Participant data were stored separately from recordings and transcripts in a firewall-protected drive. Transcripts were deidentified for analysis. Data analysis used the RADaR method.⁹ Qualitative codes were initially identified through in-vivo coding, with concepts emerging from the participants’ own words. In qualitative research, saturation is the point where information presented in the data tend to be redundant to already collected information, implying that nothing new will likely emerge by continuing to collect data.¹⁰ In this case, interviews were collected prior to initiating analysis, with a goal of 12–15 interviews, making saturation very likely. Upon reaching saturation (at 7 transcripts), a preliminary codebook was created, which was then used to code 10 additional transcripts. Two members of the team reviewed the codes, and through random selection of coded transcripts (n=5), double coded to verify interrater reliability. Thematic clusters were identified from the reduction tables and the codebook was finalized. The team used the finalized codebook and reduction tables to summarize data and identify supporting quotations. Themes and codes were collapsed through consultation with the full team for clarification and description in manuscript writing. While the team was unable to conduct member-checking interviews due to time constraints, emerging concepts were supported through the substantial number of interviews completed, well surpassing saturation.

The most significant challenge that MCs identified consistently across interviews was lack of sufficient detail to identify the most appropriate and specific ICD-10-CM code. This included issues related to variations in terminology used by healthcare providers and a lack of thoroughness in the documentation available to them.

Participants indicated that although they try to avoid it, they default to applying an “unspecified head injury” ICD-10-CM code for four main reasons: (1) LOC greater than 30 minutes not documented, (2) documentation of “closed head injury,” (3) use of “probable” or “suspicion of…” in the notes, and (4) use of the unspecified code (S06.9) anywhere in the electronic medical record (EMR). MCs stated that LOC is a necessary component of assigning an ICD code for a head injury. Participants indicated that the coding requirements are for more or less than 30 minutes, but healthcare providers may write “brief LOC.” As “brief” is a subjective term, MCs said this could be interpreted as 2 minutes or 30 minutes. If there is no specific documentation of duration of LOC, MCs may default to the “unspecified head injury” code.

ED providers may use terms such as “closed head injury” in their documentation, for which there is no direct ICD-10-CM parallel. They may also choose “unspecified head injury” as an option for diagnosis even when additional information is available. Additionally, use of terms that qualify potential injury, such as “probable” or “suspicion” are not definitive diagnoses. So even if more specificity exists in other documentation, MCs will default to the use of the unspecified identifier. The authors spoke with a physician who trains providers about coding documentation, and this provider explained:

The rules are different for the emergency department and for inpatient … A coder may not code something that says probable, likely, suspect, etc. They have to code the signs and symptoms…. So if I use the words “probable brain injury” … in the emergency department, it can’t be coded at all… the probable eliminates it. … So immediately you have what the doctor may think of as a diagnosis … being eliminated by the coder because of the coding rules about uncertainty.

Time and billing pressures are omnipresent in the work of MCs. Respondents reported feeling pressure to get claims processed and submitted as soon as possible, prioritizing speed over specificity. One respondent (MC06) noted that, “You can’t afford to have things hanging around, just because you would like a more specific (ICD-10-CM) code.” This participant described the tension between querying the provider for additional documentation versus processing it with an unspecified identifier code:

They like to use as few words as possible… physicians don’t necessarily understand that we really would like to have that information. But we also need to make sure that we get that claim out the door. So, you know, you can ask physicians questions about their documentation, but it delays the claim.

There were several observations from these interviews that elevated issues about participants’ potential bias and clinical interpretation beyond their role as an MC. While some MCs receive continuing education and adhere to all guidelines and structures within their role, these individuals are not themselves healthcare providers. In some instances, respondents described times when they interpreted what they thought was intended in documentation to achieve the best ICD-10-CM coding assignment.

There were instances of MCs making significant assumptions based on their own potential biases that would influence coding decisions and patient billing or insurance coverage. In one example, an MC stated that when they see a record for someone in their 80s who has fallen, they assume the fall was due to a dementia-related issue, gait problems, or osteoporosis. Another MC indicated that injuries among boys are more likely to be serious, so they code those injuries at a higher level than those of girls. The MCs interviewed clearly expressed these as assumptions rather than data, and in some cases, MCs explained that their assessments strayed into diagnostic interpretation. For example, one MC indicated that “history of fall” in a note indicates that the patient likely has a chronic condition that has included falling. This is inconsistent with the clinical use of the term, which can simply mean an incident prior to the current moment. Another respondent questioned the seriousness of a patient’s injury, stating, “maybe she should have come to the hospital a little bit sooner if she was so concerned about it” (MC14).

Additionally, many MCs assumed that there should be imaging for patients with a head injury and looked for that in order to apply a more specific ICD code, although such imaging is not recommended for most patients with a head injury. One respondent explained:

If it’s probable or suspicion of, it’s not a definitive diagnosis …I would just code it as unspecified head injury … until there’s a definitive diagnosis, or a CT or MRI or something, imaging that was done that was able to give me a definitive diagnosis.

Respondents were presented with two vignettes containing narrative descriptions of a 20-year-old female patient who comes to the ED after slipping in the bathroom and hitting her head. In the first vignette, four hours passed and she had not taken any medications for the headache she developed, no LOC or photosensitivity, and all other exam findings normal. The second vignette changed the same patient to 12 hours post injury, she had taken a non-opioid analgesic with no relief for the headache, presented some photosensitivity and dizziness, but still all other findings normal. The vignettes did not provide either a procedure code or ICD code of any kind, nor did they use the words concussion, trauma, or TBI. The same vignettes were presented to twenty-six ED physicians, seeking to explore discrepancies between their interpretations of the presentation with the MCs’ interpretations.

ED physicians identified a range of potential diagnostics for both scenarios (see Tables 2 and 3). MCs, however, were unable to identify much to apply an ICD-10-CM code in these narratives, with many requesