Research shows that one in five hospital patients without malnutrition at admission develop it during their stay, leading to twice the healthcare costs and more complications. A 2026 study of 3,382 patients in China found that an artificial intelligence model called XGBoost can predict which patients will develop hospital-acquired malnutrition with 88.2% accuracy using information collected when patients arrive, enabling doctors to provide early nutrition support and prevent serious complications.
Researchers in China developed an artificial intelligence system that can predict which hospital patients will develop malnutrition during their stay, even if they arrive healthy. The study tracked nearly 3,400 patients and found that over one in five developed malnutrition while hospitalized, leading to more complications, longer stays, and higher medical bills. According to Gram Research analysis, the new AI model can identify at-risk patients early so doctors can provide better nutrition support before problems develop.
Key Statistics
A 2026 multicenter study of 3,382 hospitalized patients in China found that 21.6% developed malnutrition during their stay despite arriving without malnutrition, with healthcare costs more than doubling from ¥20,640 to ¥48,980 for affected patients.
The XGBoost machine learning model achieved an area under the receiver operating characteristic curve of 0.882 (95% confidence interval: 0.857-0.906) in predicting hospital-acquired malnutrition risk at admission, demonstrating strong predictive performance in a 2026 multicenter study of 3,382 patients.
Hospital-acquired malnutrition patients experienced complications at 9.0% compared to 2.9% in those without malnutrition, and stayed an average of 1.4 days longer in the hospital, according to a 2026 analysis of 3,382 patients across 34 Chinese hospitals.
Male sex, older age, pre-admission weight loss, reduced calf circumference, decreased handgrip strength, and malignant tumors were significantly associated with developing hospital-acquired malnutrition in a 2026 study of 3,382 hospitalized patients.
The Quick Take
- What they studied: Can artificial intelligence predict which hospital patients will develop malnutrition during their stay using information collected when they arrive?
- Who participated: 3,382 adult patients admitted to 34 hospitals in China who stayed 7 to 30 days and did not have malnutrition when they arrived. About 29% of all patients studied already had malnutrition at admission and were excluded from the main analysis.
- Key finding: One in five patients (21.6%) developed malnutrition during hospitalization. An AI model called XGBoost correctly predicted which patients would develop this problem 88.2% of the time, using information available when patients first arrived at the hospital.
- What it means for you: If you or a loved one faces a week-long hospital stay, doctors may soon use this AI tool to identify nutrition risks early and provide better dietary support. This could help prevent complications and reduce time in the hospital. However, this research was conducted in China, so results may differ in other healthcare systems.
The Research Details
Researchers collected information from 34 major hospitals across China about patients who stayed in the hospital for 7 to 30 days. They looked at patients who did not have malnutrition (poor nutrition) when they arrived and tracked whether they developed it by the time they left. The researchers used machine learning, a type of artificial intelligence that learns patterns from data, to build a prediction model. They fed the computer information about each patient’s age, sex, weight loss before admission, muscle measurements, strength tests, and medical conditions. The computer learned which combinations of these factors predicted malnutrition development. They tested the model on a separate group of patients to see how well it worked.
Hospital-acquired malnutrition is a serious problem that develops during a hospital stay and can make patients sicker. By catching it early using AI, doctors can provide nutrition support before complications happen. This approach is important because it uses information doctors already collect at admission, making it practical to use in real hospitals without extra testing.
This study is a secondary analysis, meaning researchers used data already collected for another purpose. The study included nearly 3,400 patients from multiple hospitals, which makes the findings more reliable than a single-hospital study. The researchers tested their AI model on a separate group of patients to verify it worked, which is a strong research practice. However, all patients were from China, so results may not apply equally to other countries with different healthcare systems and patient populations.
What the Results Show
Among 3,382 patients without malnutrition at admission, 731 patients (21.6%) developed malnutrition during their hospital stay. This is a surprisingly high rate, showing that hospital-acquired malnutrition is common. Patients who developed malnutrition had much worse outcomes: they experienced hospital-acquired complications at a rate of 9.0% compared to 2.9% in patients who did not develop malnutrition. Their hospital stays were about 1.4 days longer on average (14.93 days versus 13.53 days), and their healthcare costs were more than double (¥48,980 versus ¥20,640, which is roughly $6,800 versus $2,850 in US dollars). All these differences were statistically significant, meaning they were not due to chance. The XGBoost machine learning model, which is a type of AI algorithm, performed best at predicting which patients would develop malnutrition. It achieved an accuracy score of 0.882 out of 1.0 (where 1.0 is perfect prediction), with a confidence range of 0.857 to 0.906. This means the model correctly identified high-risk patients about 88% of the time.
The research identified specific patient characteristics that increased malnutrition risk. Being male, older age, weight loss before admission, smaller calf muscle size, weaker grip strength, reduced muscle mass, and having a malignant tumor (cancer) were all associated with developing malnutrition during hospitalization. These factors make sense because they indicate patients already have nutritional stress or muscle loss before arriving at the hospital, making them more vulnerable to further decline.
