Intensive Care Medicine

, Volume 40, Issue 9, pp 1275–1284

Trends in admission prevalence, illness severity and survival of haematological patients treated in Dutch intensive care units

Authors

    • Department of HaematologyRadboud University Medical Center
  • Ilona W. M. Verburg
    • Department of Medical Informatics, Academic Medical CenterUniversity of Amsterdam
  • Mark van den Boogaard
    • Department of Intensive Care MedicineRadboud University Medical Center
  • Nicolette F. de Keizer
    • Department of Medical Informatics, Academic Medical CenterUniversity of Amsterdam
  • Niels Peek
    • Department of Medical Informatics, Academic Medical CenterUniversity of Amsterdam
  • Nicole M. A. Blijlevens
    • Department of HaematologyRadboud University Medical Center
  • Peter Pickkers
    • Department of Intensive Care MedicineRadboud University Medical Center
Original

DOI: 10.1007/s00134-014-3373-x

Cite this article as:
van Vliet, M., Verburg, I.W.M., van den Boogaard, M. et al. Intensive Care Med (2014) 40: 1275. doi:10.1007/s00134-014-3373-x
  • 520 Views

Abstract

Purpose

To explore trends over time in admission prevalence and (risk-adjusted) mortality of critically ill haematological patients and compare these trends to those of several subgroups of patients admitted to the medical intensive care unit (medical ICU patients).

Methods

A total of 1,741 haematological and 60,954 non-haematological patients admitted to the medical ICU were analysed. Trends over time and differences between two subgroups of haematological medical ICU patients and four subgroups of non-haematological medical ICU patients were assessed, as well as the influence of leukocytopenia.

Results

The proportion of haematological patients among all medical ICU patients increased over time [odds ratio (OR) 1.06; 95 % confidence interval (CI) 1.03–1.10 per year; p < 0.001]. Risk-adjusted mortality was significantly higher for haematological patients admitted to the ICU with white blood cell (WBC) counts of <1.0 × 109/L (47 %; 95 % CI 41–54 %) and ≥1.0 × 109/L (45 %; 95 % CI 42–49 %), respectively, than for patients admitted with chronic heart failure (27 %; 95 % CI 26–28 %) and with chronic liver cirrhosis (38 %; 95 % CI 35–42 %), but was not significantly different from patients admitted with solid tumours (40 %; 95 % CI 36–45 %). Over the years, the risk-adjusted hospital mortality rate significantly decreased in both the haematological and non-haematological group with an OR of 0.93 (95 % CI 0.92–0.95) per year. After correction for case-mix using the APACHE-II score (with WBC omitted), a WBC <1.0 × 109/L was not a predictor of mortality in haematological patients (OR 0.86; 95 % CI 0.46–1.64; p = 0.65). We found no case–volume effect on mortality for haematological ICU patients.

Conclusions

An increasing number of haematological patients are being admitted to Dutch ICUs. While mortality is significantly higher in this group of medical ICU patients than in subgroups of non-haematological ones, the former show a similar decrease in raw and risk-adjusted mortality rate over time, while leukocytopenia is not a predictor of mortality. These results suggest that haematological ICU patients have benefitted from improved intensive care support during the last decade.

Keywords

ICU outcomeHaematological malignancyProspective intensive care evaluationIllness severity

Introduction

Treatment for haematological malignancies is known to cause serious toxicity resulting in prolonged immune insufficiency and impaired mucocutaneous barriers [1]. Patients undergoing these intensive treatments are prone to infections and frequently require intensive care unit (ICU) monitoring and/or treatment. The most common indication for ICU admission is respiratory insufficiency [2]. Recent studies show that 15–28 % of recipients of induction chemotherapy for acute myeloid leukaemia need to be admitted to the ICU during treatment [3, 4]. For haematopoietic stem cell transplant recipients, ICU admission rates vary from 5 to 55 % [5].

Extremely high mortality rates underlie the reticent attitude among both haematologists and intensivists to unplanned transfers of haematological patients to an ICU [6]. This controversial attitude towards ICU admission is comparable to that of clinicians with regards to patients with solid tumours, heart failure or liver cirrhosis. During the last decade perspectives on critically ill cancer patients have changed [7] and led, together with improved survival rates of patients admitted to general ICUs [8], to less reluctance on the behalf of haematologists and intensivists to admit patients with haematological malignancies to an ICU [9, 10]. Where formerly neutropenic patients who developed organ failure were initially considered to have a dismal prognosis, nowadays the notion that neutropenia is predictive for ICU mortality is the subject of ongoing debate [11]. Recent studies have found that predominantly the need for mechanical ventilation, presence of invasive fungal infection, development of multi-organ failure and high severity of illness scores at admission are additional prognostic factors for mortality among cancer patients [2, 12].

