Case finding with incomplete administrative data: observations on playing with less than a full deck.

Area of Science:

• Health Services Research
• Health Informatics
• Biostatistics

Background:

• Disease management programs require efficient patient identification for optimal intervention.
• Predictive modeling using administrative data is a common strategy for patient stratification.
• Challenges include data incompleteness and processing delays in administrative datasets.

Purpose of the Study:

• To examine the impact of administrative data issues on disease management candidate identification.
• To evaluate proposed solutions for improving predictive model effectiveness.
• To assess the utility of different data sources and time windows for stratification.

Main Methods:

• Prospective models were built using regression analysis.
• Stratification algorithms were evaluated using R² statistics and receiver operating characteristic curves.
• Cost concentration ratios were used to assess economic efficiency.

Main Results:

• Data delays significantly reduce stratification effectiveness, with impact varying by target population proportion.
• Supplementing partial data with older, extensive data yielded inferior algorithms compared to recent, shorter data.
• Demographic data offered minimal improvement and were poor substitutes for claims data.
• Self-reported health status provided only slight improvement for high-risk patient targeting.

Conclusions:

• Timeliness and recency of administrative data are critical for effective patient stratification in disease management.
• Supplementing with older or demographic data is not cost-effective.
• Current methods for improving stratification using supplemental data show limited benefit.