Diagnosis Prediction Model for Emergency Room Decision Support

A Joint Internship at LUMC and Autoscriber.

Purpose
The objective of this joint internship is to develop a diagnosis prediction model for the
emergency room (ER) at LUMC. The model will be built using entities extracted from free-
text clinical notes by Autoscriber's software, in combination with other structured data from
the electronic health record.

Background
In the ER, rapid and accurate diagnosis is critical, as is the efficient and goal-directed
allocation of resources. The effectiveness of physicians in the ER in making a diagnosis and
the amount of additional research they need for that depends on several factors, such as
their clinical experience and the incidence of the disease. Decision support systems can
assist physicians in identifying the most likely diagnoses in a quantitave way. With this
information, physicians can augment their clinical gaze and gut feeling with quantitative
data, while at the same time being protected from cognitive biases that are frequent in
medical care, such as neglect of base rate and anchoring. Furthermore, physicians can
choose additional research that will distinguish best between the most likely diagnoses, and
forgo research that will not add much value, but may cost time and money. We expect such
a system to improve the time to a correct diagnosis and treatment, and lower costs, thereby
improving patient outcomes.

Objectives and Responsibilities
Learning Goals:

• Understand the workflow and diagnostic challenges in an emergency room setting.
• Gain experience in integrating structured and unstructured data for predictive modeling.
• Collaborate with technical and clinical teams across LUMC and Autoscriber.

Main tasks:

• Utilize Autoscriber's software to extract relevant entities from real-time recordings or free-
  text clinical notes.

• Integrate extracted entities with other structured data available in emergency room settings.

• Build, validate, and fine-tune a diagnosis prediction model.
• Evaluate the model's performance in collaboration with clinicians to ascertain its clinical
  utility.

Skills

• Strong foundational knowledge in machine learning and data science.
• Familiarity with or willingness to learn about healthcare data, especially in emergency care settings.
• Ability to collaborate with multidisciplinary teams.
• Excellent written and verbal communication skills.

Demonstrate your initiative, intuition and results from whatever you've been working on in the past. Tell us what packages you love. Tell us what makes you tick. Show us what you've been up to and we will do the same!