Beyond wellness monitoring: Continuous multiparameter remote automated monitoring of patients

Michael H. McGillion, RN, PhD, Katherine Allan, MASc, PhD, Sara Ross-Howe, BASc, MASc, PhD(c), Wenjun Jiang, Michelle Graham, MD, Maura Marcucci, MSc, Ana Johnson, PhD, Ted Scott, MAppSc, PhD, Carley Ouellette, RN, BScN, Dejan Kocetkov, Jennifer Lounsbury, RN(EC) MN, Marissa Bird, BSN, RN, Prathiba Harsha, MBBS, CCRA, Karla Sanchez, BSc, Valerie Harvey, BSc, Jessica Vincent, MSc, Flavia K. Borges, MD, PhD, Sandra L. Carroll, RN, PhD, Elizabeth Peter, RN, PhD, Ameen Patel, MD, Sverre Bergh, PhD, P.J. Devereaux, MD, PhD, FRCPC

The pursuit of more efficient patient-friendly health systems and reductions in need for tertiary health services use has seen enormous growth in the application and study of remote patient monitoring systems in cardiovascular patient care over the last two decades. [1][2][3] More recently, the coronavirus 2019 (COVID-19) pandemic has catapulted this area of clinical practice and research to new heights-health system innovators have worked feverishly to deploy non-invasive systems that facilitate remote patient surveillance and monitoring, as well as the physical and social distancing of patients and families to prevent viral spread. 4 During the COVID-19 crisis, allowances made by regulatory agencies such as the United States Food and Drug Administration (FDA) have permitted health systems to leverage legally marketed patient monitoring solutions, with modifications beyond their original indications for clinical use. 5 The expansion of the capabilities, reach, and availability of these technologies has resulted in an unprecedented number of patient biophysical parameters being monitored remotely. The widespread proliferation of consumer-grade wearable remote patient monitoring devices for wellness monitoring has also changed the face of home-based surveillance of patients' cardiovascular parameters, beyond the use of the original Holter electrocardiogram. [1][2][3]6 While there are many available consumer products to measure patient biophysical parameters, few have stood out in terms of moving these devices forward for clinical-grade application. Wellness devices are limited in the claims they can make regarding the diagnosis, prevention, or treatment of a disease or condition. 7 Manufacturers can suggest they may help to reduce the risk for chronic disease or conditions such as high blood pressure (BP) or type 2 diabetes, where a healthy lifestyle has been generally well accepted to have demonstrated risk reduction or help with living well. 7 The Apple Watch is an example of a product that has impressively featured a number of measurable biophysical parameters over the course of product releases, including heart rate (since Series 1), on-demand lead-1 ECG tracing (Series 4), and blood oxygen saturation (Series 6). [8][9][10] While Apple Watch on demand-ECG tracings are approved by the FDA for arrhythmia detection, the blood oxygen saturation sensor is J o u r n a l P r e -p r o o f for wellness monitoring only; it employs reflectance pulse oximetry-a technique that can be vulnerable to artifact from patient motion. 11 Furthermore, the regulation of consumer-grade wearable patient monitoring products varies considerably, with most available products having little to no published evaluation data. 12 The majority of wellness monitoring devices are capable of measuring a maximum of 1 to 2 episodic (e.g., ondemand ECG) or continuous (e.g., temperature) biophysical parameters simultaneously, thereby limiting clinical utility for patient surveillance, diagnoses, and timely management of changes in health status. While single parameter monitoring (e.g., heart rate) has important application in certain situations (e.g., atrial fibrillation monitoring), many patients require continuous measurement of a diversity of biophysical parameters, similar to what happens among inpatients in hospital. Innovators working with market-ready solutions are also often faced with the inherent complexities of intermingling proprietary systems, 13 each with operational independence that features unique biophysical parameters and data communication channels, as well as data storage and health system integration requirements.
As the science and practice of virtual care and remote patient monitoring evolves, clinicians and researchers can benefit from an understanding of more comprehensive solutions, capable of monitoring three or more biophysical parameters continuously and simultaneously. These systems target optimal efficiency by maximizing the metrics collected and by streamlining processes through the application of a single integrated solution. More advanced systems are also capable of collecting raw biometric waveforms, such as photoplethysmography and ECG, that can be used to derive noninvasive estimates of more complex biophysical parameters that would normally be measured continuously through invasive means (e.g., continuous BP via invasive arterial catheter). These devices, hereafter referred to as continuous multiparameter remote automated monitoring (CM-RAM) devices, have the potential to revolutionize patient monitoring through seamless integration of J o u r n a l P r e -p r o o f multiple, time-synchronized biophysical parameter signals. These technologies can also yield high frequency/high volume big data for the development of algorithms to facilitate early detection of negative changes in patient health status and timely clinician response. Figure 1 illustrates the purpose of CM-RAM devices, versus wellness monitoring and other medical grade devices.

