Continuous Monitoring Alerts: Why Usability Blocks Adoption in Singapore Wards

Continuous Monitoring Alerts: Why Usability Blocks Adoption in Singapore Wards

Singapore hospitals are piloting wearable sensors and continuous monitoring devices for deterioration alerting in general wards—National University Hospital recently announced smartwatch-based vital signs monitoring for inpatients [9]. But recent research reveals a deployment gap: usability and workflow integration, not predictive accuracy, determine whether nurses actually use these systems. A 2026 scoping review of continuous monitoring devices in non-ICU settings found that alert fatigue, device wearability, and integration with existing workflows are the primary barriers to adoption [1]. For hospital CIOs, clinical informatics teams, and AI deployment leads in Singapore, this shifts the design brief: deterioration alerting is a sociotechnical problem, not just a machine learning one.

Key takeaways

• Usability determines adoption: Recent evidence shows that alert design, device wearability, and workflow integration—not model performance—predict whether nurses use continuous monitoring systems in general wards [1][3].
• Alert fatigue is the dominant failure mode: High false-positive rates and non-actionable alerts drive alarm override behavior, undermining clinical utility even when models are well-calibrated [1][3].
• Wearable sensor adherence is fragile: Patient comfort, skin irritation, and charging logistics affect compliance; a 2022 surgical cohort study found that device adherence dropped when wearables required frequent recharging [5].
• Implementation requires co-design: Mixed-methods research in acute hospital settings found that nurse involvement in alert threshold tuning and escalation protocols improved adoption [3].
• Singapore deployment context matters: Nursing ratios, ward layouts, and EHR integration maturity vary across restructured hospitals; what works at one cluster may not transfer without workflow redesign.

Why continuous monitoring systems fail in general wards

Continuous monitoring devices—wearable sensors, wireless vital sign patches, bedside monitors with algorithmic alerting—promise early detection of clinical deterioration outside the ICU. The clinical logic is sound: intermittent manual vital signs (typically every 4–6 hours in Singapore general wards) miss the gradual physiological decline that precedes adverse events like sepsis, respiratory failure, or cardiac arrest.

But a 2026 scoping review of usability in non-critical care units identified three failure modes [1]:

1. Alert fatigue: High false-positive rates (often >80% in early deployments) lead to alarm override behavior. Nurses learn to ignore alerts, defeating the system's purpose.
2. Workflow disruption: Devices that require manual data entry, frequent charging, or separate alert dashboards add cognitive load without reducing existing documentation burdens.
3. Poor wearability: Skin irritation, bulky form factors, and patient discomfort reduce adherence, especially in elderly or cognitively impaired populations common in Singapore's aging inpatient cohorts.

A 2026 mixed-methods study in acute hospital non-ICU settings found that usability factors—device comfort, alert actionability, integration with nurse call systems—were stronger predictors of adoption than model sensitivity or specificity [3]. This aligns with our experience deploying clinical AI services in Singapore hospitals: technical validation is necessary but not sufficient; operational fit determines whether a system survives pilot phase.

What the evidence shows about wearable sensor outcomes

A 2022 pragmatic cohort study of surgical patients monitored with wearable sensors and digital alerting systems found mixed results [5]. The intervention group had earlier detection of deterioration events, but the study also documented:

• Device adherence challenges: Patients removed sensors due to discomfort or charging requirements, creating data gaps that undermined continuous monitoring.
• Alert response variability: Nurses' response times to alerts varied by shift, staffing levels, and competing clinical priorities—factors not captured in the model design.
• Integration friction: The alerting system ran on a separate dashboard from the EHR, requiring nurses to context-switch between systems.

The study used propensity-matched analysis to control for confounders, but the authors noted that real-world effectiveness depends on "sustained engagement from clinical staff and seamless integration into existing workflows" [5]—a reminder that predictive models are only one component of a functioning clinical system.

For Singapore hospitals, this has implications for procurement and pilot design. A protocol paper from 2021 outlined a real-world prospective study evaluating wearable sensors in secondary care, emphasizing the need to measure not just clinical outcomes (length of stay, ICU transfers) but also usability metrics: device adherence rates, alert response times, and nurse satisfaction [2]. We recommend Singapore hospitals adopt similar dual-outcome frameworks when piloting continuous monitoring systems.