We investigated whether these new perspectives on ICU survival actually do influence the admission prevalence and outcome of haematological patients admitted to the ICU. Therefore, the aim of this study was to explore trends over time in admission prevalence and outcomes of critically ill haematological patients and to compare these data with those of non-haematological patients admitted to the medical ICU (medical ICU patients), specifically with patients with chronic heart failure, chronic liver cirrhosis and solid tumours.

Design and methods

Patient data

All consecutive patients admitted to ICUs participating in the Dutch National Intensive Care Evaluation (NICE) database between January 2004 and January 2012 were included in this study. To diminish bias by changes in case-mix, only ICUs that submitted patient data during the complete period were included. Admissions were considered to be due to a haematological disorder(s) when a haematological malignancy was recorded in the NICE or when ICU admission was requested by a haematologist. Patients for whom the recorded APACHE-II (Acute Physiology and Chronic Health Evaluation II) score at admission made it impossible for the disorder to be related to a haematological disease were considered to be non-haematological patients. Patients admitted to the ICU for non-medical reasons (e.g. surgery, trauma) were excluded from entry into the study, as were patients admitted for diagnostic procedures generally performed in the ICU (e.g. bronchoalveolar lavage) and those with an ICU length of stay (LOS) of <24 h. Those patients who were readmitted to the ICU and patients with an unknown leukocyte count were also excluded.

The NICE registry contains information on patient admissions to >90 % of all ICUs in the Netherlands. Data from the first 24 h of ICU admission are collected, including information on acute and chronic diagnoses, mechanical ventilation, minimum and maximum physiology and laboratory values, such as number of leukocytes, and need to calculate the APACHE II score [13]. In this study, the lowest white blood cell (WBC) count was used to define subgroups of haematological patients, with WBC <1.0 × 109/L and WBC ≥1.0 × 109/L, because leukocytopenia is considered to negatively affect ICU outcome. No information was available on the ICU admission policies of the participating hospitals nor on specific haematological parameters, such as level of neutropenia, underlying haematological diagnosis, disease status and type and phase of haematological treatment, including whether or not haematopoietic stem cell transplant was performed.

We compared the haematological subgroup with other subgroups of severely ill patients for whom ICU admission is traditionally considered to be controversial, namely, those with solid tumours, chronic liver cirrhosis (defined as positive liver biopsy in combination with documented portal hypertension odds ratio OR earlier episodes of gastrointestinal bleeding caused by portal hypertension OR earlier episodes of liver failure, coma or encephalopathy) and chronic heart failure (New York Heart Association Class IV).

The NICE registry further records outcome data on LOS in the ICU and hospital, respectively, ICU readmission rate and ICU and in-hospital mortality. A number of measures have been implemented to improve data quality: the collected data are subject to several quality checks, onsite data quality audits take place on a regular basis and data collectors participate in training sessions [14]. Site visits to ICUs are performed on a regular basis to investigate the number, types and causes of errors made in in-hospital mortality registration [15].

This study was carried out in the Netherlands in accordance with current regulations pertaining to review by research ethics committees and patient informed consent.

Data analysis

We compared APACHE II scores, incidence of comorbidities [diabetes, chronic obstructive pulmonary disease (COPD), cardiovascular insufficiency, acute renal failure], the use of procedures/treatments in the ICU (mechanical ventilation, vasopressive medication) and LOS between two subgroups of haematological patients (WBC <1.0 × 109/L and WBC ≥1.0 × 109/L) and four subgroups of non-haematological medical ICU patients (solid tumours, chronic liver cirrhosis, chronic heart failure and other medical ICU admissions). APACHE-II scores were compared both with and without including the subscore for WBC. LOS is separately reported for survivors and non-survivors, and for both stay in the ICU and hospital, respectively.

