Speaker 1: Hey everyone, I'm Josh. I'm a senior data analyst at a major U.S. hospital. I've been working in this space for about eight years now, and today I'm going to give you a very high-level overview of what my work looks like in the healthcare industry. There's a lot of different ways that people work in healthcare using data analytics. Personally, I look at patient safety and quality of care in hospitals, so that's what I'm going to focus on today. Now, in some ways, being a data analyst in a hospital is a lot like being a data analyst anywhere else. At a very high level, about 20% of my job is dedicated to going to meetings. 80% of my job is where I actually get into the projects and analyzing the data. I work in the neurosciences department of a major hospital, and typically my work will be planned in two-week sprints. That sprint will be comprised of various reports or metrics that business partners will ask me to build over that period of time. These tasks will be picked from a backlog of tickets that emerge from other people working in the department, like neurosurgeons, clinicians, and practice managers who want to know things like what quality of care we're delivering for our patients. Now, the 20% of my job where I go to meetings is split 50-50. Half the time I'm meeting with my business partners in the neurosciences department, the other half of my time I'm meeting with people on my immediate clinical analytics team. Meetings with the business partners will range from understanding their request and communicating an ETA to getting feedback on the report or validating some of the data that I pulled out of our database. Whereas meeting with the analytics team might involve me reporting out my progress to my supervisor, having morning huddles to see what other analysts are working on, discussing organizational goals that we've set as a team, among other things. Now, all that is about 20% of my job. What remains is the 80% that's devoted to projects. So at a high level, some of the similarities between healthcare and other industries are that I use SQL a lot to query against a database to pull data that is clean and organized so that it can be consumed by other applications like Tableau or Python to do that actual analysis of the data. But I'm going to dive deeper and explain a little bit behind why healthcare can be so complex sometimes and why it can be difficult to do data analytics in a hospital. First, a little bit of history. Over the past 25 years, there's been this growing movement to make healthcare organizations more accountable for the healthcare that they provide to their patients. In 1999, the Institute of Medicine, or IOM, released a report that sent shockwaves throughout the healthcare community. That report was called To Err is Human. What that report revealed is that almost 100,000 people die every year as a result of avoidable medical errors. More recent estimates place that number at around 250,000 deaths annually, making it a top five leading cause of death in the United States. Now to put that into perspective, about 30,000 people die every year in the United States just from auto accidents. So it roughly ranges from being three to eight times more dangerous being a patient in a hospital than a passenger in a car. So this is a really alarming number. In 2001, the Institute of Medicine published another article called Crossing the Quality Chasm. It was this watershed article that really defined what quality is in healthcare and how we can bridge this gulf between the care that we have now and the care that we could have. Their vision for healthcare came to be known as the SEPTI model of healthcare. Patient care is safe, efficient, patient-centric, timely, effective, and equitable. These revelations about insufficient healthcare quality helped lead to evolving reimbursement methods for doctors and hospitals. To put it simply, healthcare providers were previously mostly reimbursed by a model called fee-for-service where they were paid for everything that they do. Now there are additional models in addition to fee-for-service where physicians and hospitals can get paid a bonus for providing good care and can get money taken away for providing bad care. In other words, the stakes are now higher than ever for hospitals to get healthcare quality right and to make patient care safe, which means data analysts are needed now more than ever to build reports that will help us navigate these issues, which brings us where we are today. The real meat of what I do is building reports to see how our processes in the hospital influence patient safety and the quality of care that we provide. This will typically be from the perspective of that six domains of care that I outlined in that SEPTI model. My typical framework will look like this. So first I'll meet with my business partners, usually a practice manager or clinician or administrator. Typically they will present me with some metrics that they want me to track and then visualize on a dashboard. Next I'll use SQL to look up that data. Usually that data is going to be stored in electronic medical record. This is where I spend a good chunk of time on my projects. A lot of my work will be dedicated to locating that data and preparing the data. Now one note about this, healthcare data is extremely complex. A lot of data analysts and other industries will pull data from a relational database management system, which looks like this. But the nature of many medical record systems, at least in the United States, is that they will store data in a hierarchical database. That looks kind of like a tree. The data is then often moved from that tree to a relational database management system so that it's easier to query against. Given the complexity of this process and the overwhelming amount of data that a single patient can generate, it's often hard to quickly track down where all this data maps to the various views and tables and columns. So unless you already have the training or the certifications in this medical record system that you're using, finding that data can prove to be very time consuming. Once I've found that data and pulled it and organized it in the right way, I often present a sample of that data to clinicians to validate it to ensure its accuracy. After that data has been validated, I start to visualize that data. Now this is my favorite part of the process. The dashboard that I develop will often be interactive. It will allow clinicians and providers to look at their data, see how they're performing stratified by doctor. You can see various outcomes and how well we're adhering to certain goals that we have. As we saw earlier, the things that we track in healthcare often fit into the SEPTI model of care. Here's some examples. Patient safety might look at how many times did a patient fall in our hospital and sustain an injury. Or efficiency might pertain to how long did the patient stay in the hospital measured by our average length of stay. Patient centeredness might be focused on how likely the patient is to recommend this hospital to other people. There's a lot of examples that I could explain here, but I won't go through this full list. All right, so here's where it gets more complicated. Suppose we wanted to take a category from the SEPTI model, patient safety for example. Maybe we're concerned with infections that a patient might develop after surgery, also known as a surgical site infection. The physician might not just be interested in the sheer volume of surgical site infections that we have. They might also be interested in what things they could have done to prevent those surgical site infections and how often we do those things. This is where we get into yet another framework called Donabedian's model. The idea is as follows. An outcome measure like surgical site infections is influenced by processes that we follow or fail to follow. One of these process measures might be something like practicing adequate hand hygiene before the surgery. The structure deals with whether we even have the resources to begin with, like if there's enough sinks and disinfectant at various stations available to the surgical staff before they actually scrub in for the surgery. So putting it all together, my business partner might come to me with a request to measure and monitor something that affects patient quality of care. They might be concerned with something like patient safety, like in the case of surgical site infections. They might ask me to create a dashboard that tracks not only the number of surgical site infections, but also process-based measures that we follow to curb surgical site infections. One of these process measures might also be influenced by the SEPTI model, like a timely process measure. Did we administer antibiotics within an hour of leading up to surgery? If we did that more often, we might see a decline in surgical site infections. Maybe they want to look at other process measures, like how effectively did we prep the patient for surgery? For example, the patient can use what's called a chlorhexidine wipe to wipe down the incision site prior to surgery, which also is known to prevent or at least curb surgical site infections. Now this is probably the most meaningful part of my job, contributing to a process that helps us deliver better and safer patient care. The unfortunate reality about healthcare, though, is that it's so complex that it's not uncommon to see a lot of mistakes made in healthcare. Patients can fall while they're in the hospital and they get injured. Patients can be given the wrong medication or at the wrong time or in the wrong dosage. Patients can acquire infections while they are in the hospital, like C. diff, MRSA, or surgical site infections. There are so many ways that a patient can be harmed in a hospital. In fact, there's an entire organization devoted to this called the National Quality Forum. You can see all of the metrics that they developed at their website. So the real meat of what I do as a data analyst in a hospital is working with the clinical stakeholders to understand what's being asked, and ultimately I create a visualization of that data to understand what we're doing well and what we're not doing so well when providing treatment to our patients. Now believe it or not, this is just the tip of the iceberg. That's about all I have for today. Thanks for watching and if you want to learn how to become a data analyst, I have videos for that. I'll drop them at the end of this video. Hit that like and subscribe button and I'll see you next time.