Body Sensor Network
CM-RAM systems include sensors that can be either attached directly to the body or integrated into different materials for indirect body sensing. The interconnection of these sensors forms a 'Body Sensor Network' (BSN) (see Figure 2). BSNs consist of three layers, with each layer communicating either wirelessly or through hardwired channels. 14 First is the body sensor layer, that includes wearable sensor 'nodes', each capable of measuring, sampling (i.e., digitizing) and processing multiple biophysical signals. 14 BSNs can be sub-categorized into sensors affixed to the patient versus the main processing unit where signals are digitized, filtered, and processed. The second layer is the personal area network layer, containing the coordinating device that runs the end user applications. These are typically hand-held devices (e.g., smartphone) or tablet-based computer devices (e.g., tablet) that can perform either limited or more advanced forms of local data processing to display the metrics being collected or integrate and analyze them to assist in clinical decision making. Signal transmission from the body sensor layer to the personal area network layer is typically through a wireless communication protocol, configured for short-range, low power radio frequency communication (e.g., Bluetooth, Zigbee). The third layer is the global network layer, a back-end cloud infrastructure to support data storage, analytics, and interfacing with dashboards via web portals.
[Insert Figure 2 here] Body Sensing Layer J o u r n a l P r e -p r o o f CM-RAM devices utilize fundamental principles of biomedical engineering to measure and derive clinically relevant biophysical parameters. Sensors convert physical measurements into electrical output that can be quantified and analyzed utilizing digital signal processing. 15 These signals can either be measured in an analog format as continuous measurements, or in a digital format with samples taken at a defined sampling frequencies (e.g., ECG signal sampled at 500 Hz [times per second]). Sensors are classified based on the mode of transduction, including mechanical, electrical, optical, and chemical modalities. 16 See Supplementary Table S1 for details on sensor categories by transduction method. Figure 3 presents common forms of biometric sensors and related positioning for optimal signal quality. Electrocardiography draws upon the principles of biopotentials measured through wet or dry electrodes. Wet electrodes consist of a solid conductive pad that interface with the skin via an electrolyte containing hydrogel that minimizes the electrical impedance of skin. 16 Dry electrodes do not contain any electrolyte materials and instead rely on direct skin contact. Disposable wet electrodes typically use silver-silver chloride (Ag-AgCl) contacts with electrolyte gel for conduction 17 and dry electrodes use metal plating. Specialized electrodes have been developed to support short-or long-term monitoring and different activities (e.g., resting versus stress testing).  where the LEDs and photo diodes are positioned on the same tissue surface. Oxygen saturation is measured by determining the ratio of absorption between red and IR channels. Green PPG is a movement-resilient signal, frequently used for pulse rate detection in ambulatory devices. Reflective PPG sensors are commonly placed on the forehead and wrist locations, and transmittance-based orientations are facilitated on the fingers, toes, ear lobes, and nasal cavity ( Figure 3). 19 These optical sensors are vulnerable to artifact from patient motion, as well as intrusion by ambient (environmental) light. 19 Moreover, accuracy testing in patients with a range of skin pigmentation has been limited.

