INF ECTION P R EVENTION
head start on CDIs, using smarter predictive surveillance technology to predict the risk of infections days in advance. Evidence shows that when CDI happens in a hospital, it typically occurs in clusters. This means that when the first patient contracts a CDI, it’s usually closely followed by more patients in the same hospital area. If you can detect CDIs earlier, or predict them, you can not only begin treatment earlier but also isolate patients to try and prevent spread and implement other protocols. Wolters Kluwer’s clinical surveillance tool, Sentri7, is embedded into the EMR and uses a predictive algorithm to continuously monitor the inpatient hospital population in real-time. The system works by analysing disparate data, such as a patient’s vitals, procedure documentation, patient demographics, medication orders, lab results and other inputs. The algorithm combines this data with evidence-based knowledge to deliver real-time alerts and recommendations to the point of care, with a risk score assigned to a specific patient. The application of AI to big data makes this a powerful tool, as the technology can monitor more than 500,000 patients at any given time, and process four billion lab orders and 677 million drug orders every year. By consulting the CDI risk score, clinicians can identify high-risk patients, remove modifiable risk factors and initiate evidence-based care much sooner. That time advantage can mean the difference between an isolated case and an outbreak for a hospital.
The application of this technique is showing strong results – reductions of CDIs in some hospitals are as high as 83%, with the ability to predict CDIs five days early. A change of that magnitude presents a clear opportunity for AI to out-perform traditional methods of infection risk assessment.
Challenges to overcome However, the earlier days of using automated warnings in the clinical workflow have created some challenges, notably the continued problem of alert fatigue among
clinicians. Although much of this is caused by older rules-based surveillance tools, and not the latest AI solutions, there is still work to do to overcome these challenges. The solution to the problem is two-fold.
First, data scientists like us need to validate our models and prove that the results generated by AI surveillance models are robust. Only that way can we build trust in the warnings and overcome the fatigue that has built up over time. Secondly, we need to integrate AI-powered infection- prevention tools more seamlessly into the workflow, making their use more intuitive and informative. Ultimately, AI holds huge potential in healthcare – not only in infection prevention but also in imaging areas like radiology, where it could have a huge impact – but the speed at which it takes hold will be dictated by human acceptance. Many doctors are still wrestling with what it means for them. My argument is that it should always be considered as a tool that is there to support, not replace, the doctor’s expertise. AI should never be seen as a black-box approach, and that’s very important for making sure
clinicians feel they can embrace it, interpret it, and use it to its fullest potential.
CSJ
References 1 Guest, JF, Keating, T, Gould D, Wigglesworth, N, Modelling the annual NHS costs and outcomes attributable to healthcare-associated infections in England, BMJ Open 2020;10:e033367. doi: 10.1136/ bmjopen-2019-033367
2 Manaktala S, Claypool, SR, Evaluating the impact of a computerized surveillance algorithm and decision support system on sepsis mortality, Journal of the American Medical Informatics Association, Volume 24, Issue 1, January 2017, Pages 88–95, https://
doi.org/10.1093/jamia/ocw056,
About the author
John Langton leads data science at Wolters Kluwer. He has an extensive background in machine learning and holds a Ph.D. in computer science from Brandeis University. John started his career conducting research and development for the Department of Defense in the US.
64 l
WWW.CLINICALSERVICESJOURNAL.COM MAY 2021
©Rostislav Sedlacek -
stock.adobe.com
©jorge -
stock.adobe.com
Page 1 |
Page 2 |
Page 3 |
Page 4 |
Page 5 |
Page 6 |
Page 7 |
Page 8 |
Page 9 |
Page 10 |
Page 11 |
Page 12 |
Page 13 |
Page 14 |
Page 15 |
Page 16 |
Page 17 |
Page 18 |
Page 19 |
Page 20 |
Page 21 |
Page 22 |
Page 23 |
Page 24 |
Page 25 |
Page 26 |
Page 27 |
Page 28 |
Page 29 |
Page 30 |
Page 31 |
Page 32 |
Page 33 |
Page 34 |
Page 35 |
Page 36 |
Page 37 |
Page 38 |
Page 39 |
Page 40 |
Page 41 |
Page 42 |
Page 43 |
Page 44 |
Page 45 |
Page 46 |
Page 47 |
Page 48 |
Page 49 |
Page 50 |
Page 51 |
Page 52 |
Page 53 |
Page 54 |
Page 55 |
Page 56 |
Page 57 |
Page 58 |
Page 59 |
Page 60 |
Page 61 |
Page 62 |
Page 63 |
Page 64 |
Page 65 |
Page 66 |
Page 67 |
Page 68 |
Page 69 |
Page 70 |
Page 71 |
Page 72 |
Page 73 |
Page 74 |
Page 75 |
Page 76 |
Page 77 |
Page 78 |
Page 79 |
Page 80
