Characteristics of rapid response calls in the United States: An analysis of the first 402,023 adult cases from the GWTG-MET registry

Abstract

Objective:

To characterize the rapid response team activations, and the patients receiving them, in the American Heart Association-sponsored Get With The Guidelines Resuscitation- Medical Emergency Team cohort between 2005 and 2015.

Design:

Retrospective multicenter cohort study.

Setting:

Three hundred and sixty United States hospitals.

Patients:

Consecutive adult patients experiencing rapid response team (RRT) activation.

Interventions:

RRT activation.

Measurements and Main Results:

The cohort included 402,023 RRT activations from 347,401 unique healthcare encounters. Respiratory (38.0%) and cardiac (37.4%) triggers were most common. The most frequent interventions – pulse oximetry (66.5%), other monitoring (59.6%), and supplemental oxygen (62.0%) – were non-invasive. Fluids were the most common medication ordered (19.3%) but new antibiotic orders were rare (1.2%). More than 10% of RRTs resulted in code status changes. Hospital mortality was over 14% and increased with subsequent rapid response activations.

Conclusions:

Although patients requiring RRT activation have high inpatient mortality, most RRT activations involve relatively few interventions, which may limit these teams’ ability to improve patient outcomes.

INTRODUCTION

Rapid response teams (RRTs) have been widely implemented across United States hospitals since the Joint Commission spurred their proliferation through the 2008 Patient Safety Goals (1). First described in the medical literature in the early 1990s (2), these teams quickly respond to hospitalized patients with acute clinical deterioration, generally outside of critical care areas, with the goal of preventing in-hospital cardiac arrest and mortality. While the 23-center MERIT cluster randomized trial failed to demonstrate improved outcomes in hospitals after implementation of a rapid response team (3), meta-analyses have reported decreased rates of cardiac arrest outside the intensive care unit (ICU) associated with their implementation (4, 5). However, improved hospital-wide cardiac arrest rates or mortality have not consistently been found (6).

What is clear from the current literature is that descriptions of RRTs vary widely from one study to the next, which may in part explain differences in reported outcomes. To better understand the current landscape of patients who receive such RRT interventions, we outline in this report the design of the American Heart Association (AHA)-sponsored Get With The Guidelines-Resuscitation (GWTG-R) Medical Emergency Team (MET, hereafter referred to as RRT) module and provide a contemporary description of what types of patients receive RRT interventions, their physiologic and non-physiologic triggers for activation, and their disposition after team intervention.

MATERIALS AND METHODS

Data Collection

The GWTG-R-MET registry is a voluntary component of the AHA-sponsored GWTG-R database. Launched in June of 2005, hospitals participating in the registry submit clinical information regarding the medical history, hospital care, and outcomes of consecutive patients experiencing RRT activation using an interactive online reporting form. Data accuracy is ensured by training and certification of data entry staff, use of standardized software with built-in checks for missing values or outliers, and employment of case-study methods for newly enrolled hospitals before submission of data.

All participating institutions are required to comply with local regulatory and privacy guidelines and, if required, to secure institutional review board approval. Because collected data were used primarily at the local site for quality improvement, sites were granted a waiver of informed consent under the common rule.

Case Inclusion and Exclusion

Cases were included for all patients, visitors, employees, and staff within the facility (including ambulatory or outpatient areas) for whom an RRT activation was reported at participating centers.

Data collection design of the GWTG-MET registry

Data were collected in accordance with consensus Utstein guidelines (7).

Admission and Discharge Data

The date and time that the patient entered the system were collected based on the subject type. For example, hospital inpatients were documented at the time of hospital admission, whereas visitors and employees were documented as entering the system at the time of the event. Demographic data were collected as part of admission data. Discharge data collected included the time of hospital discharge or death and whether the patient was declared Do Not Attempt Resuscitation (DNAR) during the admission.

Pre-Event Data

Vital sign values recorded in the four hours prior to the event were collected, including heart rate, blood pressure, respiratory rate, SpO₂, and temperature. Patient location 24 hours prior to the RRT call, including Intensive Care Unit (ICU), Post-Anesthesia Care Unit (PACU), or Emergency Department (ED), was noted. Conscious or procedural sedation (including general anesthesia) within 24 hours prior to the event was also recorded.

