Predictive Analytics: the future of healthcare
The era of large scale and predictive data analytics is upon us. We frequently encounter it in our daily lives, whether it is Google’s search engine anticipating your query as you type, your iPhone suggesting the next word in your text message, or targeted advertisements on websites based on your historical internet usage. And now predictive analytics is beginning to break into the healthcare sector in a meaningful way.
A few key developments over the last ten years have paved the way to data and automated predictive analysis in healthcare. First, there has been substantial progress in the adoption of electronic health records (EHR) systems, enabling the digitization of healthcare data at a rapid pace. Second, there is an increasing focus on reducing cost and measuring healthcare quality, driven by the adoption of new models for care reimbursement as well as new regulatory guidelines. And third, technology has evolved and grown to support the development of new applications and ways of analyzing data that were not possible before.
As a result, the healthcare industry is poised for significant adoption of predictive analytics to drive the next wave of digitization and development of new models of care. We are already seeing use cases in precision medicine and the leveraging of EHR and claims data for diagnosis and treatment.
Perhaps the best-known use of analytics in medicine to date has been around the use of data to diagnose and treat cancer. Increasing activity and support from the federal government as part of the precision medicine initiative as well as from private businesses in the biotech and healthcare space have resulted in the aggregation of clinical, pathological and genomic data. In turn, large repositories of available data make it possible to study specific mutations in cancer and develop targeted therapies that are highly effective at treating cancer, without causing collateral damage to healthy tissue. As our understanding of oncology improves and the repositories of data increase in information, these therapies will become even more effective. As with all technology, we expect the costs to reduce substantially over time, which will lead to the ability to use this technology for other conditions within the next five-ten years. Perhaps in twenty or twenty five years we will see it used to predict and prevent the common cold.
EHRs contain enormous amounts of data that has yet to be fully tapped. We have seen specialty medical associations begin to build large data registries by pulling EHR data from their members. This data is currently used for research, benchmarking provider results and for reporting performance against quality measures. However, there is significant potential for further leveraging this data for large scale studies, and to identify patterns in the health of populations. More recently, interesting uses of predictive analytics based on EHR data have been used to predict mortality risk as well as risks of hospital-acquired infections and other complications. Predictive analytics models are also being used to risk-stratify patient populations at the time of discharge to ensure patients receive appropriate amount of post-acute care to reduce readmissions or further complications. Moreover, predictive analytics can be used to detect fraud, waste, and abuse within the system, leading to system wide reductions in healthcare costs.
Increasingly, we see that data can be used to determine the best treatment plan for an individual, based not only on their specific ailment, but also accounting for personal and social factors, including social determinants of health. Unfortunately, this data remains siloed for use, primarily by individual health groups or specialties. There are current use cases of analytics engines scanning millions of medical articles in order to help diagnose rare diseases. If EHR data was available to combine with medical knowledge repositories, it would give providers the ability to use analytics engines to scan actual data in addition to, articles for diagnosis. Similarly, payers are now using clinical data to predict at risk patients and to engage them with targeted messages to influence change in behaviors and drive healthier outcomes and lower overall costs.
There are many potential uses for big data and predictive analytics, yet before healthcare can derive the benefits from data analytics, it must overcome several barriers.
1. Healthcare data is largely fragmented and siloed. This is in part due to late adoption of EHR technology across the industry.
2. Common data architecture is lacking. This has made data sharing difficult because EHR systems structure their data differently and often IT departments have justifiable security concerns in sharing data. Data sharing is further complicated by lack of common standards of interoperability as well as the high amount of effort it takes for IT personnel to design and build safe and effective data integration between various sources
3. Robust data governance efforts have been inadequate, leading to low confidence in the data shared from third parties. Because of this, some physicians repeat tests so they better trust their own results. There is also less reliance on external data.
If the healthcare industry can overcome these issues, it is poised to reap great benefits for patients, providers, and payers alike.