When a person of a certain age, of a certain weight with a certain family history falls ill, they seek care. He or she describes personal symptoms, personal habits or personal experiences that may have led them to this moment. That’s the key word. Personal.
Until now care has largely been delivered based on observed or reported symptoms of singular people who then work with a doctor to interpret them. Predictive analytics offer much potential to expand the set of data that can be used in treating that person beyond just individual experiences.
Expanding the scale and accuracy of medical inferences
Predictive analytics allow doctors to scale the context they currently draw on to make inferences and correlations in treating patients.
According to a Vinnie Ramesh, chief technology officer and co-founder of Wellframe, featured in a recent Rock Health report, “Predictive analytics is not reinventing the wheel. It’s applying what doctors have been doing on a larger scale. What’s changed is our ability to better measure, aggregate, and make sense of previously hard-to-obtain or non-existent behavioral, psychosocial, and biometric data.”
Only predictive analytics expands the data pool allowing physicians to infer how patients with similar traits have responded to treatments over time — in essence, learning from historical data to make informed conclusions about future events.
We heard from mobile innovators at MobCon last month. Predictive analytics were top of mind for attendees making sense of communication and engagement in an era of business defined by the sheer number of opportunities to connect.
As we begin preparations for our flagship digital health conference in April, I wanted to take a moment to explore the potential for predictive analytics and the challenges to realizing it fully.
Reducing waste through personalized healthcare
Today, major sources of waste generation in healthcare include overtreatment, methods of care delivery and the lack of coordination.
The potential applications for predictive analytics are varied from readmissions prevention to clinical decisions to chronic disease management.
Patients have it. Investors want it.
Data. From 2012 to 2020, the amount of healthcare data is expected to grow 24,500 petabytes. To put that in perspective, according to one technology officer, an MP3 encoded for mobile play is 1MB per minute, with the average song lasting 4 minutes. At that rate, a petabyte of music would play for 2,000 years without pause.
Data streams available for use include: clinical, claims, patient-generated, patient reported, research, molecular and clinical trials, though the vast majority of venture-backed companies use clinical data (some 71%).
Only 26 percent are using patient reported data. The dearth of companies using it represents a serious market opportunity for businesses looking to stand out.
Since 2011, investors have spent $1.9 billion funding predictive analytics companies, according to Rock Health.
Building high confidence algorithms
Besides focusing less on provider data and more on patient-collected inputs, building the high confidence algorithms needed to accurately allow model predictions represent another challenge. Some of the challenges organizations have noted include:
- Incorporating new data and sources
- Reliability of modeling
- Newness of data
- Speed to distill learning and make improvements
We reached out to a local expert to gather thoughts on the power of predictive analytics and how they’re used.
Dan Atkins, Senior Director of Data Science at Optum and Co-Founder of MinneAnalytics
On a weekly basis, we score populations to determine their risk for future medication nonadherence,
We know whether your mom is going to take her medication 7/10 times. We are able to predict ahead of time before she is non-adherence and have someone engage her and determine if it’s a health literacy issue, a financial issue or something else.
Behavior predicts behavior. There is no correlation between homeownership, college graduation or economics. Any behavioral tells you can collect will make any preemptive engagement more effective.
We have developed cohorts and run 15 models on each population every week. Every month you get more sophisticated at what you do and learn a little more.
For Dan, the return on the spend allocated for a patient intervention is the cornerstone.
“If you can’t engage somebody, you can’t overcome their barriers.”
Predictive analytics offers healthcare professionals the opportunity to better anticipate patient outcomes and adjust recommendations with more certainty by training one eye on the patient and the other squarely on the data.
Stay smart. Stay fresh.
Want more great content? Download eBooks on mobile project management, app store optimization and feature prioritization from our good friends and MobCon founding sponsor, MentorMate.