We studied the incidence of cardiac arrests and unexpected deaths in surgical patients before and after implementation of an RRS and the adherence of nurses and doctors to the RRS procedures. The number of patients who experienced a cardiac arrest and/or died unexpectedly declined with 50%. Unplanned ICU admissions increased significantly, but the APACHE ll scores and the LOS of those admissions remained almost unchanged. We found a significant improvement in ward physician interventions to almost 90% of the events with an observed abnormal EWS. The MET was consulted in half of the events on the first day when an abnormal EWS was observed.
Although we showed a 50% reduction in the composite end-point cardiac arrest and/or unexpected death, these results were not statistically significant probably due to the low baseline incidence. Reduction of cardiac arrests and unexpected deaths has been shown in studies with a higher baseline incidence compared to our study [19–23]. To show a statistically significant reduction of 50% in the composite end-point cardiac arrests and/or unexpected death, we should have included almost 20.000 patients.
Surprisingly, we found a significant increase of unplanned ICU admissions. Many studies have shown no effect [23–25] whereas others found a decrease in unplanned ICU admissions [19, 26]. However, in those studies no information on the adherence to the RRS was provided. The increase of unplanned ICU admissions could be explained because significantly more patients were detected as critically ill and were referred to the ICU. Disappointingly, after implementation of the RRS no significant decrease in the median APACHE II score at unplanned ICU admission or in the median unplanned ICU LOS was found, indicating that ICU referrals apparently were not done at an earlier stage of illness. Our MET dose was relatively high (56 per 1000 admissions) compared to hospitals with a mature RRS (26-56 per 1000 admissions) . However, in our study the MET was not consulted at all or consulted with a delay of one or two days in half of the events. Absent or delayed MET consults may be due to suboptimal adherence of the ward staff to the system. Furthermore, the two-tiered MET calling procedure may have delayed activation. Recent studies have shown that a delayed MET response was independently associated with greater risk of unplanned ICU admissions  and hospital mortality [12, 13, 15]. In addition, we found that in one out of five events, the MET chose to treat the patient on the ward for one or two days, while eventually the patient had to be transferred to the ICU. Therefore, it is also possible that the MET waited too long before transferring these patients to the ICU.
In the medical records of SAE patients, the number of records with reported abnormal vital signs prior to an SAE increased significantly in the after study. A likely explanation is the introduction of the EWS and the training program for nurses. However, EWS recordings were frequently incomplete which is of concern, as monitoring is essential for triage to an appropriate level of care . Adopting an RRS is a complex process that needs time to become established as an integral part of the ward care system [14, 27–29]. Even though we found a remarkable improvement in detecting and treating critically ill patients, our results show that further implementation strategies should be developed to improve adherence of the ward nurses and doctors to the RRS procedures and to stimulate the MET to refer the patient to the ICU at an earlier stage of deterioration.
Strengths and limitations of the study
The outcome ‘unexpected death’ did not take into account patients who died in the operation theatre or patients who died after surgery on the ICU. We also excluded deaths with a DNR order from the primary outcome. Therefore, the outcome ‘unexpected death’ is more informative to evaluate the effects of the RRS compared to the outcome measures ‘in hospital deaths’ or ‘hospital mortality’ used in other studies.
Our study had some limitations to take into consideration. First, in our study a single parameter track and trigger warning system was used. This system is comparable with the MET activation criteria studied by Cretikos et al., which have a positive predictive value of 10% and a sensitivity of 50% , implicating that the system would often trigger MET activation while the patient is not at risk for an adverse event. This may have been of influence on the adherence of the ward staff to the system. Second, in the medical records of SAE patients, often no exact time indication was recorded along with observed abnormal EWS. Therefore, timelines were defined in days on which ward physicians and MET were called following an abnormal EWS observation.
Third, we studied the effects of an RRS only in surgical patients as it was expected that those patients would benefit most from the RRS system. However, a recent study showed that an RRS had a greater impact on cardiac arrest and mortality in medical patients compared to surgical patients . Finally, this study was conducted in a single hospital; data may therefore be less applicable to other study populations and settings. However, implementation of an RRS poses challenges in change of behavior, and only progressive accumulation of evidence and experience from different settings and situations will fill the gaps of knowledge in order to adjust the system to the specific needs of a certain setting .