The Potential of AI to Transform Life Sciences is Huge…but Data Quality is Vital to its Success

In the rush to get concrete ROI from emerging technologies such as AI and machine learning, life sciences firms can’t neglect to focus on the veracity of data, warns Vada A. Perkins, former Senior Advisor for Regulatory Science, US Food and Drug Administration

Subscribe to our blog

Life sciences has been a relatively late starter when it comes to deploying digital innovation and the latest technologies. But despite being an industry that until fairly recently was based mostly around paper processes, this is starting to change.

The rise over the past five years of artificial intelligence (AI) and machine learning has seen life sciences firms begin to invest in these technologies in significant amounts, with some observers expecting investment levels to increase four-fold over the next three years. Organisations are attracted to AI by its potential to transform certain processes, particularly with regard to data.

Yet in the rush to get maximum value from AI investments, many businesses are overlooking a critical element in success – the quality and veracity of data

Data- Overlooked and Under cooked

People within life sciences have a tendency to look at technology investments and not fully understand how best to utilise them. There is not enough focus on the veracity of data. They look at the data that they have already got and try and get out what they can from that. But the quality of data is hugely supportive of the validity of the outputs, and the quality of data amongst life sciences firms is not always as high as it could or should be.

Data enters an organisation in many ways. Sometimes it is input by scientists, but at other times by data entry clerks. Furthermore, as it gets exchanged, data is also vulnerable to interpretation and bias, that over time can change its meaning. This all means that it can be easy for the quality of data to be compromised, at entry point and during its lifespan.

But the awareness of this lack of quality is simply not there in sufficient volume, and the detrimental consequences of poor data on AI initiatives will always be exposed at some point.

Health Authority Requirements

Any health authority is going to want to check the reliability of data before they can make an informed decision, bearing in mind provenance, quality, safety, and effectiveness. So however good the results might look from a manufacturer’s perspective, if the initial data is not robust enough, then the health authority may well reject the results.

It is vital for any manufacturer to keep this front of mind at all times, asking itself ‘how are we accruing info?’ how are we validating it?’ and ‘how are we versioning it?’ The data sets must be of sufficient quality and be able to withstand scrutiny from the peer review system. If not reproduceable, and not robust enough, substantial time and resources could be wasted.

Throughout my career I have observed people making assumptions about data. A common phrase to hear in industry is ‘assuming the data is correct’ – but that’s a massive assumption to make. You need to be sure the data is correct, otherwise the findings are not worth much at all.

As people engage in AI projects, there is no doubt in my mind that it will come back to data quality, reliability and validation. Data must meet certain criteria to ensure veracity. Then and only then, can the potential of AI and machine learning truly be realised.

I will be elaborating on all of this at BE THE EXPERT 2018 in June, and exploring in more depth just why data quality is so important when it comes to ROI in technology and innovation.  I look forward to speaking with you there.  

Published on    Last updated on 01/07/2019

#Life Sciences, #Artificial Intelligence

About the author

Formerly Senior Advisor for Regulatory Science for the U.S. Food and Drug Administration, Vada is a recognized data standards and governance expert with extensive regulatory experience in the development, interpretation, and implementation of guidance, regulations, and international standards for global data on product lifecycle management. His expertise in regulatory science complements his proactive collaboration with national and international stakeholders to promote global regulatory convergence that allows for better assessments of the safety, efficacy, quality, and performance of medicinal products worldwide. He received his degrees from John Hopkins University, the University of Southern California, and the University of Maryland.


Participate in this discussion