Impact of nighttime Rapid Response Team activation on outcomes of hospitalized patients with acute deterioration

The study was performed at The Ottawa Hospital, a Canadian tertiary care hospital network that consists of two individual campuses, each with its own tertiary-level ICU. The Ottawa Hospital is a 1163-bed facility than handles over 160,000 emergency visits, 50,000 inpatients, and roughly 35,000 surgical cases annually. Each hospital has approximately 28–30 ICU beds. These are mixed medical and surgical ICUs. We analyzed prospectively collected data contained in the Ottawa Hospital Data Warehouse, which is a health administrative database widely used in previous research [18‐20]. It contains information from several of the hospital’s information systems, including the patient registration system, clinical data repository, case-costing system, and patient discharge abstracts. Extensive assessments of data quality were performed during database development, and quality assurance initiatives to ensure completeness and accuracy of the data are conducted regularly [18].

We included all patients who met all of the following eligibility criteria: (a) over the age of 18 years; (b) admitted to either campus of The Ottawa Hospital; and (c) received RRT activation between May 1, 2012, and May 31, 2016. Patients with incomplete demographic or outcome data were excluded. We also excluded cases of routine scheduled RRT follow-up. We did not have data related to cases of cardiac arrest (because they involve a different response team). Patients were categorized by time of initial RRT activation. Patients with multiple activations during their admission were categorized on the basis of the time of their initial activation. Activations were classified as occurring during either daytime hours (0800–1659) or nighttime hours (1700–0759). At The Ottawa Hospital, RRTs during daytime hours are composed of an attending critical care specialist, a registered nurse, and a respiratory therapist. Nighttime RRT activations are managed by a resident physician on the critical care service, also with support from a registered nurse and respiratory therapist. An on-call intensivist is available but is not in-house. The same model is used at both sites. During daytime hours, most wards are under the direct supervision of an attending faculty physician. During nighttime hours, most hospital wards are under the supervision of resident physicians, with an attending physician available on-call. The RRT responds only to inpatients, outpatients experiencing distress (e.g., in radiology and endoscopy suites), or patients and family members requiring immediate care in hospital clinics or waiting rooms. The RRT does not respond to patients being assessed in the ED (who have not yet been admitted). Specific criteria for RRT activation at our center have been published previously [21]. They include concerns surrounding airway patency, tachypnea, systolic blood pressure ≤ 90 mmHg or ≥ 200 mmHg, decreased level of consciousness, and hypoxia, but healthcare providers are encouraged to activate the RRT for any reason of concern, even in the absence of objective changes in vital signs or laboratory values.

Patient information, including demographic data, comorbidities, previous ED visits, previous hospital admissions, and previous ICU admissions in the year prior to the index admission, was collected by registration clerical staff at the time of admission and stored in the data warehouse [18]. Information related to RRT activation is gathered by the RRT nursing staff at the time of patient assessment, and it is also stored in the data warehouse. This includes the most recent vital signs and laboratory values at the time of activation, reason for activation, admitting service, and whether the patient was admitted to the ICU. Data on latency time from onset of concerning symptoms or signs (e.g., hypotension, altered mental status) to RRT activation are also estimated by the RRT during assessment for quality improvement purposes. Outcome at hospital discharge (including death or disposition) and lengths of hospital and ICU stay are also stored in the data warehouse.

The primary outcome was in-hospital mortality, comparing patients assessed during daytime vs. nighttime hours. Secondary outcomes included ICU admission following RRT assessment and overall hospital length of stay.

