OBJECTIVES: Researchers frequently use claims data to create comorbidity measures.  The objective of this project was to examine how the number of months of observation data and the type of claims data can affect identification of comorbidities. METHODS: Inpatient (IP) and outpatient (OT) claims data were used to identify comorbidities for beneficiaries enrolled in a state Medicaid program.   Beneficiaries were included if they were enrolled for a continuous 24-month period between January 2002 and December 2004.  Data were used to identify the first month in which an ICD-9 code appeared for each of the 17 comorbidities included in the Charlson Comorbidity Index.  The D’Hoore scoring system was used to identify ICD-9 codes associated with each comorbidity.  Results from the IP and OT claims were combined to create overall comorbidity measures.    RESULTS:  The study included a total of 618,337 unique patients.  The OT claims could identify almost all comorbidities (low of 88.0% for myocardial infarctions, high of 99.5% for diabetes).   IP claims were much less likely to identify comorbidities (low of 11.5% for connective tissue disease, high of 45.7% for myocardial infarction).  At 12 months, less than half of the beneficiaries with comorbid events during the 24-month period had been identified for dementia (43.9%), cerobrovascular disease (45.6%), peripheral vascular disease (45.6%), and myocardial infarction (48.4%).  In contrast, 78.1% of patients with diabetes events and 69.0% of patients with chronic pulmonary disease events had been identified at 12 months.    CONCLUSIONS: The percentage of patients ascribed comorbidities will vary significantly depending on the number of observation months.  Obviously, the incidence rate for new events/diagnoses contributes to the increase over time.  However, researchers must be careful since some comorbidities may not be associated with frequent enough physician office visits to accurately detect existing comorbidities when the number of observation months is short.