Hello again, in this lesson, I will continue with what we started in the previous lesson, by reviewing open-source groupers. I will focus on the domains of co-morbidity groupers, patient risk, and drugs. At the conclusion of this lesson, you will be able to find open source groupers related to comorbidities, patient risks, and drugs. With this knowledge, you will be able to prepare an analytical plan, to map codes to more general and usable analytical categories. Let me show you what I mean. Let's get started by discussing the Charlson Comorbidity Index. The index was developed by Mary Charlson and other independent researchers in 1987, this group are similar to the H gap co-morbidity grouper that categorizes ICD-nine or ICD-10 diagnosis codes, into comorbidity categories. There are associated weights for each category that predict long term health care outcomes and mortality, these weights are the outputs from predictive modelling. The Charlson comorbidity index, has been carefully studied by many other researchers, and many find it quite useful for many analytical domains. Let's look at Charlson Comorbidity Scoring. This image illustrates how the Charlson comorbidity index is scored, each category scored one point, unless otherwise noted. If no comorbidities are found in the patient data, the patient will receive a score of zero. The higher the score for each category, the greater the probability, that no predictive outcome will result in mortality or higher resource use. Some examples of a condition with a score of one, include myocardial infarction, congestive heart failure, COPD, and connective tissue disease. There are some conditions that have two points, these include leukemia, malignant lymphoma. A few diseases have up to six points, which would include AIDS. This reflects the higher cost and higher risk of death for AIDS patients. Okay. We can now move on to the Hierarchical Condition Category grouper model, creative for healthcare finance. Hierarchical Condition Categories were developed in the late 1990s, to adjust for health differences among Medicare members. The Centers for Medicare and Medicaid Services or CMS, introduced this new model of HCCs in 2002, more recently the US Department of Health and Human Services, contracting with researchers and actuaries to modify the HCC model, for use with the health exchanges that are the foundation of the United States its Affordable Care Act. The purpose of the HCCs is to group ICD-nine, and later ICD-10 codes into a smaller number of manageable categories. There are 78 cc categories for the CMS version of the model, and over a 100 categories for the HSS version, that is used for the health exchanges. Once the groups are formed, actuaries and researchers, then assign weights to specific HCCs by using predictive models with costs as the outcome or the target variable. The weights then form the scores. For example, a person with an HCC related HIV/AIDS, has a numeric score higher than a person with an HCC for asthma. This is reflected in the higher average cost for patients with HIV/AIDS as compared to patients with asthma. Next, let's review the Chronic Illness and Disability Payment System or CDPS. This is another risk adjustment system similar to the HCC model. CDPS is an open source software that was developed by University of California at San Diego. It is a diagnostic classification system that allows Medicaid programs to apply, risk adjusted capitation payments, for specific types of Medicaid beneficiaries. The system requires claims or encounter data, and focuses on ICD-nine or ICD-10 diagnoses, and NDC drug codes to put members into condition groups. There are 20 major condition categories, and 67 medical categories. Similar to the HCC model, the CDPS model creates a set of condition groups, these groups are then used in predictive models, in this case regression models are used to understand Medicaid expenditures. The weights or the coefficients from the models are then used in the scoring system that the US. states such as California or in Nevada, can then apply later for their reimbursement rates. For example, conditioned groups with higher weights are likely to have greater expected costs, hence a state can decide to reimburse more money with patients that have these higher weighted conditions. Next, let's review RxNorm. This is a powerful group persistent for grouping National Drug Codes or NDC codes, into specific categories. RxNorm was a grouper developed by the US National Library of Medicine. The purpose of the group is to provide standard names, and metadata for clinical drugs. In addition, the system groups codes into therapeutic classes. NDC codes are codes that come on either clinical or administrative data sets. These are some of the most important data coming in health care datasets. There are various open source grouper systems that can be applied to RxNorm. These include, the National Drug File reference terminology, and therapeutic classes from the medical subject headings, and the anatomical therapeutic chemical classification system classes. I was once involved in a large project to apply RxNorm to a startup company, that was trying to group NDC or National Drug Codes. We found that RxNorm was free, so the CEO of our company really liked that, but it did take us many hours of our time to learn and implement the system. RXNorm is powerful, but it's not very easy to use. Therefore, one must weigh the relative value. As with any tool, there are advantages and disadvantages, and advantages of RXNorm is that since it is in the public domain, it is free. Well, at least it's free to access the computer code in the reference tables. It will take your staff time to actually employ these tools. Next, a system contains all the NDC codes subject to FDA review. For learners outside the USA, the FDA is the US Food and Drug Administration. It is nice to know that there are monthly updates to the system. Finally, there are links to many of the sources of metadata, three of which are highly relevant to analyst, who were trying to group NDCs into various types of categories. Next, onto the disadvantages. There are quite a few complex application program interfaces to navigate, to get to the data and group the data, some structuring is needed to make the data more usable. Finally, it's necessary to augment our RxNorm, with relevant FDA exempt NDCs. It's important to note that not all NDC codes are grouped into RxNorm. The US FDA maintains via the National Library of Medicine, an active list of all the drugs that they have approved. However, there's really no official inventory of all the National Drug Codes. Why do some NDCs go beyond these attributes and thus are exempt from FDA review? First, there are differences in drugs related to packaging and not to the drug. However, the drugs are in different packages and often get different NDC codes. Next, there are numerous supplies for administering drugs such as syringes, and these are not always tracked by the FDA. Another example are tools used for monitoring and/or dosing, these include glucometer, lancets, and test strips. These are often not covered by the FDA. Next, manufacturers are allowed to issue their own NDCs, when they're exempt from FDA review. Finally, submissions of FDA exempt NDCs to the National Library of Medicine are voluntary and part of manufacturers. That concludes our look at NDC codes and groupers, and this wraps up our two lesson overview of open-source groupers. In our next lesson, we will look at commercial groupers.
