A recently published study using electronic medical records has suggested that a prescription for the GLP-1 receptor agonist (RA) Semaglutide is associated with a decreased risk of wound healing complications, chronic non-healing “wounds,” pain, treatment for wound care, wound dehiscence and amputation. I was so excited when I saw the title! I thought to myself, “See, I knew it!” Then I read the study, and I am mortified on behalf of these authors. No one who knew anything about wound care coding was involved in the study. In fact, it is clear that no one who knew anything about CODING was involved. I will not discuss the results because once you see the codes that were used, you will understand why the conclusions are irrelevant. I am not even sure how this got past reviewers.

This retrospective cohort study utilized TriNetX US Research Network data between 2013 and 2023. The authors state that as of June 14, 2024, the TriNetX United States (US) Collaborative Network consisted of 113.5 million patients across 64 healthcare organizations (HCOs). The authors explain that TriNetX does not provide “patient-level data” but consists of encounter-level health records, pharmacy records, and insurance billing data, which were used to detect the presence or absence of specific medical, surgical, and prescription codes. In other words, no data were available on the characteristics of the ulcers themselves or whether they were noted to be healed. However, as you will see, the absence of clinical information is not the real problem with the study. The real problem is that they did not even get close to the correct diagnosis codes.

The study compared outcomes between Semaglutide users with DFU (Cohort A, N = 6329) and non-users with DFU (Cohort B, N = 118,821). Participants were matched by age, gender, race, and ethnicity. Patients with certain medical problems that might have separately impacted healing were excluded, including chronic kidney disease (ICD-10: N18), chronic obstructive pulmonary disease (ICD-10: J44, J44.9), and heart disease (ICD-10: I51, I51.9) – although these code sets were incomplete to ensure that the conditions were identified. They also excluded patients with “peripheral vascular disease”  using ICD-10 codes I73 (peripheral vascular disease unspecified) and I73  (“other” PVD). Many additional codes would have been needed to exclude patients with peripheral arterial disease.

To find patients with diabetic foot ulcers, the investigators correctly used a dual code approach, identifying patients with type 2 diabetes (strangely, they did not include patients with Type 1 diabetes who also get diabetic foot ulcers), combined with a very limited list of chronic ulcer codes (ICD-10 codes: L97, L97.5, L97.50). Wound care practitioners can already see the problems here. These chronic ulcer codes are of the foot, but it is not clear whether the analysis would have included all the related codes of L97.50X which are needed.

The study investigated six outcomes related to wound healing and the codes used here are in some cases, so inappropriate that it is hard to even comment. The researchers do not understand the difference between wounds and ulcers or the basics of wound care services or the basics of diagnosis coding.

Here is how outcomes were defined:

• “wound healing complications”: T79.8, T79.8XXA, T79.8XXD, and T79.8XXS – all codes related to early complications of trauma, none of which would not be appropriate to use in relation to a chronic ulcer.
• “rate of chronic non-healing wounds”- for which they used T81.89XA – “other complications of procedures not else where classified.” Complications of a surgical procedure will in no way identify the “rate of chronic non-healing wounds.”
• “ level of chronic pain”: G89.2, G89.28 –chronic pain not elsewhere classified and post procedural pain – neither of which are appropriate codes to use and neither of which measure pain “level.”
• “incidence of post-procedural wound care”: Z48 (encounter for other postprocedural aftercare) , Z48.02 (encounter for removal of sutures!), Z48.03 (encounter for change or removal of drains!), Z48.01 (encounter for removal of surgical dressing!), Z48.00 (encounter for removal of nonsurgical dressing) – Words entirely fail me here. None of these codes are relevant to diabetic ulcer management.
• “occurrence of wound dehiscence”: T81.3 – disruption of a wound not elsewhere classified – also not relevant to chronic foot ulcers.
• “rates of amputation”: S98.332A (is that a valid code?) and Z89.43 -acquired absence of foot.  . . .(They did not use the amputation procedure codes, of which there are many, or any other code that would be appropriate)
• “Time to Complete Wound Healing” using the following codes:  T79.8 (early complications of trauma), T79.8XXA (other early complications of trauma, initial encounter), T79.8XXD (other early complications of trauma, subsequent encounter), and T79.8XXS (Other early complications of trauma, sequela)- none of which have anything to do with wound healing!

In their own words: “Chronic Non-Healing Wounds: The code T81.89XA was used to represent chronic non-healing wounds. This code encompasses a range of complications and was selected as a general marker of non-healing, acknowledging that it may include broader wound complications. Amputation: The codes S98.332A and Z89.43 were utilized to capture amputation events. While S98.332A includes traumatic amputations, it was the closest available code to represent DFU-related amputations in the dataset. This was chosen as a surrogate for amputation due to DFU complications. Wound Dehiscence: The code T81.3 was used to capture wound disruption, despite it being associated more frequently with surgical wounds healed by primary intention. It was chosen to approximate wound dehiscence in the context of DFUs.” 

Boy is that excerpt above embarrassing. I cringe just reading it. I wondered how the investigators could have gotten this so horribly wrong. The project was funded in part by the Department of Health and Human Services through its grant to the UTMB Center of Excellence for Professional Advancement and supported by the Institute for Translational Sciences at the University of Texas Medical Branch, in part by a grant from the National Center for Advancing Translational Sciences at the National Institutes of Health (NIH).

I emailed the authors to ask how they chose the codes. The corresponding author admitted that the medical student who led the project “likely just typed key words” into an ICD10 search tool and used what he thought related to surgery. No one with wound care expertise was involved, and no one checked his results.

It is unfortunate that an opportunity to understand the impact of Semaglutide on diabetic foot ulcers was entirely lost. This study is a great example of how not to use electronic health records in wound care research. Those of us who served on the Clinician Expert Workgroup to develop a cost measure for “non pressure ulcers” can attest to the fact that even the “experts” did not know how to use wound care codes. We are going to have to create some guidance documents to prevent this sort of thing from continuing to happen. Maybe as a community we should create a methodological approach to use of diagnosis codes for wound care research. This study should serve as a warning to us. Badly designed studies using incorrect codes could cause tremendous harm in the field of wound care.

Caroline Fife, M.D.

Dr. Fife is a world renowned wound care physician dedicated to improving patient outcomes through quality driven care. Please visit my blog at CarolineFifeMD.com and my Youtube channel at https://www.youtube.com/c/carolinefifemd/videos

The opinions, comments, and content expressed or implied in my statements are solely my own and do not necessarily reflect the position or views of Intellicure or any of the boards on which I serve.