Subscribe to our Newsletter

Subscribe to our Newsletter


Predictive tool for early detection of delirium in hospitalized patients

Hospital General Universitario Gregorio Marañon

The Hospital General Universitario Gregorio Marañón (HGUGM) is a public hospital in Madrid, Spain. It is the largest hospital in the Community of Madrid and serves more than 2 million people. It has 8,000 employees and more than 1,200 beds. It is a teaching hospital, attached to the Complutense University of Madrid.

In the year 2021, the data on healthcare activity were: Total discharges 42.624, the average length of stay 7,61, total admissions 42.531, emergency admissions 26.691 total emergencies 239.076 and percentage of emergency admissions 11,34%. The incidence of delirium in patients over 65 years old ranges between 11,7% and 18,5%, depending on the type of hospitalization unit [1–3].

Since 2018, the Hospital has been equipped with an Electronic Health Record (EHR) that encompasses relevant information concerning the patient’s acute clinical conditions, baseline medical history, functional and mental state, as well as socio-demographic and clinical variables.

The available Electronic Health Record (EHR) electronically stores a patient’s medical information and care records. It enables healthcare professionals to access, review, and update patient information quickly and efficiently. Currently, with the EHR, it is possible to conduct a comprehensive review of a patient’s medical data, including previous diagnoses, treatments, test results, prescribed medications, healthcare provider notes, and other relevant data for patient care. Additionally, it allows for the registration of comorbidity using the Charlson index[4] and assessment of pain with the relevant scale depending on the cognitive level. Delirium is assessed upon admission and whenever there is a substantial change in the patient’s condition through the Confusion Assessment Method (CAM)[5]

The confusion assessment method (CAM) is a simple tool that can be used by physicians and nurses to integrate their observations and identify when delirium is the most likely diagnosis (Figure 1). In medical and surgical settings, CAM has a sensitivity of 94% to 100% and a specificity of 90% to 95%[5,6].

Delirium is a potentially preventable complication, and several different interventions have been developed over the last decade to prevent and control it. Some of these interventions involve nursing staff, while others focus on treatment, and many attempt to prevent delirium after surgery through pharmacological interventions.

We believe that efforts should be made to improve the identification of patients at risk during admission in order to establish preventive interventions and avoid complications such as falls, increased average hospital stay, and unintentional removal of devices. The CAM has become a standard screening device in clinical studies of delirium conducted in multiple settings, including emergency departments and long-term care [7]. It takes five minutes to administer and can be particularly useful when incorporated into routine bedside assessment, however current workloads and nursing shortages make it difficult to assess on a shift basis.

Since the Hospital does not have an early detection system, we consider that incorporating a delirium algorithm into a third-party analytical solution owned by the hospital with a systematic approach would be an optimal solution for this problem.


Delirium is an acute state of confusion characterized by an altered level of consciousness and impaired attention, resulting in cognitive and perceptual disturbances that cannot be explained by preexisting dementia. Its onset is rapid, typically occurring within a short timeframe of hours to days, and it tends to fluctuate throughout the day. Although some consider delirium to be a specific type of confusional state marked by heightened vigilance, increased psychomotor and autonomic activity, and symptoms such as agitation, tremors, and hallucinations. for the purposes of this project the terms “delirium” and “acute confusional state” are used interchangeably and encompass states characterized by decreased arousal, referred to as “hypoactive delirium.”[8]

The management of delirium is primarily based on expert consensus and observational studies, as conducting controlled clinical trials with cognitively impaired patients poses significant challenges. The strongest evidence supports nonpharmacologic, multicomponent approaches for primary prevention of delirium in high-risk patients [9–11].

The importance of having a predictive tool in the hospital is crucial especially in the context of nursing shortages and night shifts. The use of an early detection system for delirium integrated with the EHR can be a valuable tool for the medical team, allowing for early detection and a more effective response to delirium, which enhances the quality of care and patient safety. The solver should provide the technological solution.


Applications closed on the 24th of January 2024 at 17:00 CET.


Evidence-Based Behavior

Evidence-Based Behavior (eB2) is a technology start-up that develops Artificial Intelligence based solutions for behavior monitoring and evaluation. eB2 has a multidisciplinary team covering research, technical, clinical, business, organizational and management profiles that have developed the eB2-MindCare system for mental health and well-being.

eB2 EarlyDel

eB2 EarlyDel is an automated system that manages patients’ delirium risk throughout their hospital stay. It uses machine learning algorithms to analyze data from EHRs, assessments by nurses/physicians, and a post-discharge app to improve clinical workflow, quality of care, and patient safety. The system provides real-time updates on delirium risk and generates various risk measures. It also offers a dashboard for data visualization and guidance for managing delirium. Secure data exchange protocols ensure patient data privacy. eB2 EarlyDel is an innovative tool for early delirium detection and improved patient care.

eB2 EarlyDel has responded to Earyldel’s challenge main objective by developing and validating algorithms for predicting the risk of delirium, with the help of a machine learning expert team. The next step involves integrating these algorithms into a product that will be developed by the eB2 team for use and commercialisation once the project is completed.

The partnership with the challenger will result in an efficient and effective solution being implemented at the HGU GM, paving the way for the market entry of eB2 EarlyDel as a valuable addition to their existing product portfolio.