Alert design and the false-positive problem

The dominant usability failure in deterioration alerting is alert fatigue. Early warning scores (EWS) like NEWS2, widely used in Singapore hospitals, were designed for intermittent manual vital signs. When applied to continuous sensor data, they generate frequent alerts—many non-actionable—because physiological parameters fluctuate naturally.

The 2026 scoping review identified three design strategies to reduce false positives [1]:

1. Temporal smoothing: Require sustained threshold breaches (e.g., heart rate >120 for 10 minutes) rather than instantaneous triggers.
2. Contextual suppression: Suppress alerts during known high-variability periods (post-ambulation, during physiotherapy) or for patients with chronic baseline abnormalities.
3. Escalation tiering: Use multi-level alerts (informational, advisory, urgent) with different notification channels, reserving audible alarms for high-acuity events.

A 2024 scoping review of early warning scores in psychiatric settings—where baseline vital signs often differ from general medical populations—found that context-specific thresholds improved specificity without sacrificing sensitivity [4]. For Singapore hospitals deploying continuous monitoring in mixed medical-surgical wards, this suggests that one-size-fits-all alert thresholds will underperform; ward-specific tuning is necessary.

This intersects with clinical AI safety monitoring: continuous monitoring systems require ongoing performance tracking, not just pre-deployment validation. Alert override rates, time-to-response distributions, and false-positive audits should be part of routine governance.

Co-design and nurse involvement in threshold tuning

The 2026 mixed-methods study found that nurse involvement in alert threshold tuning and escalation protocol design improved adoption [3]. Hospitals that used iterative co-design—piloting devices on a single ward, gathering nurse feedback, adjusting alert logic, then scaling—had higher sustained usage rates than those that deployed vendor-default settings hospital-wide.

Practical co-design steps for Singapore hospitals:

• Shadow shifts: Observe nurses responding to alerts in real time; identify workflow bottlenecks (e.g., alerts that require EHR lookup before action).
• Threshold workshops: Present alert distributions (true positive rate, false positive rate, time-to-response) to ward nurses; adjust thresholds collaboratively.
• Escalation protocol mapping: Document who responds to which alert types, at what urgency level, and integrate with existing nurse call or rapid response team workflows.
• Feedback loops: Provide nurses with alert outcome data ("20% of your high-acuity alerts led to clinical intervention") to build trust and refine intuition.

This is consistent with broader clinical decision support adoption barriers we've documented: systems that fit existing mental models and workflows are adopted; those that require new cognitive routines are abandoned.

Why this matters in Singapore

Singapore's healthcare system faces three pressures that make continuous monitoring attractive—but also difficult to deploy:

1. Aging population: Elderly patients with multimorbidity are at higher risk of deterioration but also more likely to have baseline vital sign abnormalities, increasing false positives.
2. Nursing shortages: Continuous monitoring is often framed as a force multiplier for stretched nursing teams, but poorly designed systems add workload rather than reducing it.
3. Restructured hospital heterogeneity: Singapore's public hospital clusters (SingHealth, NUHS, NHG) have different EHR platforms, ward layouts, and staffing models; a system validated at one cluster may not transfer without re-tuning.

National University Hospital's smartwatch pilot [9] is a test case: consumer wearables offer better form factors than medical-grade patches, but they also introduce data quality variability (motion artifacts, inconsistent wear) and integration challenges (proprietary APIs, cloud dependencies). For hospital CIOs evaluating similar pilots, the evidence suggests that usability testing and workflow integration should precede large-scale procurement.

This also intersects with health data infrastructure: continuous monitoring generates high-frequency time-series data that most Singapore hospital EHRs are not designed to store or visualize. Deployment requires not just a predictive model but also data pipelines, alert routing infrastructure, and clinician dashboards—a platform engineering problem, not just a data science one.

What to do next

If your Singapore hospital is piloting or procuring continuous monitoring systems:

• Prioritize usability in vendor evaluation: Request pilot data on alert override rates, device adherence, and nurse satisfaction—not just sensitivity/specificity. Vendors who cannot provide these metrics have not tested in real clinical environments.
• Start with a single ward co-design pilot: Deploy on one medical or surgical ward, involve nurses in threshold tuning, measure both clinical outcomes (ICU transfers, rapid response calls) and usability outcomes (alert response time, device adherence).
• Build alert governance into your AI safety framework: Continuous monitoring systems require ongoing performance monitoring; integrate alert audits, false-positive tracking, and escalation protocol reviews into your clinical AI safety monitoring processes.
• Plan for EHR integration early: Separate alert dashboards fail; work with your EHR vendor or integration team to embed alerts into existing nurse workflows (e.g., EHR inbox, nurse call system).
• Measure workflow impact, not just clinical outcomes: Track time-to-alert-response, documentation burden, and nurse cognitive load; a system that improves outcomes but burns out staff is not sustainable.

