Random forest model evaluation results.

Objective
This study aimed to investigate the impact of different energy levels and ingredient ratios on the nasogastric tube patency of pureed diets, optimizing the formulations to meet the nutritional requirements of elderly nasogastric feeding patients while minimizing tube blockage risk.
Methods
The study followed the guidelines of the “Chinese Resident’s Balanced Diet Pyramid” and formulated five different energy levels of pureed diets (900 kcal, 1200 kcal, 1500 kcal, 1800 kcal, and 2100 kcal) using natural food groups. The diets consisted of seven major food categories: cereals and tubers, vegetables, meats, milk, oil, salt, and fruits. The liquid formulations for the above energy levels were prepared according to the concentration standards for special medical purpose foods (FSMPs). The maximum injection force required for nasogastric feeding was measured via a texture analyzer. The nutritional components of the pureed diets at different energy levels and ingredient ratios were evaluated via West China Hospital Nutrition Software. Spearman correlation analysis, multiple regression analysis, and random forest models were used to explore the relationships between energy levels, nutritional components, ingredients, maximum injection force, and tube patency.
Results
The study revealed that as the energy density increased, the maximum injection force of the pureed diets significantly increased (p < 0.05), particularly at the 2100 kcal energy level, where the “rice‒carrot‒beef” formula reached the highest value (117.59 ± 0.26 N), whereas the “FSMP” formula at 900 kcal presented the lowest injection force (9.62 ± 0.20 N). There was a significant difference in the impact of different energy levels and formulations on the maximum injection force (p < 0.05). Spearman correlation analysis indicated that carbohydrate (ρ = 0.736) and dietary fiber (ρ = 0.668) contents were significantly positively correlated with the maximum injection force (p < 0.05). Multiple regression analysis further revealed that carbohydrates were the primary factor influencing the injection force, with a regression coefficient of 0.247 (p < 0.05), suggesting that each additional gram of carbohydrate increased the maximum injection force by approximately 0.247 N, whereas the effects of protein, fat, and dietary fiber were not significant (p > 0.05). All nutritional components (energy (ρ = 0.629), carbohydrates (ρ = 0.621), protein (ρ = 0.582), fat (ρ = 0.547), and dietary fiber (ρ = 0.544)) were significantly positively correlated with tube blockage (p < 0.05). Mann‒Whitney U tests revealed that the energy, carbohydrate, protein, fat, and dietary fiber contents in the tube blockage group were significantly greater than those in the nonblockage group (p < 0.05). With respect to food categories, cereals (ρ = 0.742) and meats (ρ = 0.766) were significantly positively correlated with the maximum injection force (p < 0.05). Specifically, rice (ρ = 0.7886) and sweet potato (ρ = 0.506) were significantly positively correlated (p < 0.05), whereas rice flour (ρ = −0.411) and milk (ρ = −0.690) were significantly negatively correlated (P < 0.05). Moreover, cereals (ρ = 0.615) and meats (ρ = 0.628) were significantly positively correlated with the risk of tube blockage at all energy levels (p < 0.05), with rice (ρ = 0.660) and beef (ρ = 0.153) significantly increasing the risk of blockage, whereas rice flour (ρ = −0.350) and milk (ρ = −0.557) were significantly negatively correlated with the risk of blockage (P < 0.05). The random forest model’s feature importance analysis revealed that carbohydrates (33.33%) and dietary fiber (23.01%) were the most important factors for predicting tube blockage, with an AUC value of 0.91, indicating strong predictive ability.
Conclusion
This study explores the impact of nutritional components and ingredient characteristics on tube patency and blockage risk in nasogastric pureed diets, revealing key optimization pathways for pureed diet formulations. The energy density and ingredient selection of pureed diets significantly affect tube patency. High-energy diets provide higher nutritional density but significantly increase the injection force and blockage risk. Diet formulations should be optimized by reducing high-viscosity and high-hardness ingredients such as rice and beef, using rice flour to replace rice, and milk as the liquid component. For high-energy demands, the carbohydrate and dietary fiber contents should be controlled to reduce the injection force requirements and blockage risk. The study also developed a five-dimensional blockage risk warning model based on energy, protein, fat, carbohydrate, and dietary fiber (AUC = 0.91), classifying low-, medium-, and high-risk levels. Low-risk patients (energy≤1400 kcal/d, carbohydrates≤200 g/d, protein≤70 g/d) are recommended to use homemade formulas, whereas high-risk patients (energy≥1601 kcal/d, carbohydrates≥241 g/d, protein≥86 g/d) should use FSMP for full feeding to balance nutritional supply and tube patency. The findings of this study provide both theoretical and practical guidance for optimizing diets for dysphagia patients, emphasizing that adjusting formulations can effectively balance nutritional supply and tube patency, reduce blockage risk, and prevent malnutrition in homemade pureed feed. This has significant implications for reducing nasogastric complications and ensuring the safety of medical procedures.