Skip to main content

Building Healthcare Knowledge

Our goal for Knowledge Base is to accumulate the best information on how to work with and analyze healthcare data. We think the best approach to building this knowledge is by starting with the end in mind and then taking an approach similar to grad school research.

Good Research Questions

Starting with the end in mind means having good research questions to motivate our learning. A good question is both interesting (i.e. something we care about) and specific. Good questions motivate and focus our learning. They help us to "go deep" in our study of a topic. Here are two example questions, one good and one bad.

  • Good Question: How can I tell which patients were discharged to a SNF from an acute inpatient setting and which SNF they were discharged to?

  • Bad Question: What's the best way to build a chronic condition grouper?

The "good question" is both specific and potentially interesting, depending on your personal research interests. The main problem with the "bad question" is that we don't have a clear understanding of why we are building the grouper - we don't have an ultimate goal motivating us and therefore cannot perform our research with an "end in mind."

Good Research Practices

Beyond starting with a good question or set of good questions, there are a couple tenants of good research practices that we follow. These practices are roughly equivalent to research practices you would learn and follow in a grad school research setting.

  • Literature Review: You should deeply immerse yourself with all the available writing on the topic of interest. This typically takes the form of journal articles but often white papers can be helpful as well.

  • Analyze Multiple Datasets: You should try to run any code against as many healthcare datasets as possible. This will help to develop your intuition about how much variation exists across different patient populations. It can also provide evidence of whether your approach will generalize across datasets.

  • Work to Disprove Yourself: It's incredibly common to develop a hypothesis and then to focus on confirmatory evidence of that hypothesis. But this mindset can fool us into believing patterns or results that don't represent reality. The prepared research mind will take an approach of developing an approach or hypothesis and then looking for reasons why it won't work.

  • Don't Let Perfect be the Enemy of Good: Real-world healthcare data is seldom perfect for any analysis. Building incremental knowledge requires we focus on what truths we can be certain of in the face of uncertainty, versus focusing on inherent limitations.

Knowledge Club

A critical part of the knowledge development process is sharing what you've learned with others. Sharing what you've learned with others has two positive consequences. First, explaining your knowledge to another person or a group of people forces you to organize your thinking in a way that can expose logical weaknesses. Second, it gives people the opportunity to provide direct feedback to you, based on their personal knowledge of the topic. Both of these can lead to the formation of more robust knowledge.

In the coming weeks we'll be kicking off the Knowledge Club. This will take the form of a bi-weekly 1 hour open meeting where one or more individuals will present on a research topic. After each presentation we will open up the entire group for questions and comments. We are excited to implement this graduate-style research roundtable.