We in healthcare are uniquely positioned to be catalysts for change in the digital world. As we mobilize analytics and better understand the healthcare delivery system and patients’ needs, we’re improving the health and lives of our communities. Big Data in conjunction with artificial intelligence has the potential to enhance collaboration between insurers and doctors/hospitals, move to a value-based payment system with more accountability, increase transparency, and improved patient engagement.
As analytic capabilities mature, the shift to innovation intensifies by turning data into rich, timely information that we can use predictively and prescriptively. A high-quality and well-governed single source of truth within a healthcare organization (e.g. an Enterprise Data Warehouse) can provide data for predictive modeling, which uses data mining to forecast the probability of future health outcomes, and prescriptive modeling to advise possible preventive actions to take.
Each model is made up of a number of predictors, which are variables that are likely to influence future results. Once data has been collected for relevant predictors, a statistical model is formulated. The model may employ logistic regression or a more complex algorithm such as neural network or random forest. As additional data become available, the model needs to be continually validated or revised.
One core enabling piece, though, is the talent—data scientists focused on solving problems in healthcare. Pairing young millennial talent trained in open source tools such as R, Python and Hadoop with seasoned experts in SAS and SPSS creates a very intellectual, stimulating environment ripe for innovation.
Applications of artificial intelligence and machine learning are designed to drive process-based interventions. Within the context of emerging collaborative models such as Accountable Care Organizations (ACOs), payers cannot only share actionable data with network healthcare providers, but also offer financial incentives for coordinating care, driving improved health outcomes and quality.
One of the biggest targets is reducing hospitalizations and read missions. We have opportunities for better care coordination, better outreach, better discharge planning and better education, and the result of success is both tremendous clinical value as well as economic value. Our predictive models can show a practice that out of the 200 diabetic patients being seen there, two seem particularly at-risk of being hospitalized. That allows the physicians to do outreach, conduct more intense interventions for those two patients and be in a better position to lower their risks of hospitalization—while still meeting optimal care measures for diabetes for the other patients who are at lower risk.
Earlier this year, Blue Cross and Blue Shield of Louisiana implemented a machine-learning-based model to make member referrals to the Population Health Care Management division, by identifying members at high risk of hospitalization in the next six months. The model has had a very high level of accuracy compared to commercially available models. By having this ability to identify at-risk members in advance, case management and disease management nurses and other clinical staff can intervene sooner for health coaching, education and self-care support, which can potentially mitigate the clinical event, save lives and reduce costs.
We also have been looking at how payers can use data to drive effective coordination between customer service and clinical care teams, determining which specific interventions are effective for improving clinical outcomes and member experience.
As we evaluate the effectiveness of natural language processing and machine learning to predict whether a member contacting his or her health insurance carrier was likely to be dissatisfied, we’ve learned such predictive modeling tools have a place in the payer-provider-patient realm because they can decrease administrative and medical costs and increase member satisfaction and engagement.