Intensive Care Medicine

, Volume 30, Issue 10, pp 1908–1913

The longer patients are in hospital before Intensive Care admission the higher their mortality

Authors

    • Department of Anaesthesia and Critical Care MedicineThe Royal National Orthopaedic Hospital
  • Alistair F. McNarry
    • Anaesthetics Unit, The Royal London HospitalBarts and the London NHS Trust
  • Vassilis G. Hadjianastassiou
    • Department of Vascular SurgeryUniversity Hospital, Lewisham
  • Paris P. Tekkis
    • Department of SurgerySt. Mark’s Hospital
Original

DOI: 10.1007/s00134-004-2386-2

Cite this article as:
Goldhill, D.R., McNarry, A.F., Hadjianastassiou, V.G. et al. Intensive Care Med (2004) 30: 1908. doi:10.1007/s00134-004-2386-2

Abstract

Objective

To explore the relationship between hospital mortality and time spent by patients on hospital wards before admission to the intensive care unit (ICU).

Design

Observational study of prospectively collected data.

Setting

Participating intensive care units within the North East Thames Regional Database.

Patients and participants

Patients, 7,190, admitted to ICU from the hospital wards of 24 hospitals.

Interventions

None.

Measurements and results

Of ICU admissions from the wards, 40.1% were in hospital for more than 3 days and 11.7% for more than 15 days. ICU patients who died in hospital were in-patients longer (p=0.001) before admission (median 3 days; interquartile range 1–9) than those discharged alive (median 2 days; interquartile range 1–5). Hospital mortality increased significantly (p<0.0001) in relation to time on hospital wards before ICU: 47.1% (standardised mortality ratio 1.09) for patients in hospital 0–3 days before ICU admission up to 67.2% (standardised mortality ratio 1.39) for patients on the wards for more than 15 days before ICU. Length of stay before ICU admission was an independent predictor of hospital mortality (odds ratio per day 1.019; 95% confidence interval 1.014–1.024). There were significant differences (p<0.001) in patient age, APACHE II score and predicted mortality in relation to time on wards before ICU admission.

Conclusions

Mortality was high among patients admitted from the wards to ICU; many were inpatients for days or weeks before admission. The longer these patients were in hospital before ICU admission, the higher their mortality. Patients with delayed admission differed in some respects compared to those admitted earlier.

Keywords

Hospital outcomeInpatient managementPhysiological abnormalitiesPre-ICU care

Introduction

Events both before and after intensive care unit (ICU) admission are important in determining the outcome of critically ill patients [1]. Although fewer patients are admitted to ICU from the wards than from either the emergency department or the operating theatres/recovery, their percentage mortality following admission is much higher and they make up the biggest group of ICU deaths [1]. Indeed some 20–30% of patients admitted to ICU from the ward have deteriorated to the point where cardiopulmonary resuscitation (CPR) has been required before ICU admission [1]. Furthermore, approximately 25% of those who die following an ICU admission do so after they have survived to be discharged back to a ward [1]. Patients who are discharged prematurely from intensive care to a ward have an increased mortality [2, 3]. There is therefore good evidence that care delivered on hospital wards is relevant to the outcome of critically ill patients.

Intensive Care Outreach Services (ICORS) were introduced in the United Kingdom in 2001 partly to detect ward patients who are becoming critically ill. Evidence has been gathered by these ICORS that critically ill patients are often in hospital for several days before they are admitted to ICU or die [4, 5]. In view of the time span between hospital admission and critical care provision and the possibility of earlier intervention, we sought to investigate whether prolonged ward-based care impacted upon the hospital mortality (death while in ICU or in hospital after ICU discharge) of patients who were admitted to the ICU. The North Thames Region has amassed a collaborative ICU database that was analysed to address this.

