House In a Box

From ICU of the Future
Jump to: navigation, search

Group Name

First, Do no harm!

Group members and e-mails

Satyajit Deshmukh: sdeshmukh3@gatech.edu

Nikhil Bagewadi: nikhil.bagewadi@gatech.edu

Ameya Nerurkar : anerurkar3@gatech.edu

Problem statement - Elevator talk

• Doctors have to go through a tonne of collected data to isolate relevant and critical data at the right time to influence their medical decisions.

• This is difficult task given the overwhelming volume of medical history and contextual data collected at each of the care transitions.

• Medical errors are often not caught at the right level and when it is most critical.

Description of the solution

Our approach to solving this problem would be as follows:
• The primary motive of the system is to isolate important, relevant, critical portions of patient records (history, observations etc.) collected at different levels during patient care. The assigning of criticality and importance to data portions will be a function of the caregiver who observed/initiated the data, the relevance of the data to the current action/context, the importance/repercussions of the action suggested by medical research and historic documents.
• Our system involves a learning approach (Text analysis and relevance search engine) which will be capable of parsing through large portions of data and isolating relevant text. We envision a 3 step approach to achieve this:
1. First, we will create individual relevancy graphs (or simple relevancy lists) that relate certain words/medical terminology to a list of relevant keywords/markers. These are to be created by parsing sample medical documents/references for specific topics. The goal is to build a database of terminology relevant to a context.
2. Second, we parse the patients history/medical records continually (at each step in response to each of the suggested/proposed next steps). The goal is to find repeated occurrences of terms from any of the relevancy graphs/lists and establish a pattern suggestive of the occurrence of a certain error or warning about a possible side effect. The effect would be the validation of the correctness of the applied solution and giving warnings about possible errors (gleaned from the data collection & association in step 1). This step is based on a “story matching” approach frequently used in Artificial Intelligence based learning systems.
3. The imposition of a role-based care structure over the structure we’ve already envisioned in step 1 & 2 facilitates a more precise and consistent critical care scenario. For instance, some warnings or critical information is relevant or would ring bells for only a certain care-giver while it seems completely harmless to another. Getting the right information to the right person at the right time is the cardinal rule of our system.

Given the considerable complexity of the proposed system and the overwhelming number of factors involved in making this system workable and practical in a realistic scenario, we believe our goal will be:
• To propose and design a architecture for solving the issues stated in our problem statement
• To create a prototype of our system which involves some working parts and some amount of mock-up.

Milestones

We have identified the following milestones:
• Research the problem statement and talk about it with faculty members (Dr Ackerman)
• Meeting with Emory faculty Dr Ruth Lamm who works on systems similar to the ones we are targetting
• Research the problem statement and talk about it with faculty members (Dr Ackerman)
• Finalize the design of the solution that we are going to implement in the final demo ( finalize the errors on which we are focussing for the demo)
• Try and acquire the raw material and start preparing the final demo (building the design for the Nurse station) • Performance testing of the code on particular test cases which involve the error being propogated because of critical information being missed during transfer of care.
• Prepare and finalize the skit that we are going to enact during the final demo of the project.

Resource needs

We have received suggestions from Dr. Ackerman suggesting the use of furniture resembling -
• A nurse’s workstation
• A physician’s workstation
The above suggested resources are needed only for creating a scenario (a skit) mimicking an actual hospital critical care unit case involving Physicians and Nurses.
Person responsible to acquire the resources: Satyajit Deshmukh

Main faculty coach you request

Dr. Jeremy Ackerman

Draft outline of the final report

• Introduction

• Problem Statement

• Feasibility Study

• Implementation

• Prototype

• Usability

• Conclusion

• References