Written by Lauren Beard, Triage Consulting Group

On Tuesday, April 30th, 2019, the Philadelphia HFMA chapter held its 23rd Annual Data Analytics Seminar. The day began with opening remarks from our incoming president, Anne DelPizzo, who has been a friend and mentor to the majority of this chapter for many years… her active involvement in multiple chapter facets has left others wondering how she finds time to work for her day job, give back to other organizations outside of HFMA, and also, sleep.

Kicking off with a bang, Chad Johnson (Associate VP of Strategic Decision Support at Penn Medicine), led an insightful presentation on using data to make strategic decisions. Chad begs the question “So What?” to be front of mind for analysts delivering a message to executive staff so that critical information and issue impact is relayed on the front end through an “Answer First” approach. Chad reminded his audience that genuine curiosity, understanding, and trust is built by the top level dissecting the “answer” from end to beginning… which is where the data analysis occurs. Chad brilliantly exclaimed that “a problem well-framed is half-solved” and stressed the importance of validating the need for data to even be analyzed, by leading attendees through a few straightforward-seeming questions: “What are we determining?” and “What will convince the decision makers to act?” Keeping these inquiries central to the project requiring data gathering/analysis will inform every follow-up question and next step for action, saving time talking in circles and researching data unnecessarily. Another reminder from Chad was that trust is easy to lose and difficult to gain, which drives the ability to make an impact on our organizations; this, added with the fact that everyone plays a different role and requires a different level of detail, inspires thoughtfulness when analyzing the problem and presenting the data.

Next up was a seamless transition into using this data analysis to support alternative payment models based on value. Pamela Pelizzari (Principal and Senior Healthcare Consultant at Milliman) honed into the relationship between patient outcomes and total cost of care to help align incentives toward implementing quality measures. As we trend toward higher claim volumes following alternative payment models, Pelizzari prepared her listeners with initial steps to building a bundled payment episode definition, multiple options for data sources, and caution of associated time, resources, and financial feasibility. Pamela also stressed the importance of evaluating the patient’s entire episode of care and ensuring the lower cost, alternative therapies are included in a provider’s comprehensive value-based payment model.

Click here to see all the photos from the event

Before attending the next morning session Beyond The Buzzwords, I would find myself saying “What exactly does ‘Machine Learning’ mean? When you say ‘Big Data,’ are we talking about a large file size? Asking for a friend.” Thank goodness for Dr. Virginia Miori (Associate Professor, Department of Decision and Systems Science, Saint Joseph’s University) and her ability to break down the meaning and applicability behind data jargon we hear all too often and pretend to understand. One of the most interesting takeaways from Virginia’s presentation was the distinction she explained between Augmented and Artificial Intelligence, the former being a complement to and the latter being a replacement of, the person analyzing the data. Although Artificial Intelligence is a hot topic in the healthcare space, Miori’s audience learned that technology has still not fully approached the ability to replicate human decision making, and really Augmented Intelligence advances should be paid close attention to as we continue down the path of using data to better inform our processes, decisions, and solutions.

After receiving a solid dose of data usage education, Data Analytics Seminar attendees turned their attention to Mike Rossi (Director of Governmental Reimbursement, University of Pennsylvania Health System) and Nelson Asport (Technical Consultant of Strategic Decision Support & Finance, Penn Medicine) who spoke on how to protect this data. Mike mentioned that while CMS has required safeguards to protect PHI for years, they more recently clarified what the penalties would be for not following these guidelines. Rossi provided his expert advice on encrypting hard drives and USB drives, 2-Factor Authentication, sending secure zipped files, and most importantly, the avoidance of ever making your password the word “password”. Nelson then opened our eyes to the world of PowerShell and its infinite uses to automate time intensive, manual tasks through a process likened to macro-building in Excel or Access. Not only does the incorporation of this tool save time and maintain consistency by reducing the chance of human error, but it also trains those using it on algorithmic thinking which further expands opportunity for usage to simplify. Nelson’s audience was immediately convinced of PowerShell’s efficacy after citing an example about shortening an internal manual process to 15 minutes that previously took four hours to complete. I was lucky enough to sit by Nelson throughout the rest of the conference, and learned of the free resources that provide PowerShell training, in addition to many other nuggets of imparted IT wisdom from his large bank of experience.

