The World’s Foremost Population Health Analytics Tool | Johns Hopkins

For much of his decades-long career, Jonathan Weiner, DrPH, has worked on what has become one of the most powerful, sophisticated and, most importantly, widely used health analytics tools on the planet. Developed at the Johns Hopkins Bloomberg School of Public Health and now in use in about two dozen countries across the globe, the Johns Hopkins ACG® System has become a sophisticated yet practical tool to understand not only how a patient will fare, or a health system will prosper, but for insight into the health of entire populations.

ACGs (Adjusted Clinical Groups) represent a set of software-based algorithms for understanding both individual patients and populations of patients, whether that “population” is a set of patients in a given health plan, or a population the way a layperson might define it: a community defined by a zip code, a city, a state, or a country.

Using data found in Electronic Health Records or the administrative data of insurance companies or medical providers, ACGs are a potent tool for public health. The ACG System applies clinical, public health, and data science savvy to categorize the millions of potential diagnosis and medication combinations of every individual within a population.

The system, which can be applied in real time, is an agile vehicle for assessing and responding to health challenges, both emerging and perennial, much more quickly than in the past—think shifts in chronic disease burden, substance use disorder, or unnecessary emergency department visits or hospitalizations.

“We take clinical knowledge, public health knowledge, statistical knowledge, scientific knowledge, and make use of all the secondary electronic data that’s available in support of public health,” said Weiner, co-director with Hadi Kharrazi, MD, PhD, of the Center for Center for Population Health IT (CPHIT). “ACGs are among the best tools out there for linking the clinical, the epidemiological, and the statistical,” he added. The ACG system, the world’s leading population health analytics software, was developed at and resides in the Bloomberg School’s Department of Health Policy and Management (HPM).

Increasingly, the ACG tool has been applied to identify and adjust for risk when analyzing real world health care data, both for improving care and to speed up research and other evidence gathering. Google Scholar identifies over 14,000 published articles from around the globe where researchers have applied the ACG tool to create new knowledge. The ACG system collects, categorizes and benchmarks pretty much all useful structured information available in the computer files of providers and insurers, even if it is messy.  ACGs now also ingest information regarding the patient’s neighborhood and community to help understand environmental and socio-demographic risks they face, often referred to as the social determinants of health. Intriguingly, the ACG team has developed tools to scrape and utilize “unstructured” data—provider notes and the like in patients’ electronic charts—to give insight into the health of individuals and populations, including social determinants. These and other research and development (R&D) activities will find their way into future versions of the ACG software.

Harnessing all these new types of data for all members of a population of interest means that ACGs can help to pinpoint and rigorously document the potential lack of equity in health, both in the U.S. and in places like the United Kingdom and South Africa. “By using ACGs to assess an entire population, it lets us say ‘wait a minute, why aren’t you paying attention to this group of people?’ Just because a third of the people don’t come to the clinic, maybe because they don’t have an insurance card, that doesn’t mean they don’t deserve attention,” Weiner said.

Every week, more than 250 million lives around the world are touched by the ACG System—including many here in Baltimore and the state of Maryland. That’s partly because both the health insurance plans run by the Johns Hopkins Health System and Maryland’s Medicaid agency use it. “ACGs represent an extension of our public health practice and service roles here in the Department of Health Policy and Management,” said Weiner. “It is gratifying that we have been able to improve health and health care so broadly and with such a sustained time period through our tech transfer efforts.”

Over three decades, the tool has been refined and improved again and again—the next version is 14.0. The current ACG R&D work is housed at CPHIT, which was founded in 2012. A business management unit at Johns Hopkins Medicine helps distribute the software globally through a network of commercial partners.

The Johns Hopkins ACG® System is made available to academics or researchers around the world for a nominal fee. But the technology is also sold to hundreds of health entities, including large hospitals and integrated health systems, major insurers and payers, primary care physician groups, health care and software and tech companies, and of course government health agencies both here and abroad. That constitutes the biggest tech transfer ever at BSPH in terms of royalties and money returned to the School—tens of millions of dollars over the decades.

It’s so successful, in fact, that Keshia Pollack-Porter, PhD, chair of HPM, recently announced that Weiner and his ACG team colleagues are using a portion of the ACG royalties they’ve earned to establish a $1.6 million endowment to help support CPHIT’s ongoing mission—creating digital tools that improve the health of individuals and populations around the world.

But the ACG System is more than a moneymaker. It’s a changemaker.

Well before the current artificial intelligence (AI) tech explosion, starting in the early 1980s, the ACG project applied state-of-the-art analytics and big data to identify and predict a person’s health risks electronically. Its roots can be traced back to the landmark primary care research by the late Barbara Starfield, MD, a legend at the Bloomberg School and an early mentor of Weiner. An evangelist for primary care in an era that spawned increasing specialization, Starfield identified several key domains for measuring morbidity among children. She and Weiner, joined by dozens of other faculty and staff over the years, have built on that early observation, to create the ACGs of today, which capture, categorize and benchmark about 350 health dimensions across all ages. When all the possible permutations are combined with one another using the latest data science techniques, the system can measure a huge range of important things about the health and wellbeing of patients and populations, as well as the health care and public health systems that serve them.

As the concept of population health took hold over the last two or three decades, ACGs offered a tool for assessing, managing or financing, the health of a population within a health system or health plan. Insurers and providers used it as both a clinical and business tool. As new models of “value-based” care emerge—like accountable care organizations and health maintenance organizations—more sophisticated tools are needed; ways of amassing and interpreting data to hit targets, avoid penalties, and in general, ways to spend smarter while improving the quality of care. ACGs are commonly used to “adjust for the risk” of different sub-populations when working toward these challenging clinical and business targets.

The ACGs can detect not only what is in EHR or an insurance record but, crucially, what is not. For instance, the records may indicate a patient is on insulin, yet no diabetes diagnosis is recorded. That lack of diagnosis—picked up by the ACG tools but missed by the patient’s own care team—creates dangerous gaps both for the patient and the health system. Such gaps are also often a marker for health disparities.

For instance, Weiner recalled, when Maryland’s Medicaid agency looked at African American and white rates of health care utilization, it initially found no disparities. But when researchers applied ACGs, Weiner said, they discovered “the African American population was one-third sicker, and yet they had the same use rates. So there you go. It shows that without a case mix tool, without a morbidity assessment, you can’t really do a fair equity assessment.”

AI of course is changing health informatics, but Weiner points out AI is not the secret sauce of ACG’s success. “We are actively updating ACGs with new AI techniques, and we expect that to offer some modest improvements over non-AI advanced analytics. But the real value of the ACG analytics and predictive modeling tools we’ve applied for decades is that the system integrates clinical, organizational, social, and data science all in one meaningful and intuitive package. And, at its core, this package is intrinsically designed by leading public health and medical experts to improve health and wellbeing. AI may be a useful a tool, but alone it will never achieve all that,” said Weiner.

ACGs have gone through multiple iterations over the years, and Weiner expects that to continue. Technology keeps getting better and the ACG system is an unusually agile tool. “When we have a new development, we test it, we put it in our software, we pilot it, and within a year or two, it’s helping 250 million people,” he said.

But nothing is forever. With the rapid pace of technological change, expansion of digital data sources and advances in data science—including AI—something else may come along. That too may be birthed at CPHIT, which is currently doing R&D on tools that not only build on ACGs but also go beyond them. If that happens, it’s OK with Weiner. His mission, and the mission of the Center, isn’t to “build one specific software package or decision algorithm.” Rather, it’s to “use digital tools, whatever form they take, to improve the health of populations millions at a time.” For 30 years and counting, that is exactly what they’ve done.