Medical records processing · Production

Aetion uses Amazon Bedrock and generative AI to unlock hidden insights about patient populations

The problem

Without the right structured query and deep familiarity with complex real-world patient datasets, many trends and patterns remain undiscovered, and users without prior expertise cannot easily explore or generate hypotheses from real-world data.

Workflow diagram · grounded in source
1
Define patient population
trigger
“Users define a patient population using the Aetion Measure Library (AML) features. The AML feature store standardizes variable definitions using scientifically validated algorithms.”
2
Compute patient features
ai_action
“The AEP computes over 1,000 AML features for each patient across various categories, such as diagnoses, therapies, and procedures”
3
Train clustering model
ai_action
“The Smart Subgroups component trains a topic model using the patient features to determine the optimal number of clusters and assign patients to clusters. The prevalences of the most distinctive features within each cluster, as determine…”
4
Generate cluster names via LLM
ai_action
“A prompt engineering technique for Anthropic's Claude 3 Haiku on Amazon Bedrock generates descriptive cluster names and answers user queries. Amazon Bedrock provides access to LLMs from a variety of model providers. Anthropic's Claude 3 …”
5
User submits natural language query
trigger
“a user engages with AetionAI to probe further with inquiries expressed in natural language”
6
LLM answers user query
ai_action
“The Interpreter API uses composite prompt engineering techniques with Anthropic's Claude 3 Haiku to answer user queries”
7
Hypothesis generation output
output
“the insights enable the user to hypothesize that Dulaglutide patients might experience fewer circulatory signs and symptoms. They can explore this further in Aetion Substantiate to produce decision-grade evidence with causal inference”
Reported outcome

Users can now discover patterns and turn findings into hypotheses for decision-grade evidence generation in minutes instead of days, without needing support staff.

Reported metrics
Time to generate decision-grade evidenceminutes, as opposed to days
Support staff requirementwithout the need of support staff
Reported stack
Amazon BedrockAnthropic's Claude 3 HaikuAetionAIAetion Evidence PlatformAetion DiscoverSmart Subgroups InterpreterAmazon S3Amazon AuroraKubernetes on AWS
Source
https://aws.amazon.com/blogs/machine-learning/how-aetion-is-using-generative-ai-and-amazon-bedrock-to-unlock-hidden-insights-about-patient-populations?tag=soumet-20
Read source ↗

Frequently asked questions

What did this team achieve with this AI workflow?

Users can now discover patterns and turn findings into hypotheses for decision-grade evidence generation in minutes instead of days, without needing support staff.

What tools did this team use?

Amazon Bedrock, Anthropic's Claude 3 Haiku, AetionAI, Aetion Evidence Platform, Aetion Discover, Smart Subgroups Interpreter, Amazon S3, Amazon Aurora, Kubernetes on AWS.

What results were reported?

Time to generate decision-grade evidence: minutes, as opposed to days; Support staff requirement: without the need of support staff (source-reported, not independently verified).

How is this medical records processing AI workflow structured?

Define patient population → Compute patient features → Train clustering model → Generate cluster names via LLM → User submits natural language query → LLM answers user query → Hypothesis generation output.

Related medical records processing cases