KDnuggets Home » News » 2015 » May » Opinions, Interviews, Reports » Interview: Mark Weiner, Temple University Health System on Addressing Healthcare Data Gaps through Advanced Simulation ( 15:n15 )

Interview: Mark Weiner, Temple University Health System on Addressing Healthcare Data Gaps through Advanced Simulation


We discuss dealing with current gaps in healthcare data, challenges in using real world healthcare data, desired skills for data scientists in healthcare industry, advice, and more.



mark-weinerProf. Mark Weiner is currently the Assistant Dean for Informatics, and Professor of Clinical Sciences at the Temple University School of Medicine, charged with developing a curriculum to support the needs of the next generation of Clinical and Research Informatics leaders. He is also the Chief Medical Information Officer for the Temple University Health System, overseeing the clinical aspects of the inpatient Epic implementation, ensuring it achieves its goals of supporting the clinical and academic missions of the institution.

Throughout his career, he has conducted research and applied techniques that help bridge the gap between Health Services Research, clinical and research operations and Medical Informatics.

First part of interview

Here is second part of my interview with him:

Anmol Rajpurohit: Q6. While silos are being broken down to integrate healthcare data, it doesn't seems to be happening any time soon. Meanwhile, how can we make the most of non-integrated data from sources such as EHRs, claims, etc.?

healthcare-non-integrated-dataMark Weiner:

One of the key problems with conducting retrospective studies using any data source, regardless of how well-integrated it is, is the uncertainty with which the observational data accrual associated with clinical care can simulate the random treatment assignment and regular follow up of rigorous clinical trials.

I’d like to see efforts to simulate well-accepted clinical trials using observational data, at first focusing on patients very similar to those in the real clinical trial. If minimal differences exist between the observational data analysis and the real clinical trial, then other observational analyses, anchored on the original clinical trial, but perhaps expanded to wider groups of people are more likely to be trusted. If differences exits, we can study the causes and impact of the differences to better calibrate findings from the observational analyses.

AR: Q7. What are the most underrated challenges in using real world healthcare data?

real-world-health-dataMW: “Big data” enthusiasts often seem to view the acquisition and compilation of massive amounts of data as the major hurdle to overcome in understanding the relationship between disease outcomes and interventions. While that is an important step, the analysis still needs to be guided, or at least informed by people with clinical domain expertise. A big data analysis can find associations different between a positive and negative test, though with domain expertise, the analysis can include the clinical significance ordering the test in the first place, regardless of its result.

Even when large, longitudinal data sets are available, far too often, analyses are based on independent variables being set as the aggregate accumulation of clinical characteristics known at a point in time. The different timing with which these characteristics appeared, and the full spectrum of clinical findings over time is often aggregated into the binary presence or absence of disease, or the highest, lowest or average laboratory parameter. The pace and order of the availability of findings can often be as important as the results.

AR: Q8. What are the common myths in healthcare analytics?

MW: A great deal of resources are expended to create and purchase analytical tools that are easy enough for anyone to use. While these tools have a role in helping high level administrators become better consumers of data, the sophisticated analyses will still require engagement of people with the analytical, research and clinical domain expertise and the time to devote to the analysis.

Furthermore, often, the features that make these tools easy to use, make them unsuited toward the more sophisticated analyses that are required.

Good visualizations of data can help unmask trends and patterns that can be difficult even for sophisticated analyses to detect, but the ability to create these visualizations requires time, skill and patience too.

AR: Q9 .Which of the current trends in Healthcare Analytics are of great interest to you?

MW: While I have some reservations about “big data” in general, I am intrigued by the possibility of merging high volume and velocity data collected through passive modalities to better understand potential contributors to or to make predictions about outcomes. For years, patients have worn diagnostic Holter monitors to evaluate heart rhythm abnormalities, though patients needed to manually input what they were doing at various points over the day, and if they were feeling any symptoms.
healthcare-trends With new sensors, we can automatically capture how active the patient was at the same time as we are capturing the heart rhythm, along with external factors like temperature, and external exposures. The ability to align internal and external parameters, along with an understanding of how active a patient is over the course of a day fills in a lot of gaps in our understanding of how patients are responding to treatments.

AR: Q10. What is the best advice you have got in your career?

MW: Taking risks to alter the direction of your career, within reason, can be very rewarding, both personally and professionally

AR: Q11. What skills do you think are the most important for practitioners in the field of Healthcare Analytics?

team-workMW: People who conduct data analytics often approach it mainly from the perspective of their initial training. Computer scientists, mathematicians, statisticians, clinicians, and computational biologists all help to advance data science. It is important to know the priorities, methods, assumptions and goals of the other players in the analytical space, and know how to engage them in analytics projects.
raspberry-pi
AR: Q12. What do you like to do when you are not working?

MW: In addition to supporting the creative arts pursuits of my wife and daughters. My current hobby is amateur robotics with the Raspberry Pi

Related: