Serious adverse events (SAEs) in hospitalised patients are preceded by signs of clinical deterioration in up to 80% of the patients [1]. Therefore, changes in vital parameters such as pulse rate, respiratory rate, and level of consciousness are often considered as predictors of SAEs such as cardiac arrest, death and unplanned intensive care unit (ICU) admissions [1, 2]. To improve timely detection and treatment of deteriorating patients on nursing wards, rapid response systems (RRSs) have been introduced [3–5]. RRSs consist of two different components: an afferent limb consisting of track and trigger systems (TTS) such as Modified Early Warning Score (MEWS) and an efferent limb, a rapid intervention team (RIT) consisting of trained ICU personnel who will deliver immediate treatment to deteriorating patient at the bedside.

Some studies have demonstrated positive effects of implementing TTSs such as MEWS on patient outcomes [6]. On the basis of these results TTSs have been introduced in many hospitals to increase patient safety [7, 8]. Firstly introduced in 1997 by Morgan et al. the TTS functions as the afferent limb and is designed to detect deterioration early [9]. Since this first introduction multiple early warning bedside monitoring tools have been developed and implemented internationally [10, 11]. These TTSs are used to detect deterioration and call upon a team to monitor and treat patients to prevent further deterioration [12]. In the VU university medical center (VUmc), RRS with an afferent limb consisting of a TTS (MEWS) and an efferent limb consisting of a rapid intervention team (RIT) was introduced a few years ago. Because the afferent limb of the system (RIT) did not function optimally, it was decided to reintroduce the MEWS protocol in 2015 and (re)train the clinical staff aiming to change their mind set and improve protocol adherence.

The effectiveness of a RRS is not only decided by the quality of the RIT but also by an appropriate implementation and use of the TTS such as the MEWS [8, 13]. Unfortunately, very few prospective studies have yet been performed investigating the compliance to any TTS protocol in a real-life setting. In addition, although Smith et al. (2008) demonstrated MEWS as a predictor for clinical outcomes retrospectively, prospective studies investigating the ability of the MEWS to predict relevant clinical outcomes in a general in-hospital population are lacking [14]. In addition, no previous studies have investigated the association between MEWS and the chance of 30-day readmissions. Positive association of MEWS with these endpoints can be used to convince doctors and nurses about the value of MEWS as a prediction tool and thereby increase their protocol adherence.

Therefore, the main aim of this study was to determine the protocol adherence mainly to the afferent limb but also to the efferent limb in a real-life setting. The secondary aims were to investigate the ability of once a day MEWS measurement to predict patient outcomes: in-hospital mortality, hospital length of stay, cardiac arrests, ICU-admissions and 30-day readmissions. Ultimate goal was to provide the hospital staff more insights into the value of the MEWS in predicting outcomes in their own patient population and thereby increase the awareness and protocol adherence.

This prospective study was conducted in a large urban university medical centre (VUmc) with approximately 50,000 admissions per annum in the Netherlands.

In this prospective study conducted in a real life setting, we have demonstrated that adherence to the MEWS protocol in our hospital was good (89%). However in some cases (18%) the MEWS was calculated incorrectly because values were not added up properly, influencing the total score. Although, in the majority of the cases the nurse informed the doctor about the critical score an intervention only occurred in one-third of the cases mostly because the situation was judged as not alarming. The MEWS of 3 or higher was a strong predictor of clinical endpoints such as in-hospital mortality, 30-day readmissions, hospital length of stay. In addition, the negative predictive value of MEWS < 3 in this general hospital population was 98.1% indicating the reliability of this score as a screening tool.

