With 80 years of experience, CUF stands as Portugal’s largest private healthcare provider. Its network of 43 hospitals and clinics serves over 1.3 million customers, supported by 16,800 employees. In 2024 alone, CUF managed a massive volume of 68,000 surgeries and 3.3 million consultations, operating at a massive scale.

The CUF back-office team faced a critical challenge: managing surgical cost estimates. This highly complex, manual process was being overwhelmed by explosive growth.

• Unsustainable Volume: A team of just 11 employees had to process over 64,000 estimates in 2024.
• Rapid Growth: This volume represented a 25% year-over-year increase, putting immense pressure on the team.
• Extreme Complexity: The team had to navigate the rules of over 3,000 different contracted funding entities, all while accounting for each patient’s specific diagnosis and each surgeon’s unique clinical practice.
• SLA Failure: The strain was impacting customers. The average estimate response time hit 3.95 days, significantly missing the organization’s 2-day target.

CUF partnered with Devscope Health to build an intelligent “Smart Operations” solution on the Databricks platform. This new flow automates the entire “VIC” (estimated surgical cost) lifecycle, transforming a manual process into a predictive and automated engine.

The solution leverages a metadata-driven ETL Framework on Azure Databricks and Delta Lake to create an intelligent, end-to-end flow.

1. Unified Data Ingestion (Workflows & ETL): The system continuously gathers all necessary data. This includes historical material movements and daily pricing information, which are gathered weekly to feed the predictive model. Separately, the ETL process runs multiple times per day to ingest new surgical proposals and VIC requests (tickets).
2. AI-Powered Categorization (LLMs): As new products and materials are processed, the system uses LLMs (GPT-4o) for product naming analysis. This automatically generates the correct Category, Sub-Family, and Family.
3. Predictive “Causal Model”: A “Causal Model” was developed, using historical data to power a prediction engine. This model forecasts the specific Materials and Quantities that will be consumed during the surgery.
4. Automated (Re)Calculation: The calculation engine combines these inputs with known costs like Room Consumption, Surgical Team, Daily Inpatient Value, and Surgical Packages & Kits to generate the final VIC Value (the estimated surgical cost).
5. Integrated Frontend: The system automatically generates the final Patient Document and Company Document. This entire flow is integrated with a user interface (an MS Power App), which automates sending the final information to the client.

Beyond the core KPIs, the solution delivered transformative process improvements:

• Real-time Accuracy: Enabled real-time consultation of price lists and rules for all 3,000+ financial entities.
• Predictive Refinement: The predictive model guarantees the continuous refinement of data over time, learning from historical performance.
• Standardized Workflows: Implemented standardized management of work lists, including new urgency/priority criteria and automated templates.
• Simplified Analytics: Simplified the data needed to analyze process service levels, team performance, and estimate-to-bill accuracy.
• Seamless Integration: Fully integrated relevant clinical information from the surgical proposal and provided a custom frontend via MS Power App.

Devscope Health’s expertise in modern data platforms was instrumental in the success of this “Smart Operations” project. Leveraging Databricks as a core accelerator, Devscope’s team demonstrated a deep understanding of the complex business problem: surgical cost estimation.

The commitment to a collaborative, customized solution is evident in the project’s architecture. This included the development of a predictive “Causal Model”, the deployment of LLMs (GPT-4o) for data categorization, and the seamless integration with an MS Power App to provide a frontend “customized to our needs” . This partnership successfully transformed a highly manual process into an intelligent, automated flow.