Logo

The Data Daily

The Art Of Data Science

Last updated: 03-15-2019

Read original article here

The Art Of Data Science

CareCentrix, a health care company that manages services, therapies and resources to help patients heal and age at home.">CTO at CareCentrix, a health care company that manages services, therapies and resources to help patients heal and age at home.

While the proposed Aetna-Humana and the Anthem-Cigna mergers did not receive regulatory approval in 2017, the challenges in the U.S. health care system continue to drive merger and acquisition activity to disrupt and transform the health care ecosystem. Three recent planned M&As of note include the $69 billion purchase of Aetna by CVS, Humana's (along with two private equity firms) $4.1 billion acquisition of Kindred Healthcare and Roche's decision to buy Flatiron Health for $1.9 billion. In each of these transactions, "data" and "analytics" were cited as justifications for the deals.

To tame health care spend, many large health care companies believe that data can help integrate the siloed role each company plays in the delivery of care and unlock insights about how best to rein in health care spend. This integration will require critical thinking. The Advisory Board Company went as far as to claim that "return on data is rapidly becoming a favored metric of success" and that these deals integrate "financial, clinical and consumer data" in addition to people, process and other assets.

Spurred on by the availability of increased computing power in smaller form factors, we now have computer systems that can leverage sophisticated machine learning algorithms to make sense of the large volume and variety of data available to them and exhibit intelligent behavior. In many ways, the aggregation of data and its exploitation by learning algorithms will likely be a big factor in the transformation of health care.

However, before we get carried away, let's make sure that we are paying attention to the dialectic process of extracting meaningful insight from data and communicating this insight in a way that humans who do not understand the advanced machine learning algorithms can synthesize and apply appropriately.

One of my former colleagues, who developed a neural network model to evaluate risks for patients with pneumonia, when asked by clinicians if they should use the model, flatly replied, "No -- we don’t understand what it does inside." Commenting on Flatiron Health, one analyst wrote, "Flatiron doesn’t just suck in a bunch of numbers and spit them out -- its datasets are curated in an effort to separate statistical noise from actionable information."

Any practitioner of data science or the forerunner fields of statistics, data mining and knowledge discovery will swear that insights are not found by feeding data into a computer and then magically harvesting the insight. This is the same point a former colleague and I made in a paper we published in 1994.

Yet when I reviewed the data science or business analytics curriculum from many universities, the focus was squarely on the math and the algorithms and not on the critical thinking required to make the decisions that are necessary to ensure that these algorithms produce credible insight. To quote Michelle Dimon, a research scientist at Google Accelerated Science, "Computers are both unintelligent and lazy." Human interaction and critical thinking are required to produce insight that benefits humankind and adds to the human knowledge base.

At CareCentrix, we have charted a different approach, which we believe is well-suited for health care. Our mantra is "Math is subservient to human critical thinking." Our extensive work in predictive analytics is guided by the following principles that can be applied to your own company's next data science project:

While the proposed Aetna-Humana and the Anthem-Cigna mergers did not receive regulatory approval in 2017, the challenges in the U.S. health care system continue to drive merger and acquisition activity to disrupt and transform the health care ecosystem. Three recent planned M&As of note include the $69 billion purchase of Aetna by CVS, Humana's (along with two private equity firms) $4.1 billion acquisition of Kindred Healthcare and Roche's decision to buy Flatiron Health for $1.9 billion. In each of these transactions, "data" and "analytics" were cited as justifications for the deals.

To tame health care spend, many large health care companies believe that data can help integrate the siloed role each company plays in the delivery of care and unlock insights about how best to rein in health care spend. This integration will require critical thinking. The Advisory Board Company went as far as to claim that "return on data is rapidly becoming a favored metric of success" and that these deals integrate "financial, clinical and consumer data" in addition to people, process and other assets.

Spurred on by the availability of increased computing power in smaller form factors, we now have computer systems that can leverage sophisticated machine learning algorithms to make sense of the large volume and variety of data available to them and exhibit intelligent behavior. In many ways, the aggregation of data and its exploitation by learning algorithms will likely be a big factor in the transformation of health care.

However, before we get carried away, let's make sure that we are paying attention to the dialectic process of extracting meaningful insight from data and communicating this insight in a way that humans who do not understand the advanced machine learning algorithms can synthesize and apply appropriately.

One of my former colleagues, who developed a neural network model to evaluate risks for patients with pneumonia, when asked by clinicians if they should use the model, flatly replied, "No -- we don’t understand what it does inside." Commenting on Flatiron Health, one analyst wrote, "Flatiron doesn’t just suck in a bunch of numbers and spit them out -- its datasets are curated in an effort to separate statistical noise from actionable information."

Any practitioner of data science or the forerunner fields of statistics, data mining and knowledge discovery will swear that insights are not found by feeding data into a computer and then magically harvesting the insight. This is the same point a former colleague and I made in a paper we published in 1994.

Yet when I reviewed the data science or business analytics curriculum from many universities, the focus was squarely on the math and the algorithms and not on the critical thinking required to make the decisions that are necessary to ensure that these algorithms produce credible insight. To quote Michelle Dimon, a research scientist at Google Accelerated Science, "Computers are both unintelligent and lazy." Human interaction and critical thinking are required to produce insight that benefits humankind and adds to the human knowledge base.

At CareCentrix, we have charted a different approach, which we believe is well-suited for health care. Our mantra is "Math is subservient to human critical thinking." Our extensive work in predictive analytics is guided by the following principles that can be applied to your own company's next data science project:


Read the rest of this article here