One of the most important steps in building clinical prediction models that transform clinical practice, is ensuring that your models are actually usable by the people who will be affected by your model. Building usable models requires work and thought before, during, and after the model building process. Before you start building your model, you need to do some groundwork to understand whether you are building a meaningful and implementable model. A major challenge clinical care is getting the right information about the right patient at the right time, in the right way. Understanding the information needs of clinicians or really any end user of your model can help you identify one, what that right information is? Two, who the right patients are? And three, when your models should be predicting this information? How does this impact your model? Knowing the right information tells you what outcome you should be predicting. Knowing the right patients tells you who your model population should be trained for, which can help you identify the correct data-set you should use to build your model. Finally, knowing when this information is most useful, helps you perform appropriate feature selection and data wrangling to ensure that your model will be accurate at a meaningful moment in time. To ensure that you provide this information at the right time and in the right way, you need to understand your users workflow and the technical resources available for implementation. This information can again help you with a feature selection to make sure that the predictors you use will be available to your model when you go to put it into practice. You will want to incorporate this knowledge into your model-building process. After you have built your model, you'll move into the pre-implementation phase. Now that you have a final model in play, you will need to identify final workflow modifications and the technical implementation approach. Ideally, you will have one or two end-users who are invested in the project, who will help you design the implementation. However, you will want to talk to many users about the implementation plan to ensure that the changes you are proposing will be widely acceptable and ideally beneficial to them. During this process, you should also pay attention to what training might be necessary during implementation to make sure that the end users know about your model and how to use its findings. After you have an implementation plan in place, and implement your model in practice, there are post-implementation assessments that you will want to apply. There are two main types of assessments for you to think about, process and outcomes. Process assessments, help you understand whether people are even using your model. While outcomes focused on whether your model made a difference for either providers or patients. The thing that all of these steps have in common is that you are trying to understand people, not data. However, just like there are different modeling methods, there are also a lots of different qualitative methods that can help you better understand those people. In the long run, learning to understand how people work, process and affect change, while challenging, makes a huge difference and building models that can transform clinical practice. The goal for this week is to help you build your qualitative methods toolbox. I am going to cover the basics of a number of different qualitative methods. You'll learn some approaches and examples for how to apply these methods in the context of health care. We also will be switching things up and ask you to do a practice qualitative project this week rather than a quiz. Just like with programming, the best way to learn qualitative methods is by just jumping in and trying them out. Don't worry, you don't need to be perfect on your first try. Before we get started, I do just want to acknowledge the folks that have trained me on the importance of qualitative methods and given me the foundation that I'm passing along to you today. The content I'll be teaching this week draws heavily on content that they originally created and the lessons that I've learned putting their teachings into practice. They'll know that any mistakes or misrepresentations are entirely on me. With that, let's get started.