Event Data

The location, illness category, date, and time of each event were collected. Location categories included ambulatory/outpatient area, delivery suite, diagnostic/intervention area, general inpatient area, newborn nursery, PACU, rehab, skilled nursing or mental health (psychiatric, substance abuse) unit/facility, same-day surgical area, and telemetry or step-down unit. Illness categories included medical-cardiac, medical-noncardiac, surgical-cardiac, surgical-noncardiac, newborn, obstetric, trauma, or other for visitors and employees. Event time included the time at which the RRT was activated, the time at which the first RRT member arrived at the scene, and time at which the last RRT member departed. Vital signs at the time of the event were recorded.

Conditions that triggered the RRT activation were recorded, with no limit to the number of conditions selected as triggers. Triggers fell under the categories of respiratory, cardiac, neurological, medical, other (which included staff worry), or trigger unknown/undocumented.

Interventions

Drug and non-drug interventions were recorded, including specific medication categories which were not available in the dataset until after 2010. Non-drug intervention categories include respiratory management, ventilation, monitoring, vascular access, urgent specialty consultation, and blood product transfusion.

Outcomes

Outcomes included whether or not the patient required emergency assisted ventilation for acute respiratory compromise (ARC) or chest compressions and/or defibrillation for cardiopulmonary arrest (CPA) during the RRT event. Patient location following the event, such as transfer to ICU or cardiac catheterization lab, was also recorded. Patient survival was recorded. Additionally, a yes/no response was recorded to the question “Was RRT response scope limited by patient/family end of life decisions or physician decision of medical futility? – Indicate if therapy was limited by patient/family end of life decisions or medical futility.”

Review of RRT Response

Additional information about specific issues encountered during the RRT response was recorded. This information included lack of activation despite the presence of a trigger, delays in team response, delays in medication delivery, equipment issues, and lack of essential patient data. Issues between RRT and other providers or departments were recorded.

Statistical Analysis

Categorical variables are reported as n (%), and continuous variables are reported as median (IQR). Hospital characteristics were compared using chi-squared tests. All analyses were performed using Stata version 12.0 (Statacorp, College Station, TX).

RESULTS

Between June 2005 and February 2015, a total of 402,023 RRT activations occurred in 347,401 unique adult patients. These cases were submitted from a total of 360 primarily urban (90.3%), geographically diverse hospitals with a wide distribution of bed size and annual admissions (Table 1). Hospitals included major teaching (n=84, 24.1%), minor teaching (n=110, 31.5%), and non-teaching institutions (n=155, 44.4%).

Table 1.

Hospital characteristics

Hospitals, n	360
Teaching statusa
Major teaching	84 (24.1%)
Minor teaching	110 (31.5%)
Non-teaching	155 (44.4%)
Populationa
Rural	34 (9.7%)
Urban	315 (90.3%)
Annual admissionsb
<2500	10 (2.9%)
2500–4999	33 (9.5%)
5000–7499	42 (12.1%)
7500–9999	29 (8.4%)
10000–14999	66 (19.1%)
15000–19999	62 (17.9%)
20000–29999	65 (18.8%)
30000–39999	22 (6.4%)
40000+	17 (4.9%)
Total bedsb
<100	31 (9.0%)
100–199	72 (20.8%)
200–249	28 (8.1%)
250–299	40 (11.6%)
300–349	32 (9.3%)
350–499	70 (20.2%)
500+	73 (21.1%)
Regionc
North Mid Atlantic	59 (16.7%)
South Atlanticd	104 (29.5%)
North Central	72 (20.4%)
South Central	53 (15.0%)
Mountain/Pacific	65 (18.4%)

^aTeaching status and population type missing for 11 hospitals.

^bTotal admissions and beds missing for 14 hospitals.

^cRegion missing for 7 hospitals.

^dSouth Atlantic region includes Puerto Rico.

Most (88.1%) patients experienced only one RRT call, 9.3% experienced 2 calls, 1.8% experienced 3 calls, and 0.8% of patients experienced 4 or more RRT calls. Patient and event characteristics for the first RRT call are listed in Table 2. The majority of patients were non-cardiac medical patients, and the median age was 66 (IQR 53–78). About half of RRT calls occurred in general inpatient areas (51.2%) and one-third (34.1%) of calls occurred in telemetry/step-down units, with approximately 2% occurring in outpatient settings. A total of 68,249 (19.7%) patients had been transferred from the Emergency Department within 24 hours of the RRT call, 37,811 (10.9%) from an ICU, and 21,652 (6.2%) from a PACU. Less than one-third of RRT calls (27.5%) occurred overnight, defined as Monday through Sunday between 11:00pm – 6:59am.