Absolute and risk-adjusted hospital mortality rates were compared between haematological and non-haematological patients. The risk-adjusted mortality rate (RAMR) was calculated as the fraction of the number of deaths actually observed and the number of deaths predicted by the APACHE-II model, multiplied by the average hospital mortality rate in the entire study cohort. It can be interpreted as the hospital mortality rate after correction for case-mix, using the APACHE-II model. The RAMR has the same statistical properties as the standardised mortality ratio but allows for a better comparison of adjusted mortality rates between studies.

Three logistic regression analyses were carried out. To analyse the prevalence of haematological patients admitted to the ICU among medical ICU admissions over time, we used the mixed model logistic regression analysis for the entire cohort using haematological status (yes vs. no) as the response variable and calendar time as the covariate. To account for clustering in ICUs, we used a random coefficient model that allowed each ICU to have its own trend in prevalence over time. For comparison, trends over time were also analysed in the same way as for medical ICU patients with chronic liver cirrhosis, chronic heart failure and solid tumours.

To analyse whether trends in risk-adjusted hospital mortality over time differed between haematological and non-haematological patients, we used fixed-effects logistic regression analysis with hospital mortality as the response variable and the logit-transformed APACHE II mortality risk score (to adjust for patient case-mix), calendar time and haematological ICU admission (yes vs. no) as covariates. We also included an interaction between calendar time and admission of haematological patients to the ICU. For comparison, trends in risk-adjusted hospital mortality over time were similarly analysed for subgroups of patients with chronic liver cirrhosis, chronic heart failure and patients with a solid tumours, but without inclusion of ‘haematological ICU admission’ and its interaction with calendar time as covariates.

Finally, a similar fixed-effect logistic regression analysis was conducted within the group of haematological patients to assess the association between leukocytopenia at ICU admission and risk-adjusted hospital mortality. In this analysis, the logit-transformed APACHE II mortality risk, calendar time, leukocytopenia (modelled as WBC <1.0 × 109/L) and an interaction between calendar time and leukocytopenia were included in the model. Furthermore, to correct for volume–outcome effects, we added the number of haematological ICU admissions that took place in the same centre and same month as the ICU admission in question, as the continuous covariate to this model.

All statistical analyses were performed using the Statistical Package for Social Sciences (SPSS) version 20.01 (SPSS Inc., Chicago, IL) and R, version 2.14 [16].

Results

Patients

A total of 288,568 patients admitted to 36 ICUs participating in the Dutch NICE during the entire study period were included in our study, of which 2,935 admissions were identified as haematological patients. Following exclusion of those patients who did not fulfill our criteria (flowchart shown in Fig. 1), data of 1,741 patients with a haematological disease admitted for the first time to the ICU were included in the analyses: 1,673 of these patients (96 %) were registered in the NICE as having a haematological malignancy, and 68 (4 %) were admitted to the ICU admission upon request of the haematologist. Data on a group of 60,954 patients without a haematological malignancy admitted to the medical ICU were available for further analysis. The median [interquartile range (IQR)] WBC for haematological and non-haematological patients was 5.8 × 109/L [IQR 1.1–12.7] and 10.6 × 109/L [IQR 7.6–14.6], respectively.
https://static-content.springer.com/image/art%3A10.1007%2Fs00134-014-3373-x/MediaObjects/134_2014_3373_Fig1_HTML.gif
Fig. 1

Flowchart illustrating the inclusion and exclusion criteria for entry into the study and for classification into haematological and non-haematological patient subgroups. WBC White blood cells, ICU intensive care unit

Among all patients admitted to the medical ICU, 2.8 % were due to haematological conditions, and this prevalence increased over time at 6 % per year [odds ratio (OR) 1.06 annually; 95 % confidence interval (CI) 1.03–1.10]. The number of patients with chronic liver cirrhosis admitted to the ICU (n = 1,264; 2.0 % of all medical admissions) also increased over time (OR 1.06; 95 % CI 1.03–1.09), as did the number of admissions of patients with solid tumours (n = 737, 1.1 %; OR 1.16; 95 % CI 1.10–1.22). The number of patients with chronic heart failure admitted to the ICU (n = 3,672; 5.9 %) decreased over time (OR 0.95, 95 % CI 0.91–0.99).