After an extensive lunch buffet and (more importantly) a large assortment of desserts had been consumed, the stimulating content did not slow down as seminar attendees settled in for a thought-provoking “Dirty Data” panel, administered by the session’s lead planner Bill Cogliano, who serves as the Manager of Strategic Planning and Business Development at Cooper University Health Care. Alyssa Dahl (Manager of Healthcare Data Analytics, DataGen) broke down the process of identifying and overcoming error values across Categorical, Quantitative, and Descriptive data types (among others). While educating her audience on the differences in troubleshooting missing/outlier data in these categories, Dahl made some valuable suggestions which included recoding missing values using statistical evidence based on correct data you did receive (or have access to). Jonathan Pearce (Principal, Singletrack Analytics) discussed the aspects of a data set that characterize it as “dirty” rather than “clean”. Pearce walked through a few examples of challenging data scenarios that were missing column headers, delimiters, and other clarifying pieces that make manipulation significantly more complex, and require a more thorough understanding of the individual fields and data set as a whole. Pamela Pelizzari (who also led the morning session: Using Data to Support Alternative Payment Model Development) stressed the importance of comparing utilization data with cost variables (as related to procedures, diagnoses, reimbursement as a portion of charges, etc) which can effectively be measured by a third party. Pelizzari voiced a widely-felt concern about the added cost associated and expressed that part of the solution should involve insurers bearing a portion of the price to engage this third party firm as potential value would benefit both provider and payer with the overall goal being shared savings.

During the next session, Alyssa Dahl continued sharing her thoughtful insight on the challenges of measuring quality from an analyst’s perspective. Dahl reminded us how quickly quality measures are becoming incorporated into payment models and clinical initiatives but also cautioned us of the challenges associated with imperfect data used to evaluate this quality (such as “population” data sets not reflecting the full population and administrative hurdles when new coding sets take effect). Alyssa encouraged her listeners to keep these strategic questions in mind: Are we choosing the right measures to evaluate performance? Should they be qualitative or quantitative (or both)? How many data sources do we have and how difficult will that data be to obtain? Is the denominator remaining stable to detect performance improvement? In addition to providing her audience with a “harm avoidance” estimation process and cost savings determination method, Dahl closed her presentation with a reminder for analysts to explain “which way is good” so that productive conversations can develop out of tying data surrounding quality measures to intended impact.

For the final session of the day, a subset of “Data Problem Solvers” highlighted their expertise in Research and Decision Support through case studies, tools, and approaches to solutions that have proven effective across their previous data quandaries. Lisa Ferretti (Strategic Decision Support Analyst, Penn Medicine) and Abby Saekang (Decision Support, AtlantiCare Regional Medical Center) both touched on the importance of tools to create dashboards, track Key Performance Indicators, and determine appropriate steps in the direction of positive change based on feedback and needs of the end users. Sadegh Mikaeili (Graduate Research Associate, State University of New York at Binghamton) then took us through an incredibly interesting segment on a computer’s ability to interpret linguistics, a more advanced version of basic “text mining” that actually takes the ebbs and flows of speech into consideration when analyzing this type of data. While text mining can help providers predict the risk of a denial before submitting a claim, an adverse event occurring based on previous incident reports, and an ED admittance of a particular patient according to his/her social determinants of health, “sentiment analysis” takes this a step further by examining certain text frequencies, the presence of sarcasm and irony, and unsupervised machine learning to inform decisions on how to market to a group of people, instill progress or solve problems based on feedback.

Armed with the data analysis knowledge and tools to transform the healthcare industry as we know it, the day’s attendees gathered at Moriarty’s for a glass of wine and excitedly discussed the progress of many nearby health systems in the data analysis arena, in addition to the technical opportunity we all have for data to guide leadership decisions and sustainable change that will allow providers to thrive in their ongoing mission of anticipating and providing quality patient care.