Artificial Intelligence to the Hospital's Rescue

Digitalisation in hospitals has not translated into an increase of the time spent by medical personnel with patients. Could artificial intelligence help?

Digitalisation in hospitals has not translated into an increase of the time spent by medical personnel with patients. Could artificial intelligence help?

Article translated from the original French version.

"Healing sometimes, relieving often, listening always." This is argued as the ideal role of a healthcare worker. But the decline in personnel in European hospitals combined with the increase in their administrative load is affecting their relationship with patients. Such involuntary multi-tasking creates a mental burden that has been well described as the cause of "caregiver burnout".1

Public hospitals are struggling and there is no doubt about the impact on the quality of patient care. The mental burden on nurses cannot be overlooked. One of the factors identified – even among carers who are not "technophobic" – is the computerisation of patient records.2

Going back to the 2000s, this technological advance was supposed to save caregivers time, allowing them to return their attention to the human dimension of care. But this was not the case. Twenty years later, nurses no longer do rounds with the physicians because they have to enter vitals into the computer and monitor the medical care. Physicians also spend hours digitising clinical examinations and managing correspondence.

Could artificial intelligence provide a second chance?

If computerisation has proved so time-consuming, it is above all because it was carried out without the expertise of its main users. But the lesson has been learned. Now companies and start-ups in the sector are working hand in hand with healthcare professionals to improve their daily lives. Examples include the project PariSanté campus3 or the Digital Medical Hub4.

AI and software designed by their future users are now able to optimise the flow of care and improve physicians' productivity, while minimising the risk of human error. The financial costs are also becoming more manageable.5-7 In Europe, estimates show that about 9 in 10 hospitals are regular users of one or more AI solutions in four important areas: data automation, logistics, diagnostics, and care.8 

New generations of physicians are involved in the development of digital tools, as long as they respect four pillars: usefulness, simplicity, adaptability, and applicability.9

AI supports computerisation

The use of speech recognition in hospitals is a first major step forward. Several studies have explored the advantages of these solutions10,11 which do not appear to be as reliable as human transcription. But thanks to deep learning, these AIs continually "learn" and the algorithm gradually adapts to the voice and then autocorrects. 

Another area where AI is taking hold and developing is in medical records. The AP-HP turned to Watson, an IBM software, to facilitate the use of Orbis, which is a hospital’s tedious and controversial Electronic Patient Record. The AI analyses what happens on the user's screen and provides recommendations, or intervenes through a chatbot.12

This "intelligent agent" guides the user on request, showing them step by step how to carry out a procedure. But it can also appear by itself to indicate an error or an omission in the input, which might, for example, jeopardise the patient's safety.

A helping hand in logistics

Another way of reducing the mental burden, especially for paramedics, is through help in logistics. Turkey's Bayındır Hospital in Ankara and Odense Hospital in Denmark use AI to support multi-site inventory systems. They oversee stock and order management, the distribution of people and equipment and also the scheduling of operating theatres.

Based on the planning of surgical blocks to trigger the purchase of pre-operative materials, AI reduces inventory costs and improves organisational efficiency.7 This AI plans and assigns tasks to logistics robots as well as to human employees.

Smarter hospital stays, and follow-ups

In the European Union, the population of people aged over 80 will double from 6.1% in 2020 to 12.5% in 2060. This means that the need for hospitals will increase, and that there will be a need to better distribute the occupancy of hospital beds. 

Demand for care is a volatile variable, dependent on seasonality (especially for tourist areas) and the geographical location of a hospital. The management of patient flow can be optimised thanks to AI. Calyps, an algorithm used since 2021 at the Valenciennes hospital13, can predict patient flows one week in advance. If we reduce this time window to 48 hours, the reliability rate climbs to 95%.  

Age of the "wearable"

The rise of "wearables" (connected watches or clothing) and the systematic collection of health data are expected to make remote monitoring a reality. Again, AI is already capable of interpreting this data flow. Chronolife14 and its waistcoat for monitoring vital parameters are one example.

Algorithms such as Hillo15 are also used in endocrinology departments to monitor diabetic patients remotely. Decompensated patients only need to stay in hospital for one or two days, and the rest of the insulin adaptation is done at home thanks to the algorithm. This frees up many beds. 

In Bolzano, Italy, an AI monitors patients with diabetes and rheumatism. The algorithm schedules the various laboratory tests, medical examinations and hospital stays. In addition, the waiting time to access medical resources is cut, which has a considerable impact in terms of preventing complications and re-hospitalisation.7

The use of diagnostic AI in medical laboratories, radiology and pathology centres should also reduce the waiting time for results. This will help decrease diagnostic errors.16

As medical knowledge skyrockets, AI helps diagnosis

In 1950, the time it took for medical knowledge to double was estimated to be 50 years. This was reduced to 7 years in 1980 and to 3.5 years in 2010. As of 2020, it is estimated to be 73 days.17 Our brains are struggling to keep up.