Demographic and clinical characteristics of included patients, categorised according to subgroup (n = 6), are shown in Table 1. APACHE-II scores at ICU admission (with and without omission of the WBC value) were 28.2 ± 7.7 and 24.6 ± 7.7 points, respectively, and generally higher for haematological patients with WBC <1.0 × 109/L than for the other subgroups of patients. For the haematological subgroup, APACHE-II scores did not change over time, illustrating that on average the severity of illness at admission to the ICU did not change during the study period. Haematological patients, as well as patients with liver cirrhosis and heart failure, were more likely to suffer from acute renal failure, while a medical history of COPD and diabetes was more frequently present in the non-haematological patient groups (Table 1).
Table 1

Characteristics of patients’ subgroups

Characteristics

Haematological patient subgroups

Non-haematological patient subgroups

Haematology WBC < 1.0 × 109/L (n = 405)

Haematology WBC ≥ 1.0 × 109/L (n = 1,336)

Solid tumours (n = 602)a

Chronic liver cirrhosis (n = 1,246)a

Chronic heart failure (n = 3,628)a

Medical ICU, other (n = 55,533)

Demographics

  Age, median (IQR)

58 (47–65)

64 (54–72)

67 (58–74)

58 (50–65)

72 (63–78)

64 (51–75)

  Male gender (%)

60.5

62.0

63.3

64.3

63.7

57.3

  WBC, median [IQR]b

0.2 (0.1–0.5)

7.1 (3.2–13.6)

10.8 (7.3–15.4)

8.5 (5.6–13.1)

11.2 (8.3–14.9)

10.6 (7.6–14.6)

  APACHE-II severity of illness score(points)—with WBC (mean ± SD)

28.2 ± 7.7

25.7 ± 8.1

20.0 ± 7.2

25.6 ± 7.7

25.9 ± 7.6

19.9 ± 8.1

  APACHE-II severity of illness score (points)—without WBC (mean ± SD)

24.6 ± 7.7

23.0 ± 7.6

19.2 ± 8.1

21.0 ± 8.4

21.9 ± 8.1

17.9 ± 8.2

Chronic co-morbidity (%)

  COPDc

2.2

5.5

14.6

9.1

15.1

10.8

  Renal insufficiency

4.0

8.9

4.8

7.6

19.1

6.1

  Cardiovascular insufficiency

1.7

2.8

2.2

3.2

100.0

0.0e

  Diabetes

9.1

10.6

11.6

15.2

18.9

11.8

  Chronic dialysis

8.9

2.2

0.3

0.7

3.5

1.7

ICU treatment (%)b

  Acute renal failure

20.7

19.0

11.0

26.0

21.0

12.0

  Vasopressors

56.0

55.8

66.6

52.5

65.8

46.8

  Mechanical ventilation

69.4

66.8

44.4

62.4

65.0

60.2

 LOS ICU (days), median (IQR)

  Survivorsd

5.1 (2.1–10.7)

4.1 (2.0–9.5)

3.0 (1.8–5.9)

2.9 (1.7–6.6)

3.3 (1.9–6.6)

3.2 (1.8–7.0)

  Non-survivors

5.7 (2.4–11.2)

5.1 (2.4–11.0)

3.1 (1.8–6.8)

4.3 (2.0–9.3)

3.9 (2.0–8.0)

4.2 (2.2–9.2)

LOS in hospital (days), median (IQR)

  Survivorsd

36.5 (21.3–54.0)

22.0 (12.8–41.0)

16.0 (8.0–27.0)

15.0 (7.0–29.0)

16.0 (9.0–29.0)

15.0 (8.0–28.0)

  Non-survivors

23.3 (10.0–38.8)

15.0 (7.0–29.4)

7.0 (4.0–14.0)

10.0 (4.4–17.0)

8.0 (3.9–18.7)

8.0 (3.6–18.0)

Readmission rate (%)

14.3

12.4

5.8

7.9

9.3

7.0

IQR, Interquartile range; APACHE-II, Acute Physiology and Chronic Health Evaluation II; WBC, white blood cells; SD, standard deviation; LOS, length of stay

aOverlapping subgroups (55 patients had >1 subgroup diagnosis)

bDuring the first 24 h of intensive care unit (ICU) admission

cChronic obstructive pulmonary disease (COPD) registered from 2007 onwards

dLOS was calculated for hospital survivors

eAll medical ICU patients with cardiovascular insufficiency were included in the subgroup chronic heart failure

ICU treatment and outcome

Haematological patients and patients with liver cirrhosis were more likely to need vasopressor therapy than the other medical ICU patients, but less likely than patients with solid tumours and chronic heart failure. Mechanical ventilation was less frequently needed for patients with solid tumours (Table 1).

