Article Text

PDF
Can the NHS be a learning healthcare system in the age of digital technology?
  1. Amitava Banerjee1,
  2. Lydia N Drumright2,
  3. Andrew R J Mitchell3
  1. 1 Farr Institute of Health Informatics Research, University College London, London, UK
  2. 2 Department of Medicine, University of Cambridge, Cambridge, UK
  3. 3 Department of Cardiology, Jersey General Hospital, Saint Helier, Jersey
  1. Correspondence to Dr Amitava Banerjee, Farr Institute of Health Informatics Research, University College London, London NW1 2DA, UK; ami.banerjee{at}ucl.ac.uk

Statistics from Altmetric.com

Introduction

‘Big data’ is defined by ‘7 V’s’: volume (most frequently cited1), velocity, veracity, variety, volatility, validity and value. In healthcare, ‘big data’ is associated with a step-change in the way information is gathered, analysed and used to facilitate disease management and prevention. With greater electronic data capture, there is enthusiasm for increased safety, efficiency and effectiveness in health and social care through, for example, machine learning and other forms of artificial intelligence (AI). However, factors maintaining and widening the gap between the promise and the reality need to be addressed.

Can ‘big’ be evidence-based?

Current best practice has its foundation in evidence-based healthcare, with growth in publications, but poorly managed scientific insights, poor recording of care and poor use of evidence.2 Big data could improve the status quo and support learning health systems (LHS).3

Computational methods can contribute to evidence management with automation of literature searching, critical appraisal and guidelines.4 5 Similarly, big data already contribute to aetiological, diagnostic, prognostic and therapeutic research, from -omics to electronic health records (EHR) trials.6 Critics emphasise lack of quality and validation of routinely collected clinical data, and risk of bias in observational studies, where scale cannot compensate for poor design.7 Conversely, data-driven approaches could transform a predominantly retrospective into a prospective or real-time paradigm, across disease boundaries.

Infrastructure and analytic tools are necessary but often poorly understood and underdeveloped. Automated extraction of necessary data fields in pseudonymised/anonymised format into curated warehouses is required with robust metadata catalogues and understanding of clinical context. To be available at the point-of-care for clinicians, data will have to be extracted, cleaned and processed promptly.

Preventing excesses and addressing deficiencies

Medicine grapples with conflicting challenges of overdiagnosis and overtreatment as well as underdiagnosis and undertreatment. Widening inequalities also manifest across many health systems, despite improving medications and technologies. Efforts to optimise healthcare delivery …

View Full Text

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.