Precision Medicine for Opioid Prescribing

Precision medicine provides a framework for implementing tools and practices that incorporate granular and intersecting patient attributes into nuanced interventions that optimize for person-level health outcomes. In the domain of opioid prescribing and stewardship, first-to-market tools with single-sided, risk-based scoring frameworks and a narrow focus on behavioral compliance and metabolic efficiency threaten to crowd out a new generation of tools. Providing several advantages, the new tools assess both risks and benefits of opioid analgesic treatments—as well as the risks and benefits of withholding those treatments.

In addition to a wider scope of clinical endpoints and influences, the next generation of precision medicine supports for opioid prescribing should leverage growing research on the use of artificial intelligence (AI) and machine learning (ML) to evaluate and optimize fairness in clinical decision supports. As a hedge against historical disparities and stigmas that have undermined patient outcomes—and as an alternative to first-to-market tools designed for behavioral surveillance and control—Booz Allen strives to integrate and validate precision medicine frameworks that prioritize equity and optimize holistic health outcomes for patients embedded in communities where they live and seek to thrive. 

The National Institutes of Health (NIH) defines precision medicine as an approach to disease treatment and prevention that seeks to maximize effectiveness by taking into account individual variability in genes, environment, and lifestyle. This concept recognizes that, while normative outcomes among large populations have traditionally driven standards for medical care, nuanced and interconnected factors can and should be incorporated into clinical care and decision support systems to optimize individual patient outcomes. Implementing decision support tools that optimize individual patient health outcomes holistically requires the application of methods beyond classical statistical analyses. NIH predicts that AI/ML will play a critical role in the advancement of precision medicine over the next decade.

Early use cases based on implementations of AI/ML have shown promise in guiding the diagnosis and treatment of cancer, management of drug-resistant tuberculosis, and risk estimation for pediatric surgeries. While AI/ML applications to precision medicine have already proven useful in the prevention and prognosis of complex medical conditions, challenges related to transparency, equity, and translation to effective clinical decision support remain to be solved. Among these challenges is the consistent and careful definition of the aims of decision support tools in the context of opioid analgesic treatments, so that the framework of precision medicine is not co-opted by tools that seek to minimize narrowly defined risks without considering potential benefits, or that prioritize institutional, prescriber, and patient behavioral compliance above optimization of patients’ health outcomes.

At Booz Allen, we understand that AI-ready data is crucial to advancing precision health, and that if we are to design, implement, and scale high-quality precision medicine tools to guide opioid prescribing, we must ensure that we interrogate the focus and impact of the precision in question. This is precisely where AI/ML promises to improve the clarity of precision medicine implementations for opioid prescribing. By focusing specifically on patient health outcomes and leveraging objective analysis to identify optimal interventions independent of the healthcare system’s policies, law enforcement agencies’ prerogatives, or society’s structural biases related to access and representation in healthcare services, AI/ML-driven precision medicine implementations can provide clinicians with the insights and support they need to weigh biomedical fact against organizational policy and make decisions that best serve the patients under their care. 

Opioid prescribing and stewardship have been the targets for several early-to-market decision support implementations focused on analysis of pharmacogenetic and patient history data. Leaders from NIH and the National Institute on Drug Abuse (NIDA) have identified precision medicine as a key component of long-term strategies for addressing opioid-related public health impacts. Furthermore, the Food and Drug Administration has convened public workshops specifically focused on the design, performance, and regulation of medical devices to prevent adverse opioid-related health outcomes. Healthcare systems and providers strive to meet patients’ pain control needs while also complying with clinical and legal standards and avoiding adverse health and social outcomes associated with iatrogenic exposure to opioid medications.

However, first-generation tools to inform clinical practice have struggled to reconcile these competing aims and have been implicated in cases of patients being denied access to established standards of care based on histories of trauma. Such cases exemplify the hazards of relying on narrow definitions of risk when implementing clinical decision support for opioid prescribing. They also highlight the potential for such tools to further marginalize patients with already-recognized individual and structural vulnerabilities. As private enterprises bring opioid-related decision support products to market and regulators prepare mechanisms to review and monitor these software solutions as medical devices, it is critical that we examine and maintain the function and framing of emerging tools as implementations of “precision medicine.”

In addition to validating improvements in expected individualized health outcomes, disparities in the impact of precision medicine tools must also be considered, especially where suboptimal outcomes under normative treatment standards may be disproportionately experienced by groups already subject to structural marginalization and harm. AI/ML approaches to precision medicine may detect and reflect unfamiliar or even counterintuitive relationships among clinical care, patient attributes, and the social ecology in which healthcare operates, bringing a rigorous and transparent focus on supporting clinicians in their efforts to optimize the health and well-being of specific individuals in their care. Booz Allen has innovated in this domain with the development of the Booz Allen Ethical Authority to Operate (AI/E ATO) Assessment that provides a framework for objective, quantifiable, and ongoing evaluation of bias to ensure that outcomes are not systematically optimized for specific subpopulations and stakeholders to the detriment of others.

In cases where elevated risk of opioid-related injury is identified by decision support systems, precision medicine validations should also consider social contexts. This means recognizing that clinically prescribed and dispensed medications are not the only drugs available in patients’ lives and that restrictive medical practices driven by narrow risk-focused frameworks may give rise to nonmedical drug exposures that undermine patient and community health outcomes. A failure to consider social contexts opens clinical decision support tools to a risk of moral hazard in which a provider’s interest in avoiding law enforcement attention and complying with guidelines is prioritized over its interest in optimizing patients’ health outcomes.

As precision medicine matures and expands into new domains of clinical decision support, it is critical that the products and services marketed under the banner of precision medicine be evaluated for alignment with the promise that the construct entails. Precision medicine must remain the domain of holistic optimization of patient’s health outcomes and not be conflated with tools to mitigate risks for healthcare providers and institutions or with narrow solutions that focus on avoidance of intermediate endpoints to the exclusion of patients’ needs and wellbeing.