This study is a single-institution retrospective cohort study of a new ICU evaluation & development of nutritional support protocol (NICE-NST) protocol, introduced for medical ICU patients in a tertiary academic hospital from December, 2018. This study was conducted in accordance with the amended Declaration of Helsinki and was approved by the Institutional Review Board of Seoul National University Bundang Hospital where the study was conducted (No. B-1907-550-107). Patient consent was waived because the NICE-NST protocol was implemented as an in-hospital quality improvement project, and this study was a retrospective review of the outcomes.

The contents of NICE-NST protocol were as follows: (1) the Penn state equation was used to determine target caloric requirements [18], which was similar to the indirect calorimetry value. (2) Patients were categorized into group 1 (<5 scores, with abnormal liver function, or with abnormal renal function), and group 2 (>5 scores or <70% of the daily nutrients requirement target after 7 days of ICU day) using nutrition risk in critically ill (NUTRIC) score. Group 1 nutrition was by EN only, but group 2 nutrition was by supplementary PN with EN to meet nutritional targets. Both groups targeted 100% of the required daily calories and 1.3 g/kg/day of protein (Table 1). EN supply was through continuous feeding with tube or gastrostomy for 16 hours from 9 AM to 1 AM the next day. The GRV was measured at 9 AM, 5 PM, and 11 PM, considering the ICU nurse-working schedule. If GRV was > 250 ml, EN feeding is held, and prokinetics are given thrice daily. After 2 hours of withholding EN feeding, GRV was rechecked, and the same protocol was repeated until 1 AM the next day (Figure 1).

The group that was supplied nutrition without NICE-NST protocol was the “control” group, and the group that was supplied nutrition with NICE-NST protocol was the “test” group. Patients admitted to medical ICU from December 2017 to March 2018 and other patients admitted to medical ICU from December 2018 to March 2019 were screened as the control and test groups, respectively. Exclusion criteria include ICU discharge within one day, non-tube feeding, transfer to medical ICU from other ICU, nutrition support team consultation before ICU admission, continuous renal replacement therapy (CRRT) application within 5 days, intermittent hemodialysis (HD) before ICU admission, and long-term inpatients in medical ICU because of legal problem.

Supplying nutrition with protocol in ICU will ensure multidisciplinary discussions on nutrient supply; it will provide appropriate calorie and other nutrients to critically ill patients quickly. Therefore, we expect that the start and amount of EN will significantly improve by following the NICE-NST protocol. In addition, it is expected to improve the clinical outcomes such as ICU mortality, hospital mortality, duration of ICU admission, etc.

Age, sex, cause of ICU admission, and medical history data were collected to evaluate baseline characteristics. Initial arterial oxygen pressure divided by the fraction of inspired oxygen (P/F ratio), Acute Physiology And Chronic Health Evaluation (APACHE) II score, nutrition severity, and NUTRIC score were collected to assess disease severity at the time of ICU admission. Daily body weight, daily calorie supply of dextrose for main fluid, daily calorie and protein supply of EN, and daily calorie and protein supply of PN data during the first 5 days in ICU were collected to evaluate the primary calorie and protein supply. The total amount of calories and protein supplied throughout one day was calculated with administered total daily volume for each item. For example, suppose EN feeding in which 1 kcal per 1 ml and 0.05 g of protein per 1 ml were supplied with 32 ml per hour for 10 hours. In that case, the total calorie supply of EN feeding will be 320 kcal, and the total protein supply of EN feeding will be 16 g. The total calorie was calculated as the sum of EN calorie supply, PN supply, and supplied dextrose and albumin. Total protein was calculated as the sum of protein of EN, PN, and albumin supplied.

In addition, in order to analyze feeding side effects, data on the total daily count of EN feeding hold due to GRV or vomit, the ratio of hyperglycemia (blood glucose level >250 mg/dl) events to total daily glucose check counts, the ratio of hypoglycemia (blood glucose level <70 mg/dl) events to total daily glucose check counts, and the total amount of daily steroid use were collected. Because the total number of blood glucose checks was different according to the status of patients, a ratio was used to analyze hyperglycemia and hypoglycemia events. In addition, the total amount of daily steroid can also affect blood glucose level; therefore, the total amount of daily steroid use was analyzed.

Fisher's exact test and Pearson Chi-square were used for the comparison of categorical variables. Mann-Whitney test and Student t-test were used for the analysis of continuous variables. We used repeated-measures analysis of variance (RM-ANOVA) to analyze repeated-measured variables. Logistic regression for ICU mortality was performed with age, sex, body mass index (BMI), nutrition severity, the time duration from admission in ICU to the start of EN feeding (Admission to EN time), and mean amount of supplied EN calorie (kcal/kg) or protein (g/kg) for 5 days. The results were considered significant, with P-values less than 0.05. All analyses were conducted using the IBM SPSS version 22.0 (IBM Corp., Armonk, NY, USA).

