We performed a multi-center, retrospective cohort study to describe the association between nighttime discharge from ICU to the ward and hospital mortality and risk of ICU readmission.
Summary of major findings
We found that approximately one in five patients are discharged from the ICU at nighttime. Patients discharged at night were more likely medical, had more comorbid disease, and were more severely ill at the time of ICU admission as compared to those discharged during the daytime hours. The number of nighttime discharges increased significantly in the two tertiary hospitals during the study period, indirectly suggesting that these ICUs have increasingly been confronted by strain on ICU capacity. Importantly, we showed that nighttime discharge was independently associated with a higher risk for hospital mortality, even when we excluded patients that died early after ICU discharge. Although nighttime discharge was not independently associated with ICU readmission, our data showed readmission rates were higher in community hospitals during the study period. This may represent another proxy measure of strain on ICU capacity. Patients who were older, burdened by greater comorbid disease, had higher admission illness severity, had primary hepatic or septic diagnoses and longer ICU stay were more likely to be readmitted.
Comparison with previous studies
Prior investigations have evaluated the association between time of ICU discharge and outcomes [
8,
12‐
15]. While most studies demonstrate that nighttime discharge is associated with unfavorable outcomes [
7,
12‐
14], others have failed to confirm these findings [
1,
8,
16]. In a landmark study from the ICNARC database in the United Kingdom, Goldfrad and Rowan showed that nighttime discharge from ICU increased risk for hospital mortality [
3]. However, after adjustment for “premature discharge” in their analysis, the independent effect of nighttime discharge was lost [
3], suggesting that the attributable risk for mortality was more related to the untimely discharge rather than the specific time of day. In a subsequent study of Finnish ICU patients, Uusaro et al failed to show “out-of-office” hour discharges (defined as those occurring from 1600 h to 0800 h) were associated with post-ICU mortality [
8]. However, their definition for out-of-office hours was more liberal than ours and those of other studies. Hanane et al, in an analysis of three ICUs in a single hospital, also failed to show an association between nighttime discharge and mortality; however, nighttime discharge was associated with higher rates of ICU readmission and longer hospital lengths of stay [
1]. Interestingly, these authors adjusted for patients’ preferences for “do-not-resuscitate” and site-specific characteristics such as the availability of ICU beds, which could have represented important confounders for the association of nighttime discharge and mortality [
1]. Notably, nighttime discharged patients in this study had more comorbidities and higher disease severity [
1], findings that were reproduced in our study and others [
2,
3,
12,
17]. In general; however, the majority of studies that have examined the consequences of nighttime discharge have shown this to be an independent risk for mortality or unplanned readmission [
1,
6,
7,
14]. Of note, the magnitude of this risk of death in our study (OR 1.29; 95 % CI, 1.14–1.46; Table
3) aligns with prior data [
2,
3,
5,
17‐
19], and adds further confidence to the generalizability of this association.
Arguably, from the perspective of the patient, there is no logical reason to transition a critically ill patient from the ICU to the ward at nighttime. There are plausible explanations for why such nighttime discharges carry a higher risk of adverse outcomes for patients when compared to planned discharges during daytime hours. Nighttime discharges are often provoked by reduced ICU bed availability and the need to accommodate the next more severely ill patient [
12]. These discharges may be unplanned, chaotic, less well coordinated and may negatively impact patient safety by predisposing to adverse events and medical errors, regardless of protocolized hand-over procedures and standardized transfer orders. Furthermore, these nighttime discharges are more likely to be premature, where patients are stepped down to a setting of lower intensity monitoring at a time of day when less resources are available [
3]. Prior studies provide support for these hypotheses. Premature ICU discharge has been shown to be an independent risk for mortality [
20]. Interestingly, a triage model to predict premature ICU discharge for “at-risk” patients was found to predict a 39 % reduction in mortality if these patients remained in the ICU for further 48 h [
21].
