Can administrative data identify active diagnoses for long-term care resident assessment?

Many veterans receive rehabilitation services in Department of Veterans Affairs (VA) nursing homes. Efficient methods for the identification of active diagnoses could facilitate care planning and outcomes assessment. We set out to determine whether diagnostic data from VA databases can be used to identify active diagnoses for Minimum Data Set (MDS) assessments. We evaluated diagnoses being considered for inclusion in MDS version 3.0 and present in at least 15% of a sample of VA nursing home residents. A research nurse following a standardized protocol identified active diagnoses from the medical records of 120 residents. A clinical nurse also identified active diagnoses in 58 of these patients. Inpatient and outpatient diagnoses from the VA National Patient Care Database were identified for the past year. We calculated kappa, sensitivity, and specificity values, considering the nurses' assessments the gold standard. We found that kappa values comparing research nurses and databases were generally poor, with only 8 of the 19 diagnoses having a value >0.60. Levels of agreement between the clinical nurse and administrative data were generally similar. We conclude that VA administrative data cannot be used to accurately identify active diagnoses for nursing home residents. How best to efficiently collect these important data remains uncertain.