10 steps toward predictive analytics

10 steps toward predictive analytics
By Paddy Padmanabhan
11:47 AM
Share

With limited budgets and ever-increasing demands for efficiency and service, how can healthcare payers and providers ensure that their analytics delivers against investments?

Think data, think ahead
Predictive analytics is a powerful tool for harnessing clinical data. Used effectively, it helps cut costs, increase efficiency and maximize returns through accurate assessment and management of population health risks, among other functions. However, investing in expensive platforms without budgeting for the training of analysts, or the hire of data scientists and skilled modelers to make best use of the gathered data, can lead to significant spend against slow-burning returns.

Strategies for success
Out of the box solutions offer low set-up costs and limited downtime, and cloud-based systems reduce hardware spend. However, failure to plan for intelligent data preparation and management, or to structure a clear strategy to bring the system to life operationally, can see sustained benefits and predicted ROI go up in smoke.


Broad data mining for improved outcomes
Companies should implement an analytics framework to enable them to absorb data from diverse sources—in-house and syndicated data, social media and other unstructured sources—to address the most complex analytics problems and develop focused healthcare-specific solutions.

Complex data made simple
Data complexity shouldn’t feel complicated to an enterprise. In my experience, a three-stage model can offer a simple structure that encompasses pilot testing, development and full implementation. Taking this route maximizes investments in predictive analytics platforms. Beware also of the hype surrounding “big data” today.

Phase 1: Pilot
In this pilot phase, contract data scientists, modelers and analysts discover how to put your company’s data to best use. They should be able to demonstrate system viability through static data and interactive visuals.  

The diversity of data absorbed and process to manage it allows you to address the complex analytics problems inherent to it, and to develop focused, healthcare-specific solutions that can eliminate manual systems, minimize human error, transform efficiency and ensure operational success.

The best models for this are cloud-based, as they are non-intrusive, low cost and flexible. This makes them particularly effective tools during a pilot phase.

Phase 2: Implement, test, validate
Once a pilot solution is in place, it has to be propagated with data feeds that enable a level of operationalization of the analysis for sustained benefits. The pilot phase also typically identifies ”hotspots” that require further analysis.

In one of our client engagements, we discovered there was a high readmission rate coming out of discharges to skilled nursing facilities (SNF) that became an item of interest for additional analysis and cost reduction.

In another case, we found that there was a very high incidence of hospital acquired conditions (HAC’s) that was leading to non-reimbursable costs. These are typically identified through rigorous statistical analysis that most health systems do not have in-house staff to look into.

Any implementation should also include periodic and regular dashboards that provide updates and insights into the effectiveness of interventions, as well as information on trends, based on predictive models. This also provides opportunities to adjust the predictive models and algorithms to reflect changes in the data and trends, and improve the accuracy rates of the models for optimal use of resources.

Any pilot project should be designed with repeatability and reusability as a key objective so that newer use cases can then be bolted onto the analytics infrastructure during this phase.

Phase 3: Operationalize
During this stage, a robust and sustainable data integration and analytics solution infrastructure is implemented. This can be a cloud-based or onsite model, equipped for scalable, real-time data aggregation, clinical workflow integration and focused intervention strategies.

Mobile enablement for care management is also a powerful tool that enables update flexibility and constant anytime, anywhere access to data.

The analytics structure can be scaled up or down during this phase, and optimized to become an accessible asset to the whole enterprise, available to different functions and departments.

Increased access and reduced cost for providers
Once in place, predictive analytics gives providers the data they need to action on the dual challenges of reduced reimbursements and increased accountability for outcomes. Using these insights, providers are able to reduce costs, increase revenue and predict and solve potential problems or issues before they impact operations.

Providers looking to set up ACOs will require a robust analytics platform to manage population health and maximize incentives. A cloud-based, pay-as-you-go model aligns well under these circumstances, since it requires no upfront investments in hardware, software or expensive human capital.

Greater insight into individual risks for payers
The payer sector is rapidly transforming from a group insurance-based model to an individual consumer based model. By assessing individual health risks using a combination of internal and external data sources, payers can improve revenue, control costs and arrive at risk scores for Health Plan members.

With the enrollment of individual members in the exchanges under Obamacare, payers have an even higher incentive to understand the risk profiles of these individuals using a wide range of external data sources and predictive models to help segment and manage population health.

Take care of healthcare (data)
As data sources continue to explode, the importance of a strong data integration and data management capability, whether on-site or cloud-based, is vital to the sustainable success of analytics programs.