High occupancy and reduced ICU bed availability at the time of discharge portends a higher risk for death or ICU readmission [
4]. Moreover, the risk of early ICU readmission has been shown several-fold higher during periods of increased demand characterized by a high number of patient ICU admissions [
22]. Increased risk of ICU readmission [
23] and death [
24] associated with admission to busy ICUs has similarly been confirmed. Although we did not have data on corresponding occupancy rates and measures of patient flow in our ICUs, based on prior data [
3], we speculate that a significant proportion of nighttime discharges occur in response to increased demand for ICU beds. We believe this may be indirectly supported by the increasing trend for nighttime discharges, occurring primarily in tertiary hospitals (Table
3; Additional file
1: Table e-1); the trends for greater adjusted odds for mortality for nighttime discharges in the later years of the study, notably from 2006 onwards, and our sensitivity analyses. This trend for increasing nighttime discharges has similarly been described in other studies [
5,
7], reinforcing the theory that a major driver for nighttime discharges is limited ICU bed availability. However, this finding is not universal. Recently, the Australian and New Zealand Intensive Care Society (ANZICS) reported no increase nighttime discharges over an 8-year period [
19]. Whether this is explained by changes to policy, operations or expanded ICU capacity is uncertain. In another prospective multicenter study, patient-level factors at the time of ICU admission and discharge, including illness severity, need for complex therapy and the presence of orders for limitations in medical therapy, were integrated into an analysis of the association between nighttime discharge and hospital mortality. Both nighttime discharge rate (16.4 %) and hospital mortality (5.2 %) were lower than observed in our study. After adjusting for these factors, in particular limitations to medical therapy, nighttime discharge was no longer associated with hospital mortality (OR 1.16; 95 % CI, 0.89–1.53) [
16]. While this study is clearly an important contribution by exposing the potential confounding influence of limits to medical therapy provided, it did not specifically adjudicate whether hospital deaths were “expected” in those with care limitations and, similar to our study, operations data on occupancy and patient flow were not available, which may impact its external validity.
Study limitations
Our study has notable limitations. First, although our study involved 5 discrete hospital ICUs in a large integrated health region and utilized high-quality prospectively collected data, our findings may have limited generalizability, and due to its observational design, are potentially predisposed to bias and residual confounding. Specifically, our study does not enable a detailed evaluation of true demand on ICU resources, strain on capacity and operations that may influence the flow of patients and decisions to discharge at night. In addition, we recognize that the majority of ICU admissions were directed to the two larger tertiary institutions, where over time, nighttime discharges increased in frequency. On the other hand, our findings are importantly coherent with similar observational studies performed in varied health jurisdictions. Second, our findings may not be applicable to other health systems where ICU capacity issues are not as prevalent [
25]. Third, we did not have data on what proportion of nighttime discharged patients were designated palliative and not for ICU readmission, which may have unduly influenced our results. However, we found our effect estimate for mortality robust in sensitivity analysis after excluding those patients that died within 48 h of ICU discharge and not readmitted, which may represent a surrogate for a change in goals of care. Furthermore, ICU readmission as a quality and performance measure and outcome may be limited due to lack of standardization in criteria, inability to adjudicate whether readmission was avoidable and whether readmission was attributable to residual issues related to critical illness, ward care, or other unmeasured factors (i.e., provider decision-making). Fourth, in our multi-variable analysis, we utilized the admission APACHE II score for adjustment which may not correlate with the clinical status at the time of ICU discharge. Finally, we did not have detailed information about ICU staffing, temporary bed closures, and nuanced hospital-wide operational or bed management characteristics across the included hospitals (other than knowing all ICUs have closed intensivist models) to integrate into our modeling.
Areas of future research
Our findings, in the context of prior studies, suggest some patient-specific factors (i.e., case-mix, illness severity) may influence nighttime discharge and risk of adverse outcome; however, health system factors (i.e., ICU demand; ICU bed availability, hospital bed management priorities) likely exert a greater influence on the probability of nighttime discharge and risk of adverse outcome. The impact of differing adaptive models to ICU or ward care, quality of transition (i.e., hand off), and ICU outreach on reducing premature or unnecessary nighttime discharges or avoidable adverse events remains to be explored. Further high-quality and high-fidelity evaluative studies are needed to better examine the patient-specific and health services factors (i.e., operational models of patient flow) associated with nighttime discharges and risk of adverse events.