Data Error Detection
Through Pattern Recognition

Explainable AI approach to detect data errors

Health plans in the United States lose billions of dollars every year due to data quality errors. This highlights the need for health plans to detect and correct data errors before they cause downstream losses. Traditionally, health plans have used deterministic rule sets to detect data errors throughout their operations. However, this approach is now obsolete given that processes and systems have grown much more complex and data volumes have increased exponentially.

Data errors cause a myriad of problems for health plans, many of which have major financial and reputational consequences. Incorrect or missing information regarding a provider in a health plan’s directory can result in surprise billing, member abrasion, provider abrasion, and compliance fines. Similarly, data errors that result in the incorrect adjudication of a claim also has direct and significant impact. A solution that can proactively detect and correct data errors across the enterprise is critical.

High-Cost Claim Adjudication Monitoring

MCheck™ has been successful in identifying incorrectly adjudicated claims processed via a pricer or manual pricing method, resulting in higher claims payouts than necessary.

HEDIS Star Rating Corrections

MCheck™ also helps discover and correct data quality errors to correctly assess the quality of care and services provided to members. The platform can also evaluate provider performance, develop performance improvement initiatives, and provide better outreach to providers and members to impact care.

Missing Provider Network Affiliations

MCheck™ uses advanced NLP to extract provider network affiliations from signed contracts and compares that against network affiliations within the plan’s provider database(s). This comparison is used to proactively detect when providers are likely missing network affiliations which prevents surprise billing and directory display errors.

Our Capabilities
Anomaly detection through analysis on every possible slice of data

HiLabs’ primary anomaly detection platform, MCheck™ is a highly scalable and efficient collection of algorithms capable of detecting data anomalies arising from a variety of human and machine created errors. The platform runs analysis across every possible aspect of the data, using multi-variate analysis across all possible combinations and permutations to detect all, true positive errors.

Filter high value anomalies from billions of patterns generated

Using distinct data models and model configurations, MCheck™ can detect a variety of high impact data quality issues. Our platform has the capability to do both statistical and correlation analysis across billions of records using distributed big data technologies.

Ensemble of pattern mining and statistical models solves a variety of healthcare data issues

MCheck™ analyses various entities: groups of patients, providers, and other healthcare data and applies appropriate statistical modelling for each group. It detects anomalies by identifying data instances that deviate significantly from group properties. With MCheck™ we can detect anomalies with stunning accuracy, often with 95%+ confidence, while ensuring all high impact anomalies are found.

Explainable AI - Auditable facts support every anomaly

MCheck™ focuses on both displaying anomalies and ensuring anyone can understand why the anomaly was flagged as a likely error. With every anomaly, a set of auditable facts is provided with one click. These could be in the form of other records (claims/members/providers) with highly similar characteristics, but significant variance in the field of interest, or through charts, drilldowns, and additional statistics showing all support for the discovery.

Deriving valuable insights by detecting anomalies in real time
Request more information