OR WAIT null SECS
Data may be used to improve (or remove) a corrective action/preventive action.
Being cautious is a good thing, especially when it comes to product quality and safety. Pharmaceutical companies, in particular, have to be aware of this. Products that can save patients’ lives can adversely affect them with a relatively small change. That’s why some pharma companies immediately open a corrective action/preventive action (CAPA) when something unexpected occurs. No one could accuse these companies of not giving a quality event the gravitas it deserves. However, this method is a recipe for overworked quality managers.
A more moderate approach is warranted, but deciding which events go to CAPA and which don’t requires a risk-based approach that uses advanced analytics and digital tools to guide the quality department in its decision-making. Eventually, the use of artificial intelligence (AI) can take much of the guesswork out of determining if a CAPA is needed, finding the root cause, and even carrying out remediation.
A deviation, nonconformance, or out-of-specification result can indicate a significant problem with a product. However, it can also indicate an operator was having a bad day, which led to an isolated event. There’s a whole spectrum of how serious a quality event is, and not all of those should lead to a CAPA. That said, why do some companies seem a bit too eager to CAPA?
At least part of the problem is a lack of visibility. CAPA should be based on risk. But risk needs to be determined by frequency and severity. In some cases, severity can be straightforward just based on a single event. However, frequency by nature requires multiple data points. Without a consistent, electronic way of documenting CAPA, pharma companies can’t analyze their events to determine if a CAPA is necessary. Hence, the dependence on the CAPA process.
As with so many quality problems, this one can be solved by data. If a quality management system (QMS) makes it difficult to track and trend quality events, finding patterns or determining root causes is difficult. McKinsey & Company found that “advanced analytics is the linchpin of transforming the investigation process” (1). Without it, finding the root cause for the event and making sure the right preventive and corrective actions are in place can be a months-long process. Capturing that data in an electronic format and using a system that can connect all that data is the first step.
Technology has advanced to the point that there are multiple ways to do this. Ideally, a pharma company can use a system that has multiple components that can be used throughout its organization (e.g., a quality system that connects to the manufacturing line for consistency in standard operating procedures [SOPs] and work instructions). Because reality is often less than ideal, systems that can integrate with each other also provide that single source of truth. For example, connectors that let the enterprise resource planning (ERP) system exchange information with the QMS.
Once those data are connected, they can be analyzed. This can be done by doing a data dump into Microsoft Excel or by using a business intelligence solution. It’s becoming more common for the QMS, ERP, or other enterprise software to provide analytics capabilities as part of its solution. These programs enable quality professionals to find the relationships and causal factors and point them toward a possible root cause. A good example of this is with natural language processing (NLP), a subset of AI that can understand free text in much the same way as a human can. This is invaluable if the data collection involves those fields. In customer complaints, the same problem with a pharmaceutical could be described as “makes me nauseous,” “I feel queasy,” and “makes me sick to my stomach.” Rather than having a human go through that information, NLP quickly determines that all these complaints are related.
When a pharma company has many of the same or similar issues, it’s usually an indication that a CAPA is warranted. But already having those events connected to a single investigation expedites matters. Now the quality professional can focus on the investigation and, with AI, make more confident assertions about how to solve the problem. AI programs can narrow a person’s focus to the most likely root causes. These programs can further use prescriptive analytics to indicate what the corrective action should be, what other processes could be affected, and how to mitigate any possible negative effects from it.
AI’s potential in quality management isn’t going to replace the people, it’s going to augment them. This brings the best of both worlds to the investigation process by letting AI take care of typical leg work and leaving humans to carry out work that requires human reasoning. An expert from Cognizant pointed out: “Decision-making and strategic-thinking skills are best performed when workers are supported by the insights generated by AI and data analytics and freed from performing rote and repetitive work by intelligent automation” (2).
The more data an AI has access to, the more intelligent and accurate it becomes. It can eventually launch and populate a CAPA based on risk or a trend related to a quality event. As previously mentioned, it can tie
multiple related quality events together. Just as CAPAs touch other aspects of quality, AI’s benefits extend to all areas of quality.
Some other practical uses of AI in pharma include:
This is just a small list of examples, but it gives an accurate depiction of what quality management can look like when AI is used. All the tedious, repetitive tasks that quality professionals do right now can be handled by automation, data analytics, and AI, leaving employees to focus on more important projects.
CAPA doesn’t have to be the tedious, time-consuming process it usually is. The technology to provide these benefits is available right now. However, purchasing these solutions today doesn’t mean they could immediately perform as mentioned above. The real benefits of AI come over time as more data are included and the algorithm becomes more accurate.
Pharma companies are already experiencing benefits from data and AI. For proof, “look no further than COVID-19 vaccine development. Companies shattered previous vaccine development records due to their ability to capture, store, process, and analyze machine data” (3). This progress isn’t limited to the pharma companies that were focused on vaccine development either. It’s spread across the entire life science industry.
A report from Cognizant revealed that 73% of life sciences companies have at least started projects to use data analytics and 68% have started projects to use AI (2). If a company doesn’t plan to start similar projects, catching up is only going to get more difficult as time goes on. On the plus side, a pharma company doesn’t need to send its employees back to school or hire AI experts to start this journey. AI solutions can help quality professionals improve their decision-making now, lowering the number of quality events that turn into CAPAs, shortening the investigation time, and decreasing the chance of issues recurring.
1. T. Foster et al., “Making Quality Assurance Smart,” McKinsey & Company, Jan. 29, 2020.
2. B. Williams “The Work Ahead in Life Sciences: Cures at the Speed of Digital”Cognizant, Feb. 24, 2021.
3. “Tech Trends 2021: A life sciences perspective,” Deloitte, 2021.
Sue Marchant is senior vice president of Product at MasterControl.
Volume 46, Number 4
When referring to this article, please cite it as S. Marchant, “Put a Cap on CAPAs,” Pharmaceutical Technology 46 (4) (2022).