Haematological patients had longer ICU and hospital LOS than non-haematological patients (both for survivors and non-survivors). For haematological patients, the longest LOS in the ICU and hospital was found for patients with leukocytopenia (WBC <1.0 × 109/L) (Table 1). ICU readmission rates were higher in both haematological subgroups compared to non-haematological subgroups (Table 1).

Absolute ICU (33.8 %) and hospital (47.4 %) mortality was higher for the whole group of haematological patients than for the whole group of non-haematological patients (ICU mortality 17.9 %, hospital mortality 26.3 %) and most specific for patients with leukocytopenia (WBC <1.0 × 109/L), with a mortality of 40.0 and 52.3 %, respectively (Table 2).
Table 2

Raw and risk-adjusted mortality rates for patients’ subgroups

Mortality rates

Haematological patient subgroups

Non-haematological patient subgroups

Haematology WBC < 1.0 × 109/L (n = 405)

Haematology WBC ≥ 1.0 × 109/L (n = 1,336)

Solid tumours (n = 602)a

Chronic liver cirrhosis (n = 1,246)a

Chronic heart failure (n = 3,628)a

Medical ICU, other (n = 55,533)

ICU mortality (%)

40.0

31.9

32.4

32.6

25.5

17.7

Hospital mortality (%)

52.3

45.9

44.9

46.7

37.6

24.9

RAMRb (%, 95 % CI)

47.2 (41.0–54.0)

45.1 (41.6–48.8)

40.4 (36.2–45.1)

38.5 (35.4–41.7)

26.9 (25.5–28.4)

17.4 (17.2–17.8)

CI, Confidence interval

aOverlapping subgroups (55 patients had > 1 subgroup diagnosis)

bThe risk-adjusted mortality rate (RAMR) was used to analyse the trend in time of hospital mortality between leukocyte strata. The RAMR was calculated as the fraction of deaths observed and deaths predicted by the APACHE II model, multiplied by the average hospital mortality rate in the entire dataset

RAMR was higher for the whole group of haematological patients admitted to the ICU (RAMR 46 %; 95 % CI 42–49 %) than for the whole group of non-haematological medical ICU patients (RAMR 18 %; 95 % CI 18–18 %). Comparison of the subgroups revealed that risk-adjusted mortality was higher for haematological medical ICU patients with WBC <1.0 × 109/L (47 %; 95 % CI 41–54 %) and with WBC ≥1.0 × 109/L (45 %; 95 %CI 42–49 %), respectively, than for non-haematological patients with chronic heart failure (RAMR 27 %; 95 % CI 26–28 %) or with chronic liver cirrhosis (38 %; 95 %CI 35–42 %), but it was not significantly different from that of patients with solid tumours (RAMR 40 %; 95 % CI 36–45 %; Table 2). The Kaplan–Meier plot for hospital survival of the six subgroups is shown in Fig. 2. Differences in absolute and risk-adjusted hospital mortality per subgroup for patients receiving mechanical ventilation or not in the first 24 h of ICU admission is reflected in Fig. 3.
https://static-content.springer.com/image/art%3A10.1007%2Fs00134-014-3373-x/MediaObjects/134_2014_3373_Fig2_HTML.gif
Fig. 2

Ninety-day Kaplan–Meier plot for hospital survival for haematological and non-haematological medical patients admitted to the ICU

https://static-content.springer.com/image/art%3A10.1007%2Fs00134-014-3373-x/MediaObjects/134_2014_3373_Fig3_HTML.gif
Fig. 3

Absolute and risk-adjusted hospital mortality rates for subgroups of haematological and non-haematological medical patients admitted to the ICU without and with mechanical ventilation during the first 24 h. CI Confidence interval

For both haematological ICU patients and medical ICU patients, risk-adjusted mortality decreased over time, at 7 % per calendar year (OR 0.93; 95 % CI 0.92–0.95) (Fig. 4). This decrease over time was similar in both groups (p = 0.79). Risk-adjusted mortality also decreased over time for patients with chronic heart failure (OR 0.93; 95 % CI 0.88–0.98), patients with chronic liver cirrhosis (OR 0.90; 95 % CI 0.82–0.98) and patients with solid tumours (OR 0.93; 95 % CI 0.75–0.99). There was no significant difference in decrease over time for each subgroup compared to patients admitted to the ICU for haematological conditions. After correction for case-mix using the APACHE-II score (with WBC omitted), leukocytopenia (WBC <1.0 × 109/L) was not a predictor of mortality (OR 0.86; 95 % CI 0.46–1.64).
https://static-content.springer.com/image/art%3A10.1007%2Fs00134-014-3373-x/MediaObjects/134_2014_3373_Fig4_HTML.gif
Fig. 4

