AI in Health
Predicting ICU Transfers
AI models can be used to find patients who are likely to crash. The machine learning models use patient medical records, laboratory results, and vital signs from patients to find early warning signals of deteriorating condition. These models can then be used with existing patients in realtime to determine their risk of a crash and as part of an early warning system for clinicians so they can intervene before the ICU transfer is needed. The AI system can also provide reason codes for a particular patient, which can help clinicians understand where they should begin their treatment.
AI based solutions can be used to help clinicians make better decisions by narrowing the types of tests that are likely to be useful for a patient. AI models can be created using volumes of patient information from healthcare systems together with data from pharmaceutical companies to predict likely test results a given patients. This model is then deployed into an AI-driven application that can provide indications of which tests are likely to produce definitive or valuable results based on the patient’s medical history and current symptoms. With this knowledge, the clinician can pursue treatments with the best outcomes and minimize the number of tests, which saves time and reduces costs to the patient.
Improving Clinical Workflow
Clinicians are often overworked, and hospitals are understaffed. Various studies estimate a diagnosis error rate of 10 – 15% which has a huge impact on those patients and the providers. Early diagnosis of critical conditions like Sepsis or Intracranial Hemorrhage has a significant effect on patient outcomes.
AI based decision support and diagnosis can help clinicians make better decisions by incorporating more data into the decision-making process and by learning patterns that are outside the clinicians’ purview. With mobile devices integrated into the clinical workflow, AI-based decision support helps doctors and nurses by providing a second opinion or by pointing out information they may have missed. These additional insights help the clinician make a more informed decision and can actual save time, expense and patient discomfort by preventing unnecessary tests.
Claim Denials Management
An AI approach uses machine learning models to streamline the denials management process by finding claims that have a high likelihood of being paid and the highest potential value. By working these claims first, the providers and payers spend the time on those claims that are most likely to be valid and will yield the most value to patients and providers. AI models can also provide reason codes for the denial which streamlines the review process by allowing the investigator to focus on the key issues. Reason codes are also helpful for patients and providers because they can inform them of issues with their claim and help them to fix the claim for reprocessing or change future claims to avoid issues.
Predicting Hospital Acquired Infections
Using AI driven models, providers can predict which patients are most likely to develop central-line infections by looking and a variety of data including patient information, treatment history and staff history. With this prediction, clinicians can monitor high-risk patients and intervene to reduce risk. AI driven models can also identify the reasons for increased risk and provide reason codes that point clinicians to recommended treatments and preventative measures for future patients.
Predicting Hospital Readmissions
Patients with serious and chronic illnesses are treated in the hospital and then discharged. Unfortunately, according to multiple studies, up to 25% of these patients will be readmitted within 30 days to be treated again, often with less favorable outcomes. With a focus on value-based care, providers are trying to prevent unnecessary readmissions and improve patient care outcomes. Readmission can be significantly reduced by taking steps while the patient is still hospitalized, defining different actions during discharge, and taking steps post discharge to ensure compliance with home care regimens.
AI is ideally suited to tasks where the data inputs are complex and may elude clinicians. Readmissions risk prediction can require data about the specific patient’s recent care, their current condition, treatment, their home life and other risk factors from electronic medical records. AI models can use this information to provide a proactive assessment of their risk and notify clinicians while the patient is still hospitalized. AI can provide the reasons that will lead to readmission and also provide recommendations for the types of treatments that are most likely to be successful given the patient’s history. The reason codes are valuable to clinicians because they can pinpoint areas to focus on when developing a care plan for the patient and prevent unnecessary and costly tests.
Sepsis is the leading cause of preventable death in U.S. hospitals with mortality from sepsis increasing 8% for every hour that treatment is delayed. As many as 80% of sepsis deaths could be prevented with rapid diagnosis and treatment according to Johns Hopkins Armstrong Institute for Patient Safety and Quality. Diagnosing sepsis can be difficult because its signs and symptoms can be caused by other disorders and there are no reliable biomarkers before onset. Doctors often must order a battery of tests to try to pinpoint the underlying infection which further delays treatment.
AI is particularly suited to situations where the signals of an issue are hidden by “noise” that obscures the actual problem from clinicians. Diagnosing Sepsis is one of those situations. AI driven early diagnosis based on routine vital signs and metabolic levels from electronic medical records can highlight patients at risk for Sepsis before they are admitted to the ICU where it may already be too late. AI can even be used to predict courses of treatment or dosage levels that are likely to be most effective based on patient history. Though early detection and more precision care, the patient is able to receive less aggressive and less costly treatments which improves patient outcomes and is better for payers and providers.