Machine Learning for Real-Time Clinical Decision Support
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The challenge in selecting optimal anti-infective therapies for treating infectious diseases is exacerbated by the emergence of resistant pathogens, requiring healthcare providers to navigate complex electronic medical record systems for real-time decision support.
Innovation Solution
A method and system that utilize machine learning to provide real-time decision support by determining trigger events in electronic health records, obtaining patient-specific data, selecting pharmacokinetic models, and generating rankings of drug therapies based on predicted efficacy, thereby recommending optimal drug regimens.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If healthcare providers manually review electronic medical records to select optimal anti-infective therapies, then treatment decisions can be personalized to patient needs, but the time required for decision-making increases and real-time support is compromised
Solution Approach 1:
The system enables automated self-service by having the computer system automatically retrieve patient data from electronic medical records, apply pharmacokinetic models, calculate probability of PK-PD target attainment, and generate ranked therapy recommendations without requiring manual clinician intervention for data collection and analysis
Solution Approach 2:
The manual mechanical process of reviewing medical records and calculating therapy effectiveness is replaced with an automated computational system that uses machine learning algorithms and pharmacokinetic modeling to rapidly analyze patient data and generate treatment recommendations in real-time
2Reliability
If complex pharmacokinetic modeling is performed for each patient to determine optimal therapy, then treatment efficacy can be optimized, but the computational complexity and processing requirements increase
Solution Approach 1:
The system manages complexity by focusing on specific critical parameters (patient demographics, infection characteristics, pathogen data, creatinine clearance, weight, height) and using these to drive pharmacokinetic modeling, rather than attempting to process all possible clinical variables
Solution Approach 2:
The system introduces an intermediary computational layer that acts as a bridge between raw patient data and clinical decision-making, using automated pharmacokinetic modeling and machine learning algorithms to translate complex medical data into actionable ranked therapy recommendations
3Productivity
If real-time automated recommendations are provided to clinicians, then decision-making speed improves, but the risk of algorithmic errors or inappropriate recommendations increases
Solution Approach 1:
The system incorporates feedback mechanisms where clinician interactions with recommendations (acceptance, modification, or rejection) are captured and used to refine and improve the machine learning models over time, allowing the system to learn from real-world outcomes and continuously enhance recommendation accuracy
Solution Approach 2:
The system performs preliminary automated analysis and generates ranked therapy recommendations in advance of the clinician's final decision, providing pre-processed, evidence-based options that streamline the decision-making process while allowing clinician review and override if needed
Data Source
AI summary
A computer system, computer program product and method for determining a probability of attaining a PK-PD target associated with efficacy for a patient that includes a processor(s) determining that a trigger event related to a patient, wherein the patient record comprises an order for a current drug regimen, has occurred in an electronic health record (EHR) system communicatively coupled to the one or more processors. The processor obtains descriptive information relating to a patient. The processor selects a pharmacokinetic model. The processor applies the pharmacokinetic model and utilizes the information relating to the patient to determine, for each of the one or more drug therapies, a probability of attaining a PK-PD target associated with efficacy for the patient with the infection. The processor generates rankings for the drug therapies and determines if the current drug regimen comprises a probability above a threshold. The processor generates a new order.