Future Predictive Analytics Solutions Need the Right Level of Regulatory Governance

By Frank Wang, System Vice President, IDN Decision Support Analytics, Health First

Frank Wang, System Vice President, IDN Decision Support Analytics, Health First

In the past more than a decade or so, big data and predictive analytics has enjoyed the encouraging development and growth environment. The impact of the big data and predictive analytics to our daily lives is tremendous. Big data and predictive analytics is part of a business operation process and crucial to the outcome from targeted marketing to clinical interventions and recommending treatment pathways. The dynamic, independent, developer centric and open source style for developing the logic/algorithm for implementing the big data and predictive analytics solutions generates some scale of variations and confusions regarding to the predictive model build and the predicted outcomes. It is anticipated that the future predictive analytics solutions need the right level of regulatory governance in the following seven major areas in order to continuing grow smoothly.

"As a provider of analytics solutions or a data science team in an organization, it is important to take proactive approach to adopt the best practices in the development life cycle"

1. Share the purpose and motivation for using a specific algorithm to help building the trust and understanding of the analytical insights when adopting an analytics solution. To reduce the “black box” barrier on the end user side will help to utilize the business users’ expertise in the process improvement and improve the buy-in status from the end users. The impacted individuals need to be aware of why and how an analytics solution is used in the decision-making process.

2. Standardize the data partition in the whole development life cycle of model development, validation and testing. An industry standard for testing and valid an analytics performance needs to be adopted by data scientists. The approach for choosing the data for building a model validate a model and test a model greatly affect the precision for the prediction.

3. Make it transparent of what data is used in the analytics solution and how the data is collected and used. To explain to the consumers of an analytics solution how a data set is used will help to build trust and reach the outcome as expected. This is crucial for building the coordination between the impacted entities.

4. Authorized agencies need to help and assess the fundamental crucial features and core performance measures of an analytics system. For example, FDA need to work with the industries to provide the clinical decision support analytics guidelines.

5. Standardize the analytics system trial and validation process. The feedback loop for analytics solutions is relatively long, especially for healthcare predictive analytics solutions. It will take very long time to validate the predicted clinical outcomes in healthcare industry. A standard trial protocol needs to be agreed and followed for developing predictive analytics solutions.

6. Publish and share the performance data for major commercial analytics solutions. The authorized agencies need to maintain and share the standard performance measures for the analytics solutions to help the uses to assess the predictive modeling methodologies and pick the proper solutions.

7. Set up a center for testing the commercial analytics solution, guided by the authorized agencies. To share and public the standard performance by the center will help the users to select the right solutions as addition to their analytics tool kit.

As a provider of analytics solutions or a data science team in an organization, it is important to take proactive approach to adopt the best practices in the development life cycle to be prepared in the seven areas discussed. We will have a more detail discussion in a separate article.

Check this out: Top Sleep Disorder Care Solution Companies

big

marketing

open source