Tony Fultz (director of upstream manufacturing at Fujifilm Diosynth Biotechnologies) presented an “Ask the Expert” webinar on 23 August 2017.
Human error was responsible for about 75% of batch failures at contract manufacturing organization Fujifilm Diosynth Biotechnologies in 2016. Technicians operate in a dynamic and stressful environment, both contributing factors to human error. With multiple projects running concurrently, some employees are involved in executing several processes during a shift.
The company has several ways of reducing human error: e.g., human-performance tools, a certification program dependent on theory-based training, and a robust method of task performance. Proper task performance is a three-step process using human-performance tools to provide a consistent approach to all tasks, regardless of their complexity: Discuss it, do it, review it (DDR).
Discuss It: The first step ensures that a performer and a witness identify their roles and share an understanding of what is required before task execution. If they disagreem or have questions about a task, then they will stop and escalate the discussion to area management immediately. Discussion should be tailored to the task: If you are setting up a bioreactor or performing an inoculation, it will be more in-depth than performing a simple task such as sampling. The important thing is to discuss what you are getting ready to do.
Do It: Next is execution of the task itself. Communication is key here, with all actions verbalized focusing on value and unitive measures. The performer verbally states the action to be taken. The witness verbally acknowledges that action including related measurements. The performer does the task while verifying what is being done. This keeps both people focused on the task, and it leaves no doubt that correct action was taken. Note that verbalization also works when a performer is working alone.
Review It: Was the step executed properly? If the answer is yes, then the performer and witness sign off on the step in writing. If it was not performed correctly (or results are unexpected), then they immediately escalate to area management. This step can be missed in many settings because it is human nature to complete a task and then move on to the next one. Although reviewing might not prevent an error, it should detect one as soon as possible.
Using this system, the company has reduced deviations by 65% year-to-date. Human error always will be the main contributor to deviations/failures. But we can and should improve procedures continually and provide engineering solutions and training to our technicians. Mistakes will be made because people are fallible. If we are diligent in our use of the new DDR system, though, we can reduce human error (and thus deviations and subsequent batch failures) significantly.
Questions and Answers
How much time does DDR add to a task? On average, the method adds 15–25% of time to a task, depending on its complexity. I tell technicians in training that if we have time to repeat a task, then we can spend an extra 15–25% of the time to get it right the first time.
How many areas are involved in this across Fujifilm Diosynth? Right now it is piloting in manufacturing, but we are looking to expand it to other areas such as QC microbiology and analytical laboratories.
How much labor is added if you need a witness for key steps? We require that all critical steps be witnessed, and we have some that are “reviewed by” or “checked by” someone independently. The focus is to identify critical steps. I’m not prepared to tell you that we could do it with fewer full-time employees (FTEs).
Do witnesses rely on verbal instruction, or are they also required to read an instruction themselves? They absolutely should be reading the instruction themselves.
In the review step, do you always use both concurrent and independent verification? Typically, it is one or the other. The witness provides concurrent verification. Independent verification (our “checked by” box) can happen later. We may use both in a record but not in the same step. All critical steps are witnessed; “checked by” is used for noncritical steps.
In your experience, about how many issues bypass the first verification and are caught by the second? I cannot tell you how many issues are prevented in real time by the witness policy, but we have had issues caught by manufacturing and quality review. Our goal is to prevent errors or identify them as quickly as possible.
Is a reviewer independent or one of the two operators? The reviewer should be independent. Manufacturing review typically happens after the fact. There is no requirement or independent review of a step immediately after it is executed. That is the purpose of the witness and the DDR program, to make sure that the step is performed correctly.
You can watch Fultz’s full presentation (with slides, audio, and question-and-answer session) now.