This website is intended for healthcare professionals only.

Hospital Healthcare Europe
Hospital Pharmacy Europe     Newsletter    Login            

Automation of patient biosignal monitoring

Utilisation of validated risk prediction algorithms and wearable 

and automated systems may help to prevent serious adverse events 

Niku Oksala MD PhD DSc 

Associate Professor of Surgery,

School of Medicine, University of Tampere and Division of Vascular Surgery, Tampere University Hospital, Finland

Pekka Puska MD PhD 

Professor, National Institute for Health and Welfare (THL), Helsinki, Finland

The current healthcare system is struggling with inefficiency, limited resources, logistic errors and ageing of the population. In Europe and globally, the healthcare resources are not adequate to meet the new demands, particularly the consequences of the ageing population. Ageing populations present with multiple comorbidities. Together these factors contribute to increased risk for highly expensive serious adverse events (SAE, for example, cardiovascular events, infections) during healthcare processes, affecting patients both during hospital stays and after discharge and making patient safety an important issue. It is estimated that up to 20–30% of hospitalised patients have a high risk for SAEs at any stage of care.1

Ultimately, the productivity cannot be improved without simultaneously improving patient safety. To increase productivity, the throughput must be increased, that is, hospital stays should be shortened considerably. This is also rational from the medical viewpoint, that is, the health and wellbeing of the patient, since prolonged hospitalisation exposes the patient to new risks such as hospital infections and impaired recovery despite high caretaking costs. However, early discharge without appropriate monitoring may result in unnecessary complications and readmissions to hospital after discharge.

There is also evidence that the available resources are allocated inefficiently, that is, healthcare professional’s headcount is not utilised efficiently and there are also logistic challenges, that is, either the healthcare professional or the patient is in the wrong place, missing or in the wrong time. Therefore the productivity is not growing in line with the increased costs. Furthermore, the public healthcare budgets and funding are tighter and tighter year-by-year and the costs are not under optimised control. A significant problem is that there is no methodology available to provide patient priority order.

Taken together, the major problems with the current healthcare systems are that healthcare resources are allocated inefficiently, there is growing demand for improved patient safety, the human resources are either limited or not capable of fulfilling the task and finally, the risk of SAEs in inpatients or outpatients is not recognised early enough.

Risk prediction algorithms

One solution to these problems is development of risk prediction algorithms, which are based on the finding that alterations in vital signs occurring before severe adverse events such as cardiac arrests, death and unanticipated intensive care unit (ICU) admission. These algorithms are then used to alert the medical emergency team. The input to these algorithms could be generated from existing hospital databases, monitoring devices or generated manually. Previously, multiple early warning scores such as National Early Warning Score (NEWS),2 VitalPAC Early Warning Score (ViEWS),3 Early Warning Score (EWS)4 and Modified Early Warning Score (MEWS)5 have been published and validated. MEWS contains the same vital parameters as NEWS but also adds urine output, which is not recorded at all in NEWS. Therefore, results with MEWS cannot be directly compared with NEWS. ViEWS contains exactly the same parameters as NEWS and the subcategory scoring definitions are exactly the same. However, the algorithm resulting in final risk category differs slightly. 

In previously published literature, NEWS has been described to have the highest sensitivity and specificity overall in terms of highest Area Under the Receiver Operating Curve (AUROC) analysis for the detection of adverse events (AEs) in hospitalised patients. In a patient cohort with 35,585 patient episodes 3149 adverse events (defined as cardiac arrest, emergency ICU admission or death and any of the outcomes within 24 hours following risk evaluation NEWS) were reported. Thus the prevalence (incidence) of adverse events was 8.8%.6

For the prediction of cardiac arrest within 24 hours of admission, moderate to good AUROCs of 0.722,7 0.7226 have been demonstrated. Unanticipated ICU admission within 24 hours has been shown to be detected at excellent AUROCs of 0.857,7 0.8576 and at moderate AUROCS of 0.670.8 Death within 24 hours of admission has been demonstrated to be detected at excellent AUROCs of 0.894,7 0.909,9 0.876,10 0.886,9 0.8946 and 0.885.11 Any outcome within 24 hours of admission has been demonstrated to be detected at excellent AUROCs of 0.840,1 0.8737 and 0.873.6 ICU admission and/or mortality has been demonstrated to be detected at moderate to good AUROCs of 0.7008 and 0.742.12 Death within 48 hours has been shown to be detected at excellent AUROCs of 0.930.13 Death within 30 days has been demonstrated to be detected at moderately good AUROCs of 0.800,1 0.876,3 0.73212 and 0.810.13 Recently, Tirkkonen et al. described a good AUROC of 0.840 for predicting any adverse event and 0.800 for death within 30 days in a patient cohort of 615 patients using NEWS.1 

