System for searching similar patients as supportive tool for clinical decision
Principal Investigator
Name
Anton Kovac
Degrees
MSc.
Institution
Siemens Healthineers
Position Title
Data Scientist
Email
About this CDAS Project
Study
PLCO
(Learn more about this study)
Project ID
PLCO-807
Initial CDAS Request Approval
Jul 15, 2021
Title
System for searching similar patients as supportive tool for clinical decision
Summary
Clinicians often tend to decide on further treatment of patients based on their clinical experience and heuristics they developed through their service. But what to do if a new patient has different characteristics than previous historical patients and the doctor is not sure of the diagnosis decision? The final diagnosis or determination of the correct treatment procedure is not always easy for the doctor. Also, the population of each country or region may have different definitions of the attributes boundaries, or criteria for individual attribute values respectively. Supporting doctors in decision-making is therefore very helpful.
Nowadays, the personalized medicine becomes more popular as it focus more on particular patient and brings enhanced patient-centric approach to the clinical praxis. Our intention is to develop smart system based on statistical and machine learning modelling to provide support of clinical decision making based on a cohort of the most similar subjects to the new treated patient that doesn´t have a determined the final diagnosis yet. The summary statistics of historical data (e.g. biomarkers, laboratory values, past therapeutic procedures, etc.) of such cohort can be a valuable information for clinicians to this decision and help them in better understanding of existing knowledge. Diagnosis then becomes more personalized and is tailor-made for the new treated patient. Furthermore, we believe that our system in a proper settings can also provide predictive probabilities for the next outcomes of the therapeutic treatments.
Nowadays, the personalized medicine becomes more popular as it focus more on particular patient and brings enhanced patient-centric approach to the clinical praxis. Our intention is to develop smart system based on statistical and machine learning modelling to provide support of clinical decision making based on a cohort of the most similar subjects to the new treated patient that doesn´t have a determined the final diagnosis yet. The summary statistics of historical data (e.g. biomarkers, laboratory values, past therapeutic procedures, etc.) of such cohort can be a valuable information for clinicians to this decision and help them in better understanding of existing knowledge. Diagnosis then becomes more personalized and is tailor-made for the new treated patient. Furthermore, we believe that our system in a proper settings can also provide predictive probabilities for the next outcomes of the therapeutic treatments.
Aims
SA 1: Identify the variables of interest (outcomes) which have the most informative value in identifying similar patients to the new/currently treated patient
SA 2: Develop an algorithm to calculate the rank of the similar patients to the new/currently treated patient with respect to variables of interest
SA 3: Adjust the algorithm to calculate predictive probabilities for the next best therapeutic treatment
Collaborators
Fabiola Fernandez-Gutierrez, PhD. (Siemens Healthineers)
Ludmila Pusztova, PhD. (Siemens Healthineers)
Peter Marusin (Siemens Healthineers)
Marcel Klamarik (Siemens Healthineers)