Table 2.

Patient characteristics

Patients, n	347,401
Age, median (IQR)	66 (53–78)
Race/Ethnicity
Non-Hispanic White	233,434 (67.2%)
Non-Hispanic Black	68,307(19.7%)
Hispanic White	11,367 (3.3%)
Hispanic Other	5,944 (1.7%)
Asian/Pacific Islander	4,471 (1.3%)
Other/Unknown	23,878 (6.9%)
LOS, median (IQR)a	8 (4–15)
Sex
Male	162,688 (46.8)
Female	184,613 (53.1%)
Other/Unknown	100 (0.03%)
Illness category
Medical, Cardiac	67,968 (19.6)
Medical, Non-cardiac	204,607 (59.0)
Surgical, Cardiac	8,009 (2.3)
Surgical, Non-cardiac	50,181 (14.5)
Obstetric	3,078 (0.9)
Trauma	3,151 (0.9)
Other/Unknown	10,407 (3.0)
DNAR before index RRT	18,343 (5.3)
DNAR on/after index RRT	38,959 (11.2)
Pre-event Location within 24 hours
Previous ICU	37,811 (10.9%)
Previous PACU	21,652 (6.2%)
Previous ED	68,249 (19.7%)
Previous sedation	29,949 (8.6%)
Location
General Inpatient Area	177,579 (51.2%)
Telemetry/Step-down unit	118,267 (34.1%)
Rehab, skilled nursing facility, mental health unit/facility	12,308 (3.6%)
Diagnostic/Interventional areasb	9,129 (2.6%)
Ambulatory/Outpatient Area	8,096 (2.3%)
Surgical areasc	1,409 (0.4%)
Intensive care unitsd	3,817 (1.1%)
Delivery Suite	1,033 (0.3%)
ED	2,071 (0.6%)
Other/Unknown	13,415 (3.9%)
Time of day
Weekdaye	189,612 (54.6%)
Weekend dayf	62,142 (17.9%)
Nightg	95,647 (27.5%)

All results are shown as n (%) unless otherwise indicated.

^aNumbers are aggregated from the index RRT event, n=332,676. Only index RRT events are included.

^bIncludes cardiac catheterization lab.

^cIncludes the operating room, PACU, and same-day surgical area.

^dIncludes CCU, adult ICU, pediatric ICU.

^eMonday through Friday from 7:00am to 10:59pm.

^fSaturday and Sunday from 7:00am to 10:59pm.

^gMonday through Sunday from 11:00pm to 6:59am.

IQR, interquartile range

DNAR, do not attempt resuscitation

RRT, rapid response team

ICU, intensive care unit

PACU, post-anesthesia care unit

ED, emergency department

Events were triggered by a broad range of criteria. Of the physiologic categories, respiratory (38.0%) triggers were most common, followed by cardiac (37.4%) and neurological (30.7%) issues (Table 3). Specifically, decreased oxygen saturation (21.2%), altered mental status (20.6%) and tachycardia (17.7%) were frequently noted. General staff worry was the specific trigger in a quarter of all calls.

Table 3.

Triggers. Patients may have more than one trigger in each category or triggers across multiple categories.

Triggers	n=338,384
Respiratory	128,656 (38.0%)
Decreased oxygen saturation	73,627 (21.2%)
New onset of difficulty breathing	53,277 (15.7%)
Tachypnea	42,237 (12.5%)
Respiratory depressiona	17,483 (5.2%)
Cardiac	126,654 (37.4%)
Tachycardia	59,871 (17.7%)
Hypotension	53,103 (15.7%)
Chest pain unresponsive to nitroglycerin	15,585 (4.6%)
Bradycardia	13,241 (3.9%)
Neurological	103,735 (30.7%)
Mental Status change	69,532 (20.6%)
Acute LOC	22,143 (6.5%)
Seizure	15,019 (4.4%)
Suspected acute stroke	9,166 (2.7%)
Unexplained agitation or delirium	3,239 (1.0%)
Other	149,683 (44.2%)
Uncontrolled bleeding	5,076 (1.5%)
Acute decrease in urinary output	1,313 (0.4%)
Staff member acutely worried about patient	85,459 (25.3%)
Other, Not Specified	78,598 (23.2%)

All results are shown as n (%).

^aIncludes respiratory depression and reversal agent without immediate response. n=338,384. Only index RRT events are included.