Materials and methods

Ethics committee approval was given for this database analysis. The methods and procedures of the data collection, validation and maintenance of the database have been previously described [6]. The information collected for each patient includes demographic data, diagnostic details, physiological values on ICU admission and Acute Physiology and Chronic Health Evaluation (APACHE) II scores from the first 24 h after ICU admission. All patients admitted to ICUs between April 1992 and the end of December 2000 were selected from the Intensive Care Database. The data was contributed by 24 ICUs located in north-east London and Essex consisting of five university/teaching hospitals and 19 community/district general hospitals. All the ICUs admitted both surgical and medical patients. Although they are not necessarily representative, they are a typical group of British hospitals, most with emergency departments. If a patient had more than one admission to the ICU during their hospital stay, only the first admission was considered. We excluded all patients undergoing cardiac surgery, aged less than 16 years or where hospital outcomes were unknown. Patients were not excluded from analysis on the basis of the time they spent on the ICU before discharge or death.

The data were analysed for patients admitted to ICU from a hospital ward. The number of days in hospital before ICU was calculated by subtracting the hospital admission date from the ICU admission date. The patients were grouped by length of stay before ICU admission to allow for presentation and analysis as follows: days 0 (admitted to ICU on day of hospital admission) to 3, 4 to 6, 7 to 9, 10 to 12, 13 to 15 and 16 days or more. The mortality predicted by APACHE II [7] and the standardised mortality ratio (SMR—observed/predicted hospital mortality) were calculated. Confidence intervals for the SMRs were obtained by calculating the 95% confidence interval of observed mortality by assuming it is a Poisson variable and dividing these numbers by the predicted mortality [8]. The acute physiology score (APS) was defined as the APACHE II points derived from the physiological parameters (that is the APACHE II score minus points for chronic health and age). Admissions were classified by primary admission diagnosis into seven groups: haematological, metabolic, renal, gastrointestinal, neurological, cardiovascular and respiratory.

The statistical analysis was performed using “Statistical Package for the Social Sciences” version 9 for Windows (SPSS, Chicago, Illinois, USA) and Intercooled STATA 6.0 for Windows (STATA, College Station, Texas, USA). This focused on differences in physiology and outcome in relation to time in hospital before ICU admission. Clinically relevant factors that may have been associated with in-hospital mortality were assessed using univariate logistic regression analysis to identify any statistical relation to in-hospital mortality. Categorical variables were compared using the Mann Whitney U test, the chi-squared test for trend and the Jonckheere-Terpstra tests where appropriate. The Jonckheere-Terpstra test is similar to the Kruskal-Wallis H test in testing whether several independent samples are from the same population. The Kruskal-Wallis H test assumes that there is no a priori ordering of the populations from which the samples are drawn. When there is a natural a priori ordering of the populations, the Jonckheere-Terpstra test is more powerful.

All variables whose univariate test had a p value less than 0.05 were included in the multivariate analysis to assess independent predictors of mortality. As age, APS and chronic health status are the components of the APACHE II score, this was not entered as a separate variable. A backward method of selection was used in multiple logistic regression analysis, with in-hospital mortality as the dependent variable. Multiple logistic regression was used to provide adjusted odds ratios and 95% confidence intervals for each independent risk factor. The Jonckheere-Terpstra test was used to identify whether the APACHE II points awarded to the individual components of the acute physiology score at ICU admission changed significantly as in-patient stay before ICU admission increased. Chi-squared and the Jonckheere-Terpstra tests were used to examine differences in data between the study years.

Results

The database contained details of 50,837 patients admitted to ICUs between April 1992 and December 2000 with complete data and known ICU outcome. Excluding patients less than 16 years of age left 49,047, of whom 41,107 were first admissions to ICU. Of these, 7,266 were admitted from a hospital ward: 76 patients did not have a known hospital status (dead or alive) and were, thus, excluded from the analysis, leaving 7,190 patients for the study. The median number of patients from the units was 241, range 45–917. The five university/teaching hospitals contributed 22.4% of the cases.