Available sensitivity and specificity data from previous trials show potential for 90–95% sensitivity in predicting adverse events or mortality within the following 24 hours after patient evaluation based on existing scoring systems with manual data input while specificity could concomitantly remain at 70–75%.11 It is reasonable to assume that with automated vital sign/biosignal input, the predictive accuracy will increase further. Optimal performance of risk prediction poses a caveat of alarm fatigue due to excessive amount of false positives.

Automated ambulatory or wearable systems

The problem with manual risk prediction systems is potential human error and increased workload. Another way to accomplish this is via development of devices capable of continuously monitoring vital signs of the patient. Existing bedside monitoring devices or ambulatory, wearable devices could accomplish this. The key problem with bedside devices is that they are limited to monitoring non-ambulatory patients in hospital beds. 

According to an exhaustive product search, 34 ambulatory or wearable products are available. These devices typically contain a plethora of sensors including photoplethysmogram (PPG), temperature sensors, blood pressure sensors, electrocardiography and movement sensors and wireless connectivity via Wi-Fi or Bluetooth. Additionally, the user interface may be composed of a screen, microphone or a speaker. Location services are provided by global positioning systems (GPS). The majority of these devices rely on PPG sensors located in fingers thus limiting daily activities. Another challenge is the measurement of blood pressure, which is usually accomplished by intermittent cuff-based techniques resulting in patient discomfort and discontinuous measurement. In addition to that, another challenge is that the devices are composed of several modules interconnected with wires. Medical clearance has been provided mainly for products with roots in the hospital environment. The majority of the devices provide a centralised data processing and analysis service.

Based on our product search of medical grade ambulatory or wearable devices, currently only a few devices exist, namely the Visensia mobile (OBS Medical), ViSi mobile (Sotera Wireless) and Intellivue MX40 (Philips). 

Future challengers exist, namely VSM1 (Biovotion) and ErrS (Medieta). Both Visensia Mobile and Sotera ViSi rely on finger PPG sensors while both VSM1 and ErrS utilise PPG sensors elsewhere, namely at arm and wrist area, respectively. The Visensia lacks blood pressure measurement while ViSi mobile utilises measurement of pulse transit time with ECG and thumb PPG sensors and Intellivue MX40 utilises conventional intermittent cuff-based technique. Advantageously, ErrS utilises modified applanation tonometry and pulse decomposition analysis capable of measuring continuous systolic, diastolic, pulse pressure and central aortic systolic pressure solely from the wrist. The techniques for the determination of respiratory rate also differ: Visensia mobile and ErrS utilise PPG-based pulse wave analysis for this purpose while ViSi mobile and Intellivue MX40 utilise impedance measurement and VSM1 lacks this feature.

The impedance technique provides robust determination of respiratory rate but necessitates chest electrodes. As movement artefacts are an essential problem preventing biosignal measurement during high activity and are also needed for the detection posture, falls and seizures, all the manufacturers plan to provide accelerometers except Visensia mobile and Intellivue MX40. ViSi mobile and Intellivue MX40 include ECG in their basic setup, while ErrS provides truly wireless interconnectivity with an optional ECG module. With respect to connectivity, ErrS provides both Bluetooth and Wi-Fi while others provide either Bluetooth or Wi-Fi. Only the challengers provide measurement without sensors attached to fingers. ErrS and Intellivue provide wireless interconnectivity with the wrist unit and optional chest unit while ViSi mobile utilises cables. The operation time ranges from 8–26 hours. The existing devices rely on existing hospital data infrastructure while the challengers utilise cloud-based architecture.

Wearable devices allow for monitoring of ambulatory patients and could be potentially utilised after hospital discharge as well. The key components of the wearable solution are mobile sensors that measure patient vital signs using modern technology and are capable of communicating the data wirelessly with a centralised monitoring system that continuously analyses the data and executes an alarm when indicated by the data. The majority of the present algorithms include consciousness, temperature, heart rate, blood pressure, respiratory rate, arterial blood oxygen saturation and urinary output as key biosignals.