The afferent limb is an important component of a RRS, since an effective clinical response depends on early recognition of deterioration [8, 13]. When we implemented the RRS in our hospital a few years ago the afferent limb was implemented without a clear protocol. Therefore the TTS did not function properly. We re-trained the clinical staff and a clear protocol was implemented in 2015. In this protocol nurses were requested to always take a MEWS score in the morning. The main aim of this study was to analyse protocol adherence after this reimplementation. In addition, we aimed to analyse the value of the morning MEWS measurement in predicting clinical outcomes in this general hospital population because this has not been evaluated in a prospective study in a real life setting. The results of this study showed a high protocol adherence with nurses completing MEWS documentation in 89% of the measurements. However, a percentage (18%) of wrongly documented scores were also seen, likely due to wrong calculations in adding up separate MEWS parameters. An important finding was that due to these wrong calculations, a relatively high percentage of critical scores were missed by nurses. Twenty-eight percent of the critical scores, where a doctor was supposed to be alarmed, were not recognised by the nurses. Our study has also shown, that doctors were not contacted in one-third (31%) of the critical scores. When physicians were contacted, they only undertook an action in 28% of the cases. The main reason for not taking action was that staff judged the situation as not alarming. These findings are comparable to Jones et al. (2011) who also found a percentage of 29% [16]. Reasons for these findings, as explained in previous work, are that clinical staff feel the parameter is too rigorous in its cut-offs or the nursing staff estimate the situation as being under control [17, 18]. However, previous work has already demonstrated that changing the critical cut-off to 4 devaluates MEWS as a reliable screening tool [19].

Since creating awareness and emphasising the importance of the MEWS can increase protocol adherence a secondary aim was to validate the MEWS as a predictor for adverse events in our own hospital population. We demonstrate prospectively for the first time in a real-life setting that patients with a MEWS ≥ 3 in one of the morning measurements had an increased risk for an unplanned adverse event than the patients with a MEWS < 3. No significant increase was observed for unplanned resuscitations, likely due to the very low incidence of events. To our knowledge, this is the first study validating this MEWS protocol prospectively in a general in-hospital population in real-life setting. One other study has prospectively validated the value of MEWS in predicting adverse events in a European surgical population. This study was also performed in a real-life setting. Their results are consistent with our findings [20]. In addition, a recent publication in Africa validated the MEWS prospectively in a research setting in low-resource circumstances [21]. They too found that the MEWS was a useful triage tool to identify patients at the greatest risk of experiencing an adverse event. We also demonstrate for the first time that MEWS ≥ 3 is associated with a significantly higher readmission rate within 30 days for a critical score (10.8% vs. 18.6%). Since readmissions are known to increase mortality and are associated with functional decline, it again underlines the importance of the MEWS as a screening tool [22–24].

MEWS as part of the RRT system, was implemented in many Dutch hospitals to potentially increase patient safety [25]. MEWS is a relatively low-cost and convenient bedside monitoring tool, however critical scores can lead to a higher workload for clinical staff. This study, however, again emphasised the clinical importance of recognising patients with a MEWS higher than 3 since these patients are at high risk of developing adverse events. In addition, the negative predictive value of MEWS < 3 was 98.1 underscoring the importance of MEWS as a screening tool. Nevertheless, it is worth mentioning only 7% of the patients in our population with a MEWS ≥ 3 were transferred to the ICU. We do not know how many patients were prevented from ICU admission by early recognition and prompt treatment on the wards.

The strength of this study is its prospective study design in a real world general hospital sample in which 3290 MEWS values were analysed. In addition we personally contacted every nurse who was involved with the MEWS or vital parameter measurements to collect information daily. This is the largest prospective study conducted so far validating MEWS as a screening tool in a general in-hospital (medical and surgical) population [26, 27]. The study was conducted in a single-center which uses one specific MEWS protocol. Therefore, results might not be generalised to hospitals using another EWS protocol. Also, since our aim was to determine clinical relevance of MEWS in daily practice, a real-life hospital situation was studied. As a result, the determined MEWS and not the completeness of the vital parameter set was taken into account. This could possibly under- or overestimate the relation between MEWS and patient outcomes.

In this prospective study performed in a real-life setting we demonstrated that adherence to the MEWS protocol in our hospital is good (89%). A morning Modified Early Warning Score of 3 or higher was a strong predictor of clinical endpoints such as in-hospital mortality, 30-day readmissions, hospital length of stay. In addition, the negative predictive value of MEWS < 3 in this general hospital population was 98.1% indicating the reliability of this score as a screening tool. Therefore, it is important to keep emphasising the clinical relevance of the MEWS among clinical staff.