| Description |
Patients with elevated lactic acid have a high mortality rate. In a retrospective search of the computerized LDS Hospital patient file, from February 1976 through August 1978, it was found that 198 patients had lactic acidosis and of those 118 died. The mortality rate was about 60%, However, only 2% of the patients admitted ever had the lactic acid test. Means of predicting and diagnosing lactic acidosis from other more common tests would therefore be helpful. Studies were undertaken to predict lactic acidosis from other more common laboratory tests (the SMA-6, SMA-12, C3C, and the blood gas test). STUDY I used the first set of data after patient's admission. Three multivariate statistical methods were used* linear discriminant analysis, Bayes method with equal bin widths, and Bayes method with equal bin counts. Linear discriminant analysis was the best method if the blood gas variables were included. True postive rate was 54,2% and true negative rate was 97.5% when the predictive value (true positives divided by the sum of true positives and false positives) was 23.3%. Bayes method with equal bin widths yielded the best results when only the SMA-6, SMA-12, and G3C tests variables were used. True positive rate was 40.0% and true negative rate was 98.7% when the predictive value was 23.5%. In STUDY II, only patients who had the first high lactic acid within 24 hours of admission were included. Again the three multivariate methods were utilized. Linear discriminant analysis always yielded the best results. ''When using only variables from SfMA-6, SMA-12, and C3C tests, the result was better than in STUDY I. However, if the blood gas variables were included, the results were not as good as in STUDY I. In STUDY III, we evaluated patient information for three different time periods* from 0 to 6 hours, 6 to 12 hours, and) 2 to 24 hours before the first high lactic acid result. Linear discriminant functions obtained from the closest set of data before lactic acidosis and from STUDY II were tested and compared. In the period from 0 to 6 hours before the first high lactic acid, discriminant functions from the closest data and the STUDY II data performed about the same if the blood gas variables were included. . True positive rate was 60.0% while true negative rate was 97.6%. When only variables from the SA^-6, SMA-12, and CBC were used, the discriminant functions from STUDY II gave the better results. True positive rate was 50.8% while true negative rate was 98.7%. For the periods from 6 to 12 hours and 12 to 24 hours before the first high lactic acid, none of the discriminant functions worked well. The results of the above studies warrant a clinical trial. |
