Health-care quality improvement and emerging technologies: the potential and the pitfalls

Abstract

Health-care quality improvement and emerging technologies: the potential and the pitfalls Even with continued global efforts toward improvement of health-care delivery, equity, access, and outcomes, many countries struggled with quality improvement during the COVID-19 pandemic. The pandemic has further exposed inequities and inconsistencies in health-care delivery across the globe [1]. Over the last two decades, the health-care industry has undergone rapid changes for inclusion of emerging innovative technological advances with the potential to improve quality outcomes and reduce inequity and costs [2, 3]. Inconsistencies in care delivery drive the type of inequality that perpetuates high morbidity and mortality as well as poor outcomes. Emerging technologies under the umbrella of artificial intelligence (AI) and advanced analytical methods offer promise to raise the standard for quality improvement. Yet, the introduction, regulation, and evaluation of AI in technology are complex. Greenhill et al. [4] described AI as an umbrella term culminating in the simulation of intelligence behaviors in computers that allows them to train themselves to become better at their analysis. The AI umbrella includes machine learning and deep learning methods and tools that function based on probabilistic computer algorithms [5]. These algorithms have the capability to explore and analyze complex, large data sets [5] and glean insights that would be otherwise impossible [6]. These large complex data fall into the definition of 'big data'. Big data is described in terms of its orientation to the four Vs [7]: volume, veracity, variety, and velocity. • Volume is the amount of data. In today's world, the inter-net of things (IoT) and sensor data constantly produce an exhaust of data not seen before. • The veracity of data relates to the lack of quality and accuracy of data being produced. • Variety describes the various ways, methods, and types of data being produced. • The velocity of data describes the speed as well as the constant collection of data. Thus, big data represents a big opportunity under the right conditions. AI applications that leverage combinations of big data from unconventional sources (e.g. IoT and other sensor data) offer new perspectives [6]. While AI does not represent a new innovation, the accessibility of big data, its analyses, and application broadly in health care is. Many clinicians are leveraging AI applications to advance patient care and operations effectiveness [2, 6, 8]. For practitioners, there are a range of opportunities to apply machine learning and deep learning methods to patient safety, project management, and performance improvement. Quality improvement practitioners often collect data to evaluate the snapshot of a process. This snapshot may lack context to other processes in the organization, thereby creating a 'whack-a-mole' paradigm of improvement whereby one problem is solved but other unrecognized ones then emerge. The collection of complete data sets or streaming data offers an opportunity to view organizational quality from a more holistic perspective. One such application of AI might be around human factors and patient safety. Consider a complete de-identified surgical data set collected for a time period on high-risk surgeries. The data sets may include: the hospital surgery repository, human resources data on days off (e.g. sick time or vacation), and scheduling of operating room staff. AI offers an opportunity to look across these data sets for patterns that point to fatigue or other human factors considerations. While these data might not seem congruent, machine learning tools may uncover important dimensions of interest for improvement. Big data has been successfully used in health care in this fashion for population health management [7] and other areas. While the potential is exciting, there are challenges related to the workforce [9] and competencies among health-care leaders toward enterprise adoption [4]. The challenges with application of AI methods in health-care quality stem from the vastness of tools and complexities of data mining [10]. This may be less of an issue for quality practitioners as there are natural synergies between the statistical foundations in quality improvement and AI methodologies. For quality improvement practitioners that use advanced statistical methods such as Six Sigma (e.g. Green Belt and Black Belt levels), the methodologies may overlap. However, with any new technological introduction, there are many, key considerations for use of AI methods in health care [4]. They are improperly trained AI models and their ability to perpetuate biases in care delivery, ethics and privacy considerations from the use of sensor data, and AI decision-associated liabilities for mistakes. Quality practitioners as the guardians of safe, accountable, and transparent care must ensure that generalizable improvement is reliable and reduces harm. Emerging technologies offer a new perspective on quality improvement and patient safety as they can help us see beyond the current paradigms.