Absolute and risk-adjusted hospital mortality rates over time for the admission of haematological (n = 1,741) and non-haematological (n = 60,954) medical patients to the ICU

Volume–outcome effect haematological ICU admissions

A total of 36 centres admitted 1,741 haematological patients to their ICUs during the study period. When divided into four volume quartiles (each with 9 hospitals), these volume quartiles have (raw) hospital mortality rates of 37.1 % (median 18 admissions [IQR 15–21]), 44.7 % (median 28 admissions [IQR 26–29]), 46.2 % (median 35 admissions [IQR 34–39]) and 50.0 % (median 90 admissions [IQR 64–141]), respectively. There was no effect of haematological patient–volume on risk-adjusted hospital mortality (OR 1.01; 95 % CI 0.98–1.04; p = 0.38).

Discussion

We report here on the changes over time in ICU admission prevalence and outcomes of more than 1,700 critically ill haematological patients and these changes in more than 60,000 non-haematological medical ICU patients, all admitted to 36 Dutch ICUs during the last decade. The main findings of our study are threefold. First, the proportion of haematological patients among all medical ICU admissions increased over time, suggestive of a change in ICU admission policy. Second, the prognosis of haematological ICU patients significantly improved to a similar extent as that of several subgroups of non-haematological medical ICU patients. Third, our data convincingly confirm that a WBC count of <1.0 × 109/L was not a predictor of in-hospital mortality in our ICU patients with haematological disorders. In addition to these findings, it is important to realise that the hospital mortality rate of haematological patients admitted to the ICU was approximately twofold higher than that of the non-haematological medical ICU patients, but similar to that of ICU patients with solid tumours. When mechanical ventilation was needed, mortality was higher for all of the subgroups investigated. The higher mortality of haematological patients was likely related to their underlying disease, its treatment and the need for more rigorous ICU therapy and was also reflected by the longer LOS in the ICU and hospital by our haematological patients. Our results on both ICU and hospital mortality rates are within the ranges of those for critically ill haematological patients in ICUs in France and Belgium [17], as well as for a cohort of critically ill haematological patients in the UK [18]. Taken together, these results not only indicate that the denial of ICU treatment to haematological patients is not warranted, but they also emphasise the need for appropriate criteria for referring this specific group of patients to an ICU and that intensive supportive care for haematological patients should be considered irrespective of the degree of leukocytopenia.

The improvement over time we observed is in accordance with the results of a recent publication in which 428 neutropenic patients with sepsis were studied [19]. While the comparable improvement in hospital survival between haematological and non-haematological patients might suggest that improved ICU care accounted for this improved prognosis, changes in haematological treatment regimens may also have contributed to this trend. Less organ damage, infection and severe acute graft versus host disease related to allogeneic hematopoietic stem cell transplantation has been reported [20], and non-myeloablative therapies are associated with improved ICU survival compared to myeloablative therapies [21]. The absence in the current study of a case–volume/outcome relationship is however surprising. Two previous studies describe that survival is higher in ICUs that admit larger numbers of critically ill haematological patients with acute respiratory failure and septic shock compared to ICUs that admit fewer haematological patients [22, 23]. We found an increase in crude mortality rates for hospitals with more admissions, but after adjusting for case-mix this association was no longer apparent, indicating that larger centres treat more severely ill patients, but do not have better or worse outcome results.

Until recently, published outcomes of haematological ICU patients were generally poor despite a growing number of studies demonstrating improvement [24, 25]. The perception that transfer to an ICU may be futile and that long-term outcomes for those who survive critical illness are rather poor has led to a reluctance among both haematologists and intensivists to admit patients to an ICU. In our view, this reluctance might delay ICU admission, resulting in a self-fulfilling prophecy of a detrimental outcome for haematological patients, as illustrated by the fact that the number of failing organs in patients upon ICU admission is directly related to ICU mortality among patients with haematological malignancies [26]. Moreover, in their recent publication Azoulay et al. report that among patients who were denied ICU admission because they were considered too sick to benefit from the ICU, approximately one of four survived until hospital discharge; in contrast, among patients who were considered too well to benefit from ICU admission, one of ten patients eventually died [17]. These results emphasise the need for a better specification of ICU admission criteria for haematological patients.

