Intensive Care Medicine

, Volume 34, Issue 1, pp 148–151

After-hours admissions are not associated with increased risk-adjusted mortality in pediatric intensive care


    • Intensive Care UnitSydney Children’s Hospital
    • University of New South Wales
  • Gary Williams
    • Intensive Care UnitSydney Children’s Hospital
    • University of New South Wales
  • John Awad
    • Intensive Care UnitSydney Children’s Hospital
    • University of New South Wales
  • Barry Duffy
    • Intensive Care UnitSydney Children’s Hospital
    • University of New South Wales
Pediatric Original

DOI: 10.1007/s00134-007-0904-8

Cite this article as:
Numa, A., Williams, G., Awad, J. et al. Intensive Care Med (2008) 34: 148. doi:10.1007/s00134-007-0904-8



To examine the influence of time of admission on risk-adjusted mortality and length of stay for nonelective patients admitted to a pediatric intensive care unit (ICU) without 24-h per day in-house intensivist coverage.


Data analyzed came from a comprehensive, prospectively collected ICU database.


A 12-bed pediatric ICU located in a university-affiliated tertiary referral children's hospital.


Subjects consisted of 4,456 consecutive nonelective patients admitted over a 10-year period (1997–2006).



Measurements and results

Patients were categorized according to time of admission to the ICU as either in-hours (0800–1800 Monday–Friday and 0800–1200 on weekends), when an intensivist is present in the ICU, or after-hours (all other times), when intensivists attend only on an as-needed basis. Multivariate logistic regression was used to assess the effect of time of admission on outcome after adjustment for severity of illness using the Paediatric Index of Mortality (PIM). Patients admitted after hours had a lower risk-adjusted mortality than those admitted during normal working hours, with an odds ratio for death of 0.712 (95% confidence interval 0.518–0.980, p = 0.037). Length of stay was also significantly shorter for patients admitted after hours (44.05 h vs. 50.0 h, p = 0.001).


A lack of in-house intensivist presence is not associated with any increase in mortality or length of stay for patients admitted to our pediatric ICU; on the contrary, after-hours admission in this cohort was associated with a decreased risk-adjusted mortality and a shorter length of stay.


MortalityPediatricOutcomesIntensive careLength of stayQuality


There is an increasingly widespread belief that intensive care units (ICUs) should be staffed with in-house senior clinicians 24 h per day [13]. Reduced staffing levels after hours have been reported as being associated with worse outcomes and reduced levels of intervention in adult and pediatric patients [46]. Recently published data support consistent risk-adjusted outcomes for patients admitted throughout the day and night in ICUs with a 24-h in-house intensivist [2, 3]. However, the observation that universally good outcomes are obtained irrespective of time of admission in such units is not evidence of cause and effect—i. e., the good outcomes are not necessarily a consequence of the staffing model. This issue is important because if it can be shown that having intensivists in-house 24 h per day improves outcomes, it would be incumbent upon all ICUs to move toward this staffing model.

We hypothesized that risk-adjusted mortality and length of stay for nonelective patients admitted to our ICU—where intensivist coverage after hours is conducted primarily by telephone, with attendance only on an as-needed basis—would be unaffected by the time of admission.

Materials and methods

The Sydney Children's Hospital ICU admits approximately 700 patients per year and is one of two tertiary pediatric centers in Sydney, serving a population of 1.32 million children age 0–14 years.

Intensivists cover the ICU on a rotating weekly basis, working in-house 0800–1800 on weekdays and 0800–1200 on weekends and public holidays. Outside these hours, coverage is provided by telephone with return visits to the ICU at the discretion of the intensivists, who all live within 15 min of the hospital. Not every patient admitted to the ICU after hours is reviewed by the on-call intensivist; patients with uncomplicated illnesses in no need of immediate intervention are generally not seen until the following morning. In-house registrars have at least 3 years of training in pediatrics, emergency medicine, or anesthesia and work 13-h shifts (0800–2100 and 2000–0900); two registrars are on duty at all times. An ICU fellow is rostered Monday–Friday from 0800 to 1800 and acts as first-on-call from home (with intensivist backup) one or two weeknights per week but provides no weekend or public holiday cover. Nursing care is generally provided at a ratio of one nurse for each ventilated patient and one nurse for every two nonventilated patients.

