The CMS Chronic Condition Data Warehouse (CCW) provides researchers with Medicare and Medicaid beneficiary, claims, and assessment data linked by beneficiary across the continuum of care. The files also use ICD-9 codes and other information to identify the chronic condition each beneficiary has. The question is, how well does the CCW identify these chronic conditions?
According to Gorina and Kramarow (2011) the results are mixed. The authors examine the strengths and limitations of using CMS’s CCW algorithm with Medicare claims data to identify chronic conditions in older persons. Their methodology is as follows:
“Records from the NHANES I Epidemiologic Follow-up Study (NHEFS), including data from questionnaires, physical examinations, medical facility records, and death certificates, have been linked to Medicare claims records. We selected five conditions common among older persons: diabetes, ischemic heart disease (IHD), chronic obstructive pulmonary disease (COPD), dementia, and arthritis. We compared diagnoses for these five conditions derived from the two data sources (NHEFS and Medicare claims using the CCW algorithm).”
The authors identified the person as having the chronic disease if either i) the patient self-reported having the chronic disease or ii) the ICD code was listed in the patient’s medical records.
The percent of preexisting NHEFS cases identified during the CCW reference period for each of the 5 diseases was:
- Diabetes: 68.6%
- IHD: 62.9%
- Dementia: 39.3%
- COPD: 24.4%
- Arthritis: 17.0%
The references period the CCW uses to identify these diseases were 3 years for dementia, 1 year for COPD, and 2 years for the other three illnesses.
The authors find that the CCW is good at detecting chronic illnesses when these illnesses require frequent visits to the doctor. This is why CCW performs well in identifying IHD, but not identifying arthritis, a disease where patients may visit the doctor infrequently for this illness.
The paper does have some drawbacks. First, the authors do not examine the CCW’s specificity (the share of patients without the chronic condition diagnosis that the CCW correctly identifies as not having the disease. Second, patient self-reported disease may be unreliable. Patients with sore hands may report having arthritis when this may not in fact be the true cause of their ailment. The authors could have considered how well the CCW performed in the case where the NHEFS requirement was both a self-report and an ICD code in the patient’s records. Finally, the NHEFS sample that is linked to Medicare records is fairly small.
Nevertheless, the article does make a good point: researchers should not blindly use the CCW without fully understanding how the disease cohorts are created.
- Gorina, Y. and Kramarow, E. A. (2011), Identifying Chronic Conditions in Medicare Claims Data: Evaluating the Chronic Condition Data Warehouse Algorithm. Health Services Research, 46: no. doi: 10.1111/j.1475-6773.2011.01277.x