Navigating AI in optometry: A balancing act of innovation and ethics

Image credit: JenniferToomey/MJHLifeSciencesusingAI

Integrating artificial intelligence (AI) into optometry represents an opportunity to revolutionize patient care, yet it also necessitates a thorough examination of ethical implications and professional dynamics. This article explores AI’s role in optometry, the ethical intricacies of its application, and the proactive measures the profession must adopt to fully realize AI’s potential in eye care.

When you hear “artificial intelligence,” what comes to mind? It is a term as loaded with promises as it is with misconceptions. Many fear that AI is a harbinger of job replacements and impersonal health care. They see a world where their career and expertise become obsolete. But let’s pivot that perspective.

Adopting AI in health care signifies a critical shift toward more efficient, precise, and accessible diagnostic and treatment modalities. In optometry, its potential impact would be significant, bringing about an enhanced ability to process vast amounts of data to detect eye conditions earlier, more accurately tailor patient-centered treatment plans, and more effectively prevent vision impairment.

As AI operates based on the data it receives, optometrists—who generate large amounts of data on health care and prescribed treatment options from clinical and optical or retail environments—are uniquely positioned to thoughtfully inform and refine AI tools. Enhanced data utilization is just the beginning of our expanding professional contributions. Optometrists have a deep understanding of the complexities of eye care, from primary to rehabilitative care, integrating a focus on ocular health and its implications for overall systemic health and well-being. They possess a clinical eye and vision care expertise that must be leveraged to guide the development of future AI technologies.

Urgency in advocating for eye care providers to engage in AI development is needed. Without proactive involvement, our profession is at increased risk of marginalization by other health care sectors or technology companies who may not prioritize patient-centered care. We can shape the future of AI in eye care, ensuring that these innovations enhance patient outcomes and uphold the highest standards of equitable, evidence-based patient care.

What exactly is AI again?

AI has several branches, but machine learning is currently the main area of focus. Essentially, data are converted into a digital form that computers can process and store. Machine learning algorithms are specialized programs that analyze large amounts of data to identify patterns. These patterns help to make predictions and guide decisions.

Deep learning

A major cause of recent AI breakthroughs is deep learning, an advanced type of machine learning. Deep learning uses structures called neural networks, which are complex systems modeled somewhat like the human brain. These neural networks consist of layers filled with nodes, or neurons. Each neuron performs specific calculations that help the system improve its pattern recognition. For example, in optometry, these networks can learn from a vast number of retinal images to detect signs of eye diseases. As they encounter new data, the algorithms in these neural networks can be refined to enhance their diagnostic accuracy.

Large Language Models

Large language models (LLMs), such as OpenAI’s ChatGPT, are advanced deep learning tools trained on vast amounts of text data—essentially, almost all publicly available information on the internet. These models can accurately predict the next word in a sentence, allowing them to generate coherent and contextually appropriate text. In optometry, LLMs can greatly enhance patient care by improving communication, simplifying record-keeping, and assisting in reviewing text information.

Extending the reach of patient care

AI excels in distilling complex data into actionable insights, automating time-intensive tasks, and enhancing diagnostic precision. For optometrists, this means more time focusing on patient interactions, nuanced decision-making, and personalized care—elements at the core of our profession and essential to improving health equity and outcomes.

To truly benefit from this partnership, we need to actively engage in developing and using AI across all its applications. Embracing the few select tools or uses marketed to us is not enough to keep pace with the rapid advancements in AI technology.

If optometry limits its engagement with AI to the passive use of tools in limited areas, we risk being left behind as technology advances and spreads across all aspects of health care. This complacency could manifest in several ways—from reluctance to investing in implementation to dismissing the technology entirely.

Integrating AI into optometric care presents specific ethical challenges, including unclear liability in machine-led diagnostics, potential biases in AI models based on population data, and difficulties in maintaining patient-centered decision-making. Addressing these issues requires a deep understanding of AI and a commitment to navigate these complexities responsibly.

Pitfalls and ethical considerations

The journey toward integrating AI ethically into optometry is like walking a tightrope. On one side, overly simplistic AI solutions might be convenient but can overlook the complex, holistic nature of personalized patient care. On the other side, excessively complex AI models risk embedding existing biases, operating without transparency, and potentially leading to misinterpretations and inequalities in care delivery and outcomes.

Constructed (artificial) reality

AI systems create a “constructed reality” based on the data they process, which can be limited and may not fully capture the nuances of human health. Providers and their teams must be aware of these limitations and work to ensure AI tools are well understood and used to complement traditional clinical practice. As in other health care sectors, AI systems in optometry create a constructed reality based on the aggregated data. This reality is inherently limited by the scope of the data and the imagination (or lack thereof) of the system’s designers. The challenge to care arises when doctors believe this constructed reality is omniscient.

