January 20, 2016
Direct Consulting Associates recently had the pleasure of interviewing Yiscah Bracha, PhD, Research Health IT Scientist at RTI International.
Please tell us a little bit about yourself.
I started in “analytics” before that word was widely used, as a mathematics major in college, also in a private tutorial relationship where I learned Philosophy of Science. In my last year of college, I taught myself statistics, where the only way to compute parameters was by using the calculator bolted to the desk in the lobby of the Psychology Department. A few years later, I learned how to use MINITAB which was available through large mainframe computers with dumb terminals, and that made it possible to apply statistics in ordinary life, or at least a life where mainframe computers were ordinary. They existed at the manufacturing giant 3M, and for several years I taught their MS and PhD chemical engineers how to design experiments and analyze data using traditional statistical techniques, to improve their outcomes and processes. This is exactly what I would be doing decades later, in health care, but there were many circuitous twists along the way.
The first circuitous twist, and the one that moved me from manufacturing to healthcare, where I would remain, was serving as a master’s level statistician supporting a large, world-renowned, NIH-funded clinical trial for primary prevention of coronary heart disease in middle-aged men. While I was teaching at 3M, I had decided to go to graduate school and get a master’s degree in statistics. I figured that the degree would give me legitimacy, and a theoretical grounding in what I already was doing empirically, and also expose me to methods that I otherwise wouldn’t have stumbled across on my own. All that happened according to plan, and it wouldn’t have happened if I had gone to grad school directly after college, because both the world and I had to ripen a bit.
So I got the MS, and consulted a bit across multiple industries, and then accepted this job at the Coordinating Centers for Biometric Research, in the Division of Biostatistics at the University of Minnesota. It was blind luck, but I happened to stumble into a position that gave me opportunities to collaborate with some of the world’s leading MD/PhD clinical epidemiologists in chronic disease. I interacted with them routinely, learned how they thought, learned how to translate their clinical vocabulary into questions that could be answered with the data we had. It gave me publication opportunities, because I also could write; I not only was analyzing data, I also was preparing the manuscripts that shared what we found. The job also exposed me to these new technologies called “email” and “internet”; for years, I was the only person in my social circle who routinely used a computer for work, who was even aware of these new communication and information tools. But the really unique exposure, that informed many things that followed, was how the technology, data management and analytic functions were organized in this very high quality analytics shop. We were two hemispheres of the same brain: Tech and data management on one side, and analytics on the other. Each hemisphere had its own unique ways of processing and contributing to the world, they knew enough about each other to communicate, and together we did much more than each could do alone. I took this for granted at the time, not realizing how important it was until I got out of academic clinical trials research and into healthcare delivery.
I made the switch because I had lost interest in clinical trials, or more accurately, I had lost faith that randomized controlled clinical trials could produce information that was relevant in the real world. In the real world, the patients are the ones who walk through the door; they are not carefully selected like they are in trials. In the real world, nobody is giving patients their treatments and medications for free, nobody is following up with them assiduously to make sure that they are adhering to treatment protocols, nobody is making sure they return to the clinic for follow-up visits every few weeks or months. I was interested in the real world, and I imagined what the analytic possibilities could be if I was working with data that emerged from processes that took place in real life. It was the early 2000s, a few years after I made the switch (but many years before the passage of the Affordable Care Act), and in Minnesota, the large vertically integrated healthcare delivery organizations were one by one starting to implement electronic health record systems. The prospect of harvesting EHR data and using them for research was very exciting, and I got involved in that effort in the organization where I worked. We failed to attain our goal, which was to use these EHR data to identify and redress disparities in health outcomes and care, particularly in chronic disease. The goal was simply too aspirational for current state at the time, and ten years later, it still is aspirational in the place that first proposed it. But we did use the EHR to support a real-time, guideline based decision support tool that helped clinicians select the optimal treatment choices for asthma while they were with the patient in the clinic. That decision support tool could have been used to obtain real-life data about treatment choices and outcomes, which could have been analyzed to generate new kinds of evidence, grounded in the real world, but doing that required a sponsor, along with an acceptance of the “evidence” that emerged that way, and neither were possible while research and delivery were so distant from each other organizationally.