Common Biometric Sensors, continuous biophysical parameters, and sensor placement
Body temperature is measured with thermoresistors via conduction through a metal contact point with the skin or thermopiles as IR sensors that measure thermal levels. 20 Core body temperature is a more challenging metric to capture than skin temperature, with tympanic, forehead, and under arm sensor sites demonstrating the greatest promise for accuracy. 21 Many CM-RAM devices to date capture surface skin temperature and ongoing research is focused on the development of advanced algorithms to map these metrics to a core body temperature for clinical-grade application. 21 For clinical decision making, core temperature is the preferred measurement as skin temperature can be impacted by a number of environmental factors, such as ambient room temperature or climate. 21 Continuous respiration rate is among the most infrequently measured vital signs, yet the importance of this metric has become paramount during the COVID-19 pandemic. The gold standard J o u r n a l P r e -p r o o f for capturing continuous respiration rate includes the use of capnography and nasal cannula. Wearable CM-RAM devices commonly measure respiration rate with elastomeric plethysmography through belts placed snugly around the chest that utilize piezoelectric (i.e., electrical detection of mechanical stress) sensors to register expansion and contraction, or through impedance plethysmography where bioimpedance (i.e., estimation of body composition) is measured between chest-mounted ECG electrodes. 17 Discrete respiration measurements can also be derived from ECG or PPG signals by identifying the amplitude and frequency deviations caused by respiration. 17 Both these modalities pose challenges to patient comfort; the respiration belt can be uncomfortable for female or obese patients, and wet electrodes can cause skin irritation with extended wear. Further confounding can occur by artifact related to patient vocalizations and coughing. 17 A challenging area is the use of these sensors in patients with underlying chronic respiratory conditions that feature irregular breathing patterns, such as COPD and asthma. Advances are being made in contactless sensing systems, which show promise for accurate respiratory rate and heart rate measurement at a distance, while overcoming these types of movement and artifact-related challenges. 22 [Insert Figure 3 here]

Movement and Position
Some CM-RAM devices also track patient position and movement through inertial measurement units, which include accelerometer, gyroscope and magnetometer sensors placed on the limbs or torso. Accelerometers and gyroscopes utilize either piezoresistive or piezoelectric sensors to measure the acceleration of an object (or rate of change of velocity) and angular velocity, respectively.
A magnetometer measures the strength and direction of magnetic fields to establish patient position.
These signals have been utilized for human activity recognition (HAR), 23 activity indexes, step counting, energy output, sleep quality and fall risk assessment and detection. 24 Ballistocardiography (BCG) utilizes specially located (typically on the chest) inertial measurement units to measure small J o u r n a l P r e -p r o o f movements caused by the mechanical output of the heart and is often used to derive a heart rate. 23 Although some CM-RAM devices, such as VitalPatch (VitalConnect, California, San Jose, CA), 25 feature inertial measurement units s to capture motion and position metrics, these outputs are not subject to regulatory standards or widely adopted in clinical settings at this time. Often, these signal inputs are included in CM-RAM systems to support advanced digital signal processing (DSP) of incoming vital signs data, such as identification and mitigation of the confounding effects of patient motion artifact.

Personal Area Network Layer
Signals generated in the BNS sensor layer are typically wirelessly transmitted to a base station (i.e., tablet or smartphone) within the personal area network layer through wireless body area networks (WBAN). 26 Almost universally, CM-RAM systems use Bluetooth Low Energy (BLE) transmission 27 due to low power requirements, high speed data transmission rates of up to 1 megabit per second (Mbps), and an operating range of up to 100 metres from device to base station. 28 More advanced systems have sufficient on-board data storage to allow for delayed signal transmission if the base station is out of range. Once transmitted, the raw signals are digitized and pre-processed to remove noise artifacts and biophysical parameters metrics are derived.

