthcare setting. India faces significant challenges in critical care delivery, with an intensivist-to-population ratio of approximately 1:1,00,000 compared to optimal ratios of 1:30,000 in developed countries.¹⁷ The centralized hub model used in this study demonstrated scalability across multiple hospital sites with varying resource levels.

Successful Tele-ICU implementation requires structured training and quality assurance. In our program, Tele-ICU physicians completed 40 hours of structured training, while nurses received 20 hours, focusing on telemedicine communication skills, technology operation, and remote clinical assessment techniques. Ongoing quality assurance included monthly case reviews and performance feedback. Initial implementation of Tele-ICU required significant infrastructural investment, including a high-definition audiovisual system, integrated monitoring interfaces, and reliable data connectivity, which were standardized across all three participating hospitals.

The economic cost of Tele-ICU implementation was modest, and these costs were offset by savings from fewer complications and transfers. The cost-effectiveness was even better in comparison to Franzini et al. and Becker et al. studies, which were done in North America using proprietary Tele-ICU systems.^21,24,25 This suggests that Tele-ICU may be both clinically beneficial and economically sustainable, with clear benefits to both patients and hospitals through lower hospitalization costs and optimized resource utilization.

In the current study, a total of 350 IHCAs was recorded across all sites, with a 47% reduction in incidence compared to the pre-Tele-ICU period (8.2 vs 4.3 per 1,000 admissions, p < 0.001). This finding compares favorably with other hospital-based preventive interventions and underscores the clinical relevance of early recognition through the Tele-ICU system.

Global evidence supports our findings, with multicenter studies and registry data from the United States, Australia, and Asia demonstrating that Tele-ICU and RRSs are linked to lower ICU mortality, better survival rates, and enhanced adherence to evidence-based practices.^{4,7,8,12,19–21,24,26,27}

To our knowledge, this is the first multicenter, prospective observational study in India studying the feasibility of large-scale Tele-ICU-assisted monitoring of in-hospital hemodynamic events and its association with early intervention and IHCA prevention.

This study was conducted in three relatively well-resourced, accredited tertiary-care hospitals with established Tele-ICU infrastructure, experienced multidisciplinary teams, and consistent nurse-to-patient ratios (1:1 in ICU/ED, 1:7 in wards). Generalizability to lower-resource rural facilities, emerging Tele-ICU systems, or settings with severely constrained workforce and infrastructure may therefore be limited. However, the hub-and-spoke model and heterogeneity of participating hospitals (bed capacity 200–400, multiple specialties) suggest that similar Tele-ICU configurations may be feasible in comparable middle-income healthcare environments in South and Southeast Asia. Adaptation to more resource-limited settings may require modified alert thresholds, tailored training, or different staffing models.

Limitations

The present study has some limitations. First, as an observational design, causal inference is limited; observed temporal associations between Tele-ICU monitoring and improved outcomes do not establish causation. Second, despite adjustment for major covariates (age, APACHE II score, comorbidities), residual confounding remains possible. Unmeasured factors such as nursing quality, hospital safety culture, staff experience, or evolving treatment protocols during the study period may have influenced outcomes and could bias effect estimates.

Third, confounding by indication is possible: Patients who appeared more acutely unwell may have received earlier interventions regardless of Tele-ICU prompts. Although APACHE II adjustment mitigates some severity-related confounding, residual confounding by indication cannot be excluded. Fourth, the study lacked a concurrent non-Tele-ICU control group; pre-/post-comparisons within the same institutions may conflate secular trends in critical care practice with Tele-ICU effects. Improvements over time in vasopressor use, protocol adherence, staff training, and other supportive care elements may contribute to observed mortality reductions independent of Tele-ICU monitoring.

Fifth, the study did not systematically capture inter-hospital transfers, detailed complication profiles, or long-term pre-tele-ICU mortality trends, limiting more granular temporal and safety analyses. Additionally, data on ventilation duration, readmissions, and long-term functional outcomes were not collected, constraining full evaluation of downstream effects.

Conclusion

Tele-ICU monitoring with standardized protocols coincided with favorable clinical outcomes, timely intervention, and a lower risk-adjusted mortality compared with APACHE II predictions in this multicenter cohort. The program also coincided with an observed reduction in IHCA incidence and suggests a potentially reproducible approach for IHCA prevention initiatives in low- and middle-income countries.

Clinical Significance

This study emphasizes the role of Tele-ICU systems in strengthening the early recognition and prevention components of the IHCA Chain of Survival, particularly in settings with limited on-site intensivist coverage. The findings support further evaluation of Tele-ICU models as part of broader strategies that may be associated with decreased preventable deterioration and potentially improved outcomes in resource-constrained health systems.

Author Contributions

Moturu Dharanindra: Conceptualization, Methodology, Writing—Review and Editing, Ramesh B Potineni: Supervision, Supriya Rayana: Project administration, Writing—Original Draft, Sravani Thommandru, Karthikeya Jampala: Writing—Original Draft, Jahangeer Shaik: Resources, Lakshmi SSB Kakumanu: Validation, Sai T Uppalapati: Validation, Silpa C Nallapaneni: Investigation Software, Vamsi K Madduri: Writing—Review and Editing, Karthik C Yalavarthi: Writing—Review and Editing.

IEC Approval: Approval ID: IECRH19072025_1.

Supplementary Materials

The supplementary tables and figures are available on the website www.ijccm.com

Orcid

Moturu Dharanindra https://orcid.org/0000-0002-9446-5460

Ramesh B Potineni https://orcid.org/0000-0001-6983-0828

Supriya Rayana https://orcid.org/0000-0001-9904-2574

Sravani Thommandru https://orcid.org/0009-0002-3510-9447

Jahangeer Shaik https://orcid.org/0009-0003-3345-8827

Karthikeya Jampala https://orcid.org/0009-0003-8519-259X

Lakshmi SSB Kakumanu https://orcid.org/0009-0005-9752-3207

Sai T Uppalapati https://orcid.org/0009-0004-0286-3384

Silpa C Nallapaneni https://orcid.org/0009-0009-8018-536X

Vamsi K Madduri https://orcid.org/0009-0005-4890-1785

Karthik C Yalavarthi https://orcid.org/0009-0007-6639-9252