The project became the basis of my PhD dissertation, because along the way, I had returned to graduate school again, this time to get a PhD in Health Services Research and Policy. It seemed like a fitting discipline for how to work with real world data, learn from them, and apply the lessons back in the delivery environment. Once again I was interested in legitimacy, and theoretical grounding, and new methods for my analytic tool box, and once again it worked, except this time it took seven years! I finished the degree after I had moved to Cincinnati, to lead the Data Analytics team in the very robust and advanced quality improvement department at Cincinnati Children’s Hospital. Here I had the most direct opportunity to apply those early lessons from clinical trials about how to organize a team that produced high quality analytic results consistently, and during my tenure, I transformed our group from an undifferentiated collection of generalists to a well-organized multi-functional team comprised of specialists. We were good, and they still are.
I spent my last 18 months at Cincinnati Children’s working closely with the IT department, to design a set of technologies, staffing structures and governance models that would produce high quality analytics across the enterprise. This was the logical next step, because excellence required an organization-wide strategy for managing and governing its data assets, and IT had that responsibility. But I had been homesick for Minnesota almost from the first day that I arrived in Cincinnati, and eventually I left to come home. A few months ago, I joined a non-profit research firm, telecommuting from home. I’m back in the research world, and this time paradoxically applying some of what I learned in delivery to the research environment.
The world around analytics is accelerating very rapidly. What other “trends” are you seeing right now?
I’m seeing expectations that data from consumer wearables and medical-grand sensors will become routine elements of the health data ecosystem, and that analyses of these data will help patients, providers and researchers improve health. The data management and analytics worlds are not really ready to meet these expectations, but they are coming anyway.
What fascinating projects are you currently working on?
I’m currently part of a team that is preparing a proposal that responds to the President’s Precision Medicine Initiative (PMI). The initiative is hugely aspirational: One million Americans will be followed for many years, contributing self-reported data about their health states and disease, and also giving permission to harvest data from the EHR systems that their providers use, from the activity trackers and environmental scanners and biorhythm sensors that they wear, and from biospecimens they contribute. All these data are to be integrated and curated and made available for analysis not just to academic researchers, but also to participants and citizen scientists. The goal is to understand how lifestyle, environment, genetics, medical treatment, all affect individual states of health and disease. NIH Director Francis Collins recently gave an interview to Politico, which ran the story under the headline about the high hopes and mad schedule for PMI. Yep. The PMI is really pushing the envelope on what is currently possible, and it is great fun to learn about all the resources out there that can be leveraged to pull it together.
You have over 20 years of experience in HIT. What or who do you attribute your success to? Did you have a mentor(s)?
I’m not sure if I’ve “succeeded” in the conventional sense of the term. It is true that I’m now being invited by people like you to offer insights and opinions about data and analytics in healthcare, but that has a great deal to do with forces outside my control, such as changes in technology and public policy. I just happened to be available when the world started to ask for what I’d been seeking for a long time. For many years, I was on a lonely path, pursuing a vision that wasn’t widely shared. In that vision, we leverage data produced through ordinary life, grab hold of them, manage them, and analyze them using appropriate methods, to understand what is actually going on in ordinary life. The academic health research world did not consider this to be a legitimate vision, as it privileged data from randomized controlled trials (RCTs) above everything else, with an enormous infrastructure and huge funding streams dedicated to supporting RCTs. And the healthcare operations world did not understand what we were talking about. Less than ten years ago, the leading EHR vendors were mystified by this vision as well.