Global Network Layer
Originating from the patient at home (Body Sensing Layer), pre-processed vital signs data  Three of the most well-studied CM-RAM technologies to date are the VitalPatch, the SensiumVitals System (Sensium Healthcare Ltd, Oxford, UK), and Visi Mobile. 18,31,32 The VitalPatch and SensiumVitals systems consist of a disposable, adhesive wireless ECG patch sensor with 3-axis accelerometer (VitalPatch only) and thermoresistor affixed to the patient's chest that captures multiple biophysical parameters, including heart rate, respiratory rate (i.e, impedance pneumography) and skin temperature. Visi Mobile also includes a wrist-worn device that captures Sp0 2, BP and features a touch screen vital signs display. These systems transmit vital signs in real time to patients and clinicians.
Data are transmitted to cloud platforms for storage and further analysis. 18,31,32 All three devices have undergone clinical validation testing comparing their metrics against manual vital signs measurements taken by nurses. 18,31,32 In a feasibility study (n= 20), 18 patients admitted to internal medicine and surgical wards were monitored with the VitalPatch and Visi Mobile devices for 2 to 3 days. 18 Vital signs collected by both devices and nurses were used to calculate and compare Modified Early Warning Scores for clinical deterioration. The clinical measurements and both device measurements were in agreement within accepted limits, although wide limits of J o u r n a l P r e -p r o o f agreement were found. 18 In 15% and 25% of the Visi Mobile and VitalPatch cases, respectively, clinically relevant differences in Modified Early Warning Scores comparisons were found based on inconsistent respiratory ratemeasurements; both devices overestimated respiratory rate in comparison to nurses. 18 Technical issues also differed by device; over 50% of VitalPatch artifacts and data losses had no discernable cause while the rest were due to loss of skin contact, transmission problems or the patient leaving the ward without their mobile device. 18 33 Clinically significant decreases in transfers to ICU and failure-to-rescue events in the Visi Mobile group were also observed. By incorporating Visi Mobile data into their patient assessments, nurses reported they were able to prioritize patient care with greater accuracy, identify signs of clinical deterioration, and facilitate early intervention. 33 The accuracy and feasibility of the SensiumVital System was assessed in a series of RCTs in patients undergoing major elective general surgery. 32,33 The reference standard was nurse-recorded J o u r n a l P r e -p r o o f vital signs, factored into the National Early Warning Score for adverse event prediction. Participants were individually randomized 1:1 to receive either CM-RAM with Sensium plus National Early Warning Score monitoring or monitoring by National Early Warning Score, alone. In a small pilot RCT (n=51), comparison showed reasonable correlation between nurse and Sensium-recorded heart rate (R 2 =0.67, p<0.001), but poor correlation between these approaches for measurement of respiratory rate (R 2 =0.01, p<0.001) and temperature (R 2 =0.13, p<0.001). 32 Ambient room temperature was thought to be a confounding factor for the Sensium device, given that skin temperature measured by thermoresistor sensors can be affected by environmental factors. Data completeness for continuous vital signs recorded varied (respiratory rate , 31%, , 59.2%, skin temperature, 72.8%) and data losses were attributed to artifact from patient ambulation. 32 In a follow-up larger feasibility RCT (n=136), 33 preliminary clinical outcomes explored included time to antibiotics for sepsis cases, length of hospital stay, number of critical care admissions and rate of hospital readmission within 30 days of discharge.
Time to antibiotics was similar in both arms; participants monitored with Sensium had a shorter average length of stay, i.e., 11.6 days (95% Confidence Interval [CI], 9.5-13.7 days) vs 16.2 days (95% CI, 11.3-21.2 days). 33 The VitalPatch, SensiumVitals and Visi Mobile systems have also undergone wearability and usability assessments. 18,35,36 A pilot RCT of 90 post-surgical patients randomized to CM-RAM using Visi Mobile or VitalPatch for 2-3 days found that patients and nurses had overall positive feelings about both devices-earlier identification of clinical deterioration, shorter hospital stay and increased feelings of safety were frequently mentioned as positive benefits. 35 A feasibility and acceptability study (30 patients, 23 nurses) of SensiumVitals compared to VitalPatch and Visi Mobile in abdominal surgery patients found the majority of patients rated the Sensium sensor patch as comfortable, felt safer, and would choose to wear it again when next in hospital. 