Misleading metrics

AI often relies on proxy metrics––simplified, quantifiable indicators meant to represent complex outcomes. Although these metrics, such as diagnostic accuracy, provide a measurable snapshot of AI performance, they can be deceptive.

Proxy metrics are, by definition, stand-ins for true metrics. They are simpler, easier to measure, and provide a way to gauge success. For instance, in the context of AI-driven diagnostic tools, a proxy metric might be the accuracy rate of the algorithm in identifying specific conditions from retinal images.

AI models are often touted for their predictive capabilities, with some boasting accuracy rates of 90% or higher. However, this can be misleading without context. For example, 90% compared with what? If the answer is a proxy metric, there is a high chance of poor, unforeseen outcomes.

Data bias and quality

At the heart of any AI or machine learning system lie vast amounts of data. In health care, these can range from electronic health record and diagnostic imaging data to those generated by patients through wearable devices.

Characterized by its accuracy, completeness, and relevance, data quality serves as the foundation upon which AI systems can learn, make predictions, and assist us in supporting clinical decisions.

The effectiveness of AI in enhancing patient care depends on the quality of data, which must be accurate, complete, and representative of diverse patient demographics. Ensuring high-quality data is essential for AI systems to support effective clinical decisions and equitable care.

Comparisons of AI models with doctors often miss the fact that doctors have an intrinsic population bias to the area they practice in. Most providers do not see patients in all 50 states. AI models, however, are often created with the intent of national distribution and application, potentially missing regional differences.

Good old-fashioned reality

The stark truth is that human health, especially when it concerns something as complex and multifaceted as eye health, cannot be fully encapsulated by data sets and algorithms. Conditions that are rare, newly emerging, or manifest in atypical ways can easily fall through the cracks of AI systems. Additionally, the nuanced interplay of systemic health issues and eye health—a relationship well understood by experienced optometrists—may be oversimplified or entirely missed by AI diagnostics.

Human intelligence excels at navigating ambiguity, making nuanced judgments based on incomplete or intangible information—skills honed through years of clinical practice. AI, however, thrives in environments defined by precise, quantifiable metrics, struggling to replicate the depth of human understanding and intuition.

Charting an ethical path forward

As we chart this unexplored territory, ethical considerations come to the forefront: How can we use AI to enhance shared decision-making and include patient preferences while respecting patients’ autonomy and privacy? What mechanisms are in place to address potential biases in AI algorithms? How do we maintain the irreplaceable human element in health care amid the efficiency of automation?

The solution to these limitations is not to discard AI in optometry but to recognize and bridge the gap between AI’s constructed realities and the intricate reality of patient care. This approach entails a synergy between AI’s data-driven insights and optometrists’ contextual, holistic understanding. It involves acknowledging that the absence of evidence is not evidence of absence and designing AI systems and clinical protocols that account for this uncertainty. This balance between humans and AI is often called augmented intelligence.

Bridging the gap between AI and human insight

Optometrists can play a crucial role in AI development by providing critical feedback that contributes to more comprehensive data sets and advocating for AI systems that include mechanisms for uncertainty and the unknown. This collaborative approach can enhance AI’s potential while safeguarding against the pitfalls of constructed realities, ensuring that patient care remains grounded in the full complexity of human health.

AI policy and regulation: Follow facts and outcomes

As AI becomes increasingly embedded in health care, the need for clear, comprehensive policies and regulations governing its use becomes evident. By engaging with regulatory bodies, professional associations, and policy makers, we can advocate for frameworks that protect patient privacy, ensure equitable access to AI-enhanced care, and maintain high standards of accountability and transparency in AI applications.

This advocacy is crucial in preventing the potential monopolization of AI technologies by a few tech giants, which could lead to disparities in access and quality of care.

Key steps to adopting AI

Optometrists can champion the development of AI tools and collaborative platforms that encourage innovation and accessibility, ensuring that advancements in AI benefit the broadest possible patient population.

Be optimistic, responsible, and open

The most significant aspect of AI in optometry will be its ubiquity––a subtle yet pervasive force enhancing every facet of eye care. From clinical diagnostics to practice management, AI’s potential to elevate the practice of optometry is boundless. However, this can only be fulfilled through ethical stewardship and proactive engagement in AI’s development and refinement.

The future of optometry in the age of AI is not just about integrating new tools into our practice but about reimagining the way we approach eye and related health care and continuously refining that care. With a commitment to collaboration, contribution, and ethical oversight, we can navigate the complexities of this new era, ensuring that optometry not only adapts to but thrives in the changing landscape, setting new standards of excellence in patient care.

In essence, integrating AI into optometry is a call to action. It is an opportunity to enhance the personalized, patient-centric spectrum of care we are known for—from primary through tertiary—ensuring that as we step into the future, we are not simply following the trends but setting them. By embracing AI, we strengthen our place at the cutting edge of health care, steering eye care toward a future where technology and human insight work hand in hand to better serve our patients.