Over the last months, I have been experimenting with whether modern AI-assisted development platforms can support the creation of structured operational and budgeting systems for clinical research.
One interesting experiment was the development of a simplified Clinical Trial Budget Calculator (CTBC) using the AI-assisted platform Base44.
The result is a working prototype application:
https://ledger-flow-76f4f135.base44.app/
(The application currently requires free authorization via Google or email.)
Proposed Workflow
The idea behind this prototype is relatively simple:
1. Read and Parse Study Documents
The system should be able to read:
clinical trial protocols,
RFPs,
study outlines,
assumptions documents,
or similar operational materials.
The goal is to identify both:
protocol-driven cost drivers
(visits, patients, countries, monitoring frequency, timelines, data complexity, etc.)
and
non-protocol operational assumptions
(project management effort, vendor oversight, startup activities, reporting effort, risk assumptions, contingency logic, and operational overhead).
2. User Validation and Assumption Review
AI-generated assumptions should not be accepted automatically.
The user should be able to:
review extracted assumptions,
correct inconsistencies,
validate operational logic,
adjust effort estimates,
and update reusable budgeting rules.
This includes reviewing:
work units,
milestone structures,
role composition,
frequencies,
and calculation logic.
3. Budget Estimation
Once assumptions and rules are validated, the system generates:
workload estimates,
milestone rollups,
resource effort calculations,
and budget projections.
The objective is not full automation, but structured support for feasibility assessment and operational budgeting.
Example of a prompt for App generation:
Build a modern web application called:
Clinical Trial Budget Calculator (CTBC)
Purpose:
Create a scalable, protocol-driven clinical trial budgeting and operational planning application.
The application should transform:
protocol assumptions,
operational assumptions,
reusable work units,
project timelines,
countries/sites,
and role-based effort models
into:
detailed budget lines,
workload calculations,
milestone rollups,
dashboards,
and timeline views.
The app should support:
assumption review,
calculation validation,
scenario planning,
and future AI-assisted automation.
The design must remain understandable, modular, and scalable.
CORE CONCEPT
The budget engine calculates:
Approved Assumptions
× Work Units
× Unit Composition
× Quantity
× Locations
× Time Phases
× Bill Rates
Detailed Budget Lines
The system must support:
labor costs,
pass-through costs,
vendor costs,
investigator grants,
and future forecasting.
IMPORTANT DESIGN PRINCIPLES
Separate:
protocol facts,
operational estimates,
benchmark assumptions,
and calculated values.
Do NOT directly use extracted protocol values in calculations until they are reviewed or approved.
The application must provide:
calculation traceability,
validation,
review workflow,
and assumption transparency.
The application is NOT:
a validated CTMS,
eTMF,
ERP,
accounting system,
or regulatory system.
It is:
a budgeting and operational planning assistant.
MAIN MODULES
1. Study Setup
Fields:
Study Name
Protocol Number
Sponsor
Indication
Study Phase
Study Design
Currency
Status
Description
2. Protocol Upload
Allow upload of:
protocol PDFs
RFPs
contracts
amendments
scope documents
Fields:
Study ID
Document Name
Document Type
Upload Date
Parsing Status
Version
3. Protocol Assumptions Extraction
The system should parse uploaded protocols and extract:
enrollment
sites
treatment duration
follow-up windows
visit schedule
procedures
labs
ECG
PK sampling
imaging
questionnaires
central lab requirements
IRT/IWRS
safety follow-up
visit burden
Store extracted information in a review table.
Fields:
Assumption ID
Study ID
Source Document
Assumption Category
Extracted Assumption
Value
Unit
Source Page
Source Section
Confidence Level
Review Status
Reviewed By
Review Date
Notes
4. Operational Assumptions
Allow manual assumptions for items NOT usually found in protocols:
Examples:
countries
country allocation
site allocation
CRF pages
queries per subject
monitoring frequency
monitoring strategy
bill rates
investigator grants
travel assumptions
startup duration
closeout duration
contingency
vendor assumptions
Fields:
Assumption Name
Category
Value
Unit
Assumption Type
Rationale
Scenario
Review Status
5. Time Phases
Support:
Initiation
Startup
Enrollment
Treatment
Follow-up
Data Cleaning
Database Lock
Closeout
Fields:
Phase Name
Start Date
End Date
Duration
Scenario
Notes
6. Locations
Support:
Region
Country
Site
Site Group
Fields:
Country
Region
Site Name
Site Count
Enrollment Count
Country Multiplier
Grant Multiplier
Travel Multiplier
7. Roles and Rates
Examples:
CRA
Lead CRA
Project Manager
CTA
Data Manager
Safety Specialist
Regulatory Specialist
Medical Monitor
Biostatistician
Vendor Manager
Fields:
Role Name
Department
Seniority
Hourly Rate
Currency
8. Work Unit Library
Reusable operational units.
