by Scott Chesney –
Anyone who’s worked in Health Plans IT understands two things:
- Using data-driven analytics to improve member outcomes by increasing patient engagement and leveraging best-practice treatments is critical in the new world of value-based care.
- Few Health Plans have actually implemented cutting edge analytics or data visualizations. “Big Data” is in all the white papers, but very few of the data centers.
How should a health plan actually go about changing their business to meet the demands of the consumerization of care and the need to improve outcomes and cut costs? Traditionally, providing new data and tools to the business involves a corporate initiative, numerous projects, scores of consultants, and millions of dollars. It doesn’t have to be that way.
Business owners know what they need.
Change doesn’t require a huge data warehouse, or even a data lake. Often key health plan executives know what they’d like to see – and the business changes they’d like to implement it if they had data to support them. The key is that these executives may not have spent time thinking about what they could have or how they could get it, so their requirements remain unarticulated. They may even already have executive dashboards, but how many actually have information in front of them that is timely, actionable and presented in a way that encourages use?
Start by interviewing these stakeholders. As interviewers, include resources both who understand the business and who have expertise in data and visualization. Reassure the business that this is not another IT project that will take their time and budget and produce more disappointment than operational improvement. Ask questions like:
- What are the key challenges and opportunities facing your part of the business today?
- What metrics could you use to assess those challenges and opportunities?
- What measures do you need to evaluate the business initiatives you are considering?
A whiteboard and/or prepared mockups of possible analytics and visualizations is helpful at this point. Some business leaders may already have strong opinions on how they want to see their data and exactly what should be included in a dashboard, but many, while they may know what they are trying to learn from the data, do not have the experience with analytics or data presentation to know the best way to represent the results. Our job is to present best practices and help the business choose both the right metrics and the right visualization.
The business has more data than you realize.
Remember that the business is already running and business leaders are making strategic and tactical decisions every day. What data and what reports are they using currently? It may be just spreadsheets assembled manually by staff or reports from vendors, but these can serve as data sources to prove the value of the data and analytics. After interviewing the business leadership and identifying their data needs and data sources, follow up with the staff members who are assembling the spreadsheets or managing vendors’ data flow. We consistently find troves of useful information that are hidden in business units just because no one realized that it could be valuable in other areas.
The challenge, of course, with this approach is that there may be no universally agreed definition for many of the data sets. Even a simple term like ‘Member’ means something different to the Actuarial department than it does for Claims let alone Medical Management. Validate the data sets with the business before you spend time and effort developing visualizations and analytics with it. Simple queries and spot checks should be enough to gain confidence in the utility of the source data, without requiring a full design, build, test cycle. As a last resort, preserve the different data sets for each business unit even if that means having three different versions of the Member table. The goal, at this point, is not to solve the problem of Master Data Management or Data Governance. The goal is to show the business that IT can deliver value quickly, which gives IT the credibility to tackle the thornier issues of cross business unit data governance.
Collect the data.
At this point, you’ve identified a limited set of metrics to present and a reasonably small number of data sources that provide the data to support those metrics. Traditionally, teams would use database or ETL tools to load that data into a database which becomes the source for the reporting and analytics tools. There’s nothing wrong with this approach, if it’s endorsed by the business and supported by IT skills and licensed tools. But a key point is that you’ve already done most of the work now to enable a big-data approach to data ingestion.
With a good Hadoop based big-data ingestion framework that handles file operations, logging, monitoring, scheduling and auditing, you can take these data sources and quickly have them tagged and loaded into a mini data lake (data pond?). Just because the data volume is not huge does not mean that big data is inappropriate. The speed of delivery and savings on license costs make it a good value even when you’re dealing with only hundreds of thousands or millions of records. If the organization is not yet ready for a full ‘big-data’ implementation, this approach allows you to manage just these data sets as a proof of concept. Worst case, if no big data technology can be used, by identifying the data sets, use cases, and metrics, you’ve still laid the groundwork for a future implementation or POC.
Show the value.
Remember that data visualization and analytics are separate skill sets from understanding and manipulating the data sources or from knowing the business. Ensure the team contains experts in visualizations, as well as in the tools used. Just as important as the data is presentation in a format that means something to the end user. What data scenarios require immediate action? Ensure those are clearly visible and the action required is evident. Distinguish ‘interesting’ data from ‘actionable’ data. The immediate value in a short-term effort is with actionable data. The visualizations you provide should facilitate further research and analysis on interesting items, but your priority should be on the scenarios that drive executive business decisions.
Today’s visualization and data exploration platforms like Tableau1, Qlik2, MicroStrategy3 and open source tools and languages like R4, RapidMiner5, and KNIME6, along with the ecosystem around Hadoop, especially Cloudera7 and Hortonworks8, allow you to connect directly to your Hadoop data as well as any traditional data sources you’ve identified. A well thought out architecture and roadmap should allow you to quickly augment existing systems with modern visualizations that leverage big-data. With a team combining business, data and visualization skills working together, you can build dashboards in a matter of a couple weeks that give the business easier access to their data, more intuition about the state and direction of the business, and the ability to implement changes based on facts and evidence instead of speculation.
Works Cited
1. Tableau https://www.tableau.com/
2. Qlik https://www.qlik.com/us/
3. MicroStrategy https://www.microstrategy.com/en
4. The R Project https://www.r-project.org/
5. RapidMinder https://rapidminer.com/
6. KNIME https://www.knime.com/
7. Cloudera https://www.cloudera.com/
8. Hortonworks Data Platform https://www.cloudera.com/products/hdp.html