A systematic review of validated methods for identifying uveitis using administrative or claims data

Area of Science:

• Ophthalmology
• Health Informatics
• Biostatistics

Background:

• Administrative and claims databases are increasingly used for public health surveillance.
• Accurate identification of specific conditions like uveitis within these datasets is crucial but challenging.

Purpose of the Study:

• To systematically review and synthesize existing algorithms for identifying uveitis in administrative and claims databases.
• To evaluate the methodologies and validation approaches used in these identification algorithms.

Main Methods:

• A comprehensive literature search was conducted in MEDLINE via PubMed (1991-2012) using uveitis-related terms.
• Two independent investigators screened studies, extracted data on algorithm characteristics, and assessed methodological rigor.

Main Results:

• Seven studies met the inclusion criteria, revealing significant variability in uveitis identification algorithms and validation methods.
• Positive predictive values ranged from 24.8% to 52.6%, with different algorithms and validation approaches yielding diverse results.
• Only three studies included case validation; one utilizing text mining achieved a 52.1% positive predictive value.

Conclusions:

• Further research is essential to develop and validate robust uveitis algorithms for administrative data surveillance.
• The choice of algorithm should be guided by the specific requirements for balancing sensitivity and specificity.
• Individual code and case validation are critical for establishing reliable uveitis detection methods.