36 Results for wearability across devices have been mixed and speak to the need for CM-RAM technologies to be lightweight, unobtrusive, and J o u r n a l P r e -p r o o f low-maintenance. The VitalPatch was rated as positive due to its small size and "invisibility" under patient clothing, whereas the Visi Mobile device was felt by some patients to be 'big' or 'heavy' with many cables and a short battery life. 18 Large scale prospective observational studies of other hospital-based CM-RAM systems have also shown positive results. In a before (n=2139) and after (n=2263) study of patients admitted to two general medicine wards in the United Kingdom, Subbe et al. 37 examined the effect of automated continual vital signs monitoring, with relay of abnormal vital signs to rapid response teams. The Philips Guardian solution (Royal Philips, Amsterdam, Netherlands)-featuring cableless transmittance-based finger SpO 2 sensor, oscillometric BP cuff, and respiratory rate derived from accelerometer and gyroscope sensors applied to the patient, paired with a bedside spot-check vital signs monitor-was used to remotely monitor respiratory rate, heart rate, , BP, and SpO 2; temperature was acquired intermittently. During the intervention period on each ward notifications to rapid response teams increased significantly (from 405 to 524, p=0.001), resulting in interventions for intravenous fluid therapy, antibiotics and bronchodialators. 37 Reduction in mortality was also observed from pre (n= 173) to post (n= 147) intervention (p=0.042). 37 Bellomo et al. 38 found similar results in a before and after study conducted on a cohort of 18,305 patients across 12 general wards (10 hospitals) in Australia, the US, and Europe. The use of the Guardian system led to an increased proportion of response team calls related to detection of abnormal respirations (from 21% to 31%; difference 9.9% [95% CI: 0.1-18.5]); and improvements in survival following treatment to 90 days or discharge (from 86% to 92%; difference 6.3 [95% CI: 0.0-12.6]). 38 Guardian was also associated with significantly decreased nurse time required to record vital signs (from 4.1 ± 1.3 minutes [min] to 2.5 ± 0.5 min; difference 1.6 min [95% CI: 1.4 to 1.8]). 38 Some studies have focused on pilot testing these systems in the home setting, sometimes in combination with environmental ambient sensors. Saner et al. 39 tested a multimodal sensor system in a J o u r n a l P r e -p r o o f cohort of 24 community dwelling seniors over a period of 1-2 years. Vital signs and contextual data were collected using integrated sensors including a passive infrared motion sensing system (Domosafety, S.A., Switzerland) to detect physical activity, toileting, refrigerator use and door openings; as well as an upper armband sensor (Everion, Biovotion AG, Zurich Switzerland) to collect heart rate, heart rate variability, respiratory rate and skin temperature. An accelerometer (Axivity Ltd, Newcastle, UK) evaluated patient motion. 39 A bed sensor beneath the mattress also collected heart rate, respiratory rate and sleep quality. Data were transmitted automatically each night via cellular network to a secured cloud platform for hosting and analyses. 39 A total of 92,592 person hours were recorded by the Everion ® device over the course of the study. Several episodes of health deterioration, including worsening heart failure and heart rhythm disturbances, were captured by sensor signals from different sources, supporting the idea that multiple sensor streams holds promise for detecting patient deterioration and diagnosing health problems at home. 39 Participant feedback supported the use of contactless ambient sensors in the home, where possible, to reduce feelings of intrusion that can be associated with wearable devices. 39 More recently, Keogh et al. 40 investigated the usability of 7 wearable continuous remote patient monitoring devices, including the CM-RAM Everion solution ® , by asking 8 older adults to wear them in their home environment for a minimum of one week. Participants believed that light weight wristworn sensors were the most versatile and easy to use, and therefore more suitable for longer term use.
Most also agreed that long battery life was essential; a minimum of one week was considered ideal.
The need to charge systems daily was deemed unacceptable and was considered a barrier for extended use of the Everion ® device. 40 Participants expressed willingness to accept some device-related discomfort, inconvenience, and intrusion at home for systems perceived as useful through the provision of real-time feedback on their health status. Systems designed to only inform remote observers were perceived as less tolerable. 40 J o u r n a l P r e -p r o o f Some advances are also being made in the use of CM-RAM systems for acuity prediction in trauma and emergency patients. Meizoso et al. 41   small surface sensors placed on up to 5 patients simultaneously before wirelessly transmitting these data to monitors carried by any first responder within a 100-metre range. 41 The system incorporates an injury acuity algorithm, the Murphy Factor, which summarizes overall patient status, accounting for changes in vital signs every 30 seconds. Pulse wave transit time (PWTT) is used in place of systolic blood pressure. 41 MiniMedic sensors were applied to the forehead and finger of each patient to measure PWTT, temperature, heart rate and SpO 2 , which were recorded and displayed on a standard bedside monitor for 60 minutes. Shock index was calculated using bedside measured vital signs and compared to the Murphy Factor. The shock index categorized patients equally as "routine," "priority," and "critical," whereas the Murphy Factor over-triaged to "routine" and under-triaged to "critical". 41 The discrepancies were attributed to erroneous PWTT estimations of BP. Refinement of the algorithm thus requires improved accuracy of PWTT measurement or replacement of this metric with continuous noninvasive BP estimation. 41 Liu et al. 42

