Introduction

In recent years, hospitals have adopted rapid response systems (RRSs) to improve the detection of and response to deteriorating patients. Goals of RRSs include reducing cardiopulmonary arrests (CPA) outside of the intensive care unit, reducing mortality, and improving early rescue of deteriorating patients [1]. Failed escalation is involved in 16 % of cases of serious harm. Survival from arrest can be improved from 27 to 80 % if deterioration is recognized earlier [2]. Over the last decade, hospital policy and governance organizations in the USA, UK, Canada, and Australia have endorsed the importance of systems to recognize and respond to clinical deterioration in the hospital [3–6].

An RRS consists of four important elements [7, 8] The identification limb detects patient deterioration and makes the call to escalate care. The response limb typically works outside the primary clinical team’s chain of command to bring critical care expertise to the patient. The quality improvement limb examines practice and process improvement. Finally, the governance limb provides oversight, communication with hospital leadership, and maintenance of policies and procedures.

Identifying patient deterioration

Pediatric patients rarely deteriorate suddenly. There may be upwards of 12 h between the first sign of deterioration and the time interventions are made [9, 10]. Methods of detection include the following: early warning scores (EWS), standard clinical observations (vital signs), electronic monitoring, recognition of risk factors, clinician concern, and parental concern.

Early warning scores

The original cardiac arrest or code teams were called when deterioration was detected, often late, and after missed opportunities to rescue the patient early. Delays due to difficulty detecting deterioration and hierarchical escalation processes created the drive to develop new systems to support earlier decision making and treatment [8, 11–13]. The earliest EWS were described in adults [14] and were followed by pediatric EWS comprised of either single parameter triggers or aggregate scoring systems [12, 15–17]. Trigger criteria are good for screening larger populations and have high sensitivity, but lack adequate specificity, risking alert fatigue [18].

Early warning scores that are aggregate scores sum physiological measurement sub-scores to identify and amplify the trend of physiological deterioration [9, 12, 17, 19, 20, 21•]. For example, points may be assigned based on vital signs, physical exam findings, diagnoses, or “worrying condition” [18]. Embedding aggregate EWS within the medical record makes it easier for clinicians to see abnormal trends but validity is threatened by miscalculations and incomplete or delayed observations [9, 12]. The prevalence of pediatric EWS in the UK and USA has increased from ~20 % in 2005 to 80 % in 2012 [22–24].

Pediatric EWS performance has been tested and compared both retrospectively and prospectively in different data sets. They perform moderately well in predicting cardiopulmonary arrest or call for resuscitation with an area under the receiver operator characteristic scores of 0.73–0.91 [18, 25]. Pediatric EWS have been shown to decrease the time from trigger criteria to clinician response, increase the proportion of deteriorating patients seen by a physician, decrease time to PICU admission, and decrease in PICU length of stay [26]. One study also demonstrated “critical” pediatric EWS values a median of 11 h and 36 min prior to a deterioration event, long before care was escalated to a consultant, at a median of 80 min prior to ICU transfer [9]. However, prospective evaluations of EWS without a response team are rare [12, 17]. Qualitatively, clinicians believe that they are beneficial to provide age-appropriate thresholds for sick children, encourage systematic assessments, trigger critical thinking about deterioration management, and support effective communication [27, 28]. Variability of score types makes meta-analysis and comparison difficult, hence repeated calls for a universal score similar to the National Early Warning Score (NEWs) in the UK [29]. The bedside pediatric early warning system (PEWS) is the most thoroughly validated score to date and is undergoing further evaluation in the multi-site cluster-randomized Evaluating Processes of Care and the Outcomes of Children in Hospital (EPOCH) trial [30].

Standard clinical observations

Clinicians must see, feel, and judge effectiveness of breathing, skin perfusion, and level of consciousness to identify the deteriorating patient. EWS will only work to protect children from deterioration if frequent, competent assessments are conducted and interpreted correctly and appropriate decisions are made based on that information [31–34]. Practical standards for routine observation and monitoring are scarce as is high-quality research on normative vital signs (VS) values for children [19, 35, 36].

