How AI is delivering efficiencies for radiologists

In the last several years there have been many predictions about how Artificial Intelligence (AI) will impact healthcare for both patients and physicians. Expectations have been high and have not always reflected the challenges of adopting and implementing AI solutions.

AI for healthcare has gradually developed over the years into an increasing number of niche solutions, many of which are now safe for patient use. The advances have been recognized by the U.S. Federal and Drug Administration (FDA), which recently released guidelines to regulate AI tools as medical devices.

Healthcare executives have acknowledged the need and desire to implement AI technologies to improve patient care and to streamline workflows. However, one of the biggest challenges has been overcoming the economic question of who bears the cost of this new AI?

While the implementation of some AI tools may not qualify directly for reimbursement, certain use cases demonstrate that they can impact ROI nonetheless. For example, The CommonSpirit health system implemented a text-based, AI patient outreach and care coordination tool.

As a result, the hospital saw a 10% decrease in the average length of stay (LOS) for new mothers, along with a 37% decrease in pre-term births. Moreover, orthopedic surgery outcomes showed a 45% decrease in LOS along with a 71% drop in the 30-day readmission rate.

The lower LOS directly impacts the hospital’s patient room turnover rate and thus top-line revenue. Additionally, the lower 30-day readmission rate reduces the number of non-reimbursable Centers for Medicare & Medicaid Services (CMS) patient events that the hospital must absorb. Just as importantly, the AI tool receives good patient reviews, thus contributing to patient satisfaction, a key performance indicator for value-based care.

Another example is the anesthesiology department at Ochsner Health in New Orleans which launched an AI scheduling system. Six months later, managers saw an increase in the average anesthesiologist engagement scores by 27%, from 3.3 to 4.2. Moreover, the AI tool helped anesthesiologists to take an additional one to two afternoons off a month. The lead study author, Dr. Dhruv Choudhry, credits this benefit “to allow for increased work-life balance so they are better able to attend events important to them.”

Patients also experienced an increase in their quality of care. This was due to an improvement in coordinated scheduling for patients, especially for those with multiple appointments on the same day. Moreover, the optimization of clinician's schedules enabled the customization of time allocated to patients based on their needs.

To promote innovation in healthcare, and to increase the adoption of AI tools, the CMS created the New Technology Add-On Payment (NTAP) program. To qualify, AI tools must demonstrate “substantial clinical improvement, time to treatment, specialist engagement, [and] patient outcomes…” to name a few.

One such NTAP implementation leveraged an AI tool to detect large vessel occlusions (LVO) and to prioritize alerts to specialists for triage. Moreover, the AI tool “…demonstrated high sensitivity and specificity along with a median time-to-notification of 5 minutes and 45 seconds across all of the sites involved.” This is particularly important as the time-to-notification of stroke data promotes early treatment and is correlated with outcomes. Beyond these benefits, the AI tool also qualifies “for reimbursement of up to $1,040 per eligible patient.”

Another AI implementation to aid radiologists in reading breast cancer screening images, enhanced the accuracy of physicians. Specifically:

AI-assisted readings showed a detection rate 38.9% greater than without the tool.

The rate of detections “missed by both” the AI and radiologists were 17% lower.

The reduction for FTMs reading times has implications for efficiency by liberating time that may be spent on additional readings, or spending more time on complex patient cases.

Hospitals and practices are facing a challenging environment that has had many negative impacts on their physicians. A European Journal of Radiology study points to the rapid growth of their workload, showing that it doubled in the course of eight years. Not surprisingly, this pressure contributes to burnout. Moreover, the practicing population of radiologists is aging, with 82% of them over the age of 45 and 53% over the age of 55. As they near retirement, this only exacerbates the gap between the demand for radiologists and their supply, as a 2019 Frost and Sullivan Report shows.

The need to provide work/life balance shows that radiologist labor hours are at a premium, requiring that they be spent on value-add activities. However, this is not easily achieved, as one study showed that radiologists can spend up to 25% of their time on non-value-add activities.

Fortunately, AI tools can help counter this issue via both triage (prioritizing suspected pathology) and workflow (automating labor-intensive tasks) algorithms.

Beyond addressing productivity challenges, AI tools are being used to improve the quality of care. For instance, the use of an AI system at a university hospital network not only increased specificity by 24%, but also dramatically reduced the average reporting delay by overworked radiologists by 315%, (from 11.2 to 2.7 days).

AI tools have advanced over the last several years. So much so that significant and positive results have been achieved in documented use cases and practice settings. Consequently, executives can confidently evaluate proven AI tools for use in their own healthcare environments. The ROI and reimbursement gains of these tools help make AI solutions financially viable for implementation.

Originally published by Axis Imaging.