«

»

Evolving into a Data-Centric Healthcare Organisation

Share

In a previous post I reasoned why healthcare organisations should become more data centric. In this post I suggest a fluid and adaptable approach to mature the organisation with respect to data-centricity, in the process motivating why I would not utilise an approach recently suggested by IBM. In this post I focus my attention on tertiary / research hospitals, as representative healthcare organisations.

Recently published approach

In a 2016 IBM white paper titled “5 steps to becoming a data-driven healthcare organization”, the author lists these steps to become a data-centric healthcare organisation:

  1. Define all data sources
  2. Set data quality metrics and assess and improve the quality of proposed sources
  3. Integrate data sources
  4. Identify analytics needs
  5. Secure and manage the data lifecycle

While each of these steps has a useful role to fulfill, I would not necessarily follow them, especially not in that sequence, for the following reasons:

  • Defining all the data sources in a research / tertiary hospital is a mammoth task – there are many data sources being added and removed on an on-going basis. I would have a more adaptive and agile data strategy that drives the data-centricity, rather than trying to get an exhaustive list of data sources up-front. Especially in healthcare research, you need to be flexible with regards to data sources – you need to be able to bring new data sources on-board when operational priorities or demands require them.
  • It is good to establish data quality metrics, but I recommend only doing it once their context are clearly understood. The data quality metrics should be managed in a fit-for-purpose data quality framework. This is typically only possible once a more encompassing data strategy has been established, especially in a hospital where data requirements keep changing, and where the importance of data items’ quality vary over time.
  • Integrating “all” or even “some” data sources is quite a tall order – not all health data is always available or in a “ready to integrate” format or at the appropriate level of detail. Integrating all data sources also implies all the necessary data has been copied or moved to a storage platform where it can be integrated. In reality though, some reference data need not be copied, some data may only be required on an ad hoc basis and some research data may be so voluminous, sparse and infrequently accessed that it doesn’t make sense to move it all and maintaining it all on a centralised storage device. I would much rather propose to “provide integrated access to all identified / required data sources”, as and when needed, for example using a modern analytics platform.
  • I would like to see a more holistic data- and information strategy being developed and implemented in parallel as the organisation matures in its data-centricity. The analytical requirements, dove-tailed with the executive and operational reporting needs, should drive this agenda. So for me, business informational and insights needs must dictate the starting point; and it must be adapted dynamically according to operational requirements as the initiative progresses.

Suggested approach

So how then can you mature and improve a healthcare organisation’s data centricity? Taking the above commentary in consideration, and presuming you have the necessary executive buy-in, I would suggest the following approach:

  1. Identify analytical and informational needs
  2. Define an initial data and analytics strategy and roadmap
  3. Implement analytical projects in priority sequence
  4. Build out and live the data and analytics strategy

I obviously don’t have the space to cover these steps in a lot of detail here, but below follow some commentary on these suggested steps.

Identify analytical and informational needs

The first step is to determine what information the business needs to run effectively and improve substantially, together with the analytical insights it needs to potentially disrupt and alter its business strategy.

This step is crucial in establishing a data-centric healthcare organisation – it defines the business case for the entire program. You have to know the organisation’s analytical and informational needs and appetite, in order to turn the potential value of the insights and information into the upside of a business case, where the cost component covers the acquisition, implementation and deployment.

Define an initial data and analytics strategy and roadmap

We cannot specify the contents of even an initial data and analytics roadmap in this post – it’s enough material for an e-book. But the organisation needs to decide what analytics, based on what data, it requires to achieve the requirements identified in the previous step. In other words, it needs to start identifying “how” the “what” of the previous step is going to be achieved. Aspects like a data architecture that describes the data sources relevant to each problem, data flows, again relative to each problem, data storage areas, information exploitation, collaboration and distribution, all need to be analysed, designed and put in place.

This typically takes place in two stages:

  1. A quick forward-thinking high level definition, to ensure that what gets put in place will address the foreseeable requirements, or will at least be extendable and extensible in order to do so.
  2. For each initiative the details are then fleshed out, again with enough forward-thinking at each step to reduce as much later rework as possible. For each initiative, fit-for-purpose data quality measures are then defined as well.

Implement analytical projects in priority sequence

At this stage the analytical projects are implemented. I’m not going to present a whole analytics life cycle in detail here, but at least the following should be done:

  1. Validate the business need, as it may have changed or “disappeared” since it was initially identified. In particular, identify and define the analytical insights required to address the business need.
  2. Identify the appropriate data sources, obtain access to the required data, and provide integrated data access to the necessary data from those data sources.
  3. Determine and validate the appropriate analytical models, using an analytical model development lifecycle such as SEMMA (sample, explore, modify, model, assess).
  4. Secure and manage the data lifecycles of the input data, and very importantly, of the resulting insights.
  5. Operationalise the insights and ensure the appropriate players in the organisation act on the new insights.
  6. On an on-going basis, measure and study the organisational impact of the delivered insights. If the impact of insights are not measured and evaluated, it’s impossible to justify the business case for that model, and potentially for other similar models to follow too.

Step 5, to operationalise the model insights, may require some organisational change management, which may have had to be started earlier in the process.

For most analytical models, their performance in terms of predictability, accuracy, relevance, etc., also needs to be evaluated at frequent intervals. In fast-changing organisations, analytical models need to be recalibrated at least every six months.

Build out and live the data and analytics strategy

As one initiative after the other is implemented, or some even in parallel, given enough demand and available resources of course, the data and analytics roadmap has to be reviewed, updated and continuously evolved to stay abreast of changing requirements, technological changes, new data sources becoming available, and so forth.

A static data and analytics strategy will very quickly become outdated and will become totally useless. It will most likely also result in data and technology falling so far behind that another disruptive change may be required to get up to date again.

Concluding remarks

In today’s fast-changing world, healthcare organisations need an adaptive and more fluid data and analytics strategy, to drive the organisation to higher levels of data-centricity and maturity, while keeping abreast of a fast changing business and data landscape.

Leave a Reply