There is a lack of evidence to guide the frequency of assessments and use of behavioral and physiologic indicators of decline in young hospitalized children. Current EWS rely on relatively infrequent (4–12 h) and varied VS measurements, from a choice of up to 36 parameters, on a paper or electronic chart [18, 19, 23, 29]. Neurological observations are often excluded or inadequate and monitoring plans are often left to relatively inexperienced clinicians [37]. The effectiveness of the RRS is vulnerable to these flaws and points to opportunities for further research in these areas [38, 39].

Normal, abnormal, and normally abnormal vital signs

Physiology changes as children grow as do “normal” VS. Most EWS use expert-derived normal ranges for age [9, 40, 41]. However, recent data suggests that normal distributions for heart and respiratory rate in well and sick hospitalized children are significantly different from published ranges [41, 42•]. The thresholds for normal and abnormal heart and respiratory rate should be reconsidered. The rate of change is greatest up to the age of 3 years [42•], and age ranges could be better grouped by 0–1, 2–3, 4–6, 7–10, and 10–18 years.

Complex patients have their own specific range of normality, e.g., pulse oximetry of 75–85 % is normal in patients with cyanotic heart disease. Using standard thresholds risks over or under reporting physiological deterioration and these unique patients may benefit from customized monitoring plans [43].

Patient monitoring and alerts

The use of electronic monitoring often complements nursing care of patients. However, there may be flaws or gaps that compromise the quality of this data such as varied electronic monitoring technology, frequency of observations, automated processing of observed data, data mistakes, need for nurses to validate VS, and tailored warnings or alarms to individual patients [34, 44•, 45••, 46, 47, 48•]. Outcomes in patients with continuous electronic monitoring are varied in both pediatrics and adults [49–51]. Technical hurdles to overcome include spurious measurements due to movement, appropriate sensors and battery power, and secure data processing and storage.

Risk factors

Most children who deteriorate to cardiac arrest have co-morbidities [12, 52]. It is possible to identify diagnostic categories, high-risk medications, and laboratory values that on their own can identify which patients are at risk for deterioration or mortality even before the physiological trends are measured. These criteria, summarized in Table 1, describe susceptibility to deterioration based on background health or illness, medication, and age [12, 15, 16, 44•, 53–56, 57•, 58]. The original PEWS score includes additional diagnostic risk criteria that improves performance in comparison to the similar bedside PEWS and the PEW tool (ROC 0.85 vs. 0.75 vs. 0.73) [22, 59]. Similarly, a seven-item non-physiological score identified a group of patients with more than 80-fold higher probability of deterioration compared to the baseline risk [54]. Current EWS, however, focus on acute, generic physiological abnormalities and do not identify changes from previous observations or worrisome trends. There is a trade-off between simplicity (with better utility/completion) and complexity (with better sensitivity/specificity). Ideally, the background factors should provide a baseline risk and the physiological trends would increase/decrease on that foundation.

Table 1 Diagnostic and symptomatic indicators of risk for patient deterioration
Full size table

Parental activation

Parents of children who have suffered delayed identification of clinical deterioration want to have the ability to call for more expert help when they are seriously concerned about their child [60, 61]. A recent survey identified that this is possible in 60–69 % of US hospitals and in two states, it is a legal requirement [62–64]. Before implementing parental activation, there were concerns related to resources being overwhelmed by an excess of false alarms, increased parental anxiety, and undermining of therapeutic relationships [65]. However, at one institution with a robust program, there were 42 calls made by concerned parents over a 2-year period [61, 66]. Some programs enable parents to activate a rapid response team (RRT) while others are more comprehensive, serving as a format for parents to communicate all concerns to avoid the need for an urgent call [62, 67]. Educating all families is a challenge; in one study, only 27 % of families understood how to activate the RRT [66].

