The Covid-19 pandemic has fast-tracked use of artificial intelligence (AI) and machine learning in healthcare, with AI-powered clinical surveillance systems playing a critical role in diagnosing and tracking the disease, as well as understanding patients’ response to treatments. The pace and scope of development is just the tipping point; enhanced clinical surveillance systems are poised to transform healthcare delivery by improving patient outcomes, lowering the cost of care and delivering value.
Putting the AI in HAI
In recent years, clinical surveillance systems have evolved to meet hospitals’ surveillance, data analytic and regulatory compliance needs. In the inpatient setting, clinical surveillance plays a central role in monitoring and managing sepsis and other healthcare-associated infections (HAIs), for example.
By powering clinical surveillance systems with AI, health systems can proactively identify an expanding range of acute and chronic health conditions with greater accuracy and expediency. This will enable hospitals and communities to act before clusters, outbreaks, or critical medical emergencies escalate. The bottom line: AI-powered clinical surveillance can save lives and dollars for conditions that have proven resistant to prevention and can identify emerging viruses that might lead to epidemics or pandemics in the future.
Achieving those results will require refinement of the use of AI for clinical surveillance, culling data from inside and outside the hospital system to flag at-risk patients before an infection progresses and spreads beyond that individual patient to others throughout a hospital. Physicians can then intervene sooner by addressing modifiable risk factors, such as the use of high-risk antimicrobials and acid suppressants.
Despite its immense promise, barriers remain to the widespread and effective adoption of AI-enabled clinical support tools. Many clinicians are skeptical about the efficacy of AI in the patient setting, citing mistrust of data and concerns over the potential impact on workflow. Patients, too, are skeptical about the use of AI, concerned over privacy and safety issues that may arise from the use of AI tools to diagnose and treat their conditions. Hospitals and health systems must build trust with clinicians and patients around the use of AI, demonstrating its ability to achieve efficiencies and enhance both outcomes and the patient experience.
“Must-Haves” for Trust in AI
These are the 3 “must haves” to advance trust in AI in the clinical setting:
- Expand data access
AI’s predictive power is predicated upon the volume and variety of information available to it. As data scientists, we want as much data as possible in our data lake, as we call it. Optimizing emerging tools depends on comprehensive data access throughout the healthcare ecosystem which is no small feat since so much essential data still remains siloed, unstructured and proprietary — even in 2020. Data must undergo rigorous testing across multiple parameters to avoid gaming of results and bias. To be effective, data must reflect the patient population the AI tool is intended to serve and not discriminate based on gender, race or ethnicity, among other things.
- Foster focused collaboration
Clinicians must play a role in the development of AI tools from start to finish to ensure that they fit seamlessly into clinician workflow. Data scientists must work collaboratively with expert clinicians to validate the patterns AI finds through error analysis and feature engineering. This process will aid clinicians in making better diagnostic and treatment decisions.
- Support transparency
Related to trust, transparency is a critical success factor for AI adoption. Clinicians need “explainability”—a visual picture that confirms the AI solution is effective and illustrates how and why AI makes predictions. AI surveillance solutions should never replace clinical decision making. It should instead make the right information available at the right time to support clinicians making healthcare decisions.
With these “must-haves” in place, healthcare organizations can effectively deliver next-generation decision-support tools that combine powerful AI technology, the prevention goals of public health, and the diagnosis and treatment expertise of clinicians needed in a post-pandemic world.
Photo: shin28, Getty Images