Prolonged intensive care unit (ICU) stays for critical illness can result in acceptable mortality rates and quality of life despite significant costs 12. Only a few studies have specifically addressed prolonged ICU lengths of stay (LOS) after trauma 345. Our goals were to determine the outcomes and characteristics of trauma patients with prolonged ICU LOS. Based on previous studies of medical and surgical ICU patients, our hypotheses were that age and injury severity predicted prolonged ICU LOS in trauma patients admitted to the ICU, but that the majority of trauma patients who survived beyond 30 days in the ICU would survive to discharge.

This was a retrospective study based on the hospital trauma registry over a seven-year period (1998 to 2004) approved by the hospital Institutional Review Board with waiver of consent. In this level I trauma center, critical care services for injured patients are provided by the same trauma physician group that admits injured patients. Admission clinical characteristics, pre-existing conditions and acquired complications in the ICU were extracted from registry data. Selected definitions used for this study for pre-existing conditions and complications are based on those set by the Pennsylvania Trauma Systems Foundation [see Additional data file 1].

For the purposes of this study, the control group was designated as those patients who were admitted to the ICU for less than 30 days (ILOS<30). This group was compared with the group with ICU LOS of 30 days or greater (ILOS>30). Within the ILOS>30 group, we also compared survivors with non-survivors (Figure 1).

Data were summarized as mean ± standard deviation. To compare means, we used the independent samples t test and the Mann-Whitney U rank sum test. Logistic regression was used to identify independent predictors of prolonged LOS in the entire sample as well as independent predictors of mortality within the ILOS>30 subgroup. Correlation was assessed using Spearman's rho. Chi-squares and nested chi-squares analyses were used to explore relations between variables. Differences were considered significant at P < 0.05. SPSS version 14.0 (SPSS Inc., Chicago, IL, USA) was used to analyze the data.

Only a few studies have specifically addressed prolonged ICU stays in trauma patients. Trottier and colleagues 3 analyzed 339 trauma and burn patients with ICU LOS of more than 28 days and found similar survival rates (87%) to our study with age being the most important predictor of outcome. Compared with a control group of patients with shorter LOS, the authors demonstrated that age, injury severity, and the presence of burn injuries were determinants of prolonged ICU stay. Miller and colleagues 4 found that the overall mortality rate was 22% with the majority of patients dying from MOF. Age was the only significant predictor of mortality. In both these studies, pre-existing conditions were not analyzed. Goins and colleagues 5 reported a mortality rate of 17% for 87 trauma patients spending more than 30 days in the ICU. There was no comparison to a control group.

In contrast to the above-mentioned studies, our study was unique in that we analyzed differences in pre-existing conditions and acquired complications. We found that ILOS>30 patients constituted only a small percentage of all trauma admissions to the ICU but consumed a disproportionately large amount of ICU resources. These findings are similar to a prospective study by Martin and colleagues where in a heterogeneous population, prolonged-stay patients represented 5.6% of ICU admissions and accounted for almost 40% of bed days 6. Similarly, medical-surgical ICU patients with ICU LOS of more than 30 days accounted for 8% of total ICU admissions but 48% of occupied beds 7 in another study.

Not surprisingly, age and injury severity were associated with prolonged ICU stay and mortality, but after multivariate analysis, age was not found to be an independent predictor of prolonged stay, and neither were pre-existing conditions or admission GCS. Instead, sepsis, ARDS and other infectious complications were found to be powerful predictors.

That age or pre-existing conditions did not independently predict prolonged stay could simply be attributed to selection bias: older patients and those with significant pre-existing conditions may not have survived to the 30-day mark. This is suggested by comparing those who died before 30 days to the ILOS>30 patients: patients who died before 30 days of admission were older, and more likely to have a significant head injury, pre-existing cardiac or neurological condition and be on warfarin. Notably, in ILOS<30 non-survivors, 61% were aged 65 years or older versus 39% in the ILOS>30 group.

Within the ILOS>30 group, similar to the previous studies on trauma patients, we found that age was still an independent predictor of mortality. In addition, pre-existing renal conditions and the need for renal replacement therapy during the ICU stay also predicted mortality. The high mortality rates associated with dialysis have been reported in other institutions 278. The study by Eachempati and colleagues 8 demonstrated a mortality rate of 61% in patients requiring dialysis compared with an overall mortality of 45% for all patients with acute renal failure. Patients who required dialysis in our study had a mortality rate of 33%.

