AI-powered expert systems to reduce antibiotic consumption
By Lumed | Published on March 11, 2022
Problem and objectives
Problem
- Overuse of antibiotics promotes the emergence of resistant pathogens
- Difficulty controlling infections in the hospital due to staff movement
- High cost associated with overconsumption of antimicrobials
Objectives
- Reduce antibiotic consumption by 15%
- Reduce bacterial resistance to antibiotics
- Reduce hospital drug spending
- Reduce patient hospital stays
- Reduce the amount of precaution time required to prevent nosocomial infections
- Reduce the number of outbreaks of nosocomial infections (COVID-19, C. difficile, MRSA, VRE and about 20 others)
Proposed solution
Solution
Lumed is a health technology company involved in research to improve antibiotic treatments, monitor nosocomial infections, and favour hospital savings. The company develops expert systems, including APSS and ZINC, based on simple artificial intelligence (AI) algorithms.
The company’s key products are:
- APSS: antibiotic governance software
- ZINC: nosocomial infection surveillance software
- ONCO: oncology protocol prescription and management software
- DATA: infection reporting module
Principle
APSS is antimicrobial management expert software that automatically analyzes and reviews prescriptions. It manages over 600 pathogens and integrates nearly 1,000 dosage rules. It sends out alerts to pharmacists so that they can offer prescriptions that are less costly, more effective and more suitable for patients. It also monitors patients’ progress and detects those who would benefit from antimicrobials. 65% of the time, the professional acknowledges and accepts the alert sent by APSS.
ZINC is expert software used for nosocomial infection surveillance that identifies infections in real time. It helps locate patients who were potentially exposed to the initial case by guiding the investigation process. Creation of a list of at-risk patients who lived with infected patients over a long enough period of time.
APSS and ZINC are compliant to HIPAA and GDPR standards and have a communication channel encryption system (TLS).
User experience
The software is installed in the hospital management system and interfaced with the patient data portal so there is no manual data entry. The software analyzes the data and makes recommendations according to its area of expertise, namely antimicrobial stewardship and nosocomial infection surveillance.
Role of AI
Algorithms that are used
- 15 different algorithm families to create alerts
- Discovery algorithms to identify the reasons for rejecting the alerts
- Extensive knowledge base and specialized inference engines
Impacts
Main outcomes
Experience and patient care:
- Reduction in hospital stays by an average of 2 days per patient, a reduction of 2,500 days per year
- Improved patient safety through efficient management and infection prevention teams
- Automated patient follow-ups in real time
Medical team wellbeing:
Automating the arduous prescription analysis process (2,000 recommendations per year to the Sherbrooke CHU)
Population health:
- Reduction in antimicrobial use by 20%
- Avoid the development of pathogen resistance
Health costs (for Sherbrooke CHU):
- CAD$2.1M in antimicrobial cost savings, a 30% reduction (in 6 years of using the APSS software)
- 91% of interventions with prescribers are accepted with APSS
Generation and dissemination of new knowledge:
- Publication of four scientific papers (Journal of Antimicrobial Chemotherapy, Artificial Intelligence in Medicine, BMC Infectious Diseases)
- 20 scientific presentations (including Artificial Intelligence in Medicine, E-Health, Association of Medical Microbiology and Infectious Disease Canada)
Economic development:
3-16 additional employees between 2012-2021
Challenges takled
From 2016 to 2020, the IT moratorium in Québec prevented Lumed from selling its solutions to Québec hospitals. The slowness of the local market thus allowed the competition to develop. Lumed therefore chose to go through partners for commercialization purposes, resulting in a slowdown in business development from 2016 to 2020.
Conditions for success
Team mobilised
- 15 full-time employees
- 1 part-time employee
Diploma
- 3 PhDs
- 7 bachelor degrees
- 4 technicians
- 1 pharmacist
- 1 physician
Areas of expertise
- Medicine
- Data science and machine learnin
- Marketing and business development
Collaborations
- Lumed jointly develops its expert systems with the client hospitals for which they are designed and in which it implements them. Its solutions are now being implemented in the following hospitals:
- Centre hospitalier universitaire de Sherbrooke;
- Centre hospitalier universitaire de Nancy (France);
- Centre hospitalier de l’Université de Montréal;
- McGill University Health Centre;
- Fraser Health (British Columbia);
- Hamilton Health Sciences (Ontario).
Funding received
- Personal funds (confidential)
- FRQ grants: CAD$240K in 2005
- NSERC, FRQ, CMPA grants
- Accord grants
- Investors (confidential)
Project stages
Completion of ONCO development with CUSM (commercialization expected in 2022)
Project stages - to come
Continuous improvement of APSS, ZINC and ONCO modules
Lumed solutions exported to the U.S.
Development of new expert systems for other medical fields (e.g. opioid addiction or geriatrics)
Context of the solution
Customers
Current business model – SaaS
Lumed sells its software licences to hospitals.
Objective: development of new expert systems every 2 years
Market opportunity
AI health market projections in 2028: US$120.2B