The first commercial application of Watson AI, Watson for Oncology, focused on tailoring treatments for cancer patients. The AI compares a specific patient's information to a vast database, updated weekly, that includes millions of pages of medical literature (medical journals, guidelines, clinical trials, data from electronic medical records, treatment history of similar patients, etc.). With each new patient, Watson improves its accuracy.18 This is invaluable in defining a personalised treatment plan.

Data for the benefit of the public sector

There are many examples of partnerships between the public and the private sector. Should we be concerned that businesses or start-ups will take over public hospitals? In fact, the computerisation of patient data makes it possible to create data warehouses within public hospitals, data that can then be used to train algorithms specific to certain hospitals. 

This is the case in Bern, Switzerland, where an AI assists in the decision-making process during childbirth. The cardiotocography reading alone is enough to know whether to opt for a caesarean section. In Kuopio, Finland, AI readings of angioscans are used to predict patients at risk of coronary heart disease. The same AI is also used to determine the best treatment strategy.7

Over the past 20 years, public hospitals in Europe have undergone profound changes. Despite an undeniable desire to go digital, many attempts at computerisation have proved unsuccessful or even detrimental.  

In the face of chronic staff shortages and a surge in burnout cases, AI offers a second chance to public hospitals. If they are to seize the opportunity, healthcare workers will have to overcome both the culture shock of working with the private sector and the legitimate apprehension about new technologies. The question of whether AI alone can restore the attractiveness of the public hospital remains to be answered. It would be quite a gamble, and medical and economic studies in this field are still rare and contradictory.

About the author: Joris Galland is a specialist in internal medicine. After working at the Lariboisière Hospital (AP-HP), he joined the Bourg-en-Bresse Hospital.

References
  1. AFSOS - Association Francophone des Soins Oncologiques de Support - 2014
  2. Morquin D, Résistance légitime sans technophobie?: analyse des impacts de l’informatisation du dossier du patient sur le cœur du métier médical. Revue de Médecine interne, 2020
  3. https://parisantecampus.fr/
  4. https://www.linkedin.com/company/dmh75/
  5. Yu K-H, Beam AL, Kohane IS. Artificial intelligence in healthcare. Nat Biomed Eng. 2018 (10):719–31.
  6. American Medical Association – 3 ways medical AI can improve workflow for physicians
  7. Weng SF, Vaz L, Qureshi N, Kai J. Prediction of premature all-cause mortality: A prospective general population cohort study comparing machine-learning and standard epidemiological approaches. PLOS ONE. 2019;14(3):e0214365.
  8. Klumpp M, Hintze M, Immonen M, Ródenas-Rigla F, et al. Artificial Intelligence for Hospital Health Care: Application Cases and Answers to Challenges in European Hospitals. Healthcare (Basel). 2021 Jul 29;9(8):961.
  9. CNOM Conseil National de l’Ordre des médecins - Santé : la révolution numérique - 2022
  10. Latif S, Qadir J, Qayyum A, Usama M, Younis S. Speech Technology for Healthcare: Opportunities, Challenges, and State of the Art. IEEE Rev Biomed Eng. 2021;14:342–56.
  11. Poder TG, Fisette J-F, Déry V. Speech Recognition for Medical Dictation: Overview in Quebec and Systematic Review. J Med Syst. 2018 Apr 3;42(5):89.
  12. LeMagIt - AP-HP mise sur Watson pour faire avaler la pilule ORBIS à son personnel - 2020
  13. Centre hospitalier de Valenciennes - 2020 - https://www.calyps.ch/fr/data-science-by-calyps-temoignage-du-centre-hospitalier-de-valenciennes/
  14. ChronoLife - 2022 https://www.chronolife.net/telesurveillance-patients/?lang=fr
  15. Hillo - 2021 - https://experiences.microsoft.fr/articles/intelligence-artificielle/hillo-ia-predire-variations-glycemie/
  16. SIH Solutions - 2021 - https://www.sih-solutions.fr/la-digitalisation-des-lames-danatomie-pathologique/
  17. Ensen P. Challenges and Opportunities Facing Medical Education. Trans Am Clin Climatol Assoc. 2011;122:48–58.
  18. Sun TQ, Medaglia R. Mapping the challenges of Artificial Intelligence in the public sector: Evidence from public healthcare. Gov Inf Q. 2019 Apr 1;36(2):368–83.