Design and data source
A retrospective analysis was performed using data that was extracted from a medical claims database which included 451 hospitals and managed by the Medical Data Vision Co., Ltd. (MDV; Tokyo, Japan). The database uses the diagnosis procedure combination/per-diem payment system (DPC/PDPS), in which provider reimbursement is calculated on a flat-rate per-diem fee based on the diagnosis group. The study protocol was approved by the Ethics Committees of the Okayama University Graduate School of Medicine, Dentistry, and Pharmaceutical Sciences (No. 2108-041) and the Kurume University Graduate School of Medicine (No. 21139) and registered at the University Hospital Medical Information Network Clinical Trial Registry (UMIN000044962). Informed consent was not required, because all personal information used in this study was anonymized.
The database included information on dates of hospital admission and discharge, age at admission, sex, height, body weight, body mass index (BMI), number of beds in admission hospital, year and type of admission, primary diseases (coded using the International Statistical Classification of Diseases and Related Health Problems, 10th Revision [ICD-10]), comorbidities (used to determine the Charlson Comorbidity Index [CCI]) [9], activities of daily living (ADL) based on the Barthel Index (BI) [10], levels of consciousness based on the Japan Coma Scale (JCS) [11], malnutrition defined as having poor oral intake of at least 10 days and low body mass index according to the Global Leadership Initiative on Malnutrition (GLIM) criteria [12], medical treatments during hospitalization (using Japan-specific medical claims codes), and discharge outcome status, as well as other information not used in our study. The total daily doses of parenteral energy, amino acids, and ILE prescribed were calculated using the parenteral nutrition infusion product names and compositions along with the prescribed quantities of those products, as they appeared in the database. When recording these doses, day 1 was regarded as the day fasting started, day 2 as the second day after fasting started, and so on.
Patient population
This study included hospitalized adult patients ages 18 years or older who were fasting (receiving no oral or enteral nutrition) for more than 10 consecutive days and were managed with parenteral nutrition, between January 2011 and September 2020. Patients were excluded from the study who underwent surgery or entered the intensive care unit between the day of admission and the start of fasting, were suspected to be in the terminal disease phase (defined as prescribed mean energy doses < 10 kcal/kg or mean amino acid doses < 0.5 g/kg on days 4 through 10), or were considered to be overfed (which we based on prescribed mean energy doses ≥ 30 kcal/kg on days 4 through 10). The rationale for the use of days 4 through 10 was that the administration of parenteral nutrition usually involves a gradual increase in dose over the initial 3 to 4 days before reaching the full target dose [7, 8].
Clinical outcomes
The primary endpoint was in-hospital mortality. The secondary endpoints included intravenous catheter infection during hospitalization, deteriorated ADL at discharge, length of stay (LOS), readmission, and total medical costs. ADL at discharge, LOS, and readmission were recorded for only those patients who were discharged alive, whereas other data were recorded for all patients. Medical costs were calculated based on Japanese yen and were then converted to US dollar (US$) using the annual exchange rate of 2020 reported by the Organization for Economic Cooperation and Development (OECD) (US$1 = 107 Japanese yen) [13]. Patients were considered to have deteriorated ADL when their total BI scores were lower at the time of discharge than at the time of admission. Readmission was defined as being admitted to the same hospital again within 30 days of discharge.
Variables
The variables extracted from the database were categorized as follows: age at admission (18–59, 60–69, 70–79, 80–89, or ≥ 90 years), BMI (< 16.0, 16.0–18.5, 18.5–22.5, 22.5–25.0, or ≥ 25.0), number of beds admission hospital (< 200, 200–500, or ≥ 500), year of admission (2011–2012, 2013–2014, 2015–2016, 2017–2018, or 2019–2020), type of admission (elective or emergency), primary disease (by ICD-10 code), comorbidities (CCI of 0, 1, 2, or ≥ 3), ADL (BI of 0, 5–20, 25–40, 45–60, 65–95, or 100), levels of consciousness (JCS of 0 [alert], 1–3 [awake], 10–30 [arousable], or 100–300 [coma]), and nutritional status (malnutrition defined as BMI < 18.5 if < 70 years old or BMI < 20 if > 70 years old). Information about medical treatments (e.g., albumin infusion, blood transfusion, respirator use, dialysis, nutrition support team, and rehabilitation) ordered between the day of admission and day 10 was extracted from the database for each patient. Missing values for the type of admission, BI, and JCS were placed in an “unknown” category.