There were significant differences between patients admitted to ICU who died in hospital and those who survived to hospital discharge (Table 1). As shown in Table 2, age, APS, APACHE II score, APACHE II-predicted risk of death (ROD) and ICU stay differed significantly with time in hospital before ICU admission. Analysis of the points scored by physiological variables at ICU admission found that respiratory rate (p=0.04), oxygenation (p=0.037), pH (p=0.026), sodium (p<0.001), potassium (p=0.007), creatinine (p=0.004), haemoglobin concentration (p<0.001) and white cell count (p<0.001) were the parameters with significant differences over time before ICU admission. There was no significant difference between the main ICU admission diagnostic categories over the time periods. There was no significant difference between the years over which the study data was collected in terms of hospital mortality and the number of days in hospital after ICU admission. There was, however, a difference in several variables including ICU mortality (p=0.016), which was lowest at 37.6% in 1993 and highest at 44.3% in 2000. Severity of illness (APS, APACHE and ROD) also increased over the years whereas the percentage admitted after CPR fell from 25.3% in 1992 to 19.0% in 2000. There was a suggestion that overall hospital care was improving as SMRs generally fell over this period, being 1.18/1.20/1.30 in the years 1992/1993/1994 and 1.07/1.08/1.05 in the years 1998/1999/2000 (please see electronic supplementary material).
Table 1

Patient data (at ICU admission) in relation to status at hospital discharge

Status at hospital discharge

Alive

Dead

p value

Stay before ICU, median, (IQR), meana

2 (1–5), 5

3 (1–9), 8

<0.001

Male, number (%)

1849 (53.8%)

2081 (55.5%)

0.16

Age in years, median (IQR)

57 (38–70)

67 (55–75)

<0.001

APACHE II, median (IQR)

16 (10–23)

27 (21–34)

<0.001

APS, median (IQR)

13 (7–19)

22 (16–29)

<0.001

GCS, median (IQR)

15 (10–15)

9 (3–15)

<0.001

CPR before ICU, number (%)

331 (9.9%)

1237 (33.6%)

<0.001

ICU stay, median days (IQR)

3 (1–7)

2 (1–7)

<0.001

Predicted mortality, median (IQR)

0.21 (0.09–0.43)

0.64 (0.37–0.85)

<0.001

Stay before ICU days on ward before ICU admission, IQR interquartile range, APS acute physiology score, GCS Glasgow Coma Score, CPR cardiopulmonary resuscitation before ICU admission

A Mann Whitney U test was used for all the above variables

Values for APACHE II, APS and GCS are the worst values from the 24 h following ICU admission

aIn addition to median stay before ICU the average is also shown—this figure has not been used for statistical analysis

Table 2

Data grouped by number of days patients spent in hospital before ICU admission

Stay before ICU (days)

0–3

4–6

7–9

10–12

13–15

≥16

All

Missing

Significance

Number (%)

4304 (59.9)

934 (13.0)

515 (7.2)

353 (4.9)

238 (3.3)

844 (11.7)

7188

2

Age in years, median (IQR)

63 (45–72)

65 (49–74)

65 (52–74)

65 (52–74)

66 (52–74)

62 (47–72)

63 (47–73)

2

p<0.001

Male number (%)

2307 (53.6)

535 (57.3)

289 (56.1)

206 (58.4)

142 (59.7)

451 (53.4)

3930 (54.7)

4

p=0.33a

APS, median (IQR)

17 (10–25)

17 (10–25)

17 (11–25)

20 (13–27)

18 (11–28)

23 (16–31)

17 (10–25)

62

p<0.001

APACHE II, median (IQR)

21 (13–30)

21 (14–29)

22 (16–31)

23 (17–33)

23 (16–32)

23 (16–31)

22 (14–30)

62

p<0.001

ROD, median (IQR)

0.39 (0.15–0.72)

0.42 (0.17–0.72)

0.44 (0.20–0.74)

0.53 (0.23–0.81)

0.46 (0.22–0.80)

0.46 (0.23–0.73)

0.42 (0.17–0.73)

62

p<0.001

GCS, median (IQR)

14 (4–15)

14 (5–15)

13 (4–15)

13 (3–15)

13 (3–15)

14 (6–15)

14 (4–15)

62

p=0.89

CPR Number (%)

929 (22.0)

206 (22.5)

116 (23.0)

91 (26.2)

48 (20.7)

176 (21.4)

