A systematic review of validated methods for identifying pancreatitis using administrative data

Area of Science:

• Health Informatics
• Medical Record Analysis
• Epidemiological Research

Background:

• Administrative data are crucial for public health surveillance.
• Accurate identification of pancreatitis in these datasets is essential for research and patient care.
• Existing algorithms for pancreatitis detection require rigorous validation.

Purpose of the Study:

• To systematically review algorithms for identifying pancreatitis in administrative data.
• To focus on studies that have examined the validity of these pancreatitis-identifying algorithms.
• To assess the performance of different coding systems in administrative healthcare databases.

Main Methods:

• Conducted a comprehensive literature search using PubMed and the Iowa Drug Information Service database.
• Included studies utilizing data sources from the USA or Canada, reflecting common coding practices.
• Two independent investigators reviewed identified studies to ensure accuracy and relevance.

Main Results:

• Eight studies were analyzed, evaluating algorithm validity for pancreatitis detection in hospital and ambulatory records.
• The International Classification of Diseases, Ninth Revision, Clinical Modification (ICD-9-CM) code 577.0 demonstrated the highest performance.
• This code achieved a positive predictive value of 60%-80% and a negative predictive value typically exceeding 90%.

Conclusions:

• The ICD-9-CM code 577.0 is the most reliable for identifying pancreatitis in administrative data, offering the best positive predictive value and specificity.
• Current evidence does not support the use of ICD-9-CM codes 577.1 and 577.2 for pancreatitis case identification.
• Integrating laboratory values into administrative databases could significantly enhance the specificity of pancreatitis detection algorithms.