Prescribed doses of parenteral nutrition
The prescribed mean daily doses of energy, amino acids, and ILE for days 4 to 10 after the start of fasting were calculated for each patient based on the parenteral nutrition infusion product composition and prescribed quantity of that infusion and were based on the assumption that nutrient doses often take until day 4 to reach 100% of their target [14]. Prescribed daily doses of energy and amino acids were calculated as kilocalories (kcal) and grams (g), respectively, and reported per kilogram (kg) of body weight and prescribed daily doses of ILE were calculated and reported as both grams and the caloric percentage (%) of the total non-protein energy administered that day.
Statistical analysis
The data management and statistical analysis were performed by an independent third party (A2 Healthcare Corporation; Tokyo, Japan) in order to eliminate arbitrariness and ensure transparency. Categorical variables were summarized as numbers and percentages, and continuous variables were summarized as means and standard deviations (SD). Missing values were not included. First, patients eligible for the study were divided into 2 groups: the ILE group, who were prescribed ILEs during days 4 through 10, and the non-ILE group, who were not prescribed ILEs during days 4 through 10. Next, propensity score matching (PSM) was used to adjust for confounding factors [15]. The propensity score was estimated by multivariable logistic regression analysis with the ILE group as the objective variable and patient characteristics as the explanatory variables. PSM was conducted using a one-to-one nearest neighbor method and using the caliper width. The caliper value was 0.2, and matching was performed within the caliper values. To confirm the covariate balance between the groups, standardized differences were calculated before and after PSM. A standardized difference less than 10% was considered to represent a balanced covariate [16].
To compare the 2 groups for each outcome, both before and after PSM, the Student t-test was used for continuous variables and the chi-square test was used for categorical variables. To adjust for the differences in the prescribed mean daily parenteral energy doses between the 2 groups, even after PSM, multivariable logistic or multiple regression analyses, as appropriate, were performed, with the mean daily energy dose prescribed for days 4 through 10 added as an explanatory variable. In these analyses, odds ratios (ORs) or regression coefficients, as appropriate, along with 95% confidence intervals (CIs), were calculated, both before and after the adjustment for energy.
For in-hospital mortality, survival curves were generated for the 2 groups using the Kaplan-Meier method, and a log-rank test was performed. In addition, the Cox proportional hazard model was used to calculate a hazard ratio (HR), along with a 95% CI, of the ILE group to the non-ILE group, for in-hospital mortality. For these calculations, the patients who were discharged alive were censored on the day of discharge, and the inpatients who survived for 180 days or longer were censored on day 180. All statistical analyses were performed using SAS version 9.4 (SAS Institute Inc., Cary, NC, USA), with a two-sided significance level of 5%.
Sensitivity analysis
Before modeling, the variance inflation factors (VIFs) of patient characteristics and prescribed mean daily parenteral nutrition doses were calculated to confirm that there was no multicollinearity between variables based on multiple regression analysis or multivariable logistic regression analysis [17].
To confirm the robustness of PSM, confounding factors were adjusted by multivariable logistic regression analysis or multiple regression analysis, and an adjustment analysis consisting of 2 explanatory variable groups (model 1, model 2) was performed. In model 1, the explanatory variables were the 2 groups and the patient characteristics. In model 2, the explanatory variables were those included in model 1 as well as the prescribed mean daily parenteral energy during days 4 to 10. Either ORs or regression coefficients, along with 95% CIs, were calculated for each model.
Source link
