A massive backlog of routine and elective healthcare services has built up, as hospitals have hunkered down to care for Covid-19 patients.
This backlog will take months to clear, and data can help healthcare systems prioritise and cope. The amount of health data available has exploded, but many institutions simply don’t have the capability to store and fully use this new information.
How can data streamline healthcare delivery post-Covid? And how can organisations ramp up their capacity to take advantage of this? Cloudera explains.
Insights into patient health
The amount of data produced by health organisations has increased by a staggering 878% between 2016 to 2018, according to a 2019 Dell EMC report. And with Internet of Things (IoT) devices becoming more widespread, data collected from wearables, healthcare apps, and sensors are only set to rise.
A greater amount of medical data allows AI and machine learning to be more accurately trained to forecast health conditions. Researchers have already developed models to predict the risks of heart failure and stroke. This could help physicians identify the risk factors early and provide more preventative care.
However, healthcare data is often underused. Information such as images, videos, and sensor output is difficult to organise and file away neatly in hospital information systems. As a result, large amounts — up to 80% — of medical data remains unused after being created. This represents huge swathes of untapped potential to help prioritise healthcare where it will have the most impact.
Cloudera’s Enterprise Data Cloud works on-premise, in the cloud, as a hybrid of both, or across multiple clouds, allowing healthcare providers to store and process both structured and unstructured data types onto a single cohesive platform. Using established open data management tools and analytics software, organisations can rapidly gain insights into patient health and how to best deliver services
Such solutions have proved useful for companies such as Quest Diagnostics, which works with a third of the United States adult population. With Cloudera, Quest pulls together huge volumes of data from multiple sources into a single enterprise data hub. This has yielded insights that improve the firm’s operational efficiency and has added depth to their clinical reports.
Machine learning eases manpower crunch
Machine learning also shows huge promise in reducing manual work for healthcare practitioners. It depends on large datasets of information, which algorithms learn to process and interpret. These huge sample sizes allow them to accurately diagnose and detect certain conditions as accurately as skilled healthcare practitioners.
For example, Microsoft’s Project InnerEye uses machine learning to map the outline of a prostate tumor from radiological scans, to ensure that the treatment results in minimal damage to healthy cells. This process can take a skilled specialist up to four hours, but InnerEye technology completes it in a matter of minutes.
Amidst a global shortage of radiologists — US studies estimate that to cope with demand, radiologists must interpret one image every 3-4 seconds in an 8-hour workday — this is welcome news indeed. Due to the critical need for such services, the use of machine learning in radiology is likely to grow exponentially.
However, organisations may not always have the ability to process large amounts of machine learning data locally. Often, “organisations are really very interested to work with machine learning” but many lack the computational power to do so, notes Wing Leong Ho, Cloudera’s Technical Lead for ASEAN.
Cloudera’s software also allows companies to ‘burst to cloud’, or rapidly move applications from a private cloud to a public one to deal with spikes in IT demand. When local servers are overloaded, “data will be intelligently moved on premise to the public cloud and executed there,” Ho explains. This allows organisations to deal with a sudden increase in data, paying for extra processing capacity only when necessary instead of purchasing solutions that remain underused most of the time.
This comes in handy when organisations wish to work with machine learning, where there are tremendous fluctuations in the amount of data to be processed every month. For example, a healthcare provider may ordinarily only need to analyse one month’s worth of data for machine learning, but there are exceptional scenarios in which they need to look at 10 years’ worth of data — in this case, cloud bursting could allow for surplus data to be seamlessly computed on the public cloud.
As the applications for data in healthcare steadily increase, healthcare institutions similarly need to step up their processing power to keep up. A flexible and effective data management solution can help organisations stay ahead of the curve.