Improvements in data quality in the USRDS database: determining treatment modalities.

Past USRDS estimates of the prevalent ESRD population have exceeded the counts reported by the HCFA Annual Facility Surveys. One expects the USRDS estimates to be lower because the Facility Surveys include non-Medicare patients generally not in the USRDS database. The methodology for determining the treatment histories of patients has been modified to define lost to follow-up periods as periods of at least one year during which the patient has no dialysis data and does not have a functioning transplant. Patients are not counted as prevalent when they are in such a lost to follow-up period. This change brings the USRDS year end prevalent counts down to about 94 percent of the Facility Survey counts of total dialysis patients and slightly over the Facility Survey counts of Medicare dialysis patients. This change raises prevalent mortality rates by about seven percent over the rates reported in the 1991 USRDS Annual Data Report. We expect to make further refinements in this methodology.