Top 10 Considerations for an Optimal Data Science Strategy

2 November, 2016

Top 10 Considerations for an Optimal Data Science Strategy

By Tony Ursitti, MS, Manager, GE Healthcare Camden Group

Data science is the field of techniques, tools and frameworks used to study and make meaningful conclusions from data. Data is being collected at an accelerated rate and the techniques, tools, and frameworks available to data scientists have evolved significantly over the last several years. The growth of this field has created tremendous opportunity as well as challenges for leaders in the healthcare industry. As this field evolves, healthcare leaders will need to become more knowledgeable about what investments are required in order to make the best use of their data.

There are ten key points for healthcare executives to consider when forming a data science strategy:

1. Realize that data is an asset, but its value is directly proportional to how it is used

Data is an asset to organizations in a similar way that buildings and medical devices are assets. An increasing share of the present and future value of health systems will be derived from the type of data that is collected and how effectively it is used to meet clinical, financial, operational, and strategic goals.

2. Understand the types of available data assets and prioritize the ones to acquire

As the pressure to cut costs and demonstrate high quality care continues to mount, understanding the types of data assets available and also the ones necessary to acquire is increasingly important. Consider how data assets are positioning the organization for future success…or future difficulty. For example, if a data element needed for calculating a quality metric that will be publicly reported in two years is not yet being captured, determine how quickly it can be captured. Two years from now, when that data becomes publicly reported, there will be no ability to go back and create a historical data record to use, and the organization could be understating the quality of care they provide simply because they don’t have the historical data to prove how well they actually performed.

3. Position your organization to compete on analytics

In the coming years, some organizations will thrive and others will struggle. Organizations that make smart use of data and analytics will have a strong competitive advantage over those that do not. Increasing pressure to cut costs and improve quality means increasing pressure to understand as quickly as possible the factors or organizational behavior that are contributing to positive as well as negative results.

4. Understand the importance of data science in getting the most from data assets

When planning to acquire the appropriate data assets, anticipate how the organization will make the best use of these assets once they have been acquired. Data science helps organizations start to understand some of the important relationships between practice patterns or other organizational behavior and undesired or desired outcomes. For example, it is important for organizations to know that their average length of stay at a particular hospital is increasing, but it is at least equally important to know why this increase is occurring and what, if anything, can be done about it. Data science helps uncover the why and enables organizations to make more informed strategic decisions.
One general principle for executives to keep in mind when thinking about data science is that good decisions require good knowledge, that good knowledge requires asking good questions of data, and that it is impossible to know what the “good questions” are unless you understand the scope of questions that are answerable. In other words, 30 years ago it would have been fruitless for executives to be asking questions like “what variables contribute to readmission risk and to what extent” because methods for answering that question did not exist in the way that they do today. Today, because of advances made in data science, that type of question is not only appropriate to ask but increasingly important to answer. Because data science has advanced so rapidly, the risk is no longer asking questions that cannot be adequately answered, but not understanding the breadth of questions that can be answered and, consequently, leaving them unasked.

5. Keep a focus on the critical problems

While starting or continuing on the journey of using the increasingly advanced techniques, tools, and frameworks that data science makes available, it is important to keep a focus on the problems that are most important for the organization. Data science is something like a Pandora’s box; insights beget insights and those insights beget more. Without the proper focus from leadership, data scientists will find endless interesting insights that have little or no strategic value for the organization. Good data scientists should be able to uncover insights buried in data that would otherwise be left undiscovered, but for this discovery to be relevant requires an appropriate level of understanding between leadership and data scientists about the most important strategic pain points the organization must address over the coming years.

6. There is not a “one size fits all” approach to data science

Each organization is at a different stage in the journey to deliver lower cost and high quality care while competing in local markets with different dynamics. Data insights that are of significant strategic value for one organization may be of little or no strategic value for another. For example, large multispecialty organizations can benefit from a robust site of service strategy that not only moves volume out of inpatient settings where appropriate, but also moves elective procedures from high cost inpatient facilities to lower cost inpatient facilities. For physician groups, the data and analysis requirements to build a site of service strategy would need to be more granular to be of any value.

7. Get the technical expertise required to succeed

Organizations will need access to technical expertise to help sort out the signal from the noise. Concepts like model accuracy, overfitting, and statistical significance will be important to understand in order to make sure that the inferences being drawn from modeling efforts are well founded.

8. Data science is not a “check the box” exercise

There are increasingly advanced techniques for making sense of data, and the organizational decision-making process must continue to evolve in order to keep up with and derive full benefit from these techniques. This means that the way organizations solved problems and developed strategies 5 to 10 years ago is no longer the best way to solve problems and develop strategy. As data science continues to advance, organizations must continue to evolve and make smart use of it.

9. Data literacy will be increasingly important at all levels of management

There is an increased need for data literacy within members of executive teams and across many operational areas within health systems. Good data scientists will be able to translate most of their technical knowledge into actionable insights for non-technical leaders. However, executives and other stakeholders will need to know enough to ask relevant questions about the data science with which they are presented to ensure nothing important is lost in translation. Iterative conversation between leadership and data scientists is incredibly powerful.

10. Competition for data scientists will increase

There is and will continue to be a shortage of data scientists in the market. In 2014, Accenture found that more than 90 percent of its clients planned to hire employees with data science expertise, but more than 40 percent cited a lack of talent as their number one problem. This problem is even greater in healthcare, an industry that is uniquely complex in which knowledge is highly specialized and takes years to develop. This means it is critical to start focusing now on building internal or external teams to support your organizational decision-making through data science.

Ursitti_Tony.pngMr. Ursitti is a manager with GE Healthcare Camden Group in the Digital Health and Advanced Analytics practice. He has more than seven years of analytics and leadership experience in both the consulting and provider settings. He focuses on helping health executives make data-driven strategic decisions through value model development using statistics, predictive modeling, advanced data mining, and machine learning techniques. He may be reached at