Although this information has resided in the claims-payment systems of carriers and third-party administrators for years, the development of programs to effectively mine this data has been fairly recent.
Data mining is the process of finding trends and patterns by going through large amounts of claims, pharmacy and medical data by using statistical and mathematically based methodologies.
Predictive modeling is the most recognized form of data mining. However, predictive modeling is not the only tool available. Claims and pharmacy data can also be utilized to identify individuals who could benefit from case- or disease-management programs, to track compliance with evidence-based medicine protocols, or to identify patterns in provider utilization and performance that can then be used to establish a unique network, contracting strategies or provider-incentive strategies.
What to look for in data mining products
- Flexibility What information will be truly accessible to my company, my network provider and to the nurses or other staff that deliver medical management services to my employees? Can my HIPAA benefits administrator and or finance analyst access the data directly?
A good system allows more than one individual to have direct Web access to at least aggregate-level information and reporting. Look for a system that allows easily developed adhoc reports at no additional cost to you.
Staff providing wellness and care-management services (utilization management, case management and disease management) need more that just reports provided periodically from a data vendor. Care management staff should have direct access to the database information and reporting functions.
- Data It is important to preload the database with, ideally, 24 to 36 months of historical medical and pharmacy claims data. New data should then be added on at least a monthly basis going forward.
This provides the opportunity to fully utilize the system from day one and to ensure that your care management vendor is able to identify new opportunities for intervention as they present themselves.
- Cost Be sure you have had all costs disclosed to you. In addition to a base cost, there are often additional costs for mapping data from your current and potentially past plan administrator, loading historical claims data or updating future claims data more often than annually, implementing future upgrades to the system and/or processing requests for adhoc reports.
- Uses How should I expect this system to be used by my care management and network vendors?
It is important to know what your vendors actually do with the data. How do they utilize data to deliver the product to employees? Data should easily allow the care-management staff to routinely identify new candidates for care-management services based on utilization patterns, anticipated costs and cases that may hit stop-loss.
It should also be capable of monitoring physicians’ prescriptions and members’ compliance with condition-specific evidence-based medicine standards. The system should also allow care managers to measure outcomes and changes in any of this data resulting from their interventions.
Predictive modeling should provide overall anticipated costs for the health plan through the next 12-month period. Your network provider should be able to utilize this data to more accurately identify and report the performance of the network and plan.
This should provide you, in partnership with your network manager, opportunities to modify the plan and/or network design to more effectively manage the population.
Barbarra Eisenhardt is vice president of HealthSpan operations and director of care-management and wellness. HealthSpan is a provider-owned PPO network, serving more than 125,000 members and more than 160 self-funded employer groups throughout the tri-state region. Reach HealthSpan at (888) 914-7726.