Much has changed and is changing rapidly, partly due to changes in technology that make the previously unthinkable almost routinely possible, a phenomenon that has punctuated my career several times, from college till now. Technology can indeed drive change, but often in very unexpected and unpredictable ways. Also, the Affordable Care Act has now, finally, created a set of policy levers that provide incentives to improve quality and reduce costs. It is as if the country and legislature finally realized that if we don’t get a collective handle healthcare cost and quality, we will be spending every last dollar on medical care that doesn’t make people any healthier. The ACA provided incentives to get the house in order, and the data and the analytics based on them are essential to making the change, and there now are technologies available that make both the data and analytics more accessible than ever before. Those EHR vendors that were bewildered by this talk of using their data for analysis are now offering products of their own in data warehousing and self-service analytic tools. The vendor space has exploded with data management and analytics products and services targeted to healthcare providers and health information exchanges.
But still, I must say that even with all the changes underway, widespread institutionalized vocabulary is still rooted in the past. For example, I recently found a 2011 web site put up by the National Heart Lung and Blood Institute, called Data Coordinating Centers’ Best Practices. I was hoping to find something about best practices for managing the dirty crude data emanating from EHRs, medical sensor technologies, etc. Nope. After all these years, this was about best practices for managing data from multi-site clinical trials.
So if personal/professional “success” means that I’m now perceived as a thought leader and expert (which, by the way, is not how I perceive myself, because daily I am aware of what I still don’t know), then the attribution goes to a determination to seek truth, and follow that path wherever it leads. It also goes to friendships and professional relationships with kindred spirits I found along the way. I didn’t do it completely alone; I always found partners who shared the vision, as odd as it may have been at the time. We supported each other and had fun together, and we did some things that were wildly creative at the time, and we kept each other going. So the determination to stick to the vision, and the friendships and professional relationships, got me personally to the point where I am now, but it wouldn’t be perceived as valuable if the world hadn’t caught up as well. Now the world is kind of overtaking us, which is a giddy sensation indeed.
What personnel are required to succeed as a data-driven organization?
Organizations that wish to become data driven need both producers and consumers of actionable information.
The producers are not single individuals, but rather individuals with complementary skills working together on multi-functional teams. The data produced by EHR and other systems are like crude oil coming straight out of the ground: The crude won’t run your car or heat your house until a lot of refining, transformation, and delivery takes place. The same is true for “data”. What you get from the system is crude; what you want is refined, actionable information. It’s a multi-step process to transform that crude data into the actionable information, and different personnel with different training and skills are required along the way. You need database engineers, people who know how to store data, keep them safe, manage inflows and outflows. You need people with skills at cleaning raw data, and mapping the contents to standardized terms so that they make sense for analysis. You need people with data architecture skills, who can package data together in forms that analytic users can navigate easily. You need people who know how to work with state-of-art self-service and visualization software, to create applications that give consumers the ability to answer questions on their own. And you need people with high-end analytic skills, who know how to deploy advanced analytic methods or who can develop them, to answer questions that don’t have readily apparent answers. Especially at the end of the process, where you’re getting closest to the consumer, the people fulfilling these roles must serve as communicative bridges as well, translating customer questions and concerns into something that can be addressed with available data, and helping the customer understand and interpret the meaning of what they are seeing.
That’s on the producer side of a data-driven organization. The consumer side is equally important. Culturally, there has to be a hunger for this kind of information, a determination to really know the truth, and to do what it takes to get that truth. There have to be people who demand self-service analytic tools, because they refuse to wait for the next available “analyst” to service them. They want to poke and probe into the data themselves, going to an “analyst” only when they realize that their quest for information has exceeded their ability to acquire it independently. Ideally this mentality is prevalent at the highest levels of leadership, because leadership sponsorship is required to invest in the resources required to satisfy that demand. If the demand exists without the leadership necessary to satisfy it, frustration and chaos will ensue, as everybody scrambles independently to meet their own informational needs.
What soft skills do you look for when hiring new talent?
I look for curiosity, a deep desire to learn. I look for determination to get to the truth, and an insistence on producing the highest quality work while realizing that you can’t allow the perfect to be the enemy of the good. I look for a service mentality, a desire to help others, and make others’ lives and jobs easier. I look for both confidence and humbleness: Confidence that there is a way to solve the problem at hand, humbleness in the ever-present awareness of what we still don’t know. I look for an ability and desire to work as a member of team, but also independently, as both are required to meet analytic needs.