Weekdays were defined as Monday to Friday inclusive, with the exception of public holidays (9 days per year). Weekend days were defined as all Saturdays, Sundays, and public holidays. Time periods were defined as follows:

In-hours: 0800–1800 weekdays and 0800–1200 weekends and public holidays; times when an intensivist would invariably be present in the ICU.

After-hours: All other times; intensivist attendance only on an as-needed basis.

All patients admitted to the ICU from January 1, 1997, to December 31, 2006, were analyzed. Nonelective and elective admissions were defined using criteria from the Paediatric Index of Mortality (PIM) [7]. PIM scores and predicted mortality were calculated according to the revised PIM (PIM2) algorithm [7].

Binomial proportions were compared using normal theory methodology. Median lengths of stay were compared using the Mann–Whitney U-test. Multivariate logistic regression was performed using SPSS 11.0.4 (SPSS, Chicago, IL, USA) with outcome as the dependent variable and PIM2 parameters together with time of admission (categorized as either in-hours or after-hours) as covariables.


The ICU admitted 6,980 patients over the study period, including 4,456 nonelective patients. Observed mortality was 248/4,456 = 5.6% compared to a PIM2-predicted mortality of 259.51 = 5.8% [standardized mortality ratio (SMR) = 0.96, 95% confidence interval (CI) = 0.84–1.07]. The predicted mortality did not vary with time of admission (5.64% in-hours vs. 5.90% after-hours, p = 0.521). The ratio between the number of in-hours admissions per hour and the number of after-hours admissions per hour was 0.82, p = 0.21; that is, there was a nonsignificant trend toward nonelective patients being admitted more frequently after hours. However, because the overwhelming majority of elective patients was admitted during in-hours periods, the total number of admissions per hour was 1.70 times higher in-hours than after-hours (p = 0.0007).

The time of admission was entered as a dichotomous variable into a logistic regression model that included all PIM2 variables. All variables were statistically significant with the exception of bypass status (p = 0.272), which was excluded from the model. Only 0.4% of our (nonelective) patients were admitted following cardiopulmonary bypass, which accounts for the lack of statistical association. The time of admission was significant in the regression model, with an odds ratio for death in after-hours admissions of 0.712 (95% CI 0.518–0.980), p = 0.037 (Table 1).
Table 1

Multiple logistic regression for outcome using all Paediatric Index of Mortality (revised version) variables except bypass status and time of admission


Coefficient (SE)

Odds ratio (95% CI)



Pupils fixed to light (Y/N)

4.093 (0.504)

59.949 (22.319–161.020)

< 0.001


High-risk diagnosis (Y/N)

1.719 (0.182)

5.579 (3.908–7.965)

< 0.001


Mechanical ventilation (Y/N)

0.689 (0.182)

1.992 (1.395–2.843)

< 0.001


100 × FIO2/PaO2 mmHg–1

0.342 (0.102)

1.408 (1.153–1.719)



Absolute base excess, mmol/l

0.095 (0.012)

1.100 (1.074–1.127)

< 0.001


Absolute systolic blood pressure 120 (mmHg)

0.013 (0.004)

1.013 (1.006–1.020)

< 0.001


After-hours admission (Y/N)

–0.339 (0.163)

0.712 (0.518–0.980)



Recovery postprocedure (Y/N)

–0.625 (0.288)

0.535 (0.304–0.940)



Low-risk diagnosis (Y/N)

–2.291 (0.454)

0.101 (0.042–0.246)

< 0.001



–4.194 (0.223)


Patients admitted after hours had a slightly shorter median length of stay in the ICU (44.05 h) compared with those admitted during in-hours periods (50.0 h), p = 0.001.


