AIMS-MN

Need

In Spain, it is estimated that 4 million elderly people suffer from malnutrition (MN). This condition causes imbalances in nutrient intake, altering body function and leading to frailty, dysfunction, hospital readmissions, and increased morbidity and mortality, thus increasing the costs associated with their care. The prevalence of MN in hospitalized elderly patients is 38.7% and 15.3% in the community. The diagnosis of MN is well defined: a) screening to detect individuals at risk; b) clinical evaluation with anthropometric measurements (weight, height, and BMI), assessment of intake, and blood tests. Current screening is based on questionnaires and interviews (Mini Nutritional Assessment, Subjective Global Assessment, and NRS-2002) which are slow (15 minutes/patient), costly and inefficient as they are only performed on 30% of total patients at risk due to the large amount of human resources required. With the current system, a 400-bed hospital would need 60 hours daily to detect patients at risk. MN screening is crucial because an individual who is not detected as at risk can never be diagnosed and subsequently treated. We need systematic, automatic, rapid, and cost-efficient screening systems to improve the diagnosis of MN.

Solution

AIMS-MN is a predictive model based on machine learning, trained with data from over 5,000 hospitalized patients at CSdM. Automatically, without the need for any intervention by clinicians, it reads the information collected in the patient’s electronic health record (EHR) and displays whether the patient is at risk of malnutrition (MN) or not on the healthcare professionals’ workstation. Compared to current systems that screen patients individually, AIMS-MN allows: a) screening of 100% of hospitalized patients, b) increases diagnostic accuracy by 30%, c) screens in less than 5 seconds the same patients that a nurse would in 62 hours, which d) will multiply the number of patients screened and, therefore, diagnosed and treated.

Moreover, its integration with the EHR allows for rapid and accurate detection without the need for manual data entry. Being an artificial intelligence (AI) based system, it improves its predictive capacity with the incorporation of new cases, allowing rapid adaptation to new populations. AIMS-MN uses non-identifying and encrypted data, allowing it to function anywhere in the world. AIMS-MN allows universal and systematic screening of all elderly patients at risk of MN. Our group has demonstrated that when we diagnose a patient with MN, the Minimal-Massive Intervention (MMI) based on adaptation of fluids and textures, nutritional supplementation and improved oral hygiene improves health outcomes, general readmissions and survival of hospitalized and malnourished elderly.

Goal

Automate MN screening quickly and accurately, increasing the efficiency of the diagnostic process, allowing for primary and secondary prevention, as well as offering treatment based on MMI.