All statistical analyses were performed with commercially available statistical software packages (R version 3.3.3 [R Foundation for Statistical Computing, Vienna, Austria] and IBM SPSS Statistics version 24.0 [IBM, Armonk, NY, USA]). Data are presented as mean values (with SD) or as medians (with IQR), where indicated. Descriptive statistics were employed for between-group comparisons between daytime and nighttime hours, using Student’s t test for continuous variables and the χ² test for categorical variables. In evaluating the outcomes of in-hospital mortality and ICU admission after RRT activation, we used multivariable logistic regression modeling to adjust for potential confounders, including patient characteristics (age, sex, comorbidities, comorbidity index, previous ED visits in the past year, previous hospital admissions in the past year, and previous ICU admissions in the past year), number of RRT activations, latency to RRT activation, most recent laboratory investigations at the time of RRT activation, and vital signs at the time of RRT activation. We also performed a secondary analysis using time of day as the exposure variable (with 1200–1300 as a reference, as reported previously [22]) to evaluate changes in adjusted mortality over the course of the day. We analyzed disposition of survivors at hospital discharge using a similarly constructed multivariable logistic regression model but restricted to patients originally admitted from home, assuming that patients initially admitted from peripheral acute or long-term care centers were likely to return to those centers. Adjusted ORs with 95% CIs and adjusted P values are provided. A P value ≤ 0.05 was taken to represent statistical significance.

The RRT was activated for 6132 discrete patients during the study period. Of these, 109 patients were excluded because of incomplete data, leaving 6023 patients in the study. A total of 2656 patients (44.1%) had calls that occurred during daytime hours (0800–1659), and 3367 (55.9%) had calls that occurred during nighttime hours (1700–0759).

Characteristics of the included patients, along with between-group comparisons between daytime and nighttime RRT activation are depicted in Table 1. The mean age of patients with RRT activation during daytime hours was 67.3 years (SD 16.7), whereas the mean age of those with RRT activation during nighttime hours was 67.9 years (SD 16.7) (P = 0.18). There was no significant difference in admission source (home, acute care facility transfer, long-term care facility transfer) between groups. There were no significant differences between groups in the median number of previous ED visits, hospital admissions, or ICU admissions in the past year.

Between-group differences in characteristics of RRT activation calls are depicted in Table 2. There was a trend toward lower RRT use (number of RRT activations per 1000 hospital admissions, also termed the RRT “dose” [23]) during nighttime hours, though this difference was not statistically significant. Vital signs and laboratory values at the time of RRT activation are displayed. Reasons for RRT activation were significantly different between daytime and nighttime RRT calls (P < 0.001). Respiratory distress and tachycardia/arrhythmias were more common causes of activation at nighttime, whereas altered level of consciousness was more common during daytime hours. Calls arising from general nonspecific concerns about the patient were more common during the daytime. Nighttime RRT activation was associated with longer latency from symptom onset (P < 0.001). More RRT activations occurred within 1 h of symptom onset during daytime than during nighttime hours (81.0% vs. 73.2%, P < 0.001). There was no difference between groups in time to RRT response. A comparison of admitting services at the time of RRT activation is shown in Table 3. Some services (such as orthopedic surgery and hematology) more commonly requested RRT assistance during daytime hours, whereas others (such as neurosurgery and medical oncology) more often activated the RRT at nighttime.

Logistic regression using multivariable models with adjustment for patient variables demonstrated that nighttime RRT activation was associated with significantly higher odds of in-hospital mortality (adjusted OR 1.34, 95% CI 1.26–1.40) (Table 4). The logistic regression model, including all covariables, is depicted in Additional file 1. Daytime RRT activation was associated with significantly higher odds of ICU admission (adjusted OR 1.40, 95% CI 1.30–1.48). No differences in overall hospital length of stay (P = 0.92) or disposition of survivors (P = 0.47) were found between groups. The frequency of RRT calls, with associated risk of mortality across the day, is depicted graphically in Fig. 1. Of note, the frequency of RRT activation increases dramatically at approximately 0800, shortly after the beginning of daytime hours. Two statistically significant time periods were associated with increased adjusted odds of in-hospital mortality: 0600–0700 (adjusted OR 1.30, 95% CI 1.09–1.61) and 2300–2400 (adjusted OR 1.34, 95% CI 1.01–1.56). A sensitivity analysis excluding patients admitted to the ICU was performed (Additional file 2). There were no differences in mortality or hospital length of stay between groups.