Response limb

The efferent limb of the RRS consists of a responding team of providers with enough experience to rapidly assess and manage medical emergencies. Single-tiered systems have one type of team that responds to all escalation calls, whereas two-tiered systems include a traditional code blue team to respond to CPA and a second team to respond to calls for clinical assessment of deteriorating patients with a goal of reducing CPA outside the ICU [58, 68–70]. These teams are typically referred to as rapid response teams (RRTs) or critical care outreach teams (CCOT) if nurse led, medical emergency teams (MET) if physician led, or rover teams if the team has surveillance functions as well as response to calls [7, 11, 69, 71, 72]. We will use the term RRT. In 2012, 79 % pediatric intensive care units had an RRT, available 24 h a day, 7 days a week [63]. Physicians were involved in 77 % of their teams [62]. In 2014 UK data, 85 % of units were using PEWS and 18 % had an RRT in place, with rates as high as 52 % for RRT in tertiary hospitals [73].

Evidence for RRT

The evidence for rapid response teams is strong. A meta-analysis including hundreds of thousands of pediatric patients demonstrates statistically significant 38 % reduction of CPA and 21 % reduction in hospital mortality rates [68]. An updated systematic review by Winters et al., continued to support these findings [45••]. There is evidence of reduced respiratory arrest, reduction in the predicted need for ICU level treatments in the first 12 h of ICU care [74••, 75, 76] and a reduction in the time between deterioration and treatment from 9 h 55 min to 4 h 15 min [10]. The balancing metric is that use of an RRT may lead to an increase in ICU transfers, with one US center noting an increase from 4.5 to 5.2 per 1000 non-ICU patient days [74••]. Also, a challenge to evaluating RRTs is the worry that secular trends and other safety interventions may overstate the effect of RRTs. A study at a Canadian hospital without an RRT showed similar declines in mortality as observed in hospitals with RRTs [77]. Still, the level 1 evidence adds weight to the accumulating level 2–5 evidence and is sufficient to support implementation [10, 68, 78•, 79].

Beyond having a RRT, there is evidence in adults for a “dose-response” of RRT with 1 cardiac arrest potentially prevented for every 17 RRT calls [80]. That is, the more RRT calls, the greater reductions in CPA and mortality.

Team composition and skills

The composition of the RRT team varies according to the availability of providers and institutional needs [8, 61, 62]. Traditionally, pediatric RRTs include a combination of medical (critical care attending or fellow hospitalist, resident), critical care nurse, and respiratory therapist. Some teams for adult patients also include laboratory technicians, radiology technician, pharmacist, electrocardiography technician, phlebotomists, and physician’s assistants [81, 82]. Traditionally, pediatric critical care physicians have led RRTs with hospitalists, nurse practitioners, and nurses less commonly, particularly for systems with a second-tier team. One study reports that 29 % of hospitals with a PICU had RRTs led by nurses [61].

Another consideration is whether providers are staffed solely for the purposes of the RRS or have additional clinical responsibilities. In a survey of US children’s hospitals, the majority of the teams included nurses with other clinical responsibilities outside of the RRS [64]. However, several hospitals noted that there were significant concerns regarding adequate staffing of the ICU when nurses were out of the unit responding to calls from the wards. Optimal team composition of the pediatric RRT is not known [45••].

RRTs should have the skills and resources to assess and stabilize the patient at the bedside in collaboration with the primary team. They should be prepared to perform common interventions required of RRTs including airway clearance, oxygen administration, respiratory medications, positive-pressure ventilation by bag-valve mask, interpret chest X-rays, gain intravenous access, administer rapid fluid resuscitation, measure blood glucose, and provide other necessary medications [83, 84]. Non-technical skills of RRT members including leadership and communication skills are also important, with evidence in both adult and pediatric settings [28, 78•, 85, 86].

Qualitative benefits

The benefits of RRTs go beyond patient outcomes. RRTs empower providers to call for help and help to break down hierarchal culture [75, 87]. Nursing education may improve activation of the RRT and lead to decreases in code rates [78•, 86]. A qualitative review of the adult nursing literature on effective use of an RRT found nurses with more training or expertise felt more confident to activate but education and exposure to RRTs improves confidence in junior nurses [88]. The pediatric literature supports that RRTs promote self-efficacy, the benefits of teaming up with others, provide objective criteria to activating an RRT, provides support to inexperienced clinician, and helps overcome hierarchical norms [28].

