. Author manuscript; available in PMC: 2024 May 1.

Published in final edited form as: JPEN J Parenter Enteral Nutr. 2023 Mar 22;47(4):541–549. doi: 10.1002/jpen.2493

Table 4.

Multivariable model for association between patient characteristics and use of TPN/EN at either baseline and/or 48 weeks.

	Odds Ratio^† [95% CI]	P
BMI (per 1 kg/m² unit)	0.87 [0.81, 0.93]	< 0.001
Early Satiety Score (per 1 point)	1.79 [1.16, 2.75]	0.008
Nausea (per 1 point)	1.58 [1.20, 2.08]	0.001
Bloating (per 1 point)	0.76 [0.59, 0.98]	0.03
Narcotic use (yes v no)	2.33 [1.15, 4.69]	0.02

A total of 837 patients were included in the model, with 38 (4.5%) having ever been on TPN/EN and 799 (95.5%) not. The final model shown was selected using a forward and backward stepwise selection approach with a minimum AIC for selection criteria.

Candidate factors analyzed included acute vs. insidious symptom onset, BMI, number of hospitalizations in prior year, PAGI-QOL overall score, narcotic use, gastric retention at 4 hours, nausea/vomiting subscore, bloating subscore, early satiety/postprandial distension subscore, and upper abdominal scores.

NOTE: CI = Confidence Interval

^†

Odds ratios from a multivariable logistic regression model with use of TPN/EN as the dependent variable and all characteristics shown in the table as independent variables.