Previous research has shown that malnutrition in hospitals is common and harmful, but most studies focused on patients already malnourished at admission. This study is important because it focuses on patients who arrive healthy and then develop malnutrition, which is a preventable problem. The use of machine learning for prediction is newer than traditional statistical methods and appears to work better at identifying at-risk patients early.
This study only included patients from China, so results may not apply to hospitals in other countries with different patient populations, healthcare practices, and nutrition support systems. The study looked at patients who stayed 7 to 30 days, so findings may not apply to shorter or longer hospital stays. The research is a secondary analysis, meaning it used data collected for other purposes, which may have limited the information available. Finally, while the AI model performed well in testing, it still missed some patients who developed malnutrition (false negatives), so it should not be the only tool doctors use to assess nutrition risk.
The Bottom Line
If you are facing a planned hospital stay of a week or longer, ask your healthcare team about nutrition screening and support. Hospitals should consider using AI prediction tools like the one developed in this study to identify patients at risk for malnutrition and provide early nutritional intervention. For healthcare providers, this research suggests that machine learning models can effectively predict malnutrition risk using admission data, supporting the case for implementing such tools in clinical practice. Confidence level: Moderate to High for hospital settings similar to those studied.
This research matters most for hospitalized patients, especially those who are older, have experienced recent weight loss, have weak muscles, or have cancer. Hospital nutritionists and doctors should care about this because it gives them a tool to prevent a common and costly problem. Patients planning elective surgery or extended hospitalization should be aware that malnutrition risk exists and can be managed. This research is less immediately relevant to people who are not hospitalized or who have very short hospital stays.
The AI model can identify high-risk patients within hours of admission, allowing doctors to start nutrition support immediately. Benefits from improved nutrition support would likely appear within days to weeks as complications are prevented and patients recover better. Longer hospital stays could be shortened by several days if malnutrition is prevented or caught early.
Frequently Asked Questions
Can artificial intelligence predict if I will develop malnutrition in the hospital?
Yes, according to a 2026 study of 3,382 patients, an AI model called XGBoost can predict malnutrition risk with 88.2% accuracy using information collected when you arrive at the hospital, such as your age, weight loss history, muscle measurements, and strength tests.
How common is malnutrition that develops during a hospital stay?
Hospital-acquired malnutrition affects about one in five patients (21.6%) who arrive at the hospital without malnutrition and stay for 7 to 30 days, according to a 2026 multicenter study of 3,382 patients in China.
What are the consequences of developing malnutrition in the hospital?
Patients who develop malnutrition during hospitalization experience three times more complications (9.0% versus 2.9%), stay about 1.4 days longer, and incur healthcare costs more than double those of patients without malnutrition, based on a 2026 study of 3,382 patients.
Who is at highest risk for developing malnutrition in the hospital?
Older patients, men, those with recent weight loss before admission, reduced muscle size, weak grip strength, and patients with cancer face the highest risk for developing malnutrition during hospitalization, according to a 2026 analysis of 3,382 patients.
Can hospital-acquired malnutrition be prevented?
Early identification using AI prediction tools allows doctors to provide nutrition support before malnutrition develops, potentially preventing complications and reducing hospital stays. A 2026 study suggests this approach could help identify high-risk patients within hours of admission for timely intervention.
Want to Apply This Research?
- If you are a patient facing hospitalization, track your weight before admission and note any recent weight loss. During your hospital stay, ask staff to measure your grip strength and calf circumference at admission and discharge. Record your daily food intake and any nutrition supplements provided. After discharge, monitor your weight recovery and energy levels weekly for one month.
- Before hospitalization, work with your doctor or nutritionist to optimize your nutrition status. Eat protein-rich foods, maintain adequate calorie intake, and stay physically active if possible. During hospitalization, communicate with your care team about any difficulty eating, request nutrition support early, and ask about your malnutrition risk based on your admission characteristics. After discharge, follow up with a nutritionist to ensure continued recovery.
- Use a simple tracking app to log your weight weekly before and after hospitalization. Record any hospital complications you experience and your total length of stay. Note your energy levels and ability to perform daily activities at one week, two weeks, and one month after discharge. Share this information with your healthcare provider to assess your recovery trajectory and adjust nutrition support if needed.
This research describes a machine learning model developed and tested in Chinese hospitals and may not apply equally to all healthcare systems or patient populations. The AI model is a prediction tool to support clinical decision-making, not a replacement for professional medical judgment. If you are hospitalized or planning hospitalization, discuss nutrition screening and support with your healthcare provider. This article is for educational purposes and should not be considered medical advice. Always consult with qualified healthcare professionals regarding your individual nutrition and medical needs.
This research translation is published by Gram Research, the science division of Gram, an AI-powered nutrition tracking app.