THANK YOU FOR SUBSCRIBING
Analytics has a long history of being a rising force to improve plausible outcomes. Metrics and various analytical models bring reality to the forefront, but it begins and ends with the proper data set and informative actions. Clinical research has opened its arms (no pun intended with regards to wearables) to new methods and techniques for obtaining access to a wide variety of data sources; thereby only improving the robustness and conformity of data to be analyzed. The advent and exponential growth of IoT (mHealth, biometrics), data standards including Fast Healthcare Interoperability Resource (FHIR), new programmatic and statistical methods, and continued accumulation of patient electronic health records (EHR) has led to this propitious effect and will continue to do so for years to come. Great news for researchers and data scientists is that we are gaining health intelligence in novel ways, while maturing older sources (EHR). Although data access is one part of the equation to solving important medical questions, in this digital era, one may argue it is becoming the easiest once all patients have been enrolled and collecting too much data has its own consequence -- striking the right balance is a trial protocol’s treasure.
Data cleaning, data review, and cross-functional multivariate correlations requires tireless detailed attention to achieve correct analytical results.It is deep in understanding the data architecture, points of origin, and data flow where one can truly excel with cleaning and review tasks. Performing a degree of data forensics leads to intelligent data management. For example, medical health devices and applications require an understanding in how the device or app was calibrated to the fit-for-purpose design and how data is transmitted through subsequent locations (e.g. cloud platforms) is instrumental to building targeted data quality controls. Same holds true for EHR, having knowledge of specific unstructured data fields allows bespoke natural language programming algorithms to translate meaning to empirical scientific evidence, which ultimately fuels statistical analysis plans. In addition, emerging techniques in Machine Learning (ML) bring the promise of more efficient methods and approaches to data cleaning and review. Based on large historical data sets and recorded events, ML driven rules can be applied to aid data quality tasks. ML is and will continue to be an important growth area within data management. Finally given the voluminous amount of data required for clinical trials alongside data source variations, creating well-tuned computer processing algorithms which rely on data standards is essential. To help in this regard, developing a programmatic execution scheme that leverages distributed data processing and architectures scaled to size, such as those within Databricks, can provide the best path forward to receiving and processing data at accelerated time intervals.
Data listings and data visualizations have become an art form, but that may not be enough. In the fast-paced world of trial execution, stakeholders want answers succinctly and to drive action quickly. Historically analytical visualizations and tables such as those derived from Spotfire, Tableau, R, etc…have performed well at bringing business intelligence forward, acting as surveillance, and providing confirming/disconfirming information. A strong breakthrough is taking it a step further where analytics become part of a multi-step process in order to bring automation to life. An example of this type of action is in financial investments, where a price target can lead to the automatic purchase or sell of an asset. Similarly, in clinical trial research, where appropriate, predefined actions can be executed based on set data values. This type of execution where detect and action become seamless without manual intervention is a promising area in analytical-robotics processing and a source for increased productivity.
Analytics is a formidable force in the ability to launch promising treatments to those in need and will continue to evolve as programming standards, regulations, and technologies change to meet healthcare demands. Although, analytics won’t succeed alone -- data access, data management and action will all lead to an important differential.