Welcome back. The purpose of this lesson is to discuss in detail a number of algorithms that have been constructed to group codes and records into more meaningful groups. This lesson follows the previous lesson about clinical identification algorithms and quality measures. Once again, this lesson applies to administrative data but some of the algorithms can easily be applied to clinical data as well. I will introduce groupers and review the importance of groupers within healthcare analytics. Thus, I will ask questions such as, what are groupers? Why should we consider using groupers? At the end of this lesson, you will feel comfortable articulating how groupers could help you analyze a large sample of claims or clinical data. So, let me ask the question again. What are groupers? What are groupers? As the name implies, groupers are algorithms that group codes or sets of medical records into meaningful categories for analysis and study. In a way, we've already started on our adventure to understand groupers given our discussion about clinical identification algorithms and quality measurement. Both of these are algorithms that group data into meaningful concepts. Clinical identification algorithms group claims or encounters into specific medical conditions. Clinical measures use administrative data or clinical data to create numerators and denominators. These in turn can create rates to identify gaps in the quality of health care. The next question is, why use groupers? Most statistical and data mining algorithms will not work well with hundreds or thousands of categories. However, diagnostic codes, procedures, drug codes, and other aspects of clinical encounters have huge numbers of categories. For example, it's nearly impossible to create an interpretable regression statistical model that has many hundreds of variables. Moreover, individuals often have numerous diagnoses, treatments, prescriptions, and visit types. Therefore, each patient might have dozens or even hundreds of records associated with their care. In addition, there is often a need to transform data that captures administrative events into clinical events or episodes of care. Overall, it is critical for most analyses to start by grouping data into more manageable, yet actionable categories. There are many different types of groupers and each one was created for a specific purpose. Thus, it is important for users to carefully review documentation associated with groupers to understand what the grouper algorithm was intended to solve. I list here a few important categories of groupers based on the intended purpose but there are certainly more categories. Some of the ones shown here can further be broke down into subtypes. First there are tools that were created to help users group codes. This is an important concept because below, I contrast these to groupers that group records or rows of data. In the next lesson, I'll review some groupers that are very helpful to group codes. The Healthcare Cost and Utilization Project or HCUP is a federal state industry partnership sponsored by the Agency for Healthcare Research and Quality. The HCUP Clinical Classification Software helps create categories from ICD codes and procedure codes. The Berenson-Eggers type of service, groups, Healthcare Common Procedure Coding, or these procedure codes. Next, RxNorm is a system that groups pharmacy NDC codes, or National Drug Codes. Second, they are groupers that have the very specific objective of grouping medical records into financial payment categories. Further on, I will discuss groupers at assigned records and episodes of care. Thus, rather than simply grouping codes, the algorithms assigned records likely to be associated to one another into episodes of care. A good example of these will be the symmetry episode treatment groups. It is important to note that groupers can be used for a variety of purposes. Thus, even though a grouper may have been created for financial reimbursement, it might be possible with the group could be used as reliable tool to identify the prevalence of a medical condition. For example, HCC or Hierarchical Condition Category groupers is typically used for finance. While these might also be used to study medical conditions within claims or encountered datasets. However, it is critical to carefully review the algorithms to understand what biases or problems could emerge when using a grouper for a different purpose other than what it was initially created for. Now, let me compare open source and commercial groupers. Open source groupers have great potential to help analysts group codes. Given that these do not have a license fee, Open source groupers have the potential to offer organizations great value. Some groupers such as the clinical classification software are quite easy to implement. For example, simple cross walk text files are provided and is relatively easy to put these within databases or to use them within the offered computer programs. In addition, computer algorithms are often provided with associated documentation from any of the open source groupers, and it is relatively easy to apply them to pre-existing datasets. There are also more complicated systems such as RxNorm that are more challenging to implement, yet the financial savings may be worth the effort. Many of the commercial groupers have similar features but sometimes are more complex or tailored to specific purposes such as assessing risk of past or future costs. Regardless of complexity, the main difference relates to cost. Commercial groupers can have large licensing fees. Yet associated documentation and texts little support can make these products more powerful and usable. As I discussed these various grouper types in the next lesson, consider the value of the groupers from both the perspective of analytics along from the perspective of costs. Both the direct license fees and the staff time that's necessary to implement the groupers. I look forward to seeing you soon.
