Analytics cannot be a rear-view mirror. It has to work like the windshield in the front giving information and insights to improve the future. For this to happen, analytics should be tightly integrated within the system at the ground level,sharesUdai Kumar,CEO & President, OHUM Healthcare Solutions with Elets News Network (ENN)
Big data is a buzzword today. In 2018, big data spending is expected to reach $114 billion. Undoubtedly, it is generating a lot of hype across all industries, including healthcare. There are several applications and potential use cases of big data in healthcare and research institutes are also using it for advanced research projects.
Big data analytics, when used appropriately, can provide tremendous business value and competitive advantage. In the healthcare sector, it can offer incredible opportunities to improve the quality of care and financial performance.
Despite so much potential, there are a very few big data analytics projects thatare successful in healthcare.
Many organisations have spent huge time, money and efforts, and realised that something needs to change if they have to get real value out of big data analytics, especially in healthcare.
Let us look at some of the possible reasons behind the failures of big data analytics in healthcare –
- Lack of Right Data
Even the best business intelligence solutions can fail if the underlying data is incomplete or inaccurate. In healthcare, information resides in silos. Electronic health records (EHRs) have managed to digitise thousands of pages of data, but are not integrated into meaningful workflows.
Sharp boundaries across departments obstruct the smooth data flow required for meaningful analysis.
By definition, healthcare data is complex – Data needs to be gathered from heterogeneous patient sources, unstructured clinical notes need to be understood in the right context, there are large volumes of medical imaging data, genomic data and patients behavioural data, which needs to be captured through various interactions, communications and sensors. Unless and until all such information is rightly filled into the system, the analysis engine cannot offer meaningful insights.
- Incomplete Data
Having 80% data is not going to give you 80% accurate analysis. Incomplete data completely messes up with the analytics.
With the government mandate, hospitals have started adopting EHR at a minimal level, but the partial data is hardly any useful for analysis.
The inability of the system to gather and process the holistic data, understand which data connects to which data, is one of the major reasons for data analytics failure.
- Lack of Standardised Care
Standardisation of processes and functions helps in reduction of mistakes and thereby has the potential to improve the quality of care. There are instances where simply by building standardised imaging protocols in their EHRs, clinics have been able to decrease the frequency of MRI and CT scans by 30% to 40%. However, the implementation of standardisation protocol across hospitals is still not being widely accepted there are no standard procedures for data capture, diagnosis and drug prescription.
Lack of a standard data model which can be incorporated into the system severally limits the collection of right data in the right format and restricts the real-time processing of the data.
- Incorrect Application of Secrecy and Privacy Rules
The government privacy rules and regulations which regulate the security and disclosure of patients personal health information hinders the effective analysis.
Undoubtedly, the patients data privacy is paramount, but with things like data anonymisation prior to analytics, patient identity can be protected while leveraging the other useful disease-related information.
Healthcare analytics can be made more effective by applying privacy-preserving encryption while protecting the identity of a patient.
- Obvious Analytics
Everybody knows the epic failure of Google Flu Trends (GFT). The search giant missed the peak of the 2013 flu season by 140 percent!
This failure of GFT does not obliterate the importance of big data, but it does highlight the incorrect practices used in big data analysis.
Harvards HealthMap service made headlines for monitoring the mentions of Ebola outbreak in early 2014. But then, these seem to be obvious analytics. The harder ones, which require computation of multiple types of data, from multiple sources and in multiple formats, are more complicated to develop and most of the healthcare IT systems fail at that.
- Shallow Integration with Underlined Transaction System
The analytics engines cannot work in silos. The results and observations from analytics need to be used for improving the underlying transaction systems. Analytics cannot be a rear-view mirror. It has to work like the windshield in the front giving information and insights to improve the future. For this to happen, analytics should be tightly integrated within the system at the ground level.
The output of analytics should prevent the problem and direct the required solutions. Very few systems implement analytics at this level.
But things have started looking positive. Healthcare IT companies have started to integrate analytics at a more detailed level within the transaction systems, including the mobile devices used by doctors. Technologies like IBM Watson Health are allowing the experts to gain useful insights about patients as well as the population at large and take more informed decisions.
It is possible to overcome all the negatives mentioned above by making provisions and amending the lapses.
Several institutes have shown improvements through analytics when they implemented goal-driven analytics – with clearly defined goals around readmissions, sepsis mortality, VTE prophylaxis, etc. We are positive about the future of big data analytics in healthcare. At least the problems affecting the success have been identified and thats a great first step!
With careful attention and focus on data governance, metrics, standards and best practices, targeted solutions improvement program, actionable insights and machine learning healthcare analytics can deliver multi-fold results. In the upcoming blogs, we will discuss all these items in details.
About the Author
Udai Kumar, CEO & President, OHUM Healthcare Solutions, is ahighly successful serial entrepreneur, mentor and investor. Udai has a deep understanding of the global delivery model and specialised in global business development, human capital management, finance, software development, Six Sigma, systems & processes. He is an active member of theHealth Information Network group and is on the panel of Leadership Team at the Government of India level for healthcare technology.He is onTwitter athttps://twitter.com/_udaikumar