Dataiku > Case Studies > Physician Profiling

Physician Profiling

Customer Company Size

Large Corporate

Region

Europe

Product

Tech Stack

Machine Learning

Implementation Scale

Enterprise-wide Deployment

Impact Metrics

Cost Savings
Productivity Improvements

Technology Category

Analytics & Modeling - Machine Learning

Applicable Industries

Healthcare & Hospitals

Applicable Functions

Quality Assurance

Use Cases

Predictive Quality Analytics

Services

Data Science Services

About The Customer

The customer is a major hospital located in Western Europe. The hospital employs more than 2300 people and is committed to improving clinical outcomes while enhancing their ability to compete on cost. They were interested in adopting an Accountable Care Organization (ACO) model and were facing challenges in accurately measuring physician and healthcare organizations' performances due to uncoordinated heterogeneous data sources and poor quality data. The hospital estimated that administering the wrong care at the wrong time represented upward of $1.6M loss per year.

The Challenge

The customer, a major hospital in Western Europe, was facing challenges in accurately measuring physician and healthcare organizations' performances due to uncoordinated heterogeneous data sources, irregular and poor quality data, insufficient risk-adjustment of results, and lack of automation in physician profiling processes. They were seeking to embrace an Accountable Care Organization (ACO) model to improve clinical outcomes and compete on cost. Some clinical processes, like prescribing expensive or unnecessary drugs or recommending longer hospital stays than needed, were costly and detrimental to patient care. The customer estimated that administering the wrong care at the wrong time represented upward of $1.6M loss per year, a problem that they believed could be solved with accurate physician profiling.

The Solution

The customer's quality manager team built a data service with DSS that automatically cleans and aggregates various datasets (claims, patient, physician, and Rx data). The aggregated data enabled them to identify precisely which patient, treatments, and outcomes are linked to which physician processes. A machine learning algorithm was integrated at the center of this DSS-powered data service, enabling them to isolate patterns that reveal specific impacts on patient health outcome. When the model processes new incoming data from the various systems, practices ranging from drug prescriptions to hospitalization time are scored depending on how detrimental or beneficial they are in terms of cost and specific patient health.

Operational Impact

The hospital was able to implement best practices amongst its physicians, improving patient care while avoiding superfluous costs.
Physicians, once reticent, are now involved and frequently try to modify their personal practice choices based on this data analysis to increase care and cost.
The hospital was able to prototype and deliver an effective physician profiling data product in under 2 months.
The hospital now has a powerful ETL for all data sources, no matter the source and format.
The data-driven approach to measuring physician performance is now trusted by the physicians themselves.

Quantitative Benefit

Hospitalization average length is now below national standards.
Unnecessary drug prescriptions have significantly decreased, resulting in savings that amount to an estimated $1.3 million per year.