Barriers

There are on-going barriers to acceptance and effective implementation of a RRS. Proper escalation of care is one area of concern. A multi-centered study of adult patients in Australian hospitals reported 42 % of events with appropriate triggers did not result in an RRT call, even in settings with wide-spread institutional support [89]. Other adult literature report that some may still view RRTs as not effective or life-saving, lead to poor communication in silos of care, and create of additional workload to ICU personnel [13, 88, 90]. A pediatric study of attitudes and barriers reported that 32 % of nurse and 47 % of doctors would not activate an RRT for a patient fulfilling criteria if the patient appeared well. Yet, 30 % of doctors and 15 % of nurses had failed to activate an RRT when needed [91]. The reasons for not escalating concerns about deterioration may be lack of face validity, lack of confidence or expertise, or cultural barriers. In a pediatric study, cultural and socio-political reasons identified include the following:

  1. 1.

    Self-efficacy in recognizing deteriorating conditions and activating the RRT

  2. 2.

    Intraprofessional and interprofessional hierarchies

  3. 3.

    Expectations of adverse interpersonal or clinical outcomes from RRT activations and intensive care unit transfers (e.g., reluctance among subspecialty attending physicians to transfer patients to the intensive care unit for fear of inappropriate management) [28]

Additional barriers include low morale and lack of teamwork, fear of criticism, negative attitudes of attendees, lack of education, and workload [90, 91].

National policy statements specifically recommend identifying and managing the local barriers to implementation of systems including reliable escalation at the patient level [3, 4, 92–94]. Optimizing commitment of staff at senior level, governance arrangements, technological support, sharing of information, especially successes, and effective education and training of staff address these barriers.

Quality improvement and governance

The RRS and RRTs are complex interventions that require both improvement and governance oversight to ensure they effectively meet their safety and quality aims. This is best accomplished by an inter-professional team that includes clinical leaders: ideally, physicians both based on the ICU and in areas with a high number of RRT activations, nursing in the same areas, and respiratory therapy. In the early stages it is valuable to have a data analyst with expertise in run and control chart creation and a member with improvement science expertise. While the expertise needed to administer and improve an RRS has some overlap with the expertise needed to oversee a code team, it may be of value to have a separate team or working group charged with RRS oversight.

Measures

In both the implementation and monitoring phase, it is important to track three types of metrics to evaluate the success of the team and identify opportunities for improvement: outcome measures, process measures, and balancing measures. Outcome measures are best described in the literature and the most commonly tracked are codes outside of the ICU [2, 58]. Two important challenges are of note with codes outside the ICU as a primary outcome measure. First and most importantly, code events are a small subset of code team activations, and it may be difficult to demonstrate improvement in a single center. The code rate in a recent systematic review was 5.9 per 10,000 patients, a sufficiently low number to make it difficult to detect a downward change at any one center without many years of data [68]. Secondly, as certain types of code events (e.g., a first-time seizure that leads to apnea) are likely not preventable by an RRS, a goal of zero codes may not be achievable [67]. The use of a preventable code metric has the advantage of setting a zero goal which may be more motivating for the organization, but determining post hoc preventability is difficult, particularly in a way that might generalize. Codes outside of the ICU do represent potential sentinel events and may offer information on failure modes or risks in the activation or response of the RRT. In-hospital mortality is also a frequently used and important metric but has two challenges. Mortality is rare in children’s hospitals, and it is likely driven by factors that are amenable to RRS and those that are not (e.g., quality of ICU care).

