Medical emergency or rapid response teams were introduced to reduce the risk of cardiac arrest and unexpected death (1). Several before and after studies showed significant reductions in these adverse events (2–4). The only cluster-randomized trial of the introduction of medical emergency teams (the Medical Emergency Response and Intervention Trial [MERIT] Study) was able to show that emergency calling rates increased in the subsequent months, but there were no statistically significant changes in cardiac arrests, in-hospital mortality, or unexpected intensive care admissions in the active hospitals (5). Nevertheless, emergency calls to these teams are increasing over time, and hospitals are encouraged to attain rates of 30 calls/1,000 admissions for optimal benefit (6). The evidence that increased calling rates result in greater benefit is incomplete, and there are concerns that more calls can lead to unexpected consequences such as simultaneous calls (7), deskilling of ward staff, and reduced care of patients for whom the teams are primarily responsible (8). Many of the emergency calls are for patients with limitation of medical treatment (LOMT) orders and for whom end-of-life discussions would be more appropriate than escalation of care (9).
As it is unlikely that another randomized trial of medical emergency teams will be undertaken, we undertook an observational study within a tertiary referral hospital (St Vincent’s Hospital Melbourne [SVHM]) and subsequently across 15 hospitals within the State of Victoria. The study was designed to use additional information in order to model in-hospital mortality adjusted for demographic features, chronic comorbidities, and characteristics associated with increased mortality to help identify the independent impact of emergency calling rates.
MATERIALS AND METHODS
The study was undertaken with the approval of the Research Ethics Committee at SVHM and with the knowledge and consent of the Victorian Intensive Care Data Review Committee (Department of Health and Human Services Victoria) (10) and Australian and New Zealand Intensive Care Society Centre for Outcome and Resource Evaluation (ANZICS CORE) (11).
The first analysis involved patients discharged from SVHM which is a 400-bed tertiary referral and university-affiliated teaching hospital on the edge of the central business district of Melbourne. Services include cardiac and neurosurgery but not major organ transplantation (cardiac, lung, liver, bone marrow) or burns. The hospital participated in the MERIT Study (2002/3) (5) and continued the emergency callout service upon completion. Data were extracted from the Hospital’s Patient Master Index (Enterprise Health Solutions, DXC Technology, Tysons Corner, VA) and the Emergency Call Database which records the nature, duration, and outcome of all emergency calls from June 2004. The analysis period spanned 102 months from January 2008 to June 2016. The Emergency Call Database provided the number of calls/1,000 discharges for each month of the study period. These databases provided demographic details, nature of admission, Charlson Comorbidity Scores (12), and clinical outcomes. In addition, the clinical units under whom each patient was admitted were also known and were used as surrogate markers for clinical diagnosis and severity of illness. Other characteristics associated with mortality included the origin of admission (elective, emergency, interhospital transfer, rehabilitation, residential aged care), need for intensive care, and same day versus multiday admission.
The second analysis involved patients discharged from 15 hospitals in the State of Victoria which has a population that varied from 4.9 million in 2003 to 6.0 million in 2015 (13). Each of the 15 hospitals had both an ICU and emergency response team and had submitted data to the ANZICS CORE (11) for at least the last 5 years. One of the registries managed by Centre for Outcome and Resource Evaluation (CORE) contains information from the Critical Care Resources Survey which includes the number of emergency calls attended in a 12-month period. To find the denominator (yearly discharges from hospital), we extracted data from the Victorian Admitted Episode Dataset (VAED) (14) which contains detailed information on every person discharged from hospitals in Victoria. In addition to the number of discharges, we also extracted demographic details, comorbidity scores (12), and clinical outcomes for each patient discharged between July 2010 and June 2015 in each of the 15 hospitals. Similar characteristics associated with mortality were also extracted. The CORE and VAED data allowed calculation of the number of calls per hospital per 1,000 discharges for each of the years. The 15 hospitals consisted of the six tertiary/quaternary facilities, three metropolitan, and seven major regional hospitals.
Emergency teams were provided by the ICUs in each hospital. The teams comprised a medical practitioner usually an intensive care registrar with or without an internal medicine registrar and an intensive care nurse. Additional members from Emergency Department, Anesthesia, or Cardiology could also attend depending upon the nature and location of the call. Calling criteria are very similar at all the hospitals and included changes in heart rate, blood pressure, conscious state, oxygen saturation, respiratory rate, and seizures. All hospitals had a “worried” category to activate a call.