1566 (22.2)

151

p=0.98a

ICU stay days, median (IQR)

2 (1–7)

2 (1–6)

3 (1–8)

2 (1–8)

3 (1–6)

3 (1–8)

2 (1–7)

0

p=0.004

IQR interquartile range, Stay before ICU days in hospital before ICU admission, APS acute physiology score, ROD predicted mortality by APACHE II, GCS Glasgow Coma Score, CPR cardiopulmonary resuscitation before ICU admission, ICU stay time in ICU, Missing number with missing information, Significance significance between periods of stay before ICU with Jonckheere-Terpstra test: aexcept for Male and CPR rows where chi-squared test for trend was used

Table 3 details the results of multivariate logistic regression analysis, showing that the length of stay on a ward prior to admission to ICU is an independent predictor of mortality even when adjusted for other factors (as shown). The performance of CPR within 24 h of admission is the strongest predictor of outcome. Gender was not found to be statistically significant (p=0.16) with respect to mortality on univariate analysis and, thus, was not included in the multivariate analysis (please see electronic supplementary material).
Table 3

Factors shown by multivariate logistic regression to be independent predictors of hospital mortality

Coefficient β

Standard error

Odds ratio

95% Confidence interval

Length of stay before ICU (per day)

0.0175

0.0024

1.0177

1.0129–1.0225

Age (per year)

0.0296

0.0017

1.0300

1.0267–1.0334

APS (per point)

0.1193

0.0049

1.1267

1.1160–1.1375

CPR

0.7996

0.0803

2.2246

1.9008–2.6037

GCS (per point)

−0.0532

0.0085

0.9481

0.9325–0.9640

Chronic health (per point)

0.2258

0.240

1.2533

1.1957–1.3137

APS acute physiology score, CPR cardiopulmonary resuscitation, GCS Glasgow Coma Score, Chronic health chronic health points awarded by APACHE II

Figure 1 shows ICU and hospital mortality grouped by time in hospital before ICU admission. The percentage admissions in each time period and the SMR (standardised mortality ratio = observed/predicted hospital mortality) are also shown. There was a highly significant linear increase in hospital mortality with increased time on the wards before ICU admission (p<0.0001 chi-squared for trend, chi-squared for departure from linearity =0.7447).
Fig. 1

Hospital mortality (in ICU and after ICU) in relation to time in hospital before ICU admission. Black bars (ICU death) death during first ICU admission, white bars (after ICU) death during same hospital admission but after discharge from first ICU admission. aSMR standardised mortality ratio (observed/predicted hospital mortality by APACHE II), b−95% C.I. and +95% C.I. 95% confidence intervals for SMR, c%deaths after ICU absolute values for the white bars, dfigures above each column the total in hospital mortality for that group

Discussion

Models, such as APACHE II [7], which predict outcome for ICU patients are commonly based upon physiological abnormalities, age, chronic health and diagnosis. Our study went further to examine the importance of being on a hospital ward before ICU admission. We show that some 40% of the patients who are admitted to ICU from hospital wards are in hospital for more than 3 days before ICU, and nearly 13% (912/7,190) are in hospital for more than 2 weeks before ICU admission. Hospital mortality increases with a longer stay in hospital before ICU admission, and this is true even if mortality is case-adjusted. We can only speculate why patients might have been in hospital for days or weeks before ICU admission. It is possible that they deteriorated while on the ward and were admitted to ICU at the earliest possible time. It is also possible that some patients had limited life expectancy and quality of life and that their physicians were reluctant to consider critical care support at an early stage. Eventually, ICU care was considered necessary and appropriate. The relevant questions are whether a prolonged stay on the ward before ICU was harmful and if decisions about ICU admission could have been taken earlier.

