As we near two years since Covid-19 first struck the U.S., our healthcare system continues to grapple with staffing shortages, sicker patients, and an aging population, leading to an overburdened workforce. Luckily, there are technologies – such as machine learning algorithms (MLAs) and artificial intelligence – that continue to evolve and support care teams in making important clinical decisions, leading to optimal levels of care delivery and better patient outcomes.
In the coming year, we will see more health systems rely heavily on these tools to fill gaps in staffing and ease workflows. Here are three predictions for clinical decision support (CDS) innovation in healthcare that will have a significant impact the way our healthcare workers deliver optimal levels of care in 2022.
With the Rise of AI, What IP Disputes in Healthcare Are Likely to Emerge?
Munck Wilson Mandala Partner Greg Howison shared his perspective on some of the legal ramifications around AI, IP, connected devices and the data they generate, in response to emailed questions.
MLAs in the form of CDS tools will be recognized as integral components of care management that improve clinician workflow.
With the increasing availability of individual health data from electronic health records, genetic information, and continuous monitoring devices, CDS tools can actualize personalized healthcare for individuals. MLAs can provide risk prediction and early detection of complex diseases in patients, such as sepsis, diabetes, or alcohol use disorder. Furthermore, they can aid clinicians by stratifying patients into disease risk groups and classifying disease severity in individuals.
CDS tools are already being used in many different care settings, and their potential impact is vast. With the use of large datasets and the computational power of cloud-powered technologies, these tools can integrate and analyze multiple data input variables to produce a myriad of benefits to healthcare systems. Both MLAs and CDS tools are currently used for a variety of applications, including providing recommendations for patient care, optimizing workflow procedures, and assisting healthcare providers in addressing care setting challenges. Other applications, such as using MLAs to improve diagnostic imaging, are burgeoning as well. The strong support received from radiology and imaging providers demonstrates the clinical utility and expected market growth of these types of tools.
The looming reality of 500,000 nurses retiring by 2022 continues to exacerbate staffing shortages in already strained healthcare systems. Resource-constrained providers should utilize CDS tools to address critical patient-centered exigencies such as patient condition, potential risk of deterioration, patient status and readmission or mortality probabilities. Additionally, MLAs can effectuate solutions for issues in healthcare settings; optimize patient flow and predict anticipated discharges or bed shortages; and assist with healthcare records and workflow, among other examples. Importantly, as dynamic healthcare situations evolve – such as the Covid-19 pandemic – CDS tools can also be rapidly implemented to address novel emergent patient indications and care management practices.
A Personalized Approach to Medication Nonadherence
At the Abarca Forward conference earlier this year, George Van Antwerp, managing director at Deloitte, discussed how social determinants of health and a personalized member experience can improve medication adherence and health outcomes.
MLA developers will prove the efficacy of these tools and increase their adoption rates through clinical validation, providing transparent and understandable tools, and producing unbiased and externally validated products.
The last few years have been transformative in understanding the benefits and areas of improvement of CDS tools; comprehensive reviews in peer-reviewed journals, as well as critique and feedback from the consumers and stakeholders of products and services have been published. These tools that have been found to benefit patient safety and attenuate bias in healthcare delivery in specific settings.
However, areas of improvement have also been identified: a risk of bias in the algorithm development data and subsequent results; a lack of clinical verification in prospective studies; and the real dangers of alert fatigue, in which clinicians disregard CDS alerts due to a high number of unnecessary or clinically irrelevant alerts. With this first wave of feedback and claims of more regulations needed, MLA developers have critically thought through how to address these issues. Several solutions have been discussed to provide transparent information on the algorithm design, the training and development of data, feature inputs, and potential for biased results. With this information, healthcare providers and stakeholders can more effectively understand and choose the right CDS tool for their needs.
MLA developers are at a pivotal moment to communicate with healthcare professionals, clinicians, patients, and hospital stakeholders.
Current regulatory pathways available for MLAs in healthcare settings include software as a medical device (SaMD) products with 510(k) or de novo regulatory approval by the FDA, or CDS tools overseen by the Office of the National Coordinator for Health IT (ONC). Pursuing FDA approval as a SaMD product potentially ‘locks’ the algorithm design, inputs, alert delivery system, and intended patient population, amongst other features. Separately, producing a CDS tool overseen by the ONC could allow the developers flexibility to adapt the MLAs in response to dynamic health situations (e.g.,Covid-19). The ability to update the algorithm facilitates optimizes the performance of CDS tools for specific hospital data, patient populations, rapidly changing clinical care management practices, and specific healthcare professional preferences for alert notifications. This affects the clinical relevance of the tool, clinician adoption and usability of the product, and ultimately, impacts patient benefits and outcomes. While FDA approval is not needed to deploy CDS tools in a healthcare setting, physicians and clinicians depend on such regulatory green signals as a sign of safety and efficacy. So, the onus is on developers to provide interpretable MLAs that clinicians can understand and rely on based on verified results and external validations.
Medical knowledge is growing fast, with rapidly changing care management protocols and dynamic patient conditions. We will need to augment our human skills with machine learning to provide the best possible care. A fundamental theorem of biomedical informatics is that ‘a person working in partnership with an information resource is “better” than that same person unassisted’. In alignment with this, we can optimize clinician workflow and patient outcomes through human-computer interactions and support tools. MLA developers bring machine learning data-driven analysis to practitioners with clinical expertise – and together, they are better than either one alone.
Photo: Getty Images, AndreyPopov
Andrew Pucher is the CEO of Dascena and a member of the Board of Directors. Andrew is a strategic healthcare leader, with extensive experience in developing new product markets and scaling commercial businesses. Prior to joining Dascena, Andrew served as Chief Corporate Director at Goldman Sachs, where he worked for over a decade as a healthcare investment banker financing and advising leading biotech, pharma, and medical device companies. Andrew holds a JD/MBA from York University in Toronto, Canada.
