Our Posts

Preparing data for analysis

No Comments

And data doesn’t come for free. It’s expensive in terms of the setup of it and also the ongoing stories of that data in whatever form it is 

From “Leaders’ perspectives on learning health systems
BMC Health Services Research 2020

Since the inception of the idea that medical practice should be guided by objective and measurable patient outcomes, collected in controlled clinical trials, data has become the focus point for medical decision making and policy development. As the price and availability of data storage infrastructure plummeted, their ubiquity in the healthcare value chain has exploded. “When in doubt, store it. We’ll figure it out later” seems to be the attitude of healthcare application developers. But the unintended result is that different parts of the data are collected in a patchwork of dozens of incompatible databases, at inconsistent quality. Making sense of the data is expensive and time consuming. It requires mental agility that is hard to train an AI to perform. 

Every dark cloud has a silver lining, and in this case, it is the patterns that can be found in the noise of real world healthcare data. Patterns can be exploited by AI, but the common approach to adapt deep neural networks (highly optimised for applications such as facial and voice recognition), pre-trained on another dataset, fails to deliver the precision that healthcare applications require. Evidentli, in contrast, developed an ensemble of fit for purpose algorithms, that aggregate, cleanse and standardise data in situ (without pre-training). Developed for over a decade, dozens of machine learning, natural language processing, deduction and inference algorithms provide a discovery, transformation and verification workflow that transforms healthcare data in weeks rather than years. All while consistently achieving 99% precision or more.

Clinical Data Warehouses are significantly different to other data warehouses and accommodating these differences using fit for purpose software can determine the success of the project.

Recent Posts

Leave a Reply

Comments, questions or queries? Let us know!"