This is the latest edition of a Manatt Health series on AI for health systems, stemming from a recent Manatt Health white paper. Click here to read the white paper and here to read the previous issue.
| Overview |
- Aritficial Intelligence (AI) can help safety net institutions better manage their workforce, improve job satisfaction and enable staff to focus more on direct patient care.
- Despite its potential benefits, implementing AI technology in safety net hospitals comes with financial and operational challenges.
- Policymakers and health care system leaders must prioritize funding mechanisms, exploring both traditional and innovative options, to incentivize and support implementation of AI tools within safety net hospitals.
|
Safety net providers and health systems, including those in rural areas, deliver care to a disproportionately higher share of governmental-insured and self-pay patients relative to other institutions. As a result of this payor mix, lower reimbursement rates associated with providing care to the underserved, and systemic inequities, safety net institutions often face funding challenges that limit their ability to invest in emerging health care technologies more readily available to other providers. For example, many safety net and rural providers have yet to implement electronic health information exchange capabilities, patient self-scheduling programs, remote patient monitoring and basic virtual consultations. Delays in adopting emerging technologies among safety net institutions, including new artificial intelligence (AI) tools designed to support care delivery and health system operations, could perpetuate existing disparities and unequal access to cutting-edge care for patients with complex and chronic health needs. With adequate financial support and tailored AI adoption strategies, safety net institutions and the patients they serve could access the benefits of AI tools already being leveraged by other provider institutions.
AI can help safety net institutions better manage their workforce, improve job satisfaction and enable staff to focus more on direct patient care. Given the known shortages in the provider and ancillary workforce at most safety net institutions, the ability to build capacity and minimize burnout is critical for maintaining effective patient care. Safety net hospitals often operate under immense strain due to limited staffing, high patient volumes and complex medical needs. This environment can lead to significant burnout among health care workers, exacerbating turnover rates and further straining already scarce resources. Implementing AI technologies can play a pivotal role in alleviating these pressures by automating routine tasks, optimizing scheduling and enhancing decision-making processes.
The emergence of virtual care during the pandemic has resulted in sustained improvements in efficiency, patient experience and deferred care—advancements that AI can further amplify. AI-powered chatbots and other tools can offer patients 24/7 access to information and support, including answering questions about medications, monitoring patients through wearables/sensors, scheduling appointments and providing guidance on managing health conditions without requiring additional hospital staff resources.
Despite its potential benefits, implementing AI technology in safety net hospitals comes with financial and operational challenges. The cost of acquiring and maintaining AI systems may be prohibitive for some institutions. There may also be resistance from staff concerned about job displacement due to automation from AI. Training and integration programs are essential to address the reluctance of long-time employees and ensure smooth adoption of new systems. Safety net providers, often serving as anchor institutions and top employers of residents within their communities, should consider these factors when evaluating the clinical and budget impacts of adopting AI systems.
Safety net institutions may struggle to implement AI due to technical constraints like poor data quality, outdated IT infrastructure and limited technical expertise among staff. These challenges can hinder the effective adoption and utilization of AI tools, potentially exacerbating disparities in health care associated with vulnerable populations. Some safety net institutions have incomplete or fragmented EMRs making it difficult to train reliable AI models that require large and diverse data sets. These providers often lack the necessary IT infrastructure and technical expertise to implement and maintain AI systems and the funding to invest in hardware, software, training and personnel to sustain the various platforms.
Call to Action: Policymakers and health care system leaders must prioritize funding mechanisms, exploring both traditional and innovative options, to incentivize and support implementation of AI tools within safety net hospitals. Potential funding sources for AI projects could include federal and state supplemental payment opportunities, such as Section 1115 demonstration waivers with support for health care IT and equity investments, or partnerships with academic institutions and tech companies to offset AI implementation costs. With proper funding and support, safety net institutions can harness the power of AI to provide high-quality care to those who need it most.
| Health Equity and AI: Mitigating the Potential for Bias |
|
As AI becomes more embedded in health care decision making, addressing the potential for bias is crucial to ensuring equitable care for all patients. Bias can surface in many areas, such as in the underlying model training data, the design choices and assumptions made during model development and through unequal access in the availability and usability of AI technologies.
Biases inherent in training data can lead to downstream disparities in patient outcomes. AI systems rely on large datasets to make predictions and recommendations, though, if these datasets lack diversity or fail to represent all population groups, some patients may not benefit from AI assistance. For example, the data used to train AI algorithms could reflect historical inequities, such as clinical trial study participation, training the model to generate biased outcomes that disadvantage certain patient groups.
Additionally, biases can be introduced through the design and implementation of AI algorithms. Algorithms programmed with specific assumptions or preferences can inadvertently perpetuate existing biases in health care practices. To counteract this risk, it is essential for developers and health care professionals to collaborate closely, ensuring that AI systems are designed and implemented to promote fairness and equity. This involves using diverse and representative datasets, actively identifying and correcting biases in algorithm design, ensuring a diverse team develops the AI and continually monitoring AI performance to detect and address any disparities that arise.
Newly developed AI frameworks like A.C.C.E.S.S. AI1, which aim to advance health equity in the deployment of artificial intelligence in health care by addressing issues such as data diversity, algorithmic bias and equitable access, can be valuable tools to provide structured guidance for integrating AI technologies in a way that promotes fairness and mitigates potential disparities. By prioritizing health equity in the evaluation and deployment of AI tools, hospital leadership can ensure technology innovation aims to improve care outcomes for all patients while addressing systemic biases.
|
1 https://www.healthaffairs.org/content/forefront/c-c-e-s-s-ai-new-framework-advancing-health-equity-health-care-ai