A Mixed Model Based Method for Eliminating the Masking Effect of Adverse Drug Reaction

Abstract

The invention discloses a method for eliminating the masking effect of adverse drug reactions based on a mixed model. Based on the report of adverse drug reactions (ADR) in China, firstly, the lasso logistic regression method with the effect of masking and subtraction is used to analyze the ADR database with less correlation The drug-adverse reaction combination is filtered, and the combination with a high degree of correlation is left as the data set after the first round of data screening; secondly, the combination of the number of reports less than or equal to 4 is eliminated by using the removal report method, and the remaining combinations are used as the second screening Finally, the traditional IC algorithm is used to re-detect the signal on the new data set in order to minimize the shadowing effect. The composite reduction model can effectively improve the reduction rate of the shadowing effect, and make the detection of adverse drug reaction signals more reliable and accurate, thereby further ensuring people's drug safety and providing reference for relevant decision-making.

Description

technical field [0001] The invention relates to the field of signal detection of adverse drug reactions, in particular to a method for eliminating masking effects of adverse drug reactions based on a mixed model. Background technique [0002] Internationally, the traditional detection methods of Adverse Drug Reaction (ADR) signals are mainly based on the theory of disproportionate determination. Commonly used methods for disproportionation determination theory include PRR, ROR, MHRA, IC, etc. However, due to the large and diverse data in the report, the data of various adverse drug reactions are mixed with each other, and the patients often use combination drugs instead of single drugs, the traditional detection methods of adverse drug reaction signals are often interfered to a certain extent, resulting in many false positives. Positive and false negative results are generated. In addition, due to the large proportion of many routine adverse reactions in the spontaneous re...

AI Technical Summary

Problems solved by technology

However, due to the large and diverse data in the report, the data of various adverse drug reactions are mixed with each other, and the patients often use combination drugs instead of single drugs, the traditional detection methods of adverse drug reaction signals are often interfered to a certain extent, resulting in many false positives. Positive and false negative results produced

In addition, due to the large proportion of many routine adverse reactions in the spontaneous reporting database, such as "cough", "dizziness", etc., the traditional detection methods of adverse drug reaction signals mostly use the relative value of the observed value and the expected value from the database. Suspicious reports that are not commensurate with the selected background are "filtered out" to generate suspicious signals, but valuable false negative signals commensurate with the background are also ignored, causing these signals to hide in the background and fail to attract attention in time. data masking effect

The masking effect makes the results of data mining or research often have a certain degree of distortion, which affects the relevant decisions to guide safe medication

Images