Monday, January 15, 2024

UDLCO : AI LLM as CDSS point of care support to infuse meaningful EBM understanding of drug efficacies at the point of care

Summary : 

UDLCO Summary : 


What if we can train LLMs to utilize the EBM PICO rubric that can provide useful point of care CDSS EBM support by providing insights on the efficacy of each drug (a bad but contemporary analogy would be the current ADR point of care EMR CDSS alarms that largely turn out to be clinically non significant as they are non contextual to individual patient centered medical cognition)? On the other hand savor this: 
 
What are the evidence of efficacy for these ubiquitous prescription drugs Montek LC and pulmoclear! 

I guess this question needs to go to LLM as our human ecosystem is really averse to understanding clinical significances in simple pico formats partly due to badly done RCTs (again blame it on human resource constraints) 

@#Project LLM PaJR supported healthcare workflows 


Below posted on Jan 14, 2024:


Pending journal club for the day :






[1/12, 10:56 PM] : 3% hypertonic saline was more efficacious than mannitol in the initial 12 h and equally or more efficacious than mannitol therapy later.
Decrease in coma hours was a significant additional finding in the group treated with 3% hypertonic saline



[1/14, 8:36 PM] Rakesh Biswas: Share this in a PICO format

Help @human participants of this CBBLE 



 No human takers for this reiterating the fact that humans may not be suitable alone to establish an evidence based medicine framework but it also provides an idea to rapidly scale the LLM project in a impactful manner. 

Hypothesis : What if we can train LLMs to utilize the EBM PICO rubric that can provide useful point of care CDSS EBM support by providing insights on the efficacy of each drug (a bad but contemporary analogy would be the current ADR point of care EMR CDSS alarms that largely turn out to be clinically non significant as they are non contextual to individual patient centered medical cognition)



What are the evidence of efficacy for these ubiquitous Montek LC and pulmoclear! 

I guess this question needs to go to LLM as our human ecosystem is really averse to understanding clinical significances in simple pico formats partly due to badly done RCTs (again blame it on human resource constraints) @#Project LLM PaJR supported healthcare workflows 


[1/15, 9:39 AM] 2023 Med PG: 


S C/o Decreased urine output since 9 days - improved


C/o sob since 9 days  - no improvement 


C/o pedal edema since 6 days - no improvement 


C/ cough  with sputum since 20 days - improved 



pt conscious coherent cooperative
Temp - 100°F 
PR - 88 bpm
BP - 100/70 mmHg 
Rr - 20 cpm 
GRBS - 81 mg/dl
Spo2- 94% at room air 
Cvs - s1s2 heard ,no murmur 
R/S - b/l air entry +nt 
fine crepts in 
Rt - AA,MA
Lt - diffuse 
CNS - no fnd 
I/O -1050/1000 ml



?TB PERICARDITIS WITH MODERATE PLEURAL EFFUSION


POLYCYSTIC KIDNEY DISEASE WITH AKI 





INJ LASIX 20mg IV OD 
INJ NEOMOL 1GM IV SOS 
TAB MONTAK LC PO HS
TAB PULMOCLEAR PO BD 
TAB PCM 650MG PO TID 
NEB WITH IPRAVENT 8TH hrly 
BUDECORT -12TH HRLY 
PROTEIN POWDER 2 SCOOPS IN 1 GLASS OF WATER/MILK 
2 EGGS PER DAY




[1/15, 9:40 AM] PG Medicine 2021: Mention the input and output also



[1/15, 9:50 AM] Rakesh Biswas: Daywise serial intake output trends since admission and daywise average respiratory rate to objectively follow the static nature or increasing problem of his heart failure both due to his CAD as well as pericardial restriction

Glossary of frequently used "medical cognition" terminologies : http://userdrivenhealthcare.blogspot.com/2023/11/glossary-of-user-driven-healthcare.html?m=1


Adding review of literature resources to this project :

"case studies to understand current capabilities for applying AI/ML in the healthcare setting, and regulatory requirements in the US, Europe and China.

Methods

A targeted narrative literature review of AI/ML based digital tools was performed. Scientific publications (identified in PubMed) and grey literature (identified on the websites of regulatory agencies) were reviewed and analyzed."



From Wikipedia :

An LLM is a language model, which is not an agent as it has no goal, but it can be used as a component of an intelligent agent[35]

 Researchers have described several methods for such integrations.

The ReAct ("Reason + Act") method constructs an agent out of an LLM, using the LLM as a planner. The LLM is prompted to "think out loud". Specifically, the language model is prompted with a textual description of the environment, a goal, a list of possible actions, and a record of the actions and observations so far. It generates one or more thoughts before generating an action, which is then executed in the environment.[36] The linguistic description of the environment given to the LLM planner can even be the LaTeX code of a paper describing the environment.[37]

In the DEPS ("Describe, Explain, Plan and Select") method, an LLM is first connected to the visual world via image descriptions, then it is prompted to produce plans for complex tasks and behaviors based on its pretrained knowledge and environmental feedback it receives.[38]

The Reflexion method[39] constructs an agent that learns over multiple episodes. At the end of each episode, the LLM is given the record of the episode, and prompted to think up "lessons learned", which would help it perform better at a subsequent episode. These "lessons learned" are given to the agent in the subsequent episodes.





No comments:

Post a Comment