In addition, we feel that improved general ICU treatment options, such as those, for example, in sepsis and septic shock management [27] and in the field of mechanical (non)invasive ventilation [28], as well as more specific applicable haematological treatments, such as anti-fungal drugs and more aggressive diagnostic strategies, account for the better outcome in general medical ICU patients [8]. Such treatment options and diagnostic strategies have contributed to an awareness that ICU admission may not be futile for critically ill haematological patients. Importantly, haematological treatment and ICU admission have an irrefutable impact on the perceived quality of life, which is increasingly considered to be a major measure of outcome. However, the quality of life after a 1-year rehabilitation period has been found to be below the average of that of a generally healthy population [29], while it appears to be similar in haematological and non-haematological patients admitted to the ICU and in haematological patients who were not treated in the ICU [30].

We found that severity of illness on ICU admission, expressed by the APACHE-II score, did not change during the study period. This observation suggests that the increase in admission of patients with haematological diseases was likely the result of a change in admission policy over time and that haematological patients were not admitted earlier (and were not less severely ill) to the ICU. Recently, a significant association was demonstrated in critically ill cancer patients between early intervention prior to ICU admission before the development of severe organ failure and a decreased hospital mortality rate. This finding is in accordance with our observed association between mechanical ventilation and increased mortality. It also highlights the importance of early identification of any decline in vital organ function as the usual reason for ICU admission in critically ill cancer patients is the need to support failing organs [31]. The beneficial effects of earlier ICU admission and aggressive ICU treatment were also suggested by Hampshire et al., who found an increased hospital mortality for patients with an increasing length of hospital stay prior to ICU admission and those with severe sepsis [18]. Therefore, there may be room for improvement if guidelines for timely admission are established. An important role herein might be played by medical emergency outreach teams.

Several limitations to our study need to be addressed. Clinical registries, such as the NICE registry, provide unique opportunities to study very large cohorts over longer periods of time, as our study illustrates. Because of the large numbers, chance findings due to variations in case-mix are unlikely. However, the set of clinical variables that is recorded in the registry database must be determined beforehand; it will also often have limitations when specific subgroups, such as the haematological ICU population, are studied. A number of important confounding haematological variables, such as level of neutropenia, underlying haematological diagnosis, disease status and type and phase of haematological treatment, including whether or not haematopoietic stem cell transplant was performed, were not available for our analyses. Also, intensity and mucotoxicity of the applied chemotherapeutical treatment determines the pattern of inflammatory response, irrespective of the presence of infection [32], and these issues are unknown in our patients.

Nevertheless, we were able to compare haematological subgroups with several other groups of severely ill patients for whom ICU admission is traditionally considered to be controversial. We were able to demonstrate changes in clinically relevant outcomes of haematological patients transferred to an ICU, as well as a significant decrease in mortality over the years, and to compare this trend with those of other specific patient groups. In our view, it would be useful to evaluate ICU admission policies with the aim to facilitate decision-making regarding patient admission to ICUs and thereby help avoid situations in which patients who may benefit from intensive treatment are deprived of life-saving therapies. We advise caution in generalising the results of our study, as well as those of published studies, due to possible variations in ICU admission policy and levels of ICU care provided, as well as differences in discharge criteria and settings and timing for the implementation of end-of-life decisions [33].

Conclusions and future recommendations

The results of this observational study show that the critically ill haematological patients in our study cohort had a higher mortality rate than specific subgroups of non-haematological medical ICU patients, but a similar one to patients with solid tumours. In our study, differences between groups were most pronounced in those patients that required mechanical ventilation. We conclude that the hospital and ICU survival of haematological patients has improved over time in The Netherlands, similarly to that of non-haematological medical ICU patients and that leukocytopenia itself is not a predictor of mortality. Our results indicate that intensive supportive care for patients with haematological malignancies should be available without delay, irrespective of the degree of leukocytopenia. Further research is needed to define the best timing for transfer to the ICU in order to provide the patient with the best chance of survival and to determine the influence of specific haematological conditions, such as graft versus host disease and disease status.

Acknowledgments

We acknowledge all participating ICUs of the National Intensive Care Registry for contributing patient data for analysis.

Conflicts of interest

None.

Copyright information

© Springer-Verlag Berlin Heidelberg and ESICM 2014