A substantial proportion of the workload in intensive care occurs after hours. Intensive care patients are often most unstable immediately following admission, and it is at this time that hands-on management by a senior clinician might be expected to have maximal benefit [2, 3]. It is, however, reasonable to hypothesize that mortality would not be adversely affected by a staffing model in which the intensivist is on call from home and available within 15 min, provided there is no reluctance on the part of the (junior) in-house staff to call the intensivist and that the intensivist is prepared to attend promptly at any time if required.

In this study, patients admitted to the ICU after hours enjoyed a 29% reduction in the odds of death as well as a slightly decreased length of stay. It is difficult to explain the apparently protective effect of being emergently admitted after hours. PIM2-predicted mortality was not significantly different for in-hours and after-hours admissions, demonstrating equivalent levels of acuity. In-hours nonelective admissions compete for the intensivist's time with a substantial number of elective admissions. In our cohort, the total number of elective and nonelective admissions per hour was 1.7 times greater during the in-hours period. An association between caseload and mortality was previously noted by Arias et al., who identified an odds ratio of death of 1.28 for patients admitted to the ICU in the evening compared with daytime hours, a time when the number of admissions per hour was 28% higher than during the day [8].

This study has several limitations. Ideally, we would directly compare two competing staffing models (24-h in-house coverage versus on-call coverage from home after hours) in two ICUs staffed by similarly trained intensivists and junior staff treating comparable patients. We chose instead to use in-hours patients as a control group for our after-hours patients. However, because the majority of our patients stay at least one night in the ICU, most patients were exposed to a period with no in-house intensivist regardless of their time of admission. If there is a protective effect of having an in-house intensivist that is not limited to the initial hours following admission, our staffing model may lead to increased mortality for all patients, which would not be discernible from our data. However, there was no evidence of excess mortality compared with a widely validated standard (PIM2 SMR = 0.96, 95% CI = 0.84–1.07). It must also be conceded that the results of our study may not apply to ICUs with less generous nurse-to-patient ratios or less experienced in-house junior staff.

Two other studies have examined the effects of time of admission on mortality in the pediatric ICU. Arias et al. found no relationship between mortality and time of admission for nonelective patients admitted to 15 pediatric ICUs [8]. After multivariate analysis, three of nine diagnostic subgroups analyzed (congenital heart disease, cardiac arrest, and shock) demonstrated a significantly higher risk of death for evening but not weekend admissions. Unfortunately, no information regarding after-hours staffing was presented, making it difficult to speculate on any association between mortality and staffing models. Hixson et al., in a study of 5,968 patients admitted to a single ICU over 7.3 years, demonstrated no difference in standardized mortality rates for patients admitted on weekends or evenings to a pediatric ICU staffed by in-house consultants around the clock [3].

Several studies in adult ICUs without 24-h in-house intensivists have not demonstrated any increase in risk-adjusted mortality for patients admitted after hours. Ensminger et al. found no excess mortality in 29,084 patients admitted on weekends to either medical or multispecialty ICUs [9]. In a multicenter trial, Wunsch et al. found no association between mortality and day or time of admission in 56,250 ICU patients after risk adjustment by multivariate regression [10]. Uusaro et al. demonstrated no excess mortality for patients admitted to intensive care after hours for 23,134 emergency admissions to 18 ICUs [11]. Two studies have reported decreased risk-adjusted mortality for after-hours admissions. Morales et al. demonstrated significantly lower risk-adjusted mortality for after-hours admissions in a study of 6,034 patients (13.9% vs. 17.2%) [12], and Luyt et al. reported a risk-adjusted odds ratio for death of 0.93 (0.87–0.99) in patients admitted after hours [13].

Our ICU offers complex therapies, including high-frequency ventilation, extracorporeal membrane oxygenation, inhaled nitric oxide, and hemofiltration, and our case volume exceeds that of the average pediatric ICU in the United States and Europe [14, 15]. Intensivists live within 15 min of the ICU and attend after hours as necessary. Our data suggest there is no demonstrable disadvantage for patients admitted to an ICU with this model of staffing. The combination of relatively experienced junior staff with telephone backup and discretionary attendance from intensivists after hours is an acceptable model for care in the pediatric ICU.

Copyright information

© Springer-Verlag 2007