Diagnosing the deteriorating patient: remote monitoring on the ward and beyond (Preckel, Kalkman)
Current monitoring on the surgical ward
Measuring vital signs such as HR, respiratory rate, noninvasive BP, body temperature, urine output and peripheral oxygen saturation at regular intervals is the cornerstone of patient surveillance, not only in the operating room and ICU, but also on hospital wards. Measuring these physiologic variables and possible deviations from what is defined as ‘normal’ should alert healthcare professionals to check whether additional treatment is necessary to prevent patient deterioration.
Although the hospital ward is a perfect location for surveillance of postoperative patients, about 50% of the in-hospital cardiac arrests occur on these wards. The EUSOS study taught us that most patients who died in the hospital never received care in a high-dependency monitoring unit, for example, a medium care unit or ICU.520 Comparing hospitals with high and low peri-operative mortality rates, Ghaferi et al.521 observed similar postoperative complication rates, but the lower mortality in top-performing hospitals could be attributed to timely detection and adequate treatment of complications. In most cases of intrahospital cardiac arrest, deteriorations of physiologic parameters minutes to hours before the emergency event have been observed.522 Thus, optimising the ‘afferent arm’ of a rapid-response system by timely detection of vital signs abnormalities, summarising and interpreting deviations with the help of scores such as the NEWS,510 notifying relevant medical staff (e.g. the rapid response team),523 and adequate and timely treatment decisions to prevent further deterioration should significantly decrease the incidence of fatal events on the ward as well as transfers to the ICU. This should not only improve patients’ outcomes, but should also reduce care costs. A recent expert opinion-based Consensus Statement by the International Rapid Response Society states that cardiac arrest rates on general wards with a well performing rapid response system should be close to zero.504
So much for the theory: in practice not all elements of this theory are supported by robust evidence. But what then is the reality? For this chapter, we will focus on the monitoring part of the rescue chain, also known as the ‘afferent arm’ of the rapid response system.524
We can easily recognise that we have a significant monitoring gap on our wards, as well as in patients discharged home from hospitals (Fig. 12). In the operating room and on the ICU, we continuously measure HR, respiratory rate and peripheral oxygen saturation, and blood pressure at least intermittently every 3 to 5 min. Results are manually written down in the patient chart or registered electronically in the patient data management system. In addition, more invasive measurements such as intra-arterial or central venous pressures are performed in high-risk patients undergoing major surgery.
In the operating room, an anaesthesiologist – in Europe often together with an anaesthesia nurse – is caring for the patient. This staff-patient ratio of 2 : 1 reverses on the postoperative care unit, but then dramatically decreases further when the patient is discharged to the ward, especially during the late evening and night hours when nurse–patient ratios of 1 : 15 are common (Fig. 12). When the patient is finally discharged from hospital to home, monitoring of vital parameters at home most likely completely stops. A large European study demonstrated that both nurse staffing numbers and nurse qualifications are associated with patient outcome.525 However, with the shortage of nurses in most care systems in the world it is unlikely that we can easily improve nurse–patient ratios on most wards in the future. This urges us to find other possibilities for improving monitoring of vital parameters.
On the ward, manual measurement of physiologic variables by the nurse takes place every 8 to 12 h; this interval can be adapted according to the patient needs and monitoring will be more frequent in wards with critical patients.526 Measuring vital signs is time consuming, and adds a huge burden to nurses’ workload. Even if we could afford increasing nurse staffing to allow measuring vital parameters every 2 h, and assuming that the taking and recording of such a set of vital parameters lasts about 10 min, the patient will only be monitored for 120 min during a 24-h period. Thus, even in these optimised conditions patients on the ward are still not monitored for 22 of 24 h. Most complications occur on postoperative days 2 to 4.527 We therefore urgently need to optimise our surveillance protocols on the postsurgical wards to detect postoperative deterioration early, thereby further improving patient outcome.
Future monitoring on the ward
Conclusions on the afferent limb of the rapid response system from a consensus conference of safety experts addressed the following topics. First, to what extent do physiologic abnormalities predict risk for patient deterioration? Second, do workload changes and their potential stresses on the healthcare environment increase patient risk in a predictable manner? Third, what are the characteristics of an “ideal” monitoring system, and to what extent does currently available technology meet this need? And fourth, how can monitoring be categorised to facilitate the comparison of systems?’507 The authors reviewed the literature up to 2008, and concluded their publication as follows: ‘first, vital sign aberrations predict risk; second, monitoring patients more effectively may improve outcome, although some risk is random; third, the workload implications of monitoring on the clinical workforce have not been explored, but are amenable to study and should be investigated; and fourth, the characteristics of an ideal monitoring system are identifiable’.507
It has been shown that patients and nurses do not accept continuous monitoring with ‘wired’ systems on general wards, as these systems hamper mobilisation and are prone to false alarms from (movement) artefacts. Only 16% of patients were continuously monitored up to 72 h postoperatively; monitoring was stopped early for mobilisation reasons by nurses (37%) or by patients themselves (30%).528 Some monitoring systems that previously only used wired sensors can now be upgraded to allow complete wireless surveillance.529 In addition, fully wireless monitoring systems have recently become available.530 Several of the newer systems use an adhesive wireless ‘patch’ sensor to detect multiple vital signs such as HR from ECG, respiratory rate, and axillary or skin temperature. Some systems have an accelerometer to detect motion and patient position.531 Regarding the characteristics of an ideal monitoring system, DeVita et al.507 proposed a long ‘wish list’ of desired features, including such characteristics as evidence-based, multimodal, accurate, sensitive and specific, continuous, having the ability to trend in real time, not hindering patient mobility and being comfortable, allowing automated alerts/alarms directed to specific caregivers, and being cost-effective and upgradable at low cost with low maintenance costs.
