Health Data Analytics 2016


Health Data Analytics I had the privilege and pleasure to attend HISA’s Health Data Analytics conference in Brisbane on 11 and 12 October 2016. What follows is this particular BI and Analytics consultant’s impressions and insights from the conference in terms of the main themes covered and the messages and impressions I take away, again from my particular context. I’m sure a clinician or an allied health worker, for example, may have totally different impressions and take-aways.

Details of the program and participants can be found on the conference website. There were some excellent keynotes, panel sessions and research and industry short papers. However, this is not a review of specific papers. Neither do I reference individual speakers – even though all of them did such a great job.

Data collection and storage

There is an ever growing pool of health-related data being collected, both internationally and in Australia, and being put to use and made available for use in interesting and useful health analytics initiatives. Two particular trends stand out for me.

The first is organisations going to great trouble to amass and integrate as best they can large collections of interrelated data across functional, organisational and regional boundaries. The interesting work goes into linking related data and of course protecting individual’s privacy (more on privacy below). Most of this data is then made available for researchers and other interested parties to conduct analytics on this data or to in turn link it with their own data for contextual or comparative analytics.

The second is the area of “huge data”, if we can coin such a term. This is mostly used in the area of genome research. The storage space required to store a single patient’s cancer genome sequence was stated to be in the order of 300GB. So you can imagine the storage required to collect enough data points to perform significant comparative research or to deduce trends and detect exceptions. Organisations have come to the realisation that it is not cost-effective to maintain so much storage and processing capability in-house, so more and more are utilising cloud-based facilities. Cloud-based storage may be economical, especially for research projects of relatively short or varying durations, but of course it introduces other complexities and challenges such as uploading such enormous datasets.

Privacy and security

Patient and provider privacy is very important when it comes to healthcare data. Privacy needs to be addressed from both moral and legal perspectives. On a technical level, data privacy and security controls need to address data acquisition (data provision and entry), data at rest (i.e. data storage) and data exploitation (e.g. reporting, analysis and other analytics uses).

As an example of the issues involved, in some cases de-identifying patient data is not sufficient for patient privacy protection. Using appropriate statistical techniques, individuals can still potentially be identified, especially in sparse rural settings or in cases of patients with rare diseases or unique combinations of conditions or treatments. While the primary objective of using such data may be or may have been noble, one must always consider the potential “down the line” risk of exposure or disclosure.

Interoperability and data sharing

One of the biggest challenges in healthcare, not only in the context of analytics and research, but also in normal operational data flows both between different systems within the same organisation and between the systems utilised in different organisations, is to overcome the problem of siloes of data that don’t integrate. In other words, getting systems to communicate with each other and “share” data. You know the scenarios – sending data to medicare, sending a referral from a GP’s practice to a hospital, sending instructions from the GP to the independently run radiology organisation, getting the results back, and the list goes on and on and on, especially if we consider clinicians, allied health and other healthcare providers working across organisational boundaries.

If you operate in this space, you have to add another protocol and message format to the list that already includes HL7, Snowmed, XML, JSON and so on. Check out FHIR (pronounced like “fire”) – the Fast Healthcare Interoperability Resources standard. FHIR defines a set of “Resources” that represent granular clinical concepts. Make sure you know what FHIR is, how it works, its implications and how its implementation is being rolled out. Believe me, it’s the next big thing in healthcare interoperability. And it was all started by an Australian…

BI and quality / performance improvement

With this being a health analytics conference, the research and keynote papers shifted focus to the more advanced forms of analytics. However, I got a good sense from the discussions and comments passed around that conventional BI and reporting is far from done and dusted. BI is the enabling platform that should take the data wrangling slog-work (and potential inaccuracies) out of quality and performance management. However, many organisations are still battling to get BI and operational reporting bedded down.

Some of the complexities relate to the trends I’ve highlighted above – larger volumes of more complex data to process, more systems to integrate with within and across organisational boundaries, increased security and privacy demands, all this coupled with changing business strategies, increasing regulatory requirements, more urgent and more varying ad hoc requests for information and so on. Healthcare organisations are going to have to look at more agile BI delivery approaches, “wider” BI ecosystems and potentially utilising modern analytical platforms rather than scaling their data warehouses to the max.

For analytical research and proof of concept projects we may well use sandbox environments for analytics, but come the day we want to operationalise the analytical insights back into our operational systems, we’re going to require a production standard and quality BI system as foundation for analytics as well as for tracking and reporting the impacts out analytical insights have on the organisation.

Advanced analytics

Of course there were advanced analytics. A lot of really good stuff. If you think about it, there are so many steps along the patient journey where analytical insight can make a huge difference – from predicting readmissions, mortality and other clinical events, through staff, resource and patient flow optimisations to large-scale unstructured text analytics and root cause analysis on genome datasets. The breadth, level and type of research, as well as the results being reported are really interesting and impressive.

However, my impression is that a lot of this is being done in the research lab or by research focussed institutions. So we see valuable additions to the published pool of knowledge, but I see more of that happening in an (academic) research context, than what it is being implemented “on the floor” in healthcare organisations to directly affect patient outcomes. What we need to see more of are reports and publications of these analytical outcomes applied in field trials to investigate their impact on real operational processes and thereby on patient outcomes. It is really stimulating to see the extent and level of advanced healthcare analytics research, but how much of it was really used in the end to improve patient outcomes? Let’s state it in terms of the bottom line – how many lives are currently saved through analytics versus how many can potentially be saved through analytics?

Information management strategy

A few years ago, this was a data governance conference, then it changed to a big data conference and now it is focused on health data analytics, which is all in line with the informational maturity growth that healthcare organisations have to go through, aligned with the changes taking place in the information industry. So it is natural that the amount of material on information management and data governance would reduce, and the same applies to “big data”, except where it is related directly to health analytics projects.

However, similar to my sentiments about BI and reporting, even though it was evident from some of the papers that there are some organisations embarking on proper information management strategy journeys, I got a feeling from the conversations held and the comments dropped in presentations that few organisations are yet devising and implementing information strategies of the level and maturity that one would expect. Having said that though, I know it is a lot to expect from a healthcare provider to be increasing analytical maturity, and rolling it out as part of a serious information strategy, while still saving more lives and increasing operational efficiencies.

But that is what it is going to take to address two of my previous points. If you want to raise your BI implementation to the level that quality and performance monitoring is automated, and if you want take analytical research to the level where it can become operationalised, you have to draw up, implement and live out a progressively maturing information strategy.

Concluding remarks

You can probably gather from my enthusiasm and from the aspects I highlighted that HISA’s health analytics conference is a worthwhile event to attend if you have anything to do with healthcare data, information, BI, insights or analytics.

I came back so inspired from the event – we can do meaningful BI and analytics at every step of the entire patient journey, we can devise and implement so many analytical insights to improve operational efficiencies, we can assist so many organisations to draw up and implement information strategies, my goodness, this work can actually save some lives and result in better patient outcomes…  So I return with all these analytical projects lined up, but then back in the office the depressing realities of the lack of investment and the challenges of getting the right funding hit me squarely in the face again. Another day, another river to cross. Oh well, no matter how that all turns out, see you at health data analytics next year!

Photo credit Kara Burns @karaburns

Leave a Reply