From 17,689 patients from 592 ICUs for whom data were available, data were extracted for all patients with a primary diagnosis of head trauma (with and without other traumatic injuries). Diagnosis of head trauma was recorded in 1,045 patients, who were included for analysis. These patients were admitted to one of 341 ICUs from 31 countries with each ICU contributing an average of 3.1 (2.4) patients. The majority of patients were admitted to ICUs in the USA (30 %), Australia (14 %), and Canada (12 %). Data were collected from most patients for the entire 12 study days (60 %), with a total of 10,558 study days recorded. Patient demographics are shown in Table 1.
Most patients (863/1045; 83 %) were from an ICU where a bedside feeding protocol was used to allow the nurse to advance or withhold enteral feeds. Protocols contained algorithms for: motility agents (n = 667, 64 %); small bowel feeding (n = 506, 48 %); withholding nutrition for procedures (n = 480, 46 %); head of bed elevation (n = 698, 67 %); and gastric residual volume (GRV) thresholds (n = 823, 79 %). In those ICUs with a GRV algorithm, the median GRV threshold was 250 (range 50–500) ml and the mode threshold was 200 ml (n = 331, 40 %).
Nutritional assessment and prescription
The majority of patients (n = 871, 83 %) were admitted to an ICU that employed a dietician, of which 40 % had at least one full-time dietician. During the years that data relating to baseline nutrition assessments were recorded (2010–2013), 85 % (n = 443/519) of patients had a baseline nutrition assessment completed. This assessment included documentation of body weight in half of patients (n = 260) and of height in 46 % of patients (n = 237).
A variety of methods were used to estimate energy requirements. The most frequently utilised method was a weight-based approach, for example 25 kcal/kg, which was used in 49 % of patients (n = 508). Equations were used in 432 patients (42 %), the most popular of these being Harris Benedict and Schofield. Eleven patients (1 %) had their energy expenditure estimated through indirect calorimetry. The mean amount of energy and protein prescribed daily was 1,958 (376) kilocalories and 98.7 (26.6) grams respectively, equivalent to 25.9 (4.9) kcal/kg/day and 1.29 (0.3) g/kg/day.
At some point during the study period the majority of patients (94 %, n = 983) received enteral nutrition (EN), 13 % (n = 138) received parenteral nutrition (PN), and 20 % (n = 207) ingested nutrient orally. Sixteen patients (2 %) received no nutrition during the study period. Twenty-four percent (n = 239) of patients had EN commenced on day 1 of ICU admission, 41 % (n = 404) on day 2, and 20 % (n = 195) on day 3. The mean time from ICU admission to initiation of EN was 35.5 (32.7) hours.
Patients often received more than one concentration of EN formula; however, a 1 kcal/ml formula was the most common and was delivered for 53 % of all EN prescriptions. Twenty-five percent of prescriptions were of a concentrated enteral formula that provided 1.5 kcal/ml or greater. One-hundred and twenty-two (12 %) patients received glutamine during their ICU admission with a mean daily dose of 22 (11) grams. Glutamine was usually delivered via the enteral route (66 % of patients).
Of those patients receiving EN the location of the feeding tube was reported for 926 (94 %) patients. Gastric feeding was the most common route of EN, and used exclusively in 67 % (n = 620) of patients; 11 % (n = 101) of patients were exclusively fed via post-pyloric tubes, and 22 % (n = 205) received EN through a combination of gastric and post-pyloric routes. Gastrokinetic drugs were frequently prescribed; 70 % of patients (n = 713) received a gastrokinetic drug at some stage. The prevalence of gastrokinetic drug use varied according to day of admission, with 29 % (n = 185) of patients receiving gastrokinetics on day 1, 50 % (n = 415) on day 2, and a peak of 61 % (n = 556) by day 5. Even at nutritional data censor (i.e., day 12) 56 % (n = 320) of patients were receiving gastrokinetics. Metoclopramide was the most commonly prescribed gastrokinetic drug and was administered to 38 % (n = 400) of patients.
Of the patients who received EN, 66 % (n = 644) had interruptions to feeds at least once during the study period. Thirty percent (n = 191) had interruptions to feeding on just one day, 21 % (n = 133) had interruptions on two days, 16 % (n = 103) had interruptions on three days, and 34 % (n = 217) had four or more days where feeding was interrupted. There were various reasons for interruptions to enteral feeds (Fig. 1). From 2009 to 2013, the number of hours of interruptions to EN were collected, with a mean duration of 25.3 (range 0.2–120) hours per patient, equivalent to 2.6 (range 0.1–18.8) hours per day.
