Continued From Page 17
Systems would have to start by being intuitively built with basics across all IP programs and then tailored to each one to make sense . The best way to build these programs is by having IPs at the table to help develop what is needed and observed and assisted by professionals with human design and human-factors expertise .” accredited and regulated , the reporting requirements can be daunting . For example , an IP may be a part of a healthcare system that serves multiple counties and may be responsible for reporting independently to both and have some variation of the data depending on how the entity wants the data to be reported . Systems in place that would allow IPs to report the information in one location and to other platforms or retract it from platforms could be useful instead of having to upload multiple reports and forms .
HHM Which tasks would be more difficult for AI to handle ?
SCK : Staffing ratios is one of the challenges . While we see that facilities may set a metric for a hospital setting it does not account for non-hospital-based care settings for which IPs are responsible . Unfortunately , because some accrediting bodies use bed-ratios as a metric , trying to overcome this standard to demonstrate and reflect staffing needs for IPs with higher volumes of services , complex care settings and risk associated with external hospital settings covered by IPs -- such as hospital-at-home , nursing home settings and emerging care models ’ calculations -- fall short of forecasting staffing needs and hospital budgets . Furthermore , in evaluating staffing needs , because while similarities exist , no infection prevention and control programs are the same -- therefore it is hard to understand interdepartmental and intradepartmental support and how that is factored in . Regulatory reporting can change or increase regarding scope or location . For regulatory requirements , the automation of systems may not be able to sufficiently keep up with what is needed . Additional state requirements , quality and accreditation programs differ as well as constantly change .
HHM What is the learning curve of the average IP to be able to use AI effectively ?
SCK : Systems would have to start by being intuitively built with basics across all IP & C programs and then tailored to each one to make sense . The best way to build these programs is by having IPs at the table to help develop what is needed , and observed and assisted by professionals with human design and human-factors expertise . The IP workforce is shifting , as many are retiring and many new IPs are coming in , and knowledge is lost . For systems , it would be worthwhile to think about how investments can be made for them to preserve the knowledge of IPs , which can save them on costs and time when it comes to training . Many IPs receive on-the-job training and will face different learning curves as they navigate different healthcare organizations .
HHM Is there enough ROI for hospitals to
encourage IPs to start considering what AI can do to relieve some of the workload ?
SCK : Systems must be created that will calculate and make note of what IPs do . Systems must be intuitive and be used to help IPs become more efficient and in exchange , help organizations recognize that IPs need more support . Unfortunately , while I mention repetitive tasks as being a time-burden it is important to note that the role of an IP is complex . AI-based systems that automatically catalog the type of work IPs do and can calculate time spent on tasks will provide more of a true depiction of how time is spent , thus helping organizations to understand inefficient use of staffing and responsibilities that can be re-delegated for efficient use . IPs can spend a tremendous amount of time providing “ just-in-time ” education to correct something that is observed to be wrong or answering questions or requests via phone or email . These tasks are important , however the time to do them is unseen or underreported . AI , even when it comes to common answers , practices shared , and information that is repetitive , could be auto-populated or shared . For example , if someone calls me about something and I don ’ t pick up , if my phone could direct the caller to resources or provide an answer that I could approve or modify would be ideal when the workload is heavy . Furthermore , AI-based systems would have to consider the various roles in infection prevention and control programs such as epidemiologists , medical directors , analysts , administrative support , and pharmacy .
HHM Are you concerned about the limitations
of AI or the potential for healthcare to lose the human touch if AI is taken too far ?
SCK : I am not concerned about AI . We must remind ourselves that the use of artificial learning / machine learning , including deep learning platforms , use the information that human beings give it . It is up to us to remain vigilant on how we arrive at automation or outcomes that decision-making tools derive . Sometimes without having knowledge of how we arrive at certain outcomes reported , it is hard to track where flaws in the system may be . Even in the field of using automated machines to sterilize items , there are multiple checkpoints that require human verification and chemical agents to ensure that items are safe from infectious microbes before encountering patients .
HHM Any other observations or concerns to consider ?
SCK : As the role of IPs has expanded into the C-suite , government leadership roles and in general more leader-based roles , AI can help calculate the return on investment that comes from IPs ’ involvement in “ never events ” and prevention that could lead to bigger consequences and bad outcomes . The role of IPs may be hidden under quality or risk management departments that the significant role and savings is not often quantified singularly . As IPs continue to lead , build policies , programs , and science the concern is that the value of IPs to the development of infection prevention and control AI-based tools will be overlooked , similarly to many IPs being overlooked during the COVID-19 pandemic . Furthermore , when emergencies happen , AI tools can be assistive , but they do not consider the critical thinking that is needed in unique situations . AI is limited , in that it would not be able to move swiftly and efficiently for unique events that arise .
Reference
1 . Knighton SC , Engle J , Berkson J and Bartles R . A narrative review of how infection preventionist ( IP ) staffing and outcome metrics are assessed by healthcare organizations and factors to consider . Am . J . Infect . Control 52 , 91-106 ( 2024 ).
june 2024 • www . healthcarehygienemagazine . com •
19