Microsoft Fabric | Power BI | Dataflow Gen2 | Lakehouse | Healthcare Operations Analytics
This project demonstrates an end-to-end healthcare analytics workflow using synthetic pediatric encounter data. The goal was to show how raw encounter-level data can be ingested, transformed, validated, modeled, and presented in a Power BI report for operational decision-making.
The dataset contains 1,000 synthetic pediatric encounters from 2025 and was designed to simulate common provider-side analytics questions related to encounter volume, length of stay, admission type, discharge status, and 30-day readmissions.
Data note: This project uses synthetic data only. It does not contain patient-identifiable information, real hospital records, protected health information, or proprietary data from any healthcare organization.
Hospital leadership needs a reliable view of pediatric encounter throughput to understand:
- Which departments have the highest encounter volume
- Which departments have the longest average length of stay
- How admission types vary across departments
- What percentage of encounters result in 30-day readmission
- How discharge status and length-of-stay patterns differ by department
The project demonstrates how an analyst can convert raw encounter data into a curated reporting layer, semantic model, and executive-facing Power BI dashboard.
| Metric | Value |
|---|---|
| Total Encounters | 1,000 |
| Average LOS | 4.27 days |
| Emergency Admission % | 76.2% |
| Long Stay Rate | 31.9% |
| 30-Day Readmission Rate | 5.7% |
- Microsoft Fabric - Lakehouse workflow environment
- Dataflow Gen2 - data cleaning and transformation
- Power BI - semantic model, DAX measures, and report development
- CSV / Excel - synthetic source data
- SQL / Data Modeling Concepts - validation logic, curated reporting design, and source-to-report checks
The source dataset is encounter-level and includes:
| Field | Description |
|---|---|
| Patient_ID | Synthetic patient identifier |
| Encounter_ID | Unique synthetic encounter identifier |
| Age | Patient age |
| Admit_Date | Encounter admission date |
| Admission_Type | Elective, emergency, or urgent admission |
| Department | Emergency Department, NICU, PICU, or General Pediatrics |
| Length_of_Stay_Days | Encounter length of stay in days |
| Discharge_Status | Discharged home, transferred, or observation/follow-up |
| Readmit_30_Day_Flag | Indicator for 30-day readmission |
The project follows a simplified lakehouse workflow:
Raw CSV Source
↓
Fabric Lakehouse / Bronze Table
↓
Dataflow Gen2 Transformations / Silver Table
↓
Curated Reporting Table / Gold Table
↓
Power BI Semantic Model
↓
Executive Dashboard
Key transformations included:
- Loaded raw encounter data into a Bronze table to preserve source structure
- Created patient age groups: Infant, Child, and Adolescent
- Standardized admission type and department categories
- Mapped discharge status values into reporting-friendly groups
- Created encounter month from admission date
- Created length-of-stay categories: Same-day, Short stay, Moderate stay, and Extended stay
- Created long-stay and 30-day readmission flags for KPI reporting
- Validated row counts, department counts, data types, and derived fields before reporting
The reporting layer was designed around a curated encounter table, pediatric_encounters_gold, with source fields and derived reporting fields. The Power BI semantic model supported the following measures:
- Total Encounters
- Average Length of Stay
- Emergency Admission Percentage
- Long Stay Rate
- 30-Day Readmission Rate
- Encounter Volume by Department and Month
- Average LOS by Department
- Admission Type Mix by Department
- Encounter Distribution by LOS Category
Example DAX measures are included in docs/dax_measures.md.
- Emergency Department encounters represented the largest department volume.
- The dataset had an overall average length of stay of 4.27 days.
- Emergency admissions accounted for 76.2% of encounters.
- Long stays of 4 or more days accounted for 31.9% of encounters.
- The 30-day readmission rate was 5.7%.
- NICU and PICU had the highest average length of stay, while the Emergency Department had the lowest average LOS.
pediatric-throughput-analytics/
|
|-- README.md
|-- data/
| |-- pediatric_encounters_bronze.csv
| |-- pediatric_encounters_gold.csv
|-- screenshots/
| |-- fabric_workflow_demo.png
| |-- power_bi_report.png
|-- docs/
| |-- dax_measures.md
| |-- data_dictionary.md
| |-- validation_summary.md
| |-- project_summary.pdf
| |-- fabric_workflow_demo.pdf
| |-- power_bi_report_export.pdf
|-- scripts/
| |-- generate_gold_table.py
| |-- validate_encounters.py
- Review the workflow diagram to understand the data pipeline.
- Open the Power BI report screenshot to view the executive dashboard.
- Review the data dictionary and transformation logic.
- Review the DAX measures used for KPI reporting.
- Review the validation summary to understand how source-to-report accuracy was checked.
- The dataset is synthetic and limited to 1,000 rows.
- Results are intended for workflow demonstration, not clinical decision-making.
- The project focuses on data transformation, validation, semantic modeling, and dashboard design rather than predictive modeling.
This project was created as part of a healthcare analytics portfolio to demonstrate Microsoft Fabric, Power BI, Dataflow Gen2, data transformation, semantic modeling, validation, and healthcare operations reporting skills.