Energy and protein intake and deficit
Energy and protein were received from various sources (Fig. 2). Over half of the patients received propofol (59 %, n = 618), which provided a mean of 161 (165) kilocalories of additional energy per day. The mean amount of energy received from EN was 974 (524) kcal/day, and 86 (269) kcal/day from PN. The mean delivery of energy and protein to patients from all sources was 1154 (525) kcal/day and 52 (26) g/day, respectively; equivalent to 15.3 (7.2) kcal/kg/day and 0.69 (0.4) g/kg/day. The daily mean energy and protein deficit was 803 (527) kilocalories and 46 (30) grams, respectively. Nutrition from all sources met an average of 58 (range 0–166) % of estimated energy requirements and 53 (range 0–390) % of protein requirements. Daily intake data is shown in Fig. 3.
Eighty-nine percent (n = 926) of patients were from an ICU that contained a protocol to monitor blood glucose and administer insulin. In those protocols that contained a blood glucose target, the median lower blood glucose target was 4.5 (range 3.0–8.3) mmol/l and the upper blood glucose target was 8.3 (range 5.3–15.0) mmol/l. For 700 patients from 2009–2013 the mean highest blood glucose recorded in the first 24 hours of ICU admission was 9.8 (3.3) mmol/l and the lowest was 6.5 (1.9) mmol/l. The mean morning blood glucose during the study period was 7.5 (1.3) mmol/l. An episode of hypoglycaemia occurred in 9 % (n = 90) of patients. Insulin was provided in 59 % of cases (n = 611), of which the average daily insulin dose provided was 36.5 (36.2) units.
Of the 1,045 patients, 135 (13 %) died in ICU, 38 (4 %) died after ICU discharge in hospital, and 872 (83 %) survived to hospital discharge or were alive in hospital at day 60. Male patients were more likely to survive a head trauma than females (odds ratio (OR) 0.66; 95 % CI 0.46, 0.96; p = 0.026). The median ICU length of stay in survivors was 13.1 (IQR 7.9–21.6) days, and the median hospital length of stay was 29.7 (IQR 17.9–57.1) days. The median length of time during which patients required mechanical ventilation was 9.2 (IQR 4.8–15.4) days.
Patients from an ICU that utilised a feeding protocol had greater energy and protein intakes per body weight than those without (p <0.001, 0.002 respectively) and were more likely to survive (OR 0.65; 95 % CI 0.42, 0.99; p = 0.043; Table 3). When the feeding protocol contained guidance on motility agents and small bowel feeding there was a smaller energy and protein deficit (Table 2). Patients from an ICU with a feeding protocol that contained details on GRVs had smaller protein deficit, but no difference in energy delivery.
Earlier initiation of EN was significantly associated with a reduction in energy and protein deficit (r = 0.32 and 0.27 respectively, p <0.001). While the point estimate indicated reduced mortality when EN was commenced on day 1 when compared to days 2–4 or day 5 or later, this was not significant (Table 3). Greater duration of EN interruptions increased both energy and protein deficit (r = 0.219 and 0.218 respectively, p <0.001). Energy and protein deficits were reduced when EN and PN were used in combination, compared with EN alone (p = 0.023 and <0.001 respectively, Table 3).
There was a non-significant association between at least one recorded episode of hypoglycaemia and higher risk of mortality (OR 1.6; 95 % CI 0.96, 2.7; p = 0.073). Any hypoglycaemic event was associated with a reduced probability of being discharged alive from ICU: (hazard ratio (HR) 0.78; 95 % CI 0.60, 0.99; p = 0.043) and hospital: (HR 0.78; 95 % CI: 0.58, 1.03; p = 0.082).
A greater energy and protein deficit (OR per 100 kcal/day) was associated with longer times until discharge alive from ICU (energy: p <0.001, protein: p = 0.001) and hospital (energy: p = 0.002, protein: p = 0.024) (Table 4). A greater energy and protein deficit was also associated with longer time receiving mechanical ventilation (OR per 100 kcal/day, p <0.001; Table 4). However, when a sensitivity analysis was performed to include only those patients who stayed in ICU for a full eight days a statistically significant relationship only remained for energy deficit on time to discharge alive from hospital and length of mechanical ventilation (p = 0.001 n = 816, and p = 0.004 n = 732 respectively; Table 4). In an unadjusted analysis there was a significant protective effect of energy (per 10 kcal/kg/day), but not protein, delivery on mortality (energy: OR 0.79; 95 % CI 0.63, 0.998; p = 0.048). However, in the adjusted analysis neither energy nor protein delivery affected mortality (energy: OR 0.76; 95 % CI 0.48, 1.22; p = 0.256, protein: OR 1.01; 95 % CI 0.92, 1.118; p = 0.868).