Development of a HTA for medical devices using AI
| Project Title | Development of a HTA for medical devices using AI |
| Project dates | 10/04/2020 - 09/01/2022 |
| Project Lead | Davide Piaggio |
| Funder | Edwards Lifesciences |
| Description | Manufacturers must demonstrate the effectiveness and safety of medical devices to obtain approval under the current medical device regulations. When working with artificial-intelligence-based medical devices, assessing these concepts (safety, effectiveness, and cost-effectiveness) becomes even more challenging. Current regulations and guidelines are lacking in this regard. |
Manufacturers must demonstrate the effectiveness and safety of medical devices to obtain approval under the current medical device regulations (specifically, the European one, i.e., 2017/745). The emphasis on effectiveness was added to avoid potential scams involving devices that were indeed safe to use but produced no effective results for the user. Aside from that, it is recommended, but not required by law, that the cost-effectiveness of medical devices be evaluated, too.
When working with artificial-intelligence-based medical devices, assessing these concepts (safety, effectiveness, and cost-effectiveness) becomes even more challenging. Current regulations and guidelines are lacking in this regard. This is a growing concern, as not only is the medical device market expanding rapidly, but artificial intelligence is permeating it more and more.
In light of this, our work could serve as an example of how to conduct health technology assessment on artificial-intelligence-based medical devices.
Edwards Lifesciences is a company that mainly specializes on two fields: heart valves and hemodynamic monitoring. Our collaboration focuses on the latter. Edwards Lifesciences have been developing several solutions for the intraoperative hemodynamic monitoring, mainly aimed at predicting and preventing intraoperative hypotension.
Their latest technology is called Acumen Hypotension Prediction Index (HPI) device. This device can be used to monitor the health status of patients during surgeries and is able to predict the onset of intraoperative hypotension events up to 15 minutes before the event, allowing the surgical equip team to adjust the profile of care of the patients (e.g., the fluid management) to avoid or, at least, minimize the event.
Being able to predict and prevent hypotension is vital. In fact, intraoperative hypotension is significantly linked with serious adverse events such as acute kidney injury, major adverse cardiovascular events, post-surgical complications (e.g., surgical site infection, anastomotic leakage), and mortality. These could have a negative impact both on the quality of life of the patient and on the resource use and costs of the hospital.
The effectiveness of Acumen HPI has already been proved in different trials. Specifically, it has been proved that it achieves statistically significant reduction of intraoperative hypotension when combined with a treatment protocol in non-cardiac surgery versus standard of care. Moreover, it has also been shown that the system is accurate and has superior predictive abilities for hypotension compared to common hemodynamic parameters such as cardiac output, stroke volume, and changes in mean arterial pressure.
However, no study has analysed the cost-effectiveness of such device, yet, nor the impact of HPI on efficiency gains such as reduced complications, reduced length of stay and intensity of care. To this purpose, members of the ABSPIE Lab (Davide Piaggio, Carlo Federici, Leandro Pecchia) are currently involved to build and validate a health economics model comprising of a short-term part (Tree model) and a long-term one (Markov model). The model is being developed in collaboration with the Gemelli hospital (Rome, Italy), Altems (Graduate School of Health Economics and Management, Rome, Italy), specifically for the collection of data on gynaecological patients, and with surgeons from the Policlinico Universitario Federico II (Naples, Italy).
To be populated.