The Quiet Power of AI in Healthcare: Closing Care Gaps Before They Become Costly

A patient with diabetes requests a prescription refill during a primary care visit. The visit is documented, the prescription is renewed, the case is closed. What this visit fails to capture is that the patient is 18 months overdue for a retinal exam, her last HbA1c was 12 months ago, and she has an outstanding endocrinologist-recommended podiatry referral from 2 years ago. While some data elements are documented in 3 different systems, due to the existing operational model, self-service access to data is not available to the care team during the patient visit, and as such, the system allows care gaps that a visit is meant to resolve to continue unaddressed.

I have witnessed this in every health system I have worked in. Data gaps are the clearly documented gaps in care that continue to exist due to the lack of integration in data capture processes. The gap between evidence based care and what is delivered to the patient during a clinical visit is the most profound and preventable waste in the U.S. healthcare system.

The Agency for Healthcare Research and Quality has documented persistent gaps between recommended preventive services and actual delivery rates across virtually every screening category. Colorectal cancer, cervical cancer, diabetes monitoring, cardiovascular risk management, and immunizations all show adherence rates well below clinical guideline targets in large patient populations.

The economics follow the same pattern. The National Academy of Medicine estimates that missed prevention and care coordination failures add to the $750 billion to $935 billion in annual healthcare waste in the United States. Research published in Health Affairs demonstrated measurable reductions in emergency department utilization, inpatient admissions, and total cost of care for the diabetic population by closing gaps in preventive care over three-year measurement periods.

Clinicians know what screenings are due. The gap is infrastructural. The closing of referrals and the outreach to overdue services depend on systems at the point of care that are not built to communicate. The gaps are a result of the data integration problems that this newsletter has explored in earlier editions, and they show up most painfully in patients whose care is spread across disconnected systems.

AI-enabled care gap identification tools have emerged as one of the most practically grounded mechanisms for applying machine learning in healthcare operations. These tools collect data from electronic health records, claims, pharmacy, lab, and referral systems to generate patient level gap lists for clinical teams to review at any point of the patient care continuum.

The operational logic is based on a simple principle. Instead of using manual chart review or retrospective quality reporting to identify gaps, the system proactively and continuously monitors the population and identifies what is actionable. The clinical team gets a prioritized list of patients with gaps, ranked by risk, recency, and engagement likelihood.

KLAS Research has noted increasing adoption of these tools, particularly in primary care, value-based care programs, and population health management. The most effective tools are those that are embedded in EHR workflows because they are configured to identify a gap and close it. The distance between these two events is what ultimately dictates whether the information is acted upon. Clinical teams are the ones who close care gaps, working on the information the tool provides. The key is to make sure that the output is actionable, and that it gets to the right person at the right time, without adding to the overwhelming burden of notifications on their clinical workflow.

Closing gaps is directly linked to value-based reimbursement under the CMS quality programs related to care gap closures, including the HEDIS measures, Medicare Advantage Stars ratings, and the quality benchmarks associated with Accountable Care Organizations. Health systems that are more consistent with care gap closure are able to achieve better quality outcomes, which directly impacts reimbursement and creates a relationship between clinical quality and financial performance that is unmatched by most other AI use cases.

The value of care gap tools is derived from the certainty of improved quality metrics, resulting in a reimbursement impact that is measurable. This value is greater than what the majority of AI applications in the healthcare space can demonstrate.

The potential risk that I see is that there is a disconnect between the quality metric and the clinical value being delivered. A system that is active in addressing care gaps for patients who are readily accessible while neglecting outreach to patients who are disengaged may achieve good quality metrics without improving population health. Equity will need to be a component of the oversight framework for care gap closure, so that patients most likely to have care gaps, those who have disconnected care, lack of access, or are burdened with social determinants that limit their ability to engage, receive outreach that is commensurate with or greater than the effort applied to those who are easier to reach. The tool will be reinforcing the inequities that have existed in the healthcare system for years if the goal is simply to close the easiest gaps, leaving the patients who are most in need without the system or support.

Care gap tools can be effective only to the degree that the data systems supporting them are effective. A gap tool that does not include claims data will result in gaps that are documented outside the system but never captured. A tool that does not include pharmacy data is going to miss gaps related to medication adherence. A tool that does not include data about social determinants of health will lack the context to understand why gaps exist in the first place.

Here we connect the care gap challenge to the data integration and technical debt themes discussed at the beginning of this newsletter. The systems that excel at gap closure are those that make the most investment in data normalization, interoperability, and longitudinal patient identity management. The addition of AI adds value, but cannot make up for a lack of connection in the data layers below it.

Health systems assessing care gap tools should look at how complete the data environment is within which the tool operates, and not just the detection rate of the tool itself. A care gap tool that is 95 percent accurate is still missing a considerable portion of the picture if that accuracy is based on a data subset that represents 60 percent of the patient care history. The patients who are most affected by incomplete data are the most likely to have open care gaps.

The Agency for Healthcare Research and Quality offers data on the rate of delivery of preventive care and the gaps between evidence and practice. The National Academy of Medicine has estimated annual healthcare waste in the United States. Health Affairs has published research on the economic consequences of closing gaps in preventive care. KLAS Research has published reports on the use and effectiveness of care gap identification tools. CMS quality initiatives, including HEDIS, Medicare Advantage Stars, and ACO benchmarks, directly link reimbursement to the closure of care gaps.

Christopher Hutchins
Founder & CEO, Hutchins Data Strategy Consultants