Economic evaluation
Few studies have evaluated the effect of CM-RAM technologies on resource utilization (hospitalizations, length of stay, emergency department visits) and costs. 43,44 There are important considerations for future economic evaluations of CM-RAM. Analyses may be conducted from the healthcare system perspective (i.e., resource utilization components and costs from the healthcare J o u r n a l P r e -p r o o f system) and/or from the societal perspective (i.e., all costs and benefits are included regardless of who incurs costs or benefits, e.g., savings in patient related travel). 45 Different economic evaluation designs and several analytic approaches can be adopted. It is important to consider time frame/horizon, patient populations being targeted, and the number of vital signs parameters monitored. 46 In future evaluations, implementation costs should be subdivided into individual categories (e.g., capital costs, software licenses, maintenance, and upgrades) to measure the efficiency of CM-RAM programs. 46 Additional cost considerations may also be warranted, depending on system configuration and workflows. For example, whether clinical staff would act on system notifications generated through pre-existing health system infrastructure (e.g., workstations, tablets) or on extrinsic mobile devices will have implications for any potential cost savings.

Big data and advanced prediction modeling
Early Warning Scores (EWSs) are clinical prediction models that use measured vital signs to predict likelihood of clinical deterioration, 30,47 using pre-established likelihood thresholds to trigger a warning so that care can be escalated. While EWSs are now ubiquitous and drive several CM-RAM systems, they can have significant limitations. In a systematic review describing external validation of EWSs for adult inpatients, Gerry et al. 47 found that all 95 studies appraised were at high risk of bias. In addition to poor reporting, they identified several methodological weaknesses, including limited sample sizes and event rates, inadequate handling of missing data and regression models, and focus on discrimination (i.e., ability of the model to stratify patients according to higher or lower risks of events) rather than on calibration (i.e., correspondence between observed and predicted absolute event rates). 47 Moreover, EWSs are static systems, based on the same predictors and cut-offs across populations, which may be neither accurate nor efficient. 30,47 CM-RAM devices offer a way forward. These systems can generate Gigabytes of biophysical and raw physiological data on patients over a 24-hour period, 48 allowing for the application of machine J o u r n a l P r e -p r o o f learning (ML) models, which are increasingly being leveraged to harness the potential of patientcentric big data to develop more complex predictive models. 48,49 Our group is soon commencing the international Vascular events In noncardiac Surgery patIents cOhort evaluatioN study-2 (VISION-2) study, where we will apply the Vitaliti™ Continuous Vital Signs Monitor (CVSM) 49 by Cloud DX to 20,000 patients undergoing noncardiac surgery for the first 30 postoperative days. Vitaliti™ continuously measures 5-lead ECG, heart rate and heart rate variability, respiratory rate, core temperature (infrared sensor applied to the ear), SpO 2 , continuous non-invasive BP and pulse wave velocity. In addition, the device records multiple continuous, highfidelity biometric raw waveforms underpinning these metrics (e.g., photoplethysmography for core temperature and BP) that can be inputted directly into deep learning predictive algorithms. These signals are collected non-invasively at high sampling frequencies (i.e., ECG at 1kHz or 1000 cycles per second). 49 VISION-2 will result in a comprehensive data set suitable for deep predictive modelling research that overcomes challenges (e.g., missing data, representation biases) experienced by retrospective machine learning studies that rely on electronic health record data to train models. 50 Moreover, wide-spread deployment of 5G broadband cellular networks 51 are in progress across most continents. 5G promises increased bandwidths and low latency (i.e., minimal lag time) communications to many under-serviced areas, 51 which will help support data collection in VISION-2 across participating collection sites. Our aim is to build classification models for the prediction of postoperative serious adverse events associated with mortality including myocardial injury, major bleeding, sepsis, and infection. 52

Conclusion
Technologies that offer continuous multiparameter vital signs monitoring hold great promise for comprehensive clinical-grade remote care of patients, beyond what wellness monitoring can J o u r n a l P r e -p r o o f provide. Successful deployment and evaluation of these systems requires an understanding of the architecture of body sensor networks and related technical and operational challenges of their use in clinical and home environments. Scale and spread of these technologies will require attention to comprehensive economic evaluation. Harnessing the power of machine learning will advance the science and practice of CM-RAM for the prediction of critical adverse events, and provision of timely interventions.

Funding Sources
n/a Disclosures MHM and PJD are members of a research group that does not accept with a policy of not accepting honorariums or other payments from industry for personal financial gain. They do accept honorariums/payments from industry to support research endeavours and costs to participate in meetings. Based on study questions PJD has originated and grants he has written, he has received grants from Abbott Diagnostics, Boehringer Ingelheim, Covidien, Octapharma, Philips Healthcare, Roche Diagnostics, and Stryker. PJD has participated in a consultancy advisory board meeting for Boehringer Ingelheim. SRH is employed by Cloud DX, with stock ownership and registered/pending patent applications. DK is employed by Cloud DX, with stock ownership. All other authors: none.