We found significantly higher 1-year mortality rates in both cancer cohorts than in patients without cancer. A visual inspection showed a slight decrease in 1-year mortality in haematological patients, whereas neither an increasing or decreasing trend was observed in patients with solid cancer. We found a decrease in 1-year mortality rate in patients without cancer.
Literature regarding cancer patients in general showed a decrease in long-term mortality over the past decades (1, 4, 5). In our study, which included patients with an unplanned ICU admission, this trend was only observed in haematological patients and not in patients with a solid tumour. While no information on cancer treatment after ICU admission is available in the NICE register, it is tempting to speculate that the observed differences in 1-year mortality for haematological patients are secondary to novel and more efficient cancer treatments. The wide variety in types of solid malignancies and treatment options for those different malignancies might explain the absence of a decrease in 1-year mortality in patients with a solid tumour (19).
As noted before, large long-term mortality studies in cancer patients with an unplanned ICU admission are scarce and outdated (18–23). Potential differences in case-mix between our study and previous literature makes direct comparison difficult. For example, a large study performed in the United States with data from 2002–2011 (19) reported a decrease in 1-year mortality in patient with hematological cancer and solid cancers over time. However, in this study, both planned and unplanned ICU admissions were included. Other studies regarding long-term mortality do not describe a trend over time. For example, Puxty et al. (18) compared mortality rates of surgical patients with and without cancer admitted to the ICU. The 1-year mortality was lower than in our study. A major difference compared to our study was that the majority of the patients had a planned ICU admission after surgery. A study from Belgium (20) showed comparable mortality for patients with haematological cancers to our study, while the 1-year mortality of patients with a solid tumour was lower than in our study. Again, the study of Oeyen et al. included both planned and unplanned ICU patients. Finally, the 1-year mortality of cancer patients in our study was lower than that of studies with smaller sample size from Spain, Austria and Switzerland (21–23). Apart from sample size, there is no evident difference in case-mix between our study and these studies, supporting an improvement in long-term survival in this population.
We found a decrease in hospital mortality over the inclusion period for all three cohorts. These results are in line with several other studies that showed a decreasing trend in hospital mortality of cancer patients (9, 14–19). Relevant to note is that when cancer patients (either haematological or solid) survived the hospital admission, approximately one third died within the year following the ICU admission. In patients without cancer, this number was considerably lower (around 12%). A large recent study (27) showed similar results. Among all diagnostic subgroups of the ICU population included in that study, patients with nonsurgical cancer had the lowest cumulative 1-year survival. We assume that a considerable part of the deaths within one year is directly or indirectly related to the underlying malignancy or its treatment.
A possible and not unlikely explanation for the decrease in hospital mortality and high 1-year mortality may be the physician’s decision to discharge patients to a nursing home or hospice for the terminal phase, leading to a spurious reduction of mortality risk when using hospital death as outcome (28, 29). Unfortunately, we have no data available in the NICE register to support this hypothesis.
Two thirds of the patients who survive to hospital discharge is still alive 1 year after ICU admission. Earlier studies showed that severity of illness is more important than the underlying malignancy with regard to short-term mortality (9, 15, 18). Our study adds to the current literature that underlying cancer or its treatment is an important factor for long-term outcome in patients recovering from critical illness. However, ICU admissions of patients are often determined by short-term survival benefit, and not by long-term prognosis (30). In addition, physicians do not always communicate adequately about prognosis and benefits of treatments towards patients and relatives (31–33). Ideally, oncologists and haematologists discuss long-term outcomes after an ICU admission with patients and relatives well before an ICU admission, in order to manage expectations. All physicians should consider benefits and disadvantages of an ICU admission for each patient individually before ICU referral or admission.
A relevant observation was that haematological patients were more severely ill than the two other cohorts. These patients had a higher APACHE IV score, among other factors explained by a higher percentage of AKI, mechanical ventilation, and/or vasoactive drugs. As higher severity of illness scores influences short-term mortality (9, 15, 18), these factors are important for ICU physicians and may help physicians with treatment decisions and the management of expectations of patients and relatives.
Limitations
For interpreting the study results at least the following limitations should be considered. Firstly, since the NICE registry is originally a quality of ICU care registry in the Netherlands, not all clinically relevant factors for cancer patients are registered. For example, we were not able to include the exact type and stage of the malignancy, the continuation of cancer treatment, and the performance status prior to ICU admission.
Secondly, as with most studies regarding cancer patients with an ICU admission, the heterogeneity of the study is a limitation. We included a random intercept for hospital to correct for differences between hospitals. Further research should focus on various subgroups of this study in order to minimize heterogeneity. The actual numbers and proportions of cancer patients with an unplanned ICU admission were not different over the inclusion period. This suggests that at least the same number of cancer patients were admitted and no structural changes in admission decisions were made.
Recommendations and future research
Considering our results and limitations, critically ill cancer patients are still a vulnerable patient group. Because of the complexity of their condition and uncertain prognosis, care for critically ill cancer patients requires adequate interdisciplinary medical management (8, 34).
Linking the NICE data to the national cancer registry in the Netherlands may assist to enrich the data with clinically relevant factors, such as staging of cancer, treatment decisions, and type of malignancy. Such enrichment of the data may help to identify patients with a malignancy who benefit most from an ICU admission (6, 7). Due to the difference in case-mix in previous studies (e.g. planned vs. unplanned ICU admissions, surgical vs. non-surgical patients), direct comparison of mortality rates is difficult. For future research, it would be helpful to establish more uniform inclusion criteria to enable a viable comparison between countries.