Musings on Health Datapalooza 2018
Last month, five people from Pareto Intelligence attended Health Datapalooza in Washington, DC. Health Datapalooza is an annual gathering of over 1,000 data and technology experts with a focus on exploring ways to improve health for individuals and communities through data. Our employees enjoyed presentations from industry leaders, policy makers, clinicians, researchers, and more in both the public and private sectors. Below are the key takeaways from each of our team members that attended the conference and their thoughts on how Pareto can learn from and use these takeaways to help shape our future products.
Lee Stuck, Product Manager
Health plans, provider systems, government agencies, vendors, and researchers are using data and analytics to identify opportunities to support the Triple Aim objectives (improving patient experience, improving population health, and reducing per capita medical costs). One of the more interesting sessions we attended presented an analysis of outlier practice patterns using claims data that highlighted the variation in outcomes and costs between providers (https://www.ncbi.nlm.nih.gov/pubmed/28453605, https://www.ncbi.nlm.nih.gov/pubmed/27120446). Clearly, variation in provider practice patterns can lead to worse patient experiences, worse outcomes, and higher medical costs. While exercises like these produce compelling analyses and insights, gaps remain in how these analyses and insights are translated and operationalized into interventions that drive meaningful change for patients and the healthcare ecosystem. In fact, that theme manifested throughout Health Datapalooza as practitioners shared the latest and greatest in terms of analytical technology and methodology but lamented the challenges in deploying the insights. The analyses and insights that are successful in effecting change are deployed at specific points in healthcare processes to maximize impact. For example, analytical insights that help steer patients to more effective sites of care must be deployed earlier in the healthcare journey than when the patient checks into a sub-optimal care setting. Considering these processes and where analytical insights can be deployed should drive the design and execution of data-driven analysis to ensure maximum impact.
Causing a Disruption
David Thompson, Data Science Manager
One of the best things about Health Datapalooza is the career diversity among those in attendance. Each side of the health care industry is represented, and this includes nurses and physicians. Interestingly, many of the presenters in break-out sessions are career physicians that have taken a special interest in the innovative use of data to aid their efforts in improving health care. If we want to use our talents and insights to improve the healthcare space, we must work with nurses and physicians, rather than work to replace them or circumvent their roles in the care delivery process.
It is clear that our efforts should not further complicate the role of the clinician. Physicians are already spending more than half of their working hours entering and correcting data in EHRs. Doctors are ultimately in control of the delivery of patient care, and our efforts as data scientists and analysts are most fruitful when we consider their inevitable role and go beyond simply providing physicians with complex models and their results. In addition to these predictive models and their insights, the data-driven solutions we develop should minimally increase or even reduce the operational burden already placed on physicians as a result of EHR implementation.
Privacy and Ethics
Zain Jafri, Data Analytics Manager
One of the major themes discussed at Health Datapalooza this year was the topic of data security, privacy and ethics in healthcare data analytics—not surprising given current events and recent headlines in the industry. Facebook, Cambridge Analytica, GDPR, and not to mention many of the recent data breaches and attacks, were all examples mentioned at the conference.
Having grown up in the healthcare industry, data security and privacy are near and dear topics to Pareto as a company. We live it and breathe it every day; however, the discussion on data ethics was also a very interesting one. In this discussion, the question of “Can I build a model to predict whether an individual will develop a chronic condition?” morphed into “Should I build a model to predict whether an individual will develop a chronic condition?” Participants discussed what the appropriate uses of data and analytics are in healthcare, how the power of analytics should be balanced with the appropriate safeguards to mitigate against misuse, and more. As the healthcare industry catches up in the world of analytics, data becomes more accessible, and the models become more sophisticated, it is certainly a question that our industry will have to grapple with.
One Small Model, One Large Impact
Chris Ventura, Solutions Architect
While healthcare often lags behind other industries in its utilization of technology, Health Datapalooza covered a range of advanced technologies from blockchain to aritificial intelligence. Presenters explained the opportunities for converting medical histories to blockchains and using machine learning for predicting inpatient readmissions. However, what stood out more were the basic analyses that had the potential to greatly impact the patient and provider experience. Sreekanth Vemulapalli of Duke University showcased Natural Language Processing techniques to autofill an Electronic Health Record while a physician types notes, reducing the doctor’s time spent filling out forms while standardizing and tagging key attributes for easier future analysis. Other sessions emphasized the importance of benchmarking provider surgery outcomes and other things to allow the patient and provider to make more informed decisions. Blockchain and AI will be part of the future, but it’s important to first build a foundation of robust statistics. If implemented successfully, even simple models and tools can improve efficiency and outcomes.
Research and Development
Michelle Solomon, Data Scientist
The Health Datapalooza conference was not only about completed analyses, but it also presented a wealth of open source datasets. CMS announced a big release of Medicare Advantage encounter data. We also heard about Careset’s new graph dataset connecting the web of doctor referrals. The National Investment Center introduced us to their seniors housing dataset, which focuses on optimizing housing for middle-income seniors. We were excited about the potential insights to be found.
However, the CMS dataset is for research use only and we would not be able to use the data to improve or create commercial business models. Even so, we can still use this opportunity to hone our skills, learn new insights, and contribute to the greater healthcare data community. Other data analytic groups work together on Kaggle competitions as a team building exercise and to practice new techniques. We could apply a similar tactic to the Medicare Advantage data while educating ourselves further about healthcare issues. Additionally, open source datasets and software packages are becoming more common in the big data industry. As a growing healthcare consulting company, it’s important for us to keep up with the modern climate and help shape the health data future.