A systematic review of validated methods for identifying heart failure using administrative data.

Area of Science:

• Health Informatics
• Cardiology
• Epidemiology

Background:

• Administrative and claims data are increasingly used for health research.
• Accurate identification of heart failure (HF) is crucial for patient management and research.
• Validating algorithms for HF detection in these data sources is essential.

Purpose of the Study:

• To systematically review and assess the validity of algorithms used to detect heart failure (HF).
• To evaluate algorithms utilizing administrative and claims data for HF identification.

Main Methods:

• Systematic review of PubMed and Iowa Drug Information Service databases (1990-2010).
• Inclusion of studies evaluating the validity of HF detection algorithms.
• Review of abstracts and full articles by two independent investigators.

Main Results:

• 35 studies met the inclusion criteria for validity evaluation.
• Algorithms demonstrated acceptable to high positive predictive values (PPVs), often exceeding 90%.
• Primary hospital discharge diagnosis (ICD-9 code 428.X) yielded the highest PPV and specificity, but may lower sensitivity for outpatient cases.

Conclusions:

• Algorithms for HF detection using administrative/claims data are generally valid, particularly with primary discharge diagnoses.
• Consideration of outpatient management is important for comprehensive HF case identification.
• Incorporating outpatient codes enhances algorithm sensitivity for new HF cases.