Augmented Intelligence: A natural fit for healthcare
Caring for people in clinical settings is uniquely demanding. Today, clinicians and providers have to navigate different treatment options for their patients, track patient progress to adjust care plans over time, and educate family caregivers to support their loved ones. The amount of information they must evaluate and remember for each patient when making decisions is overwhelming.
As electronic health records have become integrated into care, there is much more information to consider when making clinical decisions. At WellSky, part of our focus is on developing machine learning technology to use all this patient information while decreasing clinicians' mental burden when making care decisions.
We view this process as a partnership, working hand in hand with providers to better serve their patients. Clinicians understand unique aspects of their patients in ways that don't show up in records or charts. They use their experience and intuition about what will make a person improve and make adjustments to care. However, at the same time, machine learning technology could enhance, not replace, these clinical decision-making processes. This is known as Augmented Intelligence.
What is Augmented Intelligence?
Augmented Intelligence is an alternative approach that puts AI applications in an assisting role to healthcare providers. These applications emphasize improving the clinical expert's understanding, which leads to better clinical decisions and patient outcomes.
Augmented Intelligence differs from more traditional Artificial Intelligence (AI) applications. In conventional AI development, a computer model is created to automate some form of decision-making or tasks that typically require human intelligence. For example, in healthcare, AI applications could include automated diagnoses, proactive medication changes to minimize complications, and automated discharge setting placement to maximize patient outcomes. However, in most healthcare applications, the AI industry is not ready to allow a computer model to automate decisions, especially ones that influence the care of an individual.
Augmented Intelligence at WellSky
Risk Modeling: Predicting risk for adverse events in healthcare remains one of the most challenging and vital tasks asked of data scientists. Today at WellSky, we created applications to predict adverse events like future hospitalizations, incorrect care setting fit, and estimating long lengths of care. We've implemented a range of machine learning models, spanning logistic regressions, random forest models, and XG Boosted decision tree classifiers.
Care Plan Suggestions: Setting the right course of care for patients that is effective and efficient is challenging. We've created an algorithm using Reinforcement Learning called Guidance that helps clinicians determine optimal clinical disciplines and visit frequencies for each patient, balancing clinical improvement without overutilizing visits.
Interpretability: Predicting adverse events is helpful clinically, but only if these predictions are actionable clinically. To help our clinical partners take actions to mitigate possible future events, we prioritize the development of all our solutions with interpretability in mind. We implement Shapley values, regression coefficients, and Integrated Gradients to help determine what actions must be taken clinically to prevent future patient hardships.
Healthcare is the perfect industry to use Augmented Intelligence, allowing for the expertise of experienced providers while using advanced analytical and predictive tools. Through this collaboration, clinicians will be empowered to make the best clinical decisions possible, improving patient outcomes and efficient care.
