Anyone who has watched a dog chase after a firetruck and wonder what he will do if he catches it will understand one of the biggest dilemmas healthcare organizations are facing right now around data and analytics.
Payers and providers are accumulating massive amounts of data about their members/patients, and many are beginning to use sophisticated analytics to try to make sense out of that data. While they may be creating reports, those organizations are still having trouble turning their reports and dashboards into action. Especially for their highest-risk members/patients – that 5% of the population who account for 50% of the cost of healthcare.
Part of the problem is the way the reporting function is typically used in healthcare organizations. It tends to be backward-looking, i.e., run a report to see how we did last month. That made more sense under a fee-for-service model where reimbursement wasn’t dependent on clinical outcomes.
But in the era of value-based care, where payers and providers are sharing the risk, that backward-looking approach is no longer sufficient. Instead, reports and dashboards need to be operationalized to show what is happening now in terms of patient health, while payers and providers can work with patients to affect health outcomes.
This is particularly true of members/patients in Medicare Advantage plans where basic payments are capitated. This particular population is most likely to fall into the high-risk 5% due to age and other factors (such as a propensity toward comorbid chronic conditions), requiring a higher level of care. Payers and providers must use data and analytics to ensure they are receiving the care they need, and that providers are properly using risk adjustment factors (RAFs) so they can be reimbursed for that additional care.
Addressing rising risk
Real-time data and analytics are also important for the next tier of members/patients down – the ones at-risk of moving into the 5%. If payers and providers can operationalize the data they are accumulating on this population, they can use it to enroll those members/patients in care management or other programs that can help forestall their higher risk factors, keeping them healthier while reducing the cost to care for them. We are beginning to see this happening, where the clinical incentives for focusing wellness are aligning with the financial incentives, but the healthcare industry as a whole still has a long way to go.
It’s not just the business model that presents an obstacle, however. One of the most vexing technical challenges of trying to operationalize the data to drive better outcomes is that many traditional analytics applications are only designed to work with discreet data. Yet much of the most important data about member/patient conditions lies outside of the check boxes.
It’s contained in physician notes, images, laboratory reports, notes from outside partners such as community-based organizations that work with social determinants of health (SDoH) and behavioral health issues, and a variety of other sources. All this data must be incorporated to create a more complete, 360-degree view of each member/patient and the population(s) they belong to as a whole in order to make treatment more effective.
The key to success will be healthcare organizations’ ability to put meaningful data into the hands of clinicians in real time. Not just vital signs or readings off certain machines, but all the factors that can affect care decisions at both the payer and provider levels.
Getting selective with AI and machine learning
Of course, we don’t want to inundate clinicians with data either – forcing them to drink from the firehose, so to speak. Instead, the goal will be to use the analytics to bubble the most important factors up to the surface, and provide actual guidance for them within the electronic health records (EHR) system and other normal workflows. This is where machine learning and artificial intelligence (AI) can be a huge difference-maker in the future.
In seconds, a machine learning application can analyze all the structured and unstructured data, compare it to known patterns for patients with similar issues and characteristics (along with the outcomes), and offer guidance on potential actions clinicians and/or care managers can take to improve outcomes. Once clinicians decide on a treatment plan, the data and outcomes from that patient are then fed back into the machine learning system, helping it make continuous improvements in its ability to recognize issues as well as recommend treatment options.
We have already seen this concept work in its basic form with Watson for Oncology. While early tests have had mixed results, the potential is there. As with anything innovative, however, it’s going to take some time to fully develop. In the meantime, machine learning and AI are already proving their value across many smaller population health management challenges.
Make data actionable
The important thing right now is to change the thinking in healthcare, at both the provider and payer levels. It isn’t enough to have all the data, or even to run fancy reports and build graphically pleasing dashboards from it. It’s about taking all the data from internal and external sources and using it to help inform clinicians so they can make better treatment decisions – and guide their members/patients to become more active in their managing their own care as well.
How have you been using data? Have you been able to operationalize your analytics to change the present and the future, or are you still using it to look backwards instead?