Of course, these systems should effortlessly interface with the electronic health record, and have failure mode recognition as well as default modes to allow for speciality specific displays.507 Expecting to see such fully mature system in our hospitals very soon is wishful thinking, but might become a reality in the near future.
Wireless monitoring systems will, of course, obviate the need to tie patients with wires and tubes to a (bedside) monitor, thereby allowing for early mobilisation and yet allow staff to be able to locate the patient in a given emergency. But whether these systems are also accurate, specific and sensitive still has to be investigated thoroughly. Surprisingly little validation data has been published on currently available wireless sensors, and a CE mark can be obtained easily by showing that the sensor correctly measures vital signs in healthy volunteers at rest. Nonetheless, current data are promising, showing reasonable accuracy in measuring different vital parameters with different systems.532–534 While detection of HR seems to be no real problem for most systems, respiratory rate is more prone to be influenced by talking and moving. A recent study noted that respiratory rate data determined by a wireless monitoring system showed less variability than simultaneously recorded respiratory rate data from the reference bedside monitor (thoracic bio-impedance via the ECG electrodes).532 Thus, for some vital signs these wireless systems might even be more accurate than our current standard measurement methods. These are open questions that have to be solved for individual systems in the future. However, final data on accuracy, sensitivity and specificity are still lacking for most new wireless monitoring systems.
In the future, ‘smart’ wireless monitoring systems will incorporate a clinical decision support engine to improve the specificity of the generated alerts to caregivers. Such systems may not only use vital signs as inputs, but could also receive inputs from the patient (via smartphone), informal carers, nurses and also use data available in the electronic medical record (e.g. new lab values). In theory, this approach could help us to save lives lost to unrecognised deteriorations. Some wireless monitoring systems can automatically calculate modified early warming scores (EWSs), albeit without the important ‘nurse only’ inputs such as ‘nurse worry’, allowing earlier detection of deteriorating patients on the ward.508
However, a recent analysis showed only minimal impact after implementing early warning scores and best practice alerts for patient deterioration, most likely because the manual EWS were not calculated correctly and consistently.535 Determination of EWS is user-dependent and prone to inaccurate determination of vital signs. For example, while it has been shown that changes in respiratory rate are the most important predictor of clinical deterioration, nurses relied more on oxygen saturation and regarded respiratory rate as the least important vital sign.536 Therefore, it is not surprising that respiratory rate is infrequently measured and often inaccurate (or even simply ‘guessed’ by medical staff).537,538 Another limitation of manually determined EWS is its intermittent character,539 which could at least partially be overcome by automation.526 A speciality-specific EWS might be advantageous, for example, in patients with chronically compromised pulmonary function or in patients with neurologic disorders. We should also keep in mind that the EWS was originally developed from recordings of vital signs in ICU patients, and has never been optimised for postsurgical patients. Specific treatments in this population, for example, changes in ventilatory patterns due to abdominal incision, residual muscle relaxation and the use of opioids for pain management, as well as occurrence of complications such as infections and sepsis, make it most likely that postoperative respiratory rate is different from the rate in patients not undergoing surgery.540,541 A recent analysis of respiratory rate measured by a wireless monitoring system in a postsurgical population showed lower rates than expected, and indicate that respiratory rate numbers in different EWSs probably need to be adapted.542
To further improve the value of EWS, not only data from vital signs, but integration of additional information is probably necessary. A score including nurse assessments and nurse judgements, as well as laboratory data recorded in the electronic patient file and updated as soon as new entries were available, showed improved sensitivity to detect deteriorating patients.543 The European Union Horizon 2020-funded Nightingale project () tries to develop this Holy Grail: inventing a system suitaly detection of patient deterioration on general wards AND sfety monitoring at home. The goal is to develop a wireless, multiparameter vital sign sensor combined with clinical decision support, and analysis of laboratory data, with patient and nurse inputs in order to prevent death and disability in general ward patients and in the early days after discharge home from hospital.