This study screened 170 patients. One patient discharged from the medical ICU within one day was excluded. Two patients who were given soft bland diet without tube feeding were excluded. Because the NICE-NST protocol was applied only to the medical ICU patients, 31 patients transferred from the other ICU, such as surgical or emergency ICU, were excluded. Since the NICE-NST protocol was applied at the time of medical ICU admission, four patients whose nutrition support team consultation was conducted before entering the medical ICU were excluded. Because diverse solvents are used in dialysis and protein loss is inevitable, the amount of calorie and protein supply could not be confirmed exactly in CRRT or HD. Thus, 45 patients who had been applied CRRT within 5 days from medical ICU admission and 4 patients who had previously received intermittent HD were excluded. In addition, three patients hospitalized for a long time in medical ICU because of legal problems were excluded. Initiating EN within at least 48 hours of ICU admission was described as early EN in previous studies [4,7-15]. Since one of the aims of this study was to determine whether early EN was achieved when the protocol was applied, only patients whose protocol application time was at least 2 days were selected and analyzed. In addition, because a feeding protocol was provided through nutrition support team consultation, only patients with the duration from admission in ICU to nutrition support team consultation (Admission to nutrition support team time) of less than 2 days were included in the test group. Finally, the data of remaining 62 patients (40 patients in the control group, 22 in the test group) were analyzed (Figure 2).

Baseline characteristics showed no difference in age, sex, BMI, underlying diseases, and the cause of admission to ICU between the two groups. Furthermore, there was no difference in disease severity, such as APACHE II score, initial P/F ratio, and the amount of steroid use converted to that of methylprednisolone (Supplementary Table 1).

The calorie and protein supply, which are the primary outcomes of this study, showed that total calorie and protein supplied were not significantly different between the two groups. However, the supply of EN calorie (4.0±1.0 kcal/kg vs. 6.7±0.9 kcal/kg, P=0.006) and protein (0.17±0.04 g/kg vs. 0.32±0.04 g/kg, P=0.002) were significantly higher in the test groups (Table 2). Further, the Admission to EN time was significantly shorter in the test groups (3.7±0.4 days vs. 2.4±0.5 days, P=0.010) (Table 3). The trends of supplying calories and protein over time are shown in Figure 3. EN calorie and protein supply analyzed by RM-ANOVA were significantly higher in the test group. However, during the early stage of ICU admission, more calories and protein were supplied in the control group. There was no difference in serum albumin level change between admission in ICU and 7 days after admission. No difference was observed in the number of feeding hold due to excessive GRV or vomiting during 5 days in ICU and the ratio of hyperglycemia events to total daily glucose check counts or the ratio of hypoglycemia events to total daily glucose check counts during 5 days in ICU (Table 3). There was no difference in ICU mortality, hospital mortality, and the time duration from admission to discharge in ICU (ICU duration) between the two groups (Table 3). The average admission to EN time of the test group exceeded 2 days (2.4±0.5 days) (Table 3).

Additional analysis was performed with 80 patients regardless of applying NICE-NST protocol (Supplementary Figure 1). The hospital mortality was significantly lower in the group that started EN feeding within 1.5 days (42.9% vs. 11.8%, P=0.018), and the ICU mortality was lower in the same group with marginal significance (28.6% vs. 5.9%, P=0.051) (Table 4). APACHE II score (32.3±0.96 vs 31.9±1.90, P=0.898) and initial P/F ratio (200.1±14.97 vs. 215.0±32.16, P=0.705) were not significantly different regardless of starting EN feeding within 1.5 days or not. On the other hand, there were significantly more patients with admission to EN time less than 1.5 days in the test group than in the control group (17.5% vs. 40.9%, P=0.044; OR, 3.26; 95% CI, 1.005–10.600).

Multivariable logistic regression was conducted to confirm whether mortality was associated with the amount of EN feeding. The mean EN calorie (OR, 0.815; 95% CI, 0.681–0.975) (Supplementary Table 2) and protein (OR, 0.017; 95% CI, 0.000–0.696) (Supplementary Table 3) per weight were inversely correlated with ICU mortality. Finally, when patients were divided according to 5 kcal/kg of EN calorie supply or 0.2 g/kg of EN protein supply, ICU mortality was significantly lower in the group supplied more than 5 kcal/kg of EN calorie or 0.2 g/kg of EN protein (7.7% vs. 34.8%; OR, 6.400; 95% CI, 1.338–30.606). Hospital mortality was not significant but tended to be lower in the same group (23.1% vs. 45.7%; OR, 2.800; 95% CI, 0.950–8.255) (Supplementary Table 4).