Data are expressed as median with interquartile ranges (IQRs) for continuous variables and number or percentage for categorical variables. Rates are expressed per 1,000 discharges. Mortality at SVHM was modeled with logistic regression. Independent variables included demographic details, Charlson Comorbidity Scores (12), type of admission, origin of admission, clinical unit, the aforementioned characteristics associated with mortality, month of discharge, and calling rates. For the Victoria data, we chose to account for the different casemix expected within regional, metropolitan, and tertiary hospitals (clinical level) and the clustering that might occur within individual hospitals by using a mixed effects logistic model (15) with hospital nested within clinical level. Statistical significance was set at a p value of less than 0.05. Data were analyzed with Stata V15 (StataCorp, College Station, TX). The complete outputs from the analyses are recorded in the supplemental data (Supplementary Digital Content 1, http://links.lww.com/CCM/D483).
The population of the State of Victoria grew from 4,900,212 in 2003 to 5,966,239 in 2015 (13). Between July 2003 and June 2015, the number of acute hospital separations (discharges) per year had increased from 689,445 to 935,655 while the mortality rate (per 1,000 discharges) decreased from 12.24 to 8.03. ( Supplementary Table E1 and Supplementary Fig. E1, Supplemental Digital Content 1, http://links.lww.com/CCM/D483).
In the study period, 441,029 patients were discharged from the hospital (Table 1; and Supplementary Table E2, Supplemental Digital Content 1, http://links.lww.com/CCM/D483). Median age was 61 years (IQR, 45–74 yr), 57.2% were men, 34.8% were emergency admissions, and 59.2% were classified same day admissions. A small percentage of patients required intensive care (2.5%) and mechanical ventilation (0.29%). Overall mortality was 0.7%. Monthly emergency calling rates varied from 9.21 to 30.69 with a median of 18.4/1,000 discharges; corresponding monthly mortality rates varied from 3.3 to 11.3 with an 8-year median of 6.91/1,000 discharges. The percentage of calls where LOMT orders were in place varied each month from 15.9% to 26.0%. There was a steady reduction in mortality during the study period, and a linear relationship existed between unadjusted mortality rate and emergency calling rate (r = –0.894; p = 0.013) (Fig. 1).
In the multivariable analysis, individual patient mortality was associated with increasing age, male gender, admission status, emergency admission, multiday status, higher comorbidity scores, year of discharge, and the clinical unit under whom the patient was admitted (Table 2). The monthly calling rate—prevalent at the time of patient discharge—was associated with an increased risk of mortality (odds ratio [OR], 1.019; 95% CI, 1.008–1.031). The full model is shown in Supplementary Table E5 (Supplemental Digital Content 1, http://links.lww.com/CCM/D483), and the marginal probability of death and emergency calling rate in Supplementary Figure E2 (Supplemental Digital Content 1, http://links.lww.com/CCM/D483).
There were 3,339,789 discharges from the 15 hospitals between July 2010 and June 2015 (Table 3). Further details on the individual hospitals are listed in Supplementary Tables E3 and E4 (Supplemental Digital Content 1, http://links.lww.com/CCM/D483). All hospitals had introduced medical emergency teams between 1999 and mid-2000s. Median age was 61 years (IQR, 43–74 yr), 51.4% were men, 58.8% were same day admissions. ICU was required in 2.60%. Median length of stay was 1 day overall and 3 days (IQR, 1–5 d) days for multiday admissions. Mortality for all patients was 0.83%. Calling rates gradually increased from 18.46 to 33.4 with a 5-year average of 25.75/1,000 discharges. Mortality rates gradually decreased from 9.41 to 7.01 with a 5-year median of 8.29/1,000 discharges. There was a similar relationship between unadjusted mortality rate and emergency calling rate (r = –0.921; p = 0.026) (Fig. 1). In the multivariable analysis (Table 4), mortality was associated with increasing age, multiday status, need for ICU, higher comorbidity scores, year of discharge, and male gender. Higher emergency calling rates in the year of discharge were associated with an increased probability of mortality (OR, 1.001; 95% CI, 1.001–1.006).
The full model is shown in Supplementary Table E6 (Supplemental Digital Content 1, http://links.lww.com/CCM/D483), and the marginal probability of death and emergency calling rate in Supplementary Figure E3 (Supplemental Digital Content 1, http://links.lww.com/CCM/D483).
The success of medical emergency or rapid response teams has led to an increasing number of calls in many hospitals. This increase has often occurred naturally but sometimes encouraged in the belief that increased calling rates lead to reduced mortality rates. In our detailed study of a tertiary Australian hospital over 8 years and involving over 400,000 patients, we found that the probability of death adjusted for age, gender, nature of admission, the clinical unit under whom the patient was admitted, the year of admission, and the Charlson Comorbidity Index was not improved (i.e., OR < 1.0) by the emergency calling rate at the time the patient was in hospital. Furthermore, in the subsequent multicenter analysis over 5 years and with over 3 million patients, the adjusted probability of mortality was also not improved by increased emergency calling rates.
