M-10: Study on Reducing Errors in Data Input to a Case Report Form
Poster Presenter
Hikari Ishii
Student
Waseda University Japan
Objectives
The objective of the study was to propose a novel derivation method to reduce errors during data input to a case report form (CRF).
Method
We classified data input errors to CRFs into error modes and identified error patterns. Based on a work model used for the data input process, we estimated the causes from the error patterns. Then we developed a method to reduce data input errors for the each estimated cause of errors.
Results
Data input errors detected as data correction operations were extracted from the audit trail data, which contain all the process of data input to CRFs, of 17 clinical trials.
A total of 17,833 errors were extracted from the audit trail data and the errors were compiled on a Form basis in CRFs. It was found that errors in the Form of 'concomitant medication' were the most frequent (25.2% of all errors), and the errors in the Field of "medication name" was the most in the form.
Next, we specified the error contents by comparing input data contents before and after an error. We analyzed the error contents to classify the errors to the error modes defined in the previous study [1]. As a result, it was found that the errors in data input to CRFs could be classified into 8 error modes (e.g. "missing input").
However, a single error mode, for example, "missing input" includes various error types such as, ‘disappearance of dosage form’ and ‘omission of dose’, etc. Therefore, in order to determine targets of error types to be reduced, the error types observed in each error mode were summarized into error patterns by their similarity. We analyzed the 521 errors occurred in the Field of "medication name" and collected 14 error patterns across the 8 error modes. We can plan countermeasures from an error pattern with high error frequency.
Further, we modeled the process of data input to a CRF. By modeling, it is possible to specify in which process an error occurred. In accordance with the previous study of medication accident analysis method [2], we estimated causes of errors for each error pattern based on the model.
Finally, with reference to the method of realizing error proofing [3], we derived a method of reducing errors for each cause of error. There are 18 ways to realize error proofing (e.g. "increasing the amount of information"). As a result, a total of 63 measures to reduce errors could be derived (e.g. "Create a Field of dosage form in CRF").
Conclusion
Conventionally, contents of errors in CRFs have been not analyzed, and countermeasures to prevent recurrence have not been taken. In this research, we proposed a novel method of reducing errors in data Input to a CRF.
A risk-based targeted sampling SDV is being extensively studied. The proposed method will be useful to be included in a sampling SDV design for grasping a tendency of errors. For example, performing sampling SDV on a Form which is prone to error is conceivable.
Since data input to CRFs is conducted at medical sites, it is difficult for sponsors to examine contents of errors and their causes. In this study, we propose a method to mitigate errors which can address each estimated cause of errors. It enables sponsors to propose countermeasures to a medical site to reduce input errors to CRFs by analyzing the audit trail data of the site.
References
[1] Takeshi Nakajo et al. (1984): "Study on fool proofing of work - Principles of fool proofing", "Quality", 14, [2], 128-135
[2] Ryosuke Hasui et al. (2019): “A Method of Identifying Process Factors for Medication Incidents”, “Quality”, 49, [1], 82-94
[3] Ikuo Ozaki et al. (2005): "Research on Reduction of Drug Accident Using Error Proofing", "Hospital Management", 42, [3], 361-373