For clinical AI deployment support, including usability testing frameworks and alert governance design, start a project with us.

FAQ

What is the main reason continuous monitoring systems fail in general wards?

Alert fatigue from high false-positive rates is the dominant failure mode. A 2026 scoping review found that usability factors—alert design, device wearability, workflow integration—determine adoption more than model accuracy [1][3]. Nurses override or ignore alerts when they are non-actionable or disrupt existing workflows.

How do Singapore hospitals reduce false positives in deterioration alerting?

Three evidence-based strategies: (1) temporal smoothing (require sustained threshold breaches rather than instantaneous triggers), (2) contextual suppression (suppress alerts during high-variability periods like post-ambulation), and (3) escalation tiering (multi-level alerts with different notification channels) [1]. Ward-specific threshold tuning with nurse input also improves specificity.

What usability metrics should Singapore hospitals track during continuous monitoring pilots?

Beyond clinical outcomes (ICU transfers, length of stay), track: device adherence rates (% of time sensor is worn), alert override rates (% of alerts dismissed without action), time-to-alert-response (median time from alert to nurse assessment), and nurse satisfaction scores. A 2021 protocol paper emphasized these dual-outcome frameworks for real-world evaluation [2].

How does continuous monitoring integration differ across Singapore hospital clusters?

Singapore's restructured hospitals (SingHealth, NUHS, NHG) use different EHR platforms (Epic, Allscripts, Cerner), have different ward layouts, and different nursing ratios. A system validated at one cluster may require workflow redesign and threshold re-tuning at another. Co-design with local nursing teams is essential; vendor-default settings rarely transfer.

Sources

[1] Pan JF, Dowding D, Wong D. The Usability of Continuous Monitoring Devices With Deterioration Alerting Systems in Noncritical Care Units: Scoping Review. Interactive Journal of Medical Research. 2026 Feb 1. https://pubmed.ncbi.nlm.nih.gov/41670042/

[2] Iqbal FM, Joshi M, Khan S. Implementation of Wearable Sensors and Digital Alerting Systems in Secondary Care: Protocol for a Real-World Prospective Study Evaluating Clinical Outcomes. JMIR Research Protocols. 2021 May 4. https://pubmed.ncbi.nlm.nih.gov/33944790/

[3] Pan JF, Wong D, Liao K. Factors Associated With the Usability and Adoption of Continuous Monitoring Devices With Deterioration Alerting Systems in Acute Hospital Non-ICU Settings: A Mixed Methods Study. Journal of Nursing Management. 2026. https://pubmed.ncbi.nlm.nih.gov/41873534/

[4] Velasquez VT, Chang J, Waddell A. The development of early warning scores or alerting systems for the prediction of adverse events in psychiatric patients: a scoping review. BMC Psychiatry. 2024 Oct 2. https://pubmed.ncbi.nlm.nih.gov/39468486/

[5] Iqbal FM, Joshi M, Fox R. Outcomes of Vital Sign Monitoring of an Acute Surgical Cohort With Wearable Sensors and Digital Alerting Systems: A Pragmatically Designed Cohort Study and Propensity-Matched Analysis. Frontiers in Bioengineering and Biotechnology. 2022. https://pubmed.ncbi.nlm.nih.gov/35832414/

[9] NUH to introduce smartwatches for inpatient vital signs monitoring. The Straits Times. 2026 Jul 14. https://news.google.com/rss/articles/CBMisgFBVV95cUxPRkY4am5xNkpYWDR5NTJaTGxCb25rU2R6YkFDQ1VrbVRzZEUxazUyUzZGdmhzb3prRG43MmkyTjlYM0duN1lGVXNxdnVjWFZBRGlLMW41NW04REYzT1ZQVkVHejVTQzVicHJyWGFjclFRNUdwWWY2NHE5bHk3RFdOUnVjT2hPYmRnMDVIWTNhcklqajhmc0FPbUxiMkdrUVZfN05RYmlzMWtyeFZaU25uY1hB?oc=5