PROTEMPA: a method for specifying and identifying temporal sequences in retrospective data for patient selection.

Area of Science:

• Biomedical Informatics
• Clinical Data Analysis
• Health Services Research

Background:

• Identifying patient populations with specific temporal disease patterns is crucial for clinical research and quality assurance.
• Existing methods often struggle with complex, poorly coded, or incompletely represented patient data.

Purpose of the Study:

• To develop and evaluate a novel data processing method, PROTEMPA (Process-oriented Temporal Analysis), for identifying and categorizing disease processes using temporal patterns.
• To retrieve specific patient populations from clinical data repositories based on these identified patterns.

Main Methods:

• PROTEMPA was developed for defining and detecting temporal and mathematical patterns in retrospective clinical data.
• A Java implementation was tested against SQL for identifying patients with HELLP syndrome and categorizing their disease severity and progression.
• Pattern detection utilized time-sequence characteristics in clinical laboratory test results, with verification by manual case review.

Main Results:

• PROTEMPA demonstrated higher accuracy than standard SQL in identifying patients with HELLP syndrome.
• The method successfully categorized patients by disease severity and progression based on temporal data, a capability lacking in SQL-only approaches.

Conclusions:

• PROTEMPA effectively identifies and categorizes patients with complex diseases by analyzing temporal relationships across multiple data types.
• This approach is valuable for patient population retrieval when features lack standard codes, are poorly expressed, or inaccurately recorded.