In this blog post, I am going to discuss one of the essential aspects of my ongoing research- steps to obtain the classification of diabetes-related foot disease using different coding system. I have mentioned about my ongoing research my very first blog post. So, I am expecting my readers to have some understanding of my research topic before they proceed to read this blog post.
I am using administrative data sets to address my research questions which comes from various sources. The sources include hospital department, emergency department, Medicare benefits schedule, pharmaceutical benefit schedule, cancer registry and mortality database. These data sources register health-related complications in many forms. I found two types of coding in the data sets provided to me. International Classification of Disease (ICD) and Systematized Nomenclature of Medicine - Clinical Terms (SNOMED- CT) were the systems that were used to assign diagnoses and procedures related to different health conditions. I came to realise that Australia uses slightly different versions of these codes compared to international practice. For example, ICD-10 (10th revision) codes used in Australia is termed as ICD-10 AM, where ‘AM’ means Australian modification. Again, SNOMED has its Australian version as well. As a result, I had to analyse ICD-10 AM, SNOMED-CT Australian version along ICD-9 Clinical Modification (CM) to find the diabetes-related foot complications from relevant data sets.
At first glance, it seems a straightforward task as one may understand it as mapping of different complications to its codes that are readily available. But this was just the start of a series of challenging tasks that I had to go through. First, I had to figure out what were the types of foot complications that were experienced by diabetes patients. To my surprise, the number of diseases I found was a way more than expected. Secondly, I had to search for relevant codes for diabetes and all foot diseases related to it in all of three systems, mentioned previously. Thirdly, I had to run a program in my analytical software to figure out which patients had diabetic foot complications in my data sets. The challenge I faced was that there were no specific codes for diabetes foot disease. As a result, I had to find the patients with diabetes followed by patients with foot complications and had to assign them as patients of the diabetic foot if and only if both conditions were satisfied. The identification of diabetes status from SNOMED made things more complex as the system did not have specific codes for diabetes, unlike ICD systems. At the same time, I had to be careful and efficient with my programming as I was working with large data sets, termed as “Big Data” nowadays. The efficient programming was required to get results in minimum possible time as I was constrained with limited memory of computer work-space held in cloud computers in a secured environment. In summary, the whole process required a rigorous literature review, supervisors’ directions, experts’ opinion along with proper choice of analytical software and efficient programming.