In the last several years, three proximate metrics have been developed to assess the efficacy of pediatric RRS: critical deterioration [57•], unrecognized situation awareness failure events (UNSAFE transfers) [44•], and transfer requiring intensive practice (TRIP) events [95]. Critical deterioration is defined as any patient who transfers from the floor to the ICU and receives vasopressors and is intubated or placed on non-invasive pressure support within 12 h. This is the best validated of the proximate measures. Critical deterioration occurs more than eight times more commonly than codes outside the ICU and is associated with a greater than13-fold increase in mortality. The initiation of an RRS at the Children’s Hospital of Philadelphia was associated with a significant reduction in the occurence of critical deterioration events [76]. As critical deterioration captures some children in whom CPAP is begun 10 h into an ICU stay, it is likely a more sensitive and less specific marker of RRS-preventable deterioration. Alternatively, UNSAFE transfers, defined as a patient transferred from the floor to the ICU who is intubated, placed on vasopressors or given three or more fluid boluses within 1 h of ICU transfer, is a more specific and less sensitive measure [44•]. A complex intervention targeting improved situation awareness at Cincinnati Children’s reduced the rate of UNSAFE transfers by 75 % [44•]. A similar proximate measure is the TRIP developed at Seattle Children’s [95]. A TRIP is a transfer from the hospital floor to the ICU who requires intubation, non-invasive pressure support, or vasopressors within 2 h of ICU transfer. Early evaluation has shown reasonable sensitivity and specificity for survival at 28 days.

Several process measures are useful in evaluating that the RRS is functioning as intended. The number of calls RRT calls placed per day or per month is a valuable metric for several reasons. First, the number of calls placed is a useful proxy measure for how well the identification limb is functioning. This includes both the function of specific triggering mechanisms such as EWS calculations and triggering as well as culture on individual units. Second, quantitative data that a unit has not called an RRT in months may be valuable in identifying an area where clinicians do not support calling the RRT. Conversely, high number of calls may identify training opportunities for unit staff. This may be considered in the context of a dose-response as hospitals with mature RRTs that report improved patient outcome following RRT introduction have a RRT dose between 25.8 and 56.4 calls per 1000 admissions. The concern relates to a low dose reflecting a dysfunctional afferent limb or safety culture. Other process measures include the proportion of RRT calls transferred to the ICU, RRT calls stratified by time of day and day of week, response time of the RRT (15–30 min is most common), and time between elevated PEWS and RRT evaluation or ICU transfer (so-called score to door time) [96]. Similar to the number of calls, the right number is uncertain and likely dependent on organizational factors such as number of ICU beds, but our hospitals consistently have 50–60 % of RRT calls transferred within 4 h of call.

The most commonly discussed balancing measures for RRTs based in the ICU is the cost of taking expert critical care clinicians away from the ICU. This cost is difficult to measure and has not been demonstrated in any studies, but it is an important issue for RRS leaders to consider when deciding the composition of the team. Since RRS function across silos, it may be unclear which departments should finance which aspect of the system. In terms of preventing adverse events, the cost-benefit of an RRT at one large US children’s hospital has recently demonstrated that a RRT composed of clinicians with concurrent responsibilities would recoup its value with preventing only 3.5 critical deteriorations [74••]. The financing and composition of the RRT issue should be addressed from the onset of the team and may require a financial investment by hospital leadership [8, 74••].

Administrative and governance

The administrative and governance limb of an RRS has the job of coordinating resources, finances, and relationships with hospital leadership, risk management, information technology, and patient safety. RRS encounter many barriers and benefit from the support of nursing and medical leadership for sustainability [8]. Interaction with hospital leadership is important for garnering buy-in, financing, and support for educational efforts, and quality improvement initiatives. A reporting structure to hospital leadership can also enhance safety through facilitating reporting of identified latent safety threats or other patient care, safety, or staffing issues revealed during RRS activations or process evaluation. A predetermined reporting structure is important to facilitating this.

Summary

Rapid response systems are comprehensive, hospital-wide systems of care that can effectively reduce harm and mortality to children. RRSs have an accepted, internationally recognized framework of identification, response, quality improvement, and governance arms and strong support from national safety bodies. Many before and after studies and two systematic reviews have demonstrated the utility of EWS and RRT to reduce cardiopulmonary arrest and mortality, new data endorses their cost-effectiveness, and emerging clinical deterioration metrics allow for further study and improvement. More evidence is needed for optimal patient monitoring, adoption of appropriate vital signs for sick inpatients, the ideal make up of an RRT, and the financial case for an RRS. Hospitals must address threats to the success of an RRS including cultural barriers, hierarchy, punitive consequences for calling the team, lack of appropriate education and training, staffing concerns, leadership, and financial support. The case for implementing an RRS in pediatric hospitals is made; how best to do this is evolving.