Future monitoring after hospital discharge
Up to the current time, there are no data available investigating (continuous) monitoring of vital parameters in patients being discharged home after surgery. But we can learn from lessons in other specialties, for example, cardiology and pulmonology. Patients with chronic diseases (e.g. patients with severe heart failure) frequently seek medical care. Studies using e-health systems have shown that patients’ self-determination of vital parameters and subsequently sending data via e-health support to the treating physician significantly reduced the need for healthcare support and hospitalisation.544 All hospitals today struggle with limited resources – both in terms of finance and nurse staffing – which increases the pressure to discharge postoperative patients from the hospital earlier. Enhanced Recovery after Surgery programmes have shown that early hospital discharge can have benefits in terms of patient outcomes. Wireless monitoring in the first 24 to 72 h – combined with caregiver contact via phone or video – might prove to be advantageous for patients sent home early who are at risk for late postoperative complications. Future technologies that may be useful in this setting are bed-based (‘under the mattress’ or infrared camera) systems and sensing systems built into comfortable ‘smart textiles’, which all might play a role for specific patients in the postoperative period at home.545
Wireless monitoring on the ward: risks to be considered
The experience with alarm fatigue from frequent false alarms in the operating room and ICU teaches us that the dense data streams from multiparameter sensors in multiple patients might generate unacceptably high rates of false alarms, leading probably to alarm fatigue.546 A monitoring system, which is improperly used, will not achieve the intended goals of improving patient outcome. New systems therefore need suitable algorithms to reduce the number of notifications and nonactionable alarms, while at the same time actionable alarms are forwarded to the respective ward staff, either nurses, or in case of emergency alarms also directly to physicians and rapid response teams.526 Hospital management, physicians, nurses, but also patients, insurance companies and medico-legal experts should realise that we are not simply transferring ICU-style monitoring – in wireless form – to the ward. There will likely be notifications or alarms that are not noticed or acted upon by the ward staff, for example when other patients are in more need of urgent attention. While this would be unacceptable on an ICU with a one-to-one nurse–patient ratio, the situation on a ward is significantly different and needs other interpretations, in particular a much higher focus on vital signs trends.
Smart monitoring systems should be able to ‘learn’ the vital parameters of a given patient: while a HR of 45 bpm might be perfectly normal in a patient on betablockers, it might represent a dangerous progressive bradycardia in another patient. Self-learning systems should be able to adapt to the individual patient, allowing recognition of deviations from individualised measurements.547 In the future, machine learning and artificial intelligence should be able to integrate other physiological parameters and laboratory values with the vital sign measurements from a patch sensor,548 and adequate filtering of artefacts would be able to reduce false alarms and associated alarm fatigue. While nurses are already able to adjust alarm limits of most wireless systems, only an intelligent self-learning system will help to decrease the nurses’ workloads, thereby increasing acceptance to use the new system.549 Direct transfer of electronically determined patient data into the electronic patient data file also supports this, without forcing nurses to make annotations themselves. As with other innovations, nurses’ acceptance will likely be higher if they are integral participants in the design and implementation of the systems.550,551 Nurses are convinced that the availability of continuously measured vital parameters will support them in taking clinical decisions.552 However, there might be a risk that patients are seen less frequently by nurses and physicians in those cases where every vital parameter is normal, thus all ‘lights are on green’.
Cost-effectiveness and patients’ experience
Implementation of new monitoring systems is expensive. Although most people support the idea that improved patient monitoring on the ward using continuous, wireless monitoring devices will improve patient outcomes, to date there is hardly any evidence to support this claim. One study investigated the cost-effectiveness of continuous monitoring implementation on a general medical–surgical/trauma ward. Assuming a 5-year Return-of-Investment model, for a single hospital the authors calculated a cost reduction of 0.6 to 2.1 million US dollar per year, with a break even point as early as 0.5 to 0.75 years.553 This calculation seems to be extremely optimistic, and we urgently need trustable data on the cost-effectiveness of implementing of wireless ward monitoring to support physicians during negotiations with the hospital management. However, the only way to achieve such data are to implement and study remote monitoring systems in a controlled fashion in several hospitals simultaneously. The on-going Shepherd trial, a two-centre controlled implementation study of a wireless patch sensor with a stepped-wedge design and a patient-centred outcome recently started in Amsterdam UMC and UMC Utrecht, the Netherlands (ClinicalTrials.gov NCT02957825).
Our patients are quite optimistic and positive regarding these new monitoring modalities: very few patients declined wearing a pulse-oximetry sensor, and patients wearing an adhesive patch for vital signs determination were positive regarding patient comfort.552,554 They appreciated not being woken up during the night for vital signs checks, but also they appreciate contact with nurses and thus advised the use of the continuous wireless monitoring in addition to normal nurse–patient contacts.552
Future clinical studies will need to address the question whether continuous monitoring on general wards, most likely by wireless monitoring systems, indeed reduces the rate of cardiorespiratory events and mortality, and also to what extent this technology can improve patient physical and mental outcomes after surgery. It is unlikely that all patients will benefit equally from additional monitoring, and the potential benefit of an individualised risk-tailored monitoring approach needs to be evaluated. Improvements in biomedical signal detection using ever smaller wearable wireless sensors, along with advances in data science and appropriate use of machine learning should lead to further improvement, in particular with respect to ‘smart’ artefact rejection and reduced rates of false alarms. Finally, all medical staff will need to familiarise themselves with a new culture of ward monitoring, including its advantages and potential drawbacks.
Please visit the following link to see this article in full on the EJA website