Evidence for a mortality benefit when medical emergency or rapid response teams are introduced comes from a number of before and after studies (3 , 4 , 16), but the only randomized trial (MERIT Study ) did not demonstrate this relationship. In that trial, staff education lasted 3 months, and the impact of the intervention was assessed during the following 6 months. The negative result may have been due to a number of reasons—insufficient power, better outcomes in control hospitals—but it is possible that it takes time to change behavior within large institutions and that, in retrospect, there had been insufficient training or too little time for staff to accept the concept of emergency calls. Subsequent analyses would support this hypothesis (2 , 17) and possibly link the calling rate to the effectiveness of implementation (18).
Emergency calling rates have been reported in a number of studies. The MERIT study had rates of 8.7/1,000 admissions in the experimental arm and 3.1/1,000 in the control arm (5). Devita et al (19) reported that an increase in the calling rate from 13.7/1,000 to 25.8/1,000 was associated with a reduction in cardiac arrest rates from 6.5/1,000 to 5.4/1,000. Similar associations have been made by other authors and summarized in a recent publication (6); the longer term SVHM data had a statistically significant negative correlation between unadjusted mortality and emergency calling rate.
However, there are problems with such simple associations. First, they ignore the gradual reduction in hospital mortality that has been evident for a number of years. In Victoria, the mortality rate in the 15 hospitals decreased from 12.24 to 8.03/1,000 discharges. It would be difficult to attribute this steady decline to emergency teams alone thereby ignoring the major advances in clinical medicine such as interventional cardiology, laparoscopic surgery, and chemotherapy to name a few. Second, given these changes, none of the studies adjusted for year in their analyses nor did they attempt to adjust for patient factors or casemix. When this was done in our study, increased calling rates did not reduce mortality.
The effectiveness of an emergency system may depend upon factors other than emergency calling rates. Team composition may be important. Within Victoria, the emergency teams generally comprise at least one senior trainee doctor (registrar) often based within an ICU and a critical care–qualified nurse (20). This is not always the case in other jurisdictions where teams may be led by nurses or respiratory therapists. There are also concerns over which calls and which deaths to include in the analyses. Most teams respond to calls within the major inpatient facility, but calls are also made to patients in outlying clinical areas (outpatients, clinics, rehabilitation, mental health) and to passers-by or visitors to the campus with the result that calling rates may seem higher in some but not other studies. With respect to mortality, it is unclear whether mortality rates relate to patients for full care only or include patients with LOMT orders or patients undergoing palliative care. The MERIT study had, by current standards, low calling rates, but the calls counted were on patients without LOMT orders and deaths were of patients for full active treatment.
Some of the increase in calling rates may be due to factors other than reversible clinical deterioration. In the St Vincent’s Hospital dataset, between 8.4% and 15.9% of patients, each month had LOMT orders at the time of first call, and between 15.9% and 26.0% of patients, each month had LOMT orders at some time during their admission. Unfortunately, the presence of such orders is not recorded in the hospital’s patient master index nor the VAED. Jones et al (9) addressed this issue in a 1-month study of 518 patients in seven Australian hospitals. 20.3% of patients had LOMT orders before the call, and a further 10.8% had LOMT orders after the emergency call. One can speculate why this occurs, but an emergency call is a guaranteed method to access clinician decisions or provide symptom relief particularly after hours and on weekends when the primary treating doctors are not within the hospital. Anecdotally, we have also seen an increasing number of calls for patients who do not meet usual calling criteria which might reflect the difficulties ward staff have in contacting members of the treating team who may be in operating theaters or in outpatient clinics.
There are concerns that higher calling rates may have negative effects. Most teams within Australia are not supernumerary, so that doctors and nurses leave their primary duties of tending to patients within ICU, emergency departments, or general wards to attend these calls (20). Researchers at Concord Hospital (Australia) attempted to quantify the rates of adverse events and incidents as a result of staff leaving their normal duties (8). No adverse patient events were seen, but there were 378 incidents (213.7 incidents per 1,000 emergency calls). Such incidents included interrupted ward rounds, delays in patient assessment, and delays in finishing clinical shifts. Many of these might be considered minor, but they were significant for the team members. Untoward incidents were higher when emergency calls ran longer than the median of 20 minutes. Our own experience is that teams spend over 2 hours per day attending to calls.
Strengths and Limitations of the Study
The current study has a number of strengths. The two datasets are very large and span a number of years allowing appropriate modeling to assess the impact of emergency calling rates. Both datasets also contain demographic and clinical information that allows better adjustments for the prediction of in-hospital mortality.