LOC, loss of consciousness

RRT, rapid response team

Interventions performed by the rapid response team were also diverse (Supplemental Digital Content: Table 1) but commonly included monitoring, such as pulse oximetry (66.5%) and supplemental oxygen (62.0%). Sixty percent of patients received one or more medication intervention, with 19.3% receiving intravenous fluids during the event. New antibiotics were rarely initiated during a rapid response team activation (1.2%). Peripheral access was obtained in more than 1/3 of the events. Statistically significant differences were seen in rates of almost all interventions across hospital type, but these differences were small in magnitude. Similar statistical differences were seen across time with regard to both trigger category and intervention category (Supplemental Digital Content: Table 2). In particular, rates of diagnostic testing and vascular access attempts increased more than 10% over the course of the study.

The majority of RRT calls (54.9%) resulted in patients remaining on the same hospital unit, whereas (30%) resulted in the patient being transferred to an intensive care unit (ICU) and 7.6% to a telemetry monitored unit (Table 4). Of those transferred to the ICU, the median length of ICU stay was 3 days (IQR 1–6). Overall, about 5% of patients had a DNAR order in place prior to the first RRT call, and an additional 11% of patients had DNAR orders placed during or following an RRT activation. Approximately 1% of rapid response scenarios escalated to cardiopulmonary arrest during the response. While less than 1% of patients died during the RRT call, mortality for patients with a rapid response team intervention was 14.3% during the hospitalization. Mortality rates were higher among those with multiple RRT activations (Figure 1; p < 0.01)

Table 4.

Event Outcomes

Patients	347,401
Discharge Disposition
Remained on unit	190,627 (54.9%)
Transferred to ICU	102,609 (30.0%)
Transferred to Telemetry/Step-down	26,209 (7.6%)
Other/Unknown	24,417 (6.9%)
Cath Lab	1,371 (0.4%)
OR	1,426 (0.4%)
Transferred to another hospital	742 (0.2%)
Required Emergency Assisted Ventilationa	13,905 (4.0%)
Cardiopulmonary Arresta	3,575 (1.0%)
Died During RRTb	1,963 (0.6%)
Had a subsequent RRT	41,306 (11.9%)
Died during the hospitalizationc	47,458 (14.3%)

All results are shown as n (%) unless otherwise indicated. Only index RRT events are included.

^an=347,150

^bn=347,091

^cn=333,032

ICU, intensive care unit

OR, operating room

RRT, rapid response team

Figure 1.

Hospital mortality by number of RRT events. p < 0.01.

RRT, rapid response team

DISCUSSION

Here we describe a sample of over 400,000 RRT calls from 360 hospitals across the United States, representing a wide range of bed size, facility type, and geography. This cohort is currently the largest standardized multi-hospital sample of RRTs in existence, and it contains important data on both the processes and outcomes of US RRTs. Results from this cohort may be particularly important to clinicians and hospital systems planning to establish new rapid response programs or adjust the structure of existing programs: knowledge of the types and relative frequencies of different RRT settings, triggers, and interventions may have implications for RRT expectations, resources and equipment, and training practices.

Our most notable finding involved a relative lack of interventions during RRTs compared to other – albeit smaller – reports. For instance, Chan et al reported 13–30% of RRTs involving individual laboratory tests (compared to under 9% for any laboratory testing in our cohort), 30% of RRTs involving chest radiographs (compared to 16% in our cohort), and twice as frequent intubation (7.4% versus 3.6%) as in our cohort (8). Similarly, Morris et al identified chest x-rays in 34% of RRTs and intubations in 10% of RRTs (9). This finding appeared to persist over time in our cohort for most intervention categories. Although it is possible that additional interventions are performed or documented after some patients are transferred to the ICU, this possibility does not account for the large number of patients remaining on the wards after RRT activation, which is similar to rates in some studies (8) but lower than others (9). Moreover, because increased monitoring can be considered an adjustment in the RRT’s afferent limb, it is particularly striking that other diagnostic and therapeutic activities (i.e., specific diagnostic tests, delivery of medications, respiratory support, or other therapies) occurred even less often. Additionally, nearly half of all encounters involved no medication orders; this finding was largely consistent across hospital categories, although major teaching hospitals tended to provide more non-drug interventions than other hospital categories. Moreover, the most common activities recorded in the registry – monitoring, supplemental oxygen, and peripheral intravenous access – did not require physician supervision, and nearly half of all encounters involved no medication orders. These findings were largely consistent across hospital categories, although major teaching hospitals tended to provide more non-drug interventions than other hospital categories. Importantly, the registry does not contain information regarding whether the responding team included a physician, and it lacks the granularity needed to determine if some RRTs with fewer interventions represent omissions of potentially beneficial diagnostics or therapies. It is possible that many of these ward-level interventions were overseen by physicians nonetheless; delegation of these tasks is likely to vary by institution. Despite this limitation, our report that many RRTs provide relatively few interventions may reflect one potential cause underlying the literature’s equivocal findings as to whether RRTs improve overall hospital mortality (3, 6), particularly given evidence of benefit within more proactive rapid response systems (10). The structure and actions of rapid response teams may need additional refinement before RRT-associated mortality improvements are realized.

