Communication between patients and clinicians

It is widely recognized that the current state of communication between patients and clinicians is not optimized for either party. Patients often find patient portals to be challenging to access or not accessible in their preferred language. Many clinicians see the marked increase in electronic communication with patients as a source of stress and burnout. There are several opportunities here for AI to contribute to reducing such challenges, including allowing for real-time translation, streamlining the routing of messages to the most appropriate staff member, and reducing the effort needed to respond to messages and enter appropriate orders.

Improved triage and diagnostic clarity

Primary care practices currently devote considerable resources towards patient triage, which could be eased by more effective AI-driven patient data collection and entry. This may serve to decrease variation and improve the sensitivity and specificity of triage information. Similarly, AI-driven tools could assist clinicians in obtaining appropriate tests and decreasing unnecessary testing by applying an algorithmic approach based on information gleaned from patients and the context of all of the prior data in their medical records. In time, this may lead to faster, more effective diagnostic approaches while reducing waste in the system.

Reduced administrative burden

Clinicians widely report that many tasks that are not directly patient-facing are contributors to burnout and occupy an increasingly disproportionate amount of time. These include coding, billing, clinical documentation, and prior authorizations. Each of these are repetitive tasks that do not inherently require a high cognitive load or the skills and education of a clinician. AI-driven solutions are strongly positioned to gain clinician buy-in, as they may be framed as interventions that may decrease the time spent by clinicians on these tasks while potentially increasing the accuracy of information that is entered into the record and shared with relevant partners, such as insurance companies.

More effective population health and preventive care strategies

Current population health strategies are labor-intensive and vary widely in terms of efficacy and sustainability, and they are complicated by the occasionally divergent goals set by health insurers and the capacity of systems to meet demand for procedures, such as colonoscopies. A more sophisticated approach that leverages AI to target patients most in need of preventive services, while optimizing outreach modalities and efficiency for staff, may increase the overall success of such campaigns and reduce the costs of administering them. The use of automated, AI-driven patient outreach may also result in the accessing of services in a more patient-centered way, such as through enhancing shared decision-making and scheduling, potentially impacting overall reliability of and trust in the system at a population level.

Barriers to the full potential of artificial intelligence

There are several systemic and professional barriers that are vitally important to consider while embarking on efforts to implement AI-backed systems in primary care. The marketplace is currently extraordinarily overcrowded with competing products and companies that offer similar and generally undifferentiated services that have not yet been demonstrated to function reliably on a broad scale. While consolidation in this space is inevitable, that process is likely to take several years, and this runs counter to the sense of urgency that has instilled the health care system with an impression that such technologies must be implemented in short order. This systemic impatience may offer some competitive advantages, but also risks devoting scarce and expensive IT and other resources to suboptimal – and at times predictably doomed – projects and products.

The current commercial AI ecosystem is populated by a wide array of companies, large and small, which look to AI as a potential new or enhanced revenue stream. In some cases, such as improving billing and coding, there is a clear association between the impact of AI implementation and finances, such that structuring a cost for these services may be relatively straightforward. However, the benefits of several of the applications described above (focusing on improving clinical care itself or the work of clinicians) are not immediately or inherently revenue-generating, thus making it less likely that resource-strapped organizations will invest in these technologies. An additional and powerful countervailing influence on efforts to incorporate AI systems is that several of their potential functions, such as administrative tasks and patient-clinician communication, are currently generally integrated into the work of clinicians by necessity — and are being done without additional compensation for the time required to do them. This may generate a reluctance of health care systems to invest money in something that at present comes without a direct financial cost to the system.

Ultimately, it is clear that AI will become increasingly present throughout primary care, facing both patients and clinicians. The health care industry has historically been conservative in that it is slow to adopt new technologies (leading to ongoing utilization of fax machines and pagers well after their use more broadly all but disappeared). In the case of artificial intelligence, outside market and commercial forces have pushed systems to invest in and adopt these tools at a pace that is both exciting and risky. The potential long-term benefits are considerable, but there is also a risk of further disrupting and discouraging the already vulnerable primary care workforce with implementations that are too complex to be widely adopted by patients and frontline clinicians, lack financial sustainability, or otherwise fail to meet expectations.

How to advance successfully

As health care systems explore AI technologies, it is increasingly important to incorporate stakeholder engagement early and often. In addition to clinicians themselves, staff from IT, administration, legal, and finance, as well as patients and other relevant groups, should be included in discussions regarding whether to invest in an AI integration. These tools should align with systemic priorities and should not be implemented simply because they exist. A long-term strategic vision, regularly updated by a consistent, empowered, and diverse team, can help guide where scarce financial and IT resources are invested. This may assist in disregarding much of the distracting noise that is caused by the confluence of established technology companies and new entrants all racing to push out new AI-backed products to extract money from this perceived bonanza.

There is a lot about artificial intelligence that is exciting and carries immense potential. With effective leadership and a collaborative, deliberative approach, health care systems can be well-positioned to leverage these technologies to revolutionize how care is delivered and help rescue a primary care system that is struggling to serve patients while retaining staff and functioning sustainably. It is incumbent on everyone operating in this system to approach AI with the same care that we would any of our patients, trying to offer the best, most effective management without inflicting harm.

About the author

Aaron Hoffman, DO, MPH serves as the Director of the Phyllis Jen Center Procedure Clinic at Brigham and Women’s Hospital, Co-Director of the Program in Family Medicine at the Harvard Medical School Center for Primary Care, Medical Director for the PA Studies Program at the MGH Institute of Health Professions, and as Instructor in Population Medicine at Harvard Medical School. He previously served as the Chief Clinical Innovation Engineer at Atrius Health.

*Feature photo obtained with standard licenses on Shutterstock.

Interested in other articles like this? Subscribe to our newsletter.

Interested in contributing to Perspectives in Primary Care? Review our submission guidelines.