The study does have limitations. First, the St Vincent’s dataset comes from a single tertiary referral hospital; however, the patients and treating teams are similar to peer Victorian hospitals as are calling rates in Australia, and although the many data elements come from an administrative dataset, each individual is linked to clinical information around treating units and the emergency call database permitting calculation of calling rates at the time of each patient’s hospital admission. Second, there are 22 hospitals with ICUs in Victoria, but only 15 had submitted calling rates to ANZICS CORE. These 15 hospitals did include all the tertiary facilities and the major regional centers. Third, clinical adjusters were more limited in the Victoria dataset, and the calling rates were for the year and not specific month of care for each patient. Fourth, within the administrative datasets (PMI, VAED), information on LOMT orders was not recorded and needs to be available in future analyses if we wish to define subsets of patients who might benefit from emergency calls. Fifth, all the hospitals were within a single state where patient profiles and emergency care are relatively homogenous; extrapolation to other countries should be undertaken carefully and possibly after similar analyses on local data. Finally, another randomized trial of emergency teams is unlikely to occur, so that researchers will have to rely on observational analyses such as ours to assess the impact of changes to emergency care on clinical outcomes.
In conclusion, this analysis of two large datasets over a number of years did not observe a reduction in adjusted in-hospital mortality with increased emergency calling rates. Efforts to increase calling rates do not seem warranted.
We wish to thank the members of the Victorian Intensive Care Data Review Committee and the staff of Australian and New Zealand Intensive Care Society Centre for Outcome and Resource Evaluation for their helpful suggestions.
1. Jones DA, DeVita MA, Bellomo RRapid-response teams. N Engl J Med 2011; 365:139–146
2. Santamaria J, Tobin A, Holmes JChanging cardiac arrest and hospital mortality
rates through a medical emergency team takes time and constant review. Crit Care Med 2010; 38:445–450
3. Buist MD, Moore GE, Bernard SA, et alEffects of a medical emergency team on reduction of incidence of and mortality from unexpected cardiac arrests in hospital: Preliminary study. BMJ 2002; 324:387–390
4. Bellomo R, Goldsmith D, Uchino S, et alA prospective before-and-after trial of a medical emergency team. Med J Aust 2003; 179:283–287
5. Hillman K, Chen J, Cretikos M, et alMERIT study investigators: Introduction of the medical emergency team (MET) system: A cluster-randomised controlled trial. Lancet 2005; 365:2091–2097
6. Jones D, Bellomo R, DeVita MAEffectiveness of the medical emergency team: The importance of dose. Crit Care 2009; 13:313
7. Flabouris A, Mesecke MRapid response team calls that overlap in time: Incidence, consequences and patient outcomes. Crit Care Resusc 2017; 19:214–221
8. Cheung W, Sahai V, Mann-Farrar J, et alConcord Medical Emergency Team (MET) Incidents Study Investigators: Incidents resulting from staff leaving normal duties to attend medical emergency team calls. Med J Aust 2014; 201:528–531
9. Jones DA, Bagshaw SM, Barrett J, et alThe role of the medical emergency team in end-of-life care: A multicenter, prospective, observational study. Crit Care Med 2012; 40:98–103
12. Charlson ME, Pompei P, Ales KL, et alA new method of classifying prognostic comorbidity in longitudinal studies: Development and validation. J Chronic Dis 1987; 40:373–383
15. Santamaria JD, Duke GJ, Pilcher DV, et alDischarge and Readmission Evaluation (DARE) Study: The timing of discharge from the intensive care unit and subsequent mortality. A prospective, multicenter study. Am J Respir Crit Care Med 2015; 191:1033–1039
16. Foraida MI, DeVita MA, Braithwaite RS, et alImproving the utilization of medical crisis teams (Condition C) at an urban tertiary care hospital. J Crit Care 2003; 18:87–94
17. Tobin AE, Santamaria JDMedical emergency teams are associated with reduced mortality across a major metropolitan health network after two years service: A retrospective study using government administrative data. Crit Care 2012; 16:R210
18. Cretikos MA, Chen J, Hillman KM, et alMERIT Study Investigators: The effectiveness of implementation of the medical emergency team (MET) system and factors associated with use during the MERIT study. Crit Care Resusc 2007; 9:205–212
19. DeVita MA, Braithwaite RS, Mahidhara R, et alMedical Emergency Response Improvement Team (MERIT): Use of medical emergency team responses to reduce hospital cardiopulmonary arrests. Qual Saf Health Care 2004; 13:251–254
20. Jones D, Drennan K, Hart GK, et alANZICS-CORE MET dose Investigators: Rapid Response Team composition, resourcing and calling criteria in Australia. Resuscitation 2012; 83:563–567
hospital mortality; hospital rapid response team; risk adjustments; risk factors; time factors
Supplemental Digital Content
Copyright © by 2018 by the Society of Critical Care Medicine and Wolters Kluwer Health, Inc. All Rights Reserved.