We also found that new antibiotic administration were extremely rare in this sample – under 2% of all RRT activations, and even more rare at non-teaching hospitals – which is far lower than others’ findings (11). This low observed rate could be underreported if anti-infectives are ordered after the RRT leaves (e.g., on arrival to the ICU) and could also be influenced by case mix, the absence of ordering physicians at some RRTs, or patients deteriorating despite already being on appropriate antibiotic treatment. However, because sepsis may contribute to over 20% of RRT activations in other samples (11, 12), and because of poor outcomes associated with delayed initiation of appropriate antimicrobial therapy, the possibility of undertreatment should also be considered. The RRT is an important opportunity to re-evaluate potential sepsis and provide timely interventions for this large group of at-risk patients (13).

Despite the low level of interventions reported, our findings may have implications for the makeup and training of these teams. Similar to other reports (14), cardiopulmonary triggers were the most common activation categories for RRT calls. Respiratory rate has previously been shown to be the strongest vital sign predictor of adverse outcomes in ward patients (14) – even though it is the least-accurately recorded vital sign (15, 16). In addition, prior work in this sample suggests that respiratory rate and systolic blood pressure are the most accurate predictors of in-hospital mortality for patients who receive a MET call (17). Furthermore, respiratory MET triggers (desaturation and respiratory depression) were the two activation triggers that most strongly predicted mortality in this analysis. Team members with training and expertise in cardiac and respiratory pathology – such as providers certified in Advanced Cardiac Life Support (ACLS) and respiratory therapists – may be well-suited to address these issues.

An additional team need suggested by this analysis involves capabilities for triaging patients with changing clinical status. Approximately 1 in 3 RRTs in the sample resulted in transfer to the ICU, and another 7.6% involved care escalation to intermediate-care units. The frequency of these important outcomes in the database underscores the role played by rapid responders in re-evaluating triage decisions based on evolving clinical status and the potential resource needs of these patients.

Additionally, we found a 14% mortality rate for RRT patients in the registry, with increasing risk for patients experiencing repeat activations. This suggests that patients experiencing RRT activation face hospital mortality more than ten times that of general ward patients (14). We also found a higher-than-expected rate of new DNAR orders during or after the RRT activation (11.2% in our study, compared to less than 8% in most recent studies) (6, 8, 18). Consequently, it is imperative that rapid responders – be they physicians or nurses – be prepared to engage patients and their representatives in goals of care conversations, which may require dedicated training by palliative care specialists. Additionally, rapid responders should be trained to undertake these efforts in communication with the patient’s primary team of caregivers (19). Beyond these efforts, structural and systematic approaches to improve end-of-life care for patients may still be worthwhile, including proactive palliative care outreach teams.

Limitations of our study include the voluntary nature of participation; as such, our cohort may not be representative of all US hospitals or all RRT calls at any particular hospital. In addition, the GWTG-R-MET database does not have on-site validation of data recording or entry. Further, the GWTG-R-MET database does not track team composition or patient underlying disease or severity of illness, precluding comparisons across these categories. Similarly, the database does not contain patient outcomes past hospital discharge, which prevents determination of intermediate- and long-term patient mortality and functional outcomes. Finally, because the database consists only of patients for whom the RRT was activated, it does not permit assessment of patients who fulfilled calling criteria but did not have the RRT called. This phenomenon is common (20) and may represent an important opportunity to increase RRT utilization, which has been linked to improved outcomes (21).

CONCLUSIONS

In a large sample of over 400,000 RRT calls from 360 diverse United States hospitals, we found that patients requiring an RRT activation have higher inpatient mortality, and that a large number of RRT activations involved relatively few interventions, which may influence these teams’ ability to improve patient outcomes. These findings may have implications for the staffing, training, and supplies for RRT teams.