Project: Computerized Physician Order Entry System with Integrated Artificial Intelligence Models

As the volume of available data generated daily increases, the responsibility to manage and deliver value-based care by healthcare providers has become essential. Medical professionals require access to relevant patient records on a timely basis. Due to innovations in machine learning (ML) and access to technology within the health industry, the appropriate design and development of decision support systems are quickly becoming vital. The lh-toolkit web components have made tremendous progress in the last 3 years, with nearly all components available for an electronic health record (EHR) platform. The project will develop web components and implement a computerized physician order entry (CPOE) system that is human-centered, saves clinician time, and interfaces with AI/ML models.

Clinicians spend a large amount of their time maintaining patient records and making decisions based on these records. Recent studies suggest more than 52% of visit time is spent on documentation. The quantity of information that is required by a clinician to be read and understood is getting untenable and each decision made in this environment can lead to harmful patient outcomes. Therefore, to deliver precise knowledge that enhances patients’ health clinical decision support is necessary. The prime purpose of a clinical decision support system (CDSS) is to enhance the decision-making approach through evidence-based practices. Each characteristic of an individual patient is linked to the computerized medical knowledge database that produces a patient-specific recommendation which is then presented to the physician to make an improved decision.

The traditional clinical decision support tools are often integrated into electronic health records and patient health records to streamline the workflows. This also allows the system to take advantage of existing patient data sets to provide a flexible and focused medical summary. Although these systems curb preventable mistakes many organizations are experiencing significant issues when it comes to creating a user-friendly interactive application with effective protocols. Poorly integrated CDSS can lead to the generation of unwanted alerts and continuous monitoring of critical alarms has resulted in burnouts and fatigue amongst the nurses and administrative staff.

The deliverables of the project are as follows:

  • Develop web components to build a robust CPOE system
  • Assemble the components in a SPA app for ordering drugs and tests.
  • Develop web components that allow clinicians to select relevant EHR fields to build ML classification models.
  • Integrate outputs of these models into the SPA for appropriate drug and test orders.

A developer working on this project needs to have skills in HTML, CSS, JavaScript frameworks like React/Angular/Polymer, Flask web services skills, and Python for model development.

Mentors: @sunbiz

1 Like

@r0bby @sunbiz I wanted to contribute on this project. For the same I was testing the lh-toolkit-webcomponents from last few days and I noticed that there are small UI improvements (eg. consecutive elements need same gap b/w them). So can I work on them?

In this project the CDSS has a main role of giving alerts to the physician while entering the prescription (correct me if I’m wrong). What else are the outcomes of the project? Do we need to keep track of the medical history of patient for which we will need to add more features and corresponding labels to ML model?

Yes, please feel free to create issues and send MRs.

We already have patient records tracking in lh-toolkit. We want to build a UI on top that is able to get the output of ML models and show it on the patient dashboard.

The project also includes making of ML model right? Also can you please elaborate this point. I was confused that do we need to create ML model as per clinician’s requirement or it would be a single model same for all image

Clinicians can select the EHR fields that they want to build the models.

FYI to all:

Where can I see the patients record collected by lh-toolkit? So that I could design the wireframes for GSOC proposal accordingly.

@r0bby @sunbiz I would like to contribute to this project , I do not have much experience in AI but know very basic coding. I am a 2nd year BTech student in VIT - VELLORE . Please do allow me to learn and contribute as much as I can.

Everything you need is in this thread.

One of the things we look at is if you can use available information to get started. We do not want to have to hold your hand throughout the whole project.

@sunbiz Where can I see the patients record collected by lh-toolkit? So that I could design the wireframes for GSOC proposal accordingly.

May I suggest downloading and deploying it on your machine from the codebase to see the structure of how data is collected in the database?

@arpandesai0

Yes sure will do it🙌

How are things going? You have until 2022-04-19T18:00:00Z to get this finalized. Finalized means the starter task is completed and ready for review. Failure to do the starter task is an immediate rejection. We do occasionally (though rarely) give extensions for the starter task, we have no control over the Google side of things.

There will be a reminder sent tomorrow(Saturday) at 1800 UTC as it is 72 hours from the close of the application period.

Get your proposal in now since we have no wiggle room with that. We cannot accept you without a proposal submission. You may continue to work on your starter task but link the work somewhere, please.