| Description |
In 1997, a report from the national organ-sharing registry stated that Latter-Day Saint (LDS) Hospital's kidney transplant survival rate was significantly less than expected based on data-collection forms submitted by the hospital to the national registry. Due to subjective data-collection methods and a manual reporting system, it was hypothesized that the registry may have incurred errors that affected the expected survival rates. The purposes of this study were: (1) to discover the amount of error in the data used to calculate expected survival; and (2) to evaluate the ability of a transplant information system to improve the quality of the data. We compared data from 214 transplant recipients at three points in the reporting process: (1) the initial paper data-collection forms, (2) a copy of the 1998 registry database, and (3) 1999 verification forms from the national registry. Each field was compared with a nurse's review from the transplant center for accuracy, precision and availability. The expected survival was also calculated for each of the three samples and compared to the nurse's review using a two-tailed paired sample t-test. We then loaded the registry data into a clinical information system that used a series of data quality rules to cross-check and validate the data. We compared the original registry data with the information system data and calculated expected survival rates for both data sets. We found a 10.14% error rate in the data-collection forms, a 5.99% error rate in the registry database, and a 3.10% error rate in verification forms. The errors from the paper data-collection forms made a small difference in the expected survival rate, increasing the rate from 92.6% to 93.3%. The small difference, however, did not support our hypothesis that errors in the registry compromised the expected survival calculation. The data quality rules in the information system reduced the error rate in the registry data from 5.99% to 5.17% and increased the expected survival rate from 92.6% to 93.0%, but again we concluded that the reduction of errors did not considerably alter the expected survival. We hypothesized two reasons that the errors made little impact on expected survival: (1) many errors occurred in fields that were not used to calculate expected survival rate, and (2) some errors from fields used to predict survival increased the survival rate whereas others decreased it, thus mitigating the effect of each other. Based on our findings, we recommend three ways to improve the quality of data: (1) integrate the data from the transplant information system with the current hospital information system, (2) improve the flow of information from the clinicians to the data entry personnel, and (3) electronically transfer the data from the transplant centers to the national registry. |
