Population health analytics is the hot new tool for hospitals, and there are many variations available. Some can really help you identify patients at risk and offer actionable insight for helping them achieve better health. Others, however, will miss patients who are likely to develop a serious condition in the near future.
The difference between the really useful analytics and the ones that provide less insight is mainly in the type of data used. Systems that rely on claims data alone will only show you those patients who have conditions that have required services in the past. That leaves out a whole bunch of people who really need intervention to prevent, not just treat, disease. These are patients whose risk is on the rise.
This is an important difference. For a hospital or health system in a value-based reimbursement environment, not knowing about the patients with rising risks means you are missing a huge opportunity to prevent suffering and reduce costs.
Use multiple sources of data, including your image archives
So what data sources should you be using? Several. Claims data should be one of the sources, but it shouldn’t be the only source. You should also be analyzing the data in your EHR, as well as a variety of socio-economic data to spot patients who have not yet been diagnosed with a condition but are starting to show subtle indications that something is developing. Another important data source that is newly available for analytics is your diagnostic image archive. Your archived images hold an enormous treasure trove of untapped data, because every X-ray, MRI, and other digital image shows not only the diagnostic target, but also surrounding tissue and organs. This data is largely ignored by the radiologist and attending physician, because they are focused on evaluating the part of the image related to a current set of symptoms.
But with machine vision analytics, you can automatically screen those images (both archived and current) for clinical indications for a variety of diseases. Often, subtle physiological changes can be seen long in advance of symptoms, but you have to be looking for them. It’s possible to do this solely with a human review, and, indeed, it’s common for all kinds of problems to be discovered in images created for an entirely separate diagnostic need. But automation makes it possible to look backward and scan thousands, even millions, of images at relatively low cost, giving new purpose to images that would otherwise just cost you storage.
Automated scanning of new studies can look for undiscovered disease while the radiologist focuses on the diagnostic target. Current technology can screen for liver failure, cardiac disease, osteoporosis, and respiratory disease, but researchers are continually gathering data and adding new analytic algorithms to expand this menu.
One advantage of the image scanning is that it offers objective data with well-defined standards. While human evaluation is still needed before treating the conditions uncovered, there is little about this process that is subjective. That means accuracy generally is high. There may be a few false positives and missed opportunities, but these will be fewer than with other types of population analytics.
The bottom line is, you need a wide array of data sources to get a comprehensive view of the health risks in your patient population.
Once you have a true, comprehensive, and stratified risk profile for your population, you’ll have a starting place for your population health program.
Analytics important, but only within a larger program
While finding that starting point with analytics is important, there is much more to population health than just identifying risk. Once you have that insight, you need to figure out how to effectively and efficiently address the health risk.
There are very few providers or payers (possibly none) who have created a true population health improvement program. There are lots of useful projects, but they are not created within a broader context. There are two reasons: First, much of the technology that enables a comprehensive approach is new, and we are all just figuring out how to use it effectively. Second, almost no one is starting at the true entry point: governance and planning.
For that, you need to identify your stakeholders and champions, and get them to the table to talk about how to change the culture of your organization from episodic, provider-focused care to preventive, patient-focused care. And then you need to carefully review current workflows and design protocols and mechanisms to make it easy for physicians and other caregivers to act effectively on the data they receive. That is a complex process. Even figuring out who gets the data and when requires in-depth knowledge of the organization and peoples’ roles within it.
So while you are gathering your insights and considering how to deliver this new style of care, take the time to put a team in place to look at the big picture, not just the current diagnostic target. By doing so, your organization will be better prepared to optimize the health and well-being of the populations served.
