Analytics in Action News

How Health Systems Implement Ethical Artificial Intelligence

Sanford Health’s artificial intelligence ethical principles are in line with AI guidance released by WHO.

Artificial intelligence

Source: Getty Images

By Erin McNemar, MPA

- As artificial intelligence capabilities in healthcare continue to advance, regulations and policies need to be put in place to ensure health equality when exploring this new method.

This past June, the World Health Organization (WHO) released the first global report on artificial intelligence in healthcare along with six principles to avoid health disparities. The report, Ethics and Governance of Artificial Intelligence for Health, results from two years of consultations organized by a panel of WHO-appointed international experts.

While artificial intelligence is a growing method in streamlining efficiently in healthcare, the WHO report cautions AI users of the pitfalls to avoid when implementing the technology.

Considering the guidance, many organizations and hospitals have begun implementing WHO’s recommendations. According to Sanford Health Vice President of Data Analytics Doug Nowak, the health system’s AI ethical principles already reflect many of the proposals in the WHO report.

The six principles published by the World Health Organization are consistent with how our data analytics team approaches this work at Sanford Health. We are focused more specifically on augmented intelligence, recognizing its immense potential to inform clinical decision-making and meaningfully improve the health of our patients while also being careful about its limitations,” Nowak told HealthITAnalytics.

To prevent the proliferation and perpetuation of health disparities, the WHO report released six guiding principles to ensure that algorithms are trained correctly and that AI is meeting the community’s needs:

  1. Protecting human autonomy
  2. Promoting human well-being and safety and the public interest
  3. Ensuring transparency, explainability, and intelligibility
  4. Fostering responsibility and accountability
  5. Ensuring inclusiveness and equity
  6. Promoting AI that is responsive and sustainable

According to WHO, these principles will promote and support the full potential of AI in healthcare and prevent care disparities.

At Sanford Health, the organization is working to follow all six of the principles. However, principles one and four are paramount for the health system to protect patient privacy and safety while ensuring the provider has various tools to support the clinical decision-making process, according to Nowak.

“Sanford Health goes to great lengths to protect patient information and privacy through an in-depth review process. Additionally, the data analytics team at Sanford Health engages physicians in both practicing and leadership roles as soon as we begin developing AI algorithms to ensure proper training and use of these tools,” Nowak said.

To ensure success in their augmented intelligence practices, Sanford Health created an AI oversight committee. The health system created an oversight committee at Sanford Health to accelerate opportunities for innovation in artificial intelligence. The committee comprises Sanford Health leaders and researchers who review and consider proposals from providers.

“If the committee decides that the proposed solution could help to meet the needs of our patients, the provider or subject matter expert partners with a data scientist to design the new AI tool,” Nowak explained.

Sanford Health has used AI in electronic medical records (EHRs) to improve patient outcomes by identifying potential behavioral health issues, breast cancer risk, sleep apnea, and diabetes. Additionally, Sanford Health implemented AI to assist with the health system’s response to the COVID-19 pandemic.

In the report, WHO discusses the use of AI for predictive analytics in the healthcare space. With predictive analytics, medical providers can determine a patient’s disease development or health outcome.

“We see tremendous value in applying predictive models and risk stratification to our patient population if done carefully and with the patient at the center. For example, we can leverage the data to deliver more personalized care through predictive models that take into account several aspects of a patients’ health histories,” Nowak said.

By using augmented intelligence to sort through EHRs at Sanford Health, providers can identify patients who may be at increased risk for developing chronic diseases.

“Based on the information provided through the algorithm, a physician may develop a breast cancer screening care plan for his or her patient, or prioritize access to behavioral health services, or intervene earlier in a potential diabetes diagnosis to minimize more serious complications — ensuring a more personalized approach to patient care and better health outcomes,” Nowak said.