On March 14thThe Digital Health Rewired Conference was held in London and the attendees along with the presenters had highlighted the technologies and techniques for improving the efficiency of healthcare delivery globally.
The three most important themes that came up during the conference were:
In 2018, the NHS had experienced an 8.5% vacancy rate i.e. 100,500 vacancies including those for doctors and nurses. It was the highest vacancy rate recorded in any public sector service of UK. Hence the healthcare providers were unable to deploy digital transformation projects.
However, there was an unprecedented level of citizen interaction with the NHS through digital channels. It was highlighted that the Coronavirus epidemic created a ‘must act’ environment which is emboldening healthcare innovators and forcing teams to make decisions which never made the agenda before.
For example, Matt Neligan, Director of Primary Care Commissioning Transformation for the NHS England, shared a case study from the NHS Whitley Medical Centre in Reading. They suffered from an overworked provider staff, unable to fill two open GP positions. They knew something had to change because they didn’t have enough time to manage their complex cases properly. By deploying digital patient messaging, appointment scheduling and electronic consultation (GP at Hand) services, they decreased face to face consultation volume by 43%, while maintaining 87% patient satisfaction.
Information, the commodity that reduces uncertainty, is crucial for fueling the digital transformation of healthcare.
However, the information doesn’t equal knowledge. While tremendous resources go into researching and publishing clinically relevant knowledge, mobilizing this and applying it to improve patient outcomes is a huge task.
Amazon, Google, Microsoft, and other technology giants are working to apply natural language processing (NLP) to back clinical knowledge out of the body of research published online. However, these approaches are reliant on the effectiveness of algorithms. As these algorithms differ across cloud computing providers, the clinical knowledge could vary depending on which NLP algorithm processed the report.
In contrast, researchers could publish two versions of their studies, one for human reading and the other encoded in language designed for computers to index and analyze. In this paradigm, downstream NLP would not be necessary.
Unfortunately, there are challenges on the horizon. Publishing two versions of reports would be more expensive, and it is unclear who would pay. Also, the taxonomy researchers use to classify their findings would have a significant impact on the utility of computer-processed knowledge. How the research community would make these decisions was not shared in detail. It seems that finding a method that is normalized enough to be useful yet sufficiently flexible for the diversity of use cases will be a non-trivial endeavor.