Related Experiment Video
Updated: May 25, 2026
Comparison of Predictive Performance of Three Lymph Node Staging Systems in Colorectal Signet Ring Cell Carcinoma Based on Machine Learning Model
Published on: April 18, 2025
A systematic review of validated methods for identifying lymphoma using administrative data.
Ronald A Herman1, Bradley Gilchrist, Brian K Link
1Division of Drug Information Services, The University of Iowa College of Pharmacy, Iowa City, IA 52242-4710, USA. Ronald-A-Herman@uiowa.edu.
This review found that while algorithms using administrative or claims data can identify lymphoma cases with high specificity, their sensitivity and positive predictive values (PPVs) are often low to moderate. Improving algorithms by requiring multiple codes or specific code combinations enhances accuracy.
Area of Science:
- Health Informatics
- Medical Record Analysis
- Oncology Research
Background:
- Administrative and claims data are increasingly used for health outcome research.
- Accurate identification of specific conditions like lymphoma within these datasets is crucial for epidemiological studies and healthcare management.
- Existing algorithms for identifying lymphoma in claims data vary in their reported validity.
Purpose of the Study:
- To systematically review published studies on algorithms identifying lymphoma in administrative or claims data.
- To assess the validity of these algorithms in accurately identifying lymphoma cases.
Main Methods:
- A systematic literature search was conducted using PubMed and the Iowa Drug Information Service database.
- Studies from the USA and Canada using administrative or claims databases were included.
- Two investigators reviewed search results for reported and validated algorithms for lymphoma identification.
Main Results:
- One study reported and validated an algorithm, while ten others reported unvalidated algorithms.
- The validated algorithm achieved high specificity (>99%) but varied in sensitivity and positive predictive value (PPV).
- Algorithms requiring multiple diagnostic codes or specific code combinations (e.g., two codes within 2 months, or diagnostic and procedure codes on the same day) demonstrated improved PPVs and sensitivity compared to single-code algorithms.
Conclusions:
- International Classification of Disease, Ninth Revision (ICD-9) codes for lymphoma offer high specificity but limited sensitivity and PPVs in administrative data.
- Combining diagnostic and procedure codes or requiring multiple codes across visits can increase the PPV of lymphoma identification algorithms.
- Sole reliance on a single registry is insufficient for confirming true positive lymphoma cases.