Currently, patient monitoring outside operation theatres and intensive and postoperative care units, that is, within regular hospital wards is accomplished by intermittent serial measurements by healthcare professionals. The standard of care of early discharged patients is considerably worse and based on telephone check-ups at the most. Patient monitoring by human measurements consumes a high proportion of the limited resources since a single blood pressure, heart rate and respiratory rate measurement may take up to five to ten minutes and cannot be monitored continuously. Another recognised issue is that the amount of signals measured by humans is too often misinterpreted and the status of the patient is not estimated correctly.

Technological solutions make it possible to free limited human resources to be allocated to the most critical tasks. At the same time they minimise human error, provide continuous objective and automated monitoring and thus improve patient safety and increase hospital efficiency and throughput. Thus they are practical solution for the current problems. Specifically, there is need to develop an easy-to-use, wearable, ambulatory, remote and continuous vital sign measurement device in conjunction with cloud-based analysis and alarm solution. The system should support, upgrade and develop the existing healthcare processes in hospitals and be implemented in the existing wireless hospital and home infrastructure. This way it could improve efficiency, facilitating more effective resource allocation according to patient risk level priorities and thereby provide improved patient safety by avoidance of serious adverse events and operational savings with improved patient care. More efficient patient throughput and shortened hospital stay provides savings also by reducing the amount of hospital infections, other complications and improving patient recovery. 

A serious concern is that the automation of vital signal analysis results in excessive amount of false positives resulting in alarm fatigue. 

Therefore, the current risk prediction algorithms should be validated in prospective trials utilising novel devices capable of providing continuous and automated vital sign data. There is also need for optimal wearable devices capable of providing the required input for the risk prediction algorithms.


Niku Oksala is a shareholder and a member of the board of Medieta Ltd. Pekka Puska is the chair of the advisory board of Medieta Ltd.


  1. Tirkkonen J et al. Medical emergency team activation: performance of conventional dichotomised criteria versus national early warning score. Acta Anaesthesiol Scand 2014;58(4):411–9.
  3. Opio MO et al. Validation of the VitalPAC™ Early Warning Score (ViEWS) in acutely ill medical patients attending a resource-poor hospital in sub-Saharan Africa. Resuscitation 2013;84(6):743–6.
  4. Alam N, et al. The impact of the use of the Early Warning Score (EWS) on patient outcomes: a systematic review. Resuscitation 2014;85(5):587–94.
  5. Gardner-Thorpe J, et al. The value of Modified Early Warning Score (MEWS) in surgical in-patients: a prospective observational study. Ann R Coll Surg Engl 2006;88:571–5.
  6. Smith GB et al. The ability of the National Early Warning Score (NEWS) to discriminate patients at risk of early cardiac arrest, unanticipated intensive care unit admission, and death. Resuscitation 2013;84(4):465–70.
  7. Badriyah T et al. Decision-tree early warning score (DTEWS) validates the design of the National Early Warning Score (NEWS). Resuscitation 2014;85(3):418–23.
  8. Corfield AR et al. Utility of a single early warning score in patients with sepsis in the emergency department. Emerg Med J 2014;31(6):482–7.
  9. Opio MO et al. In-hospital mortality of acutely ill medical patients admitted to a resource poor hospital in sub-Saharan Africa and to a Canadian regional hospital compared using the abbreviated VitalPAC Early Warning Score. Eur J Intern Med 2014;25(2):142–6.
  10. Eccles SR et al. CREWS: improving specificity whilst maintaining sensitivity of the National Early Warning Score in patients with chronic hypoxaemia. Resuscitation 2014;85(1):109–11.
  11. Prytherch DR et al. ViEWS–Towards a national early warning score for detecting adult inpatient deterioration. Resuscitation. 2010;81(8):932–7.
  12. Jo S et al. Modified early warning score with rapid lactate level in critically ill medical patients: the ViEWS-L score. Emerg Med J 2013;30(2):123–9.
  13. Kellett J et al. Validation of an abbreviated Vitalpac Early Warning Score (ViEWS) in 75,419 consecutive admissions to a Canadian Regional Hospital.Resuscitation 2012;83(3):297–302.