|
Abstract:
|
Multivariate categorical data are common in health care research. Typically, statistical consultants generate multiple cross-tabulation tables and test associations with the chi-square statistic. However, Correspondence Analysis (CA) and Hierarchical Log Linear Regression (HiLogLinear) provide more depth and clarity about multivariate categorical data. Asthma and COPD patients were identified in the US HealthCore Integrated Research Database between 1/1/2006 and 10/31/2014. The inclusion criteria were age?40 years, ?2 ICD-9 diagnoses (?30 days apart) for asthma, and/or ?2 diagnoses (?30 days apart) for COPD, ?2 ICD-9 procedure, CPT, or HCPCS codes (?30 days apart) for related procedures, ?3 prescription fills (?30 days apart) for asthma/COPD medication, and ?2 CPT codes for spirometer test. This procedure found 20,459 patients dually diagnosed as Asthma-COPD Overlap Syndrome (ACOS). CA identified two distinct ACOS ICD-9 asthma-COPD subtype dimensions: chronic bronchitis comorbid with extrinsic asthma and chronic airway disease dimension and a small emphysema with multiple asthma diagnoses dimension. HiLogLinear indicated the diagnostic patterns did not vary across age cohorts.
|
