Overview

Northglen Health System is a multi-hospital regional health system in the US offering acute, outpatient, and specialty care. Its EHR, claims, and operational data lived in separate systems, so a patient’s full history existed as fragments no team could see together at once. Readmission risk surfaced only after discharge, sometimes only after a patient was already back in the emergency department, leaving care teams no window to intervene. Population health and cost-of-care reporting depended on manual pulls that took days to compile and were outdated before they reached decision-makers. Bacancy Technology unified clinical, claims, and operational data into a governed Snowflake warehouse, then applied predictive analytics to flag readmission risk ahead of discharge and gave care and finance teams real-time, role-based visibility into clinical, population, and cost-of-care metrics.

Technologies Used

python

Project Highlights

checkmark

Fragmented EHR, claims, and operational records unified into a single patient profile

checkmark

Predictive readmission risk scoring applied ahead of discharge, not after

checkmark

Regulated patient data protected with role-based access and dynamic masking

checkmark

Real-time clinical, population health, and cost-of-care dashboards delivered to business teams

The Challenges

1

Patient information was split across separate EHR, claims, and operational systems, with each system maintaining its own records independently. To assemble a complete picture of a patient, care coordinators had to query each system separately and cross-reference manually, which delayed access to information exactly when it mattered most, at the point of a discharge decision.

2

Readmission risk had no early warning system attached to it. Care teams learned a patient was high-risk only when that patient returned to the emergency department, often within days of being discharged, and by then the window to intervene with a follow-up call, medication check, or home health referral had already closed.

3

Population health and cost-of-care reports were built through manual pulls and spreadsheet consolidation that took several days to complete each cycle. By the time a report reached department heads or finance, the numbers were already outdated, which left trend analysis and budget forecasting running behind the organization's actual state rather than ahead of it.

4

Patient data access could only be granted broadly, not scoped to specific tables or fields, so staff routinely saw more than their role required. Sensitive fields could not be masked, and the systems kept no consistent log of who accessed which record, leaving no reliable way to demonstrate how protected health information was being handled during a compliance review.

Solutions by Bacancy

1

Our data analytics services team unified EHR, claims, and operational data into a single Snowflake warehouse, modeling each source into conformed patient-level tables so a full record could be pulled in one query instead of three. Care coordinators now open one patient profile at the point of care rather than reconciling records across systems by hand.

2

Our data science and predictive analytics team trained a scikit-learn readmission risk model on clinical history, prior admissions, and comorbidities, then scheduled it to score every patient automatically ahead of their planned discharge. Care teams now receive a ranked, explainable risk list days before discharge instead of finding out a patient was high-risk only after an emergency department return.

3

We modeled the reporting layer in dbt and connected it to Power BI dashboards that refresh automatically on current warehouse data, maintained day to day by a dedicated Snowflake developer from our team. Population health and cost-of-care numbers that once took days to compile are now available same day, giving department heads and finance current figures instead of a week-old snapshot.

4

Our healthcare IT services team implemented role-based access and dynamic data masking at the schema level in Snowflake, so each user sees only the tables and fields their role permits, with sensitive PHI fields masked by default. Every access event is logged automatically, giving Northglen a complete, queryable audit trail for HIPAA review instead of no record at all.

Core Features

checkmark

Unified patient data warehouse across EHR, claims, and operational systems

checkmark

AI readmission risk scoring applied ahead of discharge

checkmark

Real-time population health and cost-of-care dashboards

checkmark

Role-based access with masking and explainable model outputs

No. of Resource

5

No. of Resource

Time Frame

March 2025 – September 2025

Time Frame

Project Snapshot

northglen-health-dashboard

Outcomes

High-risk patients flagged 5 to 7 days before discharge, not after an ED return

Roughly 20% drop in 30-day readmissions across the flagged high-risk patient segment

Population health reports now delivered same day, down from multi-day manual pulls

Three to four fragmented patient records collapsed into one single unified profile

Cost-of-care visibility moved from a quarterly cycle to near real-time reporting

All patient PII and PHI masked with role-based access and explainable model outputs

Technical Stack

Cloud data warehouse Snowflake
Data ingestion Fivetran
Data transformation dbt
Orchestration Apache Airflow
Machine learning PythonScikit-learn
Model tracking MLflow
BI/dashboards Power BI
Security & governance Snowflake RBACDynamic data masking
Version control & CI/CD GitGitHub Actions

Experience With Bacancy

2500+ Projects Experienced Innovation with Bacancy!

Get access to an experienced team of developers and engineers from Bacancy, handpicked to ace your goals. Kickstart within 48 hours, no-risk trial.

Book a 30 min call

14+

Years of Business Experience

1458+

Happy Customers

12+

Countries with Happy Customers

1050+

Agile Enabled Employees