Frequently Asked Questions
Who are the users of Signal 1’s AI solutions?
We currently offer solutions for use in general Medicine and Surgery departments of acute care hospitals. The primary users of our solutions are doctors, nurses and clinical managers.
What data do Signal 1’s AI models use?
Our AI models use patient data that already exists within your EHR.
How does Signal 1 integrate with my EHR?
Our interoperability platform provides seamless integration with modern and legacy EHRs through HL7, FHIR, and databases. Our solution enables the exchange of accurate and timely patient information between EHRs and our AI solutions.
Do Signal 1’s solutions require any manual data entry by my staff?
No, our solutions automatically pull data from your EHR and use it in the format in which it exists. There is no additional data entry or transformation required on your part.
Does my hospital need any special hardware or software to implement Signal 1’s solutions?
No, our AI applications run on the cloud and connect directly with your on-premise and cloud infrastructure.
How is Signal 1 able to predict ‘changes in patients’ care needs’ before they arise?
We train our models on large samples of historical data from your hospital. We identify a target for the model – such as, being clinically stable or clinically deteriorating. From your historical data, the machine learning algorithm uncovers patterns in patients’ data that are associated with that target. Then, we test the accuracy of the model on another historical sample to see how well the model’s predictions compare to what actually happened.
I’ve heard that AI models are ‘black boxes’. What information about the model will Signal 1 share with me?
We produce a detailed ‘model validation report’ for every hospital in which we deploy a solution. This report explains what data the model uses, how the model was constructed and trained, the clinical and operational features that the model learns from and a comprehensive evaluation of the accuracy, robustness and fairness of the model.
Where / how do my staff view the predictions from Signal 1’s AI models?
The outputs of our models are directly integrated into existing IT and communication systems such as your EMR, emails or text messaging and existing hardware you have in your unit.
How does Signal 1 ensure the privacy and security of my data?
How long will it take to get Signal 1’s solutions deployed at my hospital?
We can typically train our models and integrate our predictions into your workflows within 60 days depending on the setup and scope of deployment. We can define this scope in a few short discussions.
What hospitals are currently using Signal 1’s products?
Our products are currently in use at a number of hospitals in Canada including St. Michael’s Hospital and Grand River Hospital.
What research has been carried out on Signal 1’s solutions?
We are commercializing AI applications that were originally developed by researchers at St. Michael’s Hospital in Toronto, Canada. The team at St. Michael’s Hospital has published multiple papers on their experience deploying AI into clinical use, including:
Verma AA, Murray J, Greiner R, et al. Implementing machine learning in medicine. CMAJ. Aug 2021, 193(34): 1351-1357. https://www.cmaj.ca/content/193/34/E1351
Cohen J., Cao T, Viviano J, et al. Problems in the deployment of machine-learned models in health care. CMAJ. Sep 2021, 193(35): 1391-1394. https://www.cmaj.ca/content/193/35/E1391
Antoniou T, Mamdani M. Evaluation of machine learning solutions in medicine. CMAJ. Sep 2021, 193 (36): 1425-1429. https://www.cmaj.ca/content/193/36/E1425
Pou-Prom C, Murray J, Kuzulugil S, et al. From compute to care: Lessons learned from deploying an early warning system into clinical practice. Front. Digit. Health. 2022 4:932123. https://scholar.google.com/citations?view_op=view_citation&hl=en&user=lqbus3sAAAAJ&sortby=pubdate&citation_for_view=lqbus3sAAAAJ:KlAtU1dfN6UC
Verma AA, Pou-Prom C, McCoy LG, et al. Developing and validating a prediction model for death or critical illness in hospitalized adults, an opportunity for human-computer collaboration. Critical Care Explorations, forthcoming.