We will progress beyond state of the art. Our current state is the art is an advanced machine-learning (ML) algorithm (TRL level 4) that re-uses available EHR (Electronic Health Record) data to predict post-operative infections 5 days prior to the average medical team diagnosis, which achieves an accuracy of 80% (training set 40,000+ patient operating room data records). Through increasing the dataset and taking on more types of data, the ultimate goal is to ultimately achieve >90% accuracy at 5 days before the infectious symptoms on average are actually detected within the patient.
Some 50M people need to undergo inpatient surgery per year in Europe. Yet, despite intensive research efforts, around 25% of surgical patients will experience a complicated recovery with some type of infection on their path: pneumonia, urinary tract infections, wound infections, abdominal infections and bacteremia. Of course, the risk varies depending on patient and procedure, but despite efforts like biomarkers, risk scores or devices for early warning/detection, and preventive methods like ‘Enhanced Recovery After Surgery’ (ERAS), the overall risk still leaves almost one in four patients to get an infection within 30-days after surgery.
If we look beyond the personal suffering of patients and relatives, the cost of an infectious complication is estimated at around €10,000 per patient. Hence, total cost in only the Netherlands alone can be calculated at up to €3.5B per year, as approximately 350,000 Dutch patients go through an infection after surgery yearly. Total EU costs are even more shocking, at an estimated €125B per year.
Once our algorithm moves beyond 90% accuracy and is fully certified we expect to make a dent in this leading to advantages for everybody: reduction in personal suffering, reduction in overhead, reduction in bedding, reduction in staff needed, reduction in costs. Time will prove what reduction we can achieve but given the staggering absolute numbers we expect this to be significant.
