VentureBeat presents: AI Unleashed - An exclusive executive event for enterprise data leaders. Network and learn with industry peers. Learn More

I imagine every data scientist has experienced this at least once during her career: You start with an interesting question, a hunch, a hypothesis. You tame the data: engineer interesting features, identify anomalies and biases inherent in the data, make some scoping decisions, normalize it, and impute the missing values in some clever way.

After experimenting with many model forms, finally, you get to that “eureka moment” at two in the morning. There it is! You knew it all along. Your hypothesis was right and now you have the data to prove it. Or data may have surprised you and led you to a discovery that is unexpected and delightful. Now you put your “story telling” hat on and present your findings to the executives.

It is surprising how often the story ends there. Executives are impressed, the data scientist is patted on the back for the creativity, but the model is left in that PowerPoint deck to collect dust.

At companies where there is no framework for operationalization of the models, PowerPoint is where models go to die!

In the healthcare system, there is a massive amount of invaluable data that is waiting to be unearthed. We now have the computing power and the algorithms to sift through patient data and biomedical literature and surface what matters and recommend a course of action. According to PubMed, in the last 5 years alone, there have been 3,573 studies on hospital readmission, 9,745 papers on comparative effectiveness, 39,230 studies on drug-drug interaction, and 132,241 studies on hospital morbidity. But only a handful of those models are in production.

Contributing to the biomedical literature and thus expanding our understanding of diseases, risk factors, and effectiveness of treatments is very admirable. This is how science progresses: making incremental discoveries while standing on the shoulders of those who came before you. But it is unreasonable to expect physicians to consume this expansive body of work apply the learnings to their everyday practice. When we visit our doctors we should all expect to benefit from the latest discoveries without requiring our physician to be a superhero.

Editor’s note: Our upcoming DataBeat/Data Science Summit, Dec. 4-Dec. 5 in Redwood City, will focus on the most compelling opportunities for businesses in the area of big data analytics and data science. There are just a few seats left, so be sure to register today!

I hope at this point you all agree with me that dissemination of the insights to the decision makers is as important as extracting the insight from the data. This is why an application developer is a data scientist’s best friend. Regardless of how insightful the model is, if it is not put in the hands of the people who make everyday decisions, then it is of little impact.

Where does a company start?

The three basic steps involved in creating a more data driven organization are:

  1. Invest in a distributed computing platform to take advantage of all the data available: structured and unstructured.
  2. Recruit data science talent and train existing team members.
  3. Define a path for operationalization.

Even with the right technology, people, and processes in place, it may still be hard to decide which opportunities to go after first. That’s where analytics road mapping comes into the picture; with reports showing health app adoption is stagnant, it really becomes crucial to go after opportunities that will have the highest impact. To build the most impactful tools, data scientists and application developers need to understand who, when, what, and how aspects of decision-making. Let me elaborate on these criteria by considering a healthcare provider.

Who: Identify all possible target user groups for the analytics/application.

For a healthcare provider, there are three groups of people to target with applications: hospital admins, care providers (physicians, nurses, PAs) and patients.

What: Identify what decisions these user groups make.

Hospital admins make very crucial resource allocation decisions like staffing, bed management, care coordination, revenue assurance, and identifying gaps in care for potential preventive health initiatives.

Physicians on the other hand make the most important decisions at the point of care: making correct diagnoses and providing the best treatment for the patient.

Patients play an increasingly important role in the delivery of healthcare as they participate more in the decision making for treatment and control of their chronic conditions. They make many everyday decisions that have an impact of their well-being: whether or not to adherence to a drug, diet, and activity regimen or conduct high-risk behavior.

When : Understand when these decisions are made and insert analytics into existing workflows without additional effort.

Ideally, analytics should inform the decision maker at the moment the decision is being made. For instance, if a hospital is concerned about over-utilization of lesser diagnostic tests, then a point-of-care solution that offers more appropriate tests at the time they are ordered is much more effective than a retrospective physician benchmarking exercise.

Hospital admins already use tools like scheduling applications to coordinate elective surgeries. Analytics could be run in the background in order to understand the number of new cases expected by care units, predict the length of stay for existing patients, and quantify the available capacity.

Modifying patient behavior at the time those everyday decisions are being made is the hardest challenge. For instance, an ideal nutrition guidance tool would modify the patient’s behavior at the time he is deciding on where to have lunch.

How: Understand user groups’ attitude towards predictive analytics and come up with a presentation that is most appealing to them.

Understanding user expectations, experiences, and attitudes toward predictive analytics is important.

Hospital admins in general have a very positive attitude towards predictive analytics. The types of decisions they make like identifying gaps in care require sifting through large amounts of data and they appreciate what big data analytics can accomplish for them.

Physicians are more likely to adopt a tool that supports their decision making than a prescriptive tool that tries to do their job for them.

Patients constitute the most heterogeneous group with quantified-self on one end and people with chronic diseases who rather not think about their health on the other. According to a study, 26 percent of the applications downloaded in 2010 were used just once. Adoption is key to success and is extremely hard to achieve.

Many of the health applications that target patients aim to collect better/granular data in between visits to be leveraged later by the physician or to alarm the patient when they need to reach out to a physician. These applications have a better chance of being adopted if the data collection is effortless.

For example, researchers at University of Toronto studied a selection of Agatha Christie’s novels written while she was between the ages of 28 and 82. They found that the vocabulary size decreased by 15 to 30 percent as she neared the end of her life, while repetition of phrases and indefinite word usage in her novels increased significantly, suggesting that towards the end of her life she may have suffered from Alzheimer’s.

Though anectodal, this example is inspirational. Similarly, an analytics powered application can follow our daily online activities: our browsing habits, the language we use in our emails, types of news we are attracted to and infer our mood, any deterioration in our cognitive skills and allow early diagnosis of mental diseases. We already see similar approaches being used in preventing suicide among teenagers.

In conclusion, an effective partnership of big data analytics and rapid application development will disrupt many industries including healthcare. Companies must get the data scientists and application developers sitting on the same floor in order to give itself a competitive advantage in today’s market.

Hulya Farinas is a senior principal data scientist at Pivotal where she is the lead for health care vertical. She holds a Ph.D. in Operations Research from the University of Florida.

VentureBeat's mission is to be a digital town square for technical decision-makers to gain knowledge about transformative enterprise technology and transact. Discover our Briefings.