Examples:
PSV
SIV
RMV
COV
Weekly Sponsor Call
Query Management
SAE Review
TMF QC
Vendor Oversight
Database Lock
Fields:
Work Unit Name
Milestone
Deliverable
Description
Unit Type
Default Quantity Logic
Linked Assumption
TMF Deliverable
WBS Code
9. Work Unit Composition
Each work unit can contain multiple components.
Example:
RMV Trip Report:
CRA II Visit → 8h
CRA II Travel → 2h
CRA II Report Writing and Filing → 2h
Weekly Sponsor Call:
PM → 1h
Project Director → 1h
CTL → 1h
Data Manager → 1h
Medical Monitor → 1h
Vendor Manager → 1h
Fields:
Work Unit
Role
Activity
Hours per Unit
Fixed Cost
Notes
10. Milestones
Hierarchy:
Study
→ Phase
→ Milestone
→ Work Unit
→ Budget Line
Milestones should roll up:
work units
hours
costs
timelines
deliverables
Fields:
Milestone Name
Phase
Start Date
End Date
Status
Budget
Hours
Deliverables
11. Deployment Plan
Deploy work units across:
phases
countries
sites
years/months
Fields:
Study
Scenario
Work Unit
Phase
Country
Site
Year
Month
Quantity
12. Budget Engine
Generate detailed budget lines dynamically from:
Approved Assumptions
× Work Units
× Unit Composition
× Deployment Plan
× Locations
× Rates
Generated fields:
Budget Line ID
Study ID
Scenario
Phase
Milestone
Country
Site
Year
Month
Work Unit
Role
Quantity
Hours per Unit
Total Hours
Bill Rate
Labor Cost
Pass-through Cost
Total Cost
Source Assumption
Validation Status
Budget line IDs should support hierarchy:
1.1.0.5.4.3
13. Pass-through Costs
Support:
investigator grants
central labs
ECG vendor
PK bioanalysis
imaging
eCOA
EDC
IRT/IWRS
travel
translations
regulatory fees
insurance
archiving
drug supply
IP cost
If exact values are unknown:
mark as:
Requires Manual Input
Sponsor Supplied
Excluded
14. Validation and Review
Very important.
The app must validate:
missing assumptions
inconsistent deployment
missing rates
missing locations
duplicate work units
invalid dates
empty milestones
Validation dashboard:
Errors
Warnings
Missing Inputs
Review Required
Add:
Draft
Reviewed
Approved
Locked
calculation statuses.
15. Scenarios
Support:
Base Case
Low Case
High Case
Amendment Scenario
Change Order Scenario
16. Dashboards
Show:
total budget
labor vs pass-through
cost by country
cost by phase
cost by milestone
cost by role
cost by vendor
cost by year/month
top cost drivers
17. Timeline / Gantt View
Add simple Gantt-style visualization:
phases
milestones
start/end dates
budget
hours
Filter by:
study
scenario
country
phase
18. EXPORTS
Support export to:
Excel
CSV
Google Sheets
Power BI friendly tables
19. USER EXPERIENCE
Workflow:
Create Study
→ Upload Protocol
→ Review Extracted Assumptions
→ Add Operational Assumptions
→ Define Deployment
→ Generate Budget
→ Validate
→ Review Dashboard
20. FUTURE SCALABILITY
Design architecture so future versions can support:
TMF integration
CTMS integration
ERP integration
forecasting
invoicing
scenario simulations
AI-generated work units
amendment comparison
protocol version comparison
resource forecasting
earned value management
21. DESIGN STYLE
The application should:
look modern and professional
be visually clean
support editable tables
support filters
support drill-down
support dark/light mode if possible
support scalable database architecture
Keep the interface understandable even for non-technical users.
Current Limitations and Open Discussion
At the current stage, freely available AI-assisted development platforms are still insufficient to fully support the complexity of clinical trial budgeting and operational modeling.
While prototypes can already be generated surprisingly quickly, major challenges remain around:
validation,
consistency,
reusable rule structures,
protocol interpretation,
scenario management,
and reliable downstream calculations.
This experiment is therefore only an early exploration of what may become possible in the future.
The broader conceptual direction behind this work is described here:
Exploring AI-H Budgeting Framework for Clinical Trial Financial Feasibility: A Conceptual Analysis
https://doi.org/10.5281/zenodo.15651378
Anyone interested in:
AI-assisted budgeting,
operational modeling,
protocol-driven planning,
or clinical trial financial feasibility
is welcome to share ideas or comments below this blog post.
No comments:
Post a Comment