The widespread adoption of technology has enhanced the quality of human life in virtually all aspects. From a healthcare perspective, machine learning (ML) has had a profound impact on the evolution of service delivery within the sector. For instance, advances have been made in medical imaging using machine learning algorithms able to process tremendous amounts of information at mind-boggling speeds.
A case in point, Zebra Medical Vision’s platform, Profound, is able to detect every sign of potential conditions such as breast cancer and osteoporosis with a 90% accuracy rate. Another notable advancement is the increased use of Clinical Decision Support (CDS) tools. These tools assist practitioners in suggesting a tailored best treatment course; and even warn of any potential dangers resulting from previous conditions.
Technology has largely unexplored operational efficiency in healthcare
The immense potential of ML to improve healthcare operations has not been explored fully yet. Oleg S. Pianykh et al., highlight the ungodly difference between the contributions of ML to improve operational efficiency in healthcare compared to other areas such as imaging.
One of the major hurdles of healthcare systems is in managing patient flow – from entry to the point of discharge. If managed sub-optimally, patient flow can have debilitating ripple effects on the operational efficiency of a hospital. Some of these adverse effects include delay of patients in receiving proper care, occupation of the emergency department (ED) by inappropriate patients, increase in left-without-being-seen patients due to overcrowding, and delays in appropriate patient transfers.
Efficient patient flow in a hospital allows newly admitted patients to get to the right place as soon as they enter the hospital, current patients seamlessly transition to the right unit and patients who are ready for discharge leave the hospital with as little delay as possible. When hospitals manage hospital patient flow effectively, the health system and the patients win – hospitals don’t keep patients longer than necessary and patients spend the minimum amount of time at the hospital, making room for new patients who need care. This calls for encoding clinical guidelines or existing protocols through a rules-based system. The encoded protocols can then be augmented by models that will learn from the data generated.
Incorporating technology in healthcare the right way
Operational efficiency calls for more than reveling practitioners in the bells and whistles of robust machine learning models.
Stakeholders need to collaborate
To begin with, private players and healthcare professionals need to work together from the early design stage. These two parties need to collaborate as they have conflicting objectives. Private players want to scale their solutions fast, learn and iterate. However, healthcare practitioners insist that technological solutions provided should be backed by clinical evidence of quality and effectiveness. There is also the need for practitioners to understand how these models work. As a result, they will be able to address biases that maybe embedded in the algorithms used.
Have the end-user in mind when developing healthcare solutions
Design thinking is a key methodology to adopt when mapping the problems to be solved in order to build realistic models for the healthcare sector. There is little use in building tools that healthcare sector practitioners will not utilize effectively. When the end user is able to derive value, they are more likely to contribute accurate data that will in turn build more accurate machine learning models. To build a data-driven organization to enhance organizational productivity and efficiency, it is acutely critical to have the end-users understand the need to be data-driven.
Upgrade the healthcare sector curriculum to reflect new professional needs
Digital skills are important to the success of using ML in the healthcare industry. However, such skills are currently secluded or included minimally in the healthcare sector curriculum in Africa. Healthcare facilities need leaders who are dedicated and well-versed in both the biomedical and data science fields. These individuals will be able to spearhead the practical applications of machine learning into improving operational efficiency.
Moreover, the existing workforce needs to be upskilled via continuous professional development and also be incentivized to learn more on their own. Apart from the generalists’ trainings, the healthcare system needs to further build capacity by finding, attracting and retaining data science talent. These professionals will be able to develop tools hand in hand with the medical workforce which will be able to interpret the insights generated.
Strengthen data privacy and protection
Data governance and security is one of the potential roadblocks to the adoption of machine learning solutions. Patients are concerned about the use of their personal healthcare data that will be collected using the digital solutions developed for the sector. To curb this fear, healthcare providers need to set up strict governance around data management practices. These measures should be accompanied by data-sharing policies that are comprehensive, giving the data subject control of their healthcare data and anonymizing any personal identifiers where applicable.
Educate healthcare practitioners on the advantages of technology adoption
Change is sticky, even in deploying ML solutions in the healthcare industry. The clinical leadership is pivotal in making sure that use cases deployed compliment practitioners, instead of antagonizing them. The new ML solutions should aid them to deliver the best possible care, and not substitute them. These solutions should incentivize practitioners into spending more time with the patients while minimizing a lot of administration work.
In a nutshell, health systems need to address the standardization of electronic medical records (EMR), set proper data governance policies and upskill the workforce in order to unleash the benefits ML has to offer in improving operational efficiency. In addition, a conducive environment needs to be set to allow the use of ML solutions that are safe and effective while concurrently minimizing risks to both patients and practitioners.
Author: Dickson Ndoro