Over 20% of ICU admissions received CPR on a ward before ICU. These patients survived long enough after their arrest to be admitted to ICU. The mortality of this subgroup of patients was 79% and they constituted one third of all ICU deaths. Several studies report that the majority of ward patients who suffer a cardiac arrest have recorded physiological abnormalities in the hours preceding the arrest [9, 10, 11, 12]; it is desirable to prevent these arrests if possible. Patients admitted from a ward to ICU are also commonly found to have physiological abnormalities recorded before admission [13]. A point prevalence study of in-patients in our own hospital found that some 11% of those outside of the ICU had three or more physiological abnormalities [5]. This was associated with a 30-day hospital mortality of 21.3%. In addition, a further 20.1% had two abnormalities and their 30-day mortality was 9.2%. There was only one death within 30 days among those patients who had no recorded abnormalities, and this death occurred 21 days after the study.

Medical emergency teams (METs), responding primarily to patients with profoundly deranged physiological values, have been shown to be effective in decreasing the incidence of cardiac arrest and its associated mortality [14, 15]. ICORS have been widely introduced over the last few years in England and some other parts of the United Kingdom. Evidence is accumulating that they can be effective in identifying and supporting ward patients who are, or who may become, critically ill [16].

There is circumstantial evidence to suggest mortality is increased for patients cared for in a location that is unable to support their critical care needs [5]. Although critical care provision (staffed beds) has increased in the United Kingdom, it is still inadequate and remains less than that available in many other countries. This may account for the worse outcome seen in some groups of patients [17] as well as influencing when patients were considered for ICU admission.

In order for medical staff to be able to intervene effectively, patients must be accessible and there must be time to assess the patient and alter their management. Our analysis focuses on hospital in-patients who are accessible. Our data suggest that a prolonged stay in hospital before ICU admission may be detrimental. Many of these long-stay patients will have profoundly deranged physiological values before ICU admission. We do not know how long their physiology was abnormal but experience suggests it could be for hours to days before eventual ICU admission. Physiologically-based early warning scores may be one important method of identifying these patients [5]. Once patients are identified then early measures may be able to improve outcome [14, 15, 16].

We are not aware that any of the ICUs had unusual admission or discharge policies. Median ICU stays were short for those who survived but were even shorter for those that died. This may reflect British ICU care, where there is great pressure on scarce beds. It may also suggest that by the time patients were admitted there may have been little that could be done for many of them. The changes in our data over time suggest that patients on ICU admission from the wards were becoming sicker. However, the decrease in the percentage requiring CPR before ICU and the fall in SMRs suggest that management in some areas was improving. Our database did not contain details allowing identification of patients who merited ICU admission but, because of bed shortages or inappropriate clinical decisions, were denied access. We cannot comment on the appropriateness of ICU admission when a decision to admit was eventually taken, but we have previously highlighted some of the dilemmas associated with admission decisions when the prognosis is very poor [18].

Early identification allows time to discuss treatment limitation with the patient, their carers and the medical team—something that is not possible if the patient has a cardiopulmonary arrest. A delay before ICU admission may result in a prolonged ICU stay, exacerbating the bed shortage. Although a patient who dies quickly on a ward will cost the hospital less than one who is admitted to ICU and dies later, there is preliminary data to suggest that early intervention may improve survival and decrease rates of readmission to ICU [16]. Sub-optimal treatment also exposes the hospital to the risk of litigation and the costs of further treatment for the prolonged effects of major morbidity. There can be little argument against attempting to provide patients with appropriate treatment at the earliest possible time in the right location.

In summary, our data suggests that a considerable percentage of patients admitted to ICUs from hospital wards have been in-patients for days or weeks before admission. The longer these patients are in hospital before ICU admission, the higher their hospital mortality. There is evidence to suggest that the degree of physiological derangement may be worse in those with longer ward stays before admission. The patients who receive a longer period of ward-based care prior to ICU admission may comprise a different population from those admitted earlier in their hospital stay. It should be possible to identify patients who are ‘at risk’, and the use of physiologically based early warning scores may be one way of doing this. When validated, these scores may provide a method of studying the effect of delays in ICU admission when critical illness is identified. There is some evidence that early intervention is beneficial. Early recognition and intervention have the potential to improve outcome for these patients.

Supplementary material

supp.pdf (21 kb)
(PDF 22 KB)

Copyright information

© Springer-Verlag 2004