Healthcare Analytics with Cortana Intelligence

  • Article

This post is authored by Shaheen Gauher, PhD, Data Scientist at Microsoft.

U.S. healthcare spending is expected to reach $4.8 trillion in 2021, accounting for one-fifth of the U.S. economy! The total annual cost of healthcare for a typical family of four with employer-provided PPO insurance coverage, as estimated by Milliman, is about $25,000 in 2016. Meanwhile, medical debt and bankruptcies still threaten the solvency of many American families. According to some estimates, waste-spending constitutes one-third to nearly one-half of all US health spending. There is an urgent need to disrupt the status quo and find smart ways to reduce costs and optimize resources while still focusing on better patient care.

Using the power of the cloud and advanced predictive analytics, it is possible to use the vast amounts of rich data we have available to us to provide tremendous insights and discover solutions to some of these problems. Healthcare

Healthcare solutions are needed in two primary areas – managing finances and managing patient health. 

  • Finance solutions would address problems such as managing the revenue cycle, predicting claim amounts, when claims will be paid, and also resource allocation and inventory management problems, such as forecasting the demand for medicines or equipment to ensure adequate availability while cutting down on waste.
  • Managing patient health can include solutions for reducing emergency visits, predicting patient readmissions, predicting crises situations, such as patient cardiac arrests when in an ICU, predicting the propensity for diseases such as diabetes or breast cancer, and even predicting whether or not surgery is advantageous or required in some situations.

Using comprehensive patient data – from hospital visits, clinical data, lab results, integrated Electronic Medical Records, and more – we can build models that can learn from past treatment regimes, detect patterns from symptoms and detect potential issues, recommend preventive screening and provide proactive treatment measures. A shift from a reactive paradigm to a proactive one using machine learning and advanced data analytics can improve patient health and reduce costs across the spectrum. As they say – a stitch in time saves nine!

What's more, we can supplement these models with additional data sources such as behavioral and social data feeds to build models for personalized patient care, and that can help in many ways, including reduced needs for office visits or even early warnings around potentially life-threatening issues.

A primary requirement at the start of a data science project is to identify a sharp question that needs to be addressed, and then identify the data that can help answer that question (and then ask more questions!) For example, if the question we are trying to answer is, "When will this claim be paid?" we need to know what the definition of a claim being paid is, for the company. Is the claim closed and considered paid if the balance is paid in full or is below a certain amount? Is the starting point for a claim the day the patient visited the doctor or the day when the claim was submitted? Is the data consistently collected and reported by all providers? Do we want to know when this claim will be paid by primary insurance or secondary insurance? In summary, the more clarity we have around the end objective, the better the feature engineering work we can do, and the more accurate the model and end results.

Machine Learning in Healthcare – Some Use Cases

Predict At-Risk Patients in ICU, at Cleveland Clinic Cleveland Clinic, a non-profit academic medical center providing clinical and hospital care, teamed up with Microsoft to use predictive and advanced analytics to identify potential at-risk patients under ICU care. Using the data collected from monitoring units in ICUs over a period of time, they use Azure Machine Learning to predict if a patient will need to be administered a vasopressor in the near future to prevent a cardiac failure. A timely prediction can mitigate a crisis and facilitate timely intervention by medical professionals. Given the shortage of primary care physicians and nurses, ML models like these can provide a helping hand and an extra set of eyes for overworked professionals, especially in a crisis situation. Accelerate Claim Automation & Revenue, at GAFFEY Healthcare GAFFEY has targeted workflow processes, helping its customers speed up their payment collections, while keeping labor costs lower by eliminating non value-added touches.  Identify Students at Risk for Dyslexia, at Optolexia Using a repository of eye-tracking data and an analytical engine built with cloud-based Microsoft Azure ML, Optolexia aims to help schools identify students at risk for dyslexia significantly earlier than current screening tests, ensuring that students can receive appropriate treatments early which helps boost their learning skills and improves academic performance.  Identify Asthma, at Aerocrine Using Microsoft Azure, Aerocrine captures and manages real-time data about FeNO devices, used by hospitals and asthma clinics around the globe to identify asthma and monitor patients' progress in controlling the disease. The company hopes to use the solution to achieve its ultimate goal – helping physicians diagnose asthma and assisting patients in managing their symptoms.  Deliver Personalized Healthcare, with Dartmouth-Hitchcock ImagineCare Dartmouth-Hitchcock has ushered in a new age of proactive, personalized healthcare using Cortana Intelligence Suite. Their ImagineCare solution is built on Cortana Intelligence Suite, Microsoft's machine learning, big data and perceptual intelligence offering. It will change the way people interact with the healthcare system, putting patients at the center and ultimately changing the way we all think about our health.  Predict Real-Time Patient Risk Factors, at Medical Information Records Medical Information Records LLC, a leading provider of medical software technology is using Azure ML to deliver real-time predictive analytics capabilities. Anesthesiologists are able to predict real-time patient risk factors to proactive take preventive steps. 

While there are tremendous opportunities for using ML and data science in the healthcare domain, the challenges are equally pressing. One of the main challenges has to do with the quality and consistency of data collection. Healthcare data comes from various sources and, often times, the data collected is incomplete or inconsistent between different sources. Combining these data sources meaningfully to create a complete timeline of events often results in chunks of missing data. There is also a need for open and honest sharing of information while still maintaining data anonymity and privacy requirements.

As more healthcare providers realize the potential of ML and analytics, awareness around data requirements is growing. Providers are investing in relevant data capture, data warehousing and data cataloging efforts now more than ever. As a fully managed big data and advanced analytics suite, Cortana Intelligence can help organizations transform data into intelligent action. Right from ingesting data from various sources to building custom ML models, retraining models with the latest data, consuming models via automatically generated Rest APIs, getting real-time predictions and state of the art visualizations, we have all the tools needed to create and deploy end to end solutions and with fast turnaround times.

We are excited by our partnership with leading-edge customers and partners in our push for data-driven transformation in healthcare, and we can see how this will lead to better outcomes, both for patients and for healthcare providers.

Shaheen
@Shaheen_Gauher

 

References: