Section 5-1 : Case Study openEMR

AI Case Study

This tutorial is based on case study, where AI data solution is used as in a real-world example of a customer using it.

Please review important aspects and key considerations on How To Use AI Data Solutions in previous chapter.

Healthcare App openEMR Case Study

Of: openEMR App for healthcare

Summary Highlights

The case study sought to leverage the power of AI models to unlock the full potential of openEMR health system data with EasyManage help.

Successfully achieved key objectives:

1.	Get Data modeling suggestions from AI	✅
2.	Ask AI For Sensitive Data Analysis	✅
3.	Make openEMR data AI-Ready in 10 Min	✅
4.	Build, deploy and Customize MCP Server	✅
5.	AI Model Analysis performed on openEMR data	✅
6.	Ask AI What Insights and Trends Can Get	✅
7.	Ask AI What AI Workflows Can Implement	✅

Brief Synopsis

Leverage opportunities for AI data solutions such as AI-driven insights & trends and AI workflow automation for healthcare App openEMR as demo installation.

Steps

Following steps were taken towards implementing solutions using EasyManage AI.

Import Tables And Download Schema

• Import Tables

  • App Tables from MySQL/mariaDB cloud database imported into EasyManage.

• Search tables, listed

• Tables Imported

• Search metadata, search tables for field patient_id

• Download Table Schema from EasyManage.

• Downloaded SQL file.
  • TableSchemaDnld_openEMR.sql
  • Markdown format here : openEMR Table Schema Downloaded

Get Data modeling suggestions from AI

• Open Cursor and start new project workspace at e.g. C:\openEMR
• Copy TableSchemaDnld_openEMR.sql to project root.
• Using above table schema SQL file as context, Ask AI model

Give suggestions for data modeling i.e. table joins up to 4 nested levels, specify one-to-many or one-to-one

• Context given

• Asked for data modeling suggestions

• Response for data modeling suggestions

• Please see response doc:
  • Data modeling suggestions for table joins

Configure Data Modeling

• Define and configure data modeling, table joins in EasyManage Builder Studio for

  • Medication and Prescription Data Model
    • Level 1: Patient Medications
    • patient_data | 1:M | prescriptions | 1:M | drugs
    • Refer to tutorial on Data Modeling

• Data modeling configured for Patient Medications

  • patient_data ←→ prescriptions (via patient_id)
  • prescriptions ←→ drugs (via drug_id)

• In this order: 1. prescriptions ←→ drugs

• Then: 2. patient_data ←→ prescriptions

Ask AI For Sensitive Data Analysis

• Giving as context TableSchemaDnld_openEMR.sql, Ask AI model

Study columns and mark columns , put comment on right side, for columns that come under Sensitive data, PII (Personally Identifiable Information), Confidential Business Information, Data revealing sensitive personal informatio

OR

Analyze the OpenEMR database schema to identify columns that contain sensitive data, PII, confidential business information, or data revealing sensitive personal information

• Tip: For healthcare databases also check for PHI fields i.e. sensitive PHI identifiers under HIPAA.

• Chat input

• Chat Response Progress

NOTE: Using the 1st prompt, AI was not able to edit file, but decided to create a new file:

• Please see response doc:
  • Identify and mark sensitive data columns

info

• How to implement the sensitive data recommendations?
  • Before generating code, use EasyManage field hide option to exclude columns, having sensitive data, from API and MCP Layers, so that AI models will not have access to them. Use via menu links:
    • Data Table Explorer → navigate to folder having table → Select option on right side of table listed → Field Prop - Hide → Set for required columns → Field Prompt Properties → Hide [Yes] and Data Display Properties → Hide [Yes]

EasyManage Builder Studio Project Configure Settings

MCP Server Module Bundling for V1

It is recommended to keep 5-10 tables in one MCP Server module. Along with data modeling of those tables.

• Select 16 tables for AI solution module building.
• Bookmarked tables for generating code
• Build Selection

• Build/Generate : Included will be MCP Server REST with APIs REST.
  • This will be version V1.

• Download generated code.

• Create Folder structure for easymanage and cursor workspace.

• Copy below projects from extracted code to workspace openEMR_workspace.

  • backend\spring-java\emapi
  • mcp\spring-java\emmcp

• APIs REST : Open project, Build

• Run APIs - CMD Prompt
  • cd openEMR_workspace
  • java -jar emapi\app\dbrest\target\dbrest-1.0-SNAPSHOT.jar

• Run APIs - Check Swagger Home

• Run APIs - Check Swagger Data Modeling endpoint

• MCP Server REST : Open project, Build

Use MCP Server V1

• Configure and Use MCP Server V1 in Cursor.

• Sample mcp.json

{
  "mcpServers": {
    "easymanage": {
      "command": "C:/ProgFiles/Java/jdk-17/bin/java.exe",
      "args": 
      ["-jar", 
      "path_to_workspace/openEMR_workspace/emmcp/app/mcprest/target/mcprest-1.0-SNAPSHOT.jar"
      ],
      "env": {
       "em.mcp.apiUrl": "http://127.0.0.1:9080/emdbrest",
       "em.mcp.authType": "None",
       "em.mcp.authToken": ""
      }
    }
  }
}

Exceeding Total Tools Limit

• Warning is displayed as total tools 102 are exceeding 40.

Try chat input:

easymanage get and study all table schema and suggest what insights and trends can be derived from tables

• AI model sees no relevant tool call available and tries to suggest insights from static SQL file available in workspace:

• Please see response doc:

  • Analyze table schemas for insights - From SQL File

• Now delete TableSchemaDnld_openEMR.sql from project root

  • We want AI model to call tool get_all_table_schema once available later.

Customize MCP Server as V2

• Customize MCP Server REST V1, disable some table level tools globally.

  • Please see section: How to Enable/Disable Tools in README-mcprest.md

• Verify and Disable Tools as indicated below and try out.

  • Table Level - Disable
    • TOOL-ViewAll
    • TOOL-ViewAllPaged
  • Table Level - Keep
    • TOOL-SelectWhere
    • TOOL-Query
  • AI Transactions - Disable All
    • AI-TRAN-ADD
    • AI-TRAN-EDIT
    • AI-TRAN-DEL
  • Table Joins Data Modeling Level - Disable
    • TOOL-DM-ViewAll
    • TOOL-DM-ViewAllPaged
  • Table Joins Data Modeling Level - Keep
    • TOOL-DM-SelectWhere
    • TOOL-DM-Query

• Disable Tools using VS Code

• Build/Generate MCP Server REST.
  • This will be version V2.
• Configure MCP Server in Cursor.
  • Shows: Found 40 Tools

AI Model Analysis

Results from AI analysis is given in response docs below.

Presented below and listed are few analysis and solutions being recommended by AI models.

Ask AI What Insights and Trends Can Get

easymanage get and study all table schema and suggest what insights and trends can be derived from tables

• Chat input

• Chat Response Progress 1

• Chat Response Progress 2

Please refer to response doc

• Analyze table schemas for insights - From Tools

Ask AI What AI Workflows Can Implement

easymanage get and study all table schema and suggest what AI workflows can be implemented based on tables

• Chat input

• Chat Response Progress 1

• Chat Response Progress 2

• Chat Response Progress 3

Please refer to response doc

• Analyze table schemas for AI Workflows

• Also see the generated markdown files by AI model in this response:
  • Below 3 documents generated.
  • AI Workflows Analysis
  • AI Workflows - AI Implementation Guide
  • AI Workflows README

Next Steps

After AI Transaction Tools enabled, can ask AI What AI Transactions Can Implement.

Enable AI Transaction Tools

• Customize MCP Server REST V2 to build V3, where in enable AI transaction table level tools globally.
  • Please see section: How to Enable/Disable Tools in README-mcprest.md

Ask AI What AI Transactions Can Implement

easymanage get and study all table schema and suggest what AI transactions can be implemented