The mortality rate in the ILOS>30 trauma patients (12%) was consistent with the previously published studies on trauma patients. This finding could be used to support families who may be discouraged by the length of time their family member is in the ICU, as well as to illustrate to health care providers that their efforts are not in vain. In a prospective observational study 9, there were discordant predictions with regard to futility of survival and quality of life between doctors and nurses in 21% of ICU patients. Only 9 to 15% of survivors of ICU stay where health care professionals had considered treatment futile actually reported bad quality of life six months later. On the other hand, physician estimates of ICU survival can be powerful predictors of ICU mortality when compared with illness severity, organ dysfunction and the use of inotropic drugs, possibly by contributing to more 'do not resuscitate' directives in instances of cardiac arrest, and more likely withdrawal of dialysis, pharmacological support, and mechanical ventilation 10.

That patients aged 65 years and older accounted for almost 40% of the ILOS>30 group was reflective of our admission population, where these elderly patients comprised 28% of all trauma admissions to our institution. Older trauma patients have been recognized as having a higher risk of dying when chronic medical conditions exist compared with those without chronic conditions, and this relation between mortality and pre-existing medical conditions is more apparent when these patients sustain less severe injuries 11. Studies in non-trauma ICU cohorts support the conclusion that age in and of itself does not predict poor outcome 121314. Higgins and colleagues 14 determined that the need for ventilation at 24 hours, trauma and emergency surgery admissions, severity of illness, and prolonged pre-ICU stays were independent predictors of prolonged stay, and not age in itself. Pre-hospital functional status has also been found to be an important predictor of poor outcome in ICU patients 151617.

There were several limitations of this study. One was the lack of data on long-term outcome and pre-injury functional status. We also did not have prospective information on prognostic indicators of ICU survivability or measures of organ dysfunction with time in the ICU. Also, we could not assess the degree of adherence to evidence-based practices known to reduce ICU morbidity and mortality such as glycemic control, sedation protocols, ventilator practices, and transfusion and phlebotomy practices 18. Further, ICU LOS was influenced to a certain extent by discharge planning arrangements with insurance payers and transfer facilities. The lack of prospective time-dependent data regarding organ dysfunction and the degree of adherence to evidence-based guidelines makes it difficult to determine to what extent the acquired ICU complications were a result of sub-optimal ICU care rather than nature of disease due to the injuries sustained on admission.

Finally, the definitions of certain pre-existing conditions such as cardiac and pulmonary disease lacked objective criteria. This was because these criteria were frequently not available for trauma patients admitted as emergencies to the ICU. As these cardiac and pulmonary conditions were factors that were entered into the logistic regression analysis (Table 4), it is conceivable that were the definitions modified by including objective criteria, they could have emerged as independent risk factors predicting prolonged ICU stay.

Trauma patients who spent 30 days or more in the ICU consumed a disproportionate amount of ICU resources. For those who survived to 30 days, acquired pulmonary and infectious complications were important predictors of prolonged stay. Although injury severity was found to be an independent predictor of ICU LOS of 30 days or more, partly confirming our hypothesis, age was not. Age, however, did predict mortality in the patients with LOS of 30 days or more, together with pre-existing renal disease and the development of renal failure in the ICU requiring renal replacement therapy. The majority (88%) of these prolonged-stay patients also survived to discharge, confirming our second hypothesis. These findings imply that resources should continue to be directed at infection prevention and surveillance in trauma ICU patients, and also underscores the necessity of adhering to evidence-based guidelines that may decrease ICU LOS. We feel that this study suggests associations between variables in a broad spectrum of trauma patients and communicates important trauma outcomes, and that the data could provide a framework for the generation of hypotheses about prolonged ICU stay in a trauma patient population.

• Trauma patients who have ICU LOS of 30 days or more constituted only 4% of all trauma ICU admissions but accounted for a disproportionate usage of ICU resources.

• 88% of these patients survived to hospital discharge.

• Infectious complications, sepsis, ARDS were independent predictive factors for ICU LOS of 30 days or more.

• Mortality in these prolonged-stay patients was influenced by age, development of renal failure requiring renal replacement therapy, and pre-existing renal dysfunction.

ARDS: acute respiratory distress syndrome; GCS: Glasgow Coma Score; ICU: intensive care unit; ILOS<30: patients with ICU length of stay less than 30 days; ILOS>30: patients with ICU length of stay greater or equal to 30 days; ISS: Injury Severity Score; LOS: length of stay; MOF: multiple organ failure.

The authors declare that they have no competing interests.

AO analyzed the data, participated in the study design and drafted the manuscript. LO conceived of the study, participated in the study design and helped draft the manuscript. DV participated in the study design and analyzed the data, and helped draft the manuscript. BG collected the data, participated in the study design and analysis of the data. JP collected the data, participated in the study design, and helped draft the manuscript. EJ conceived of the study with LO, collected the data and participated in the study design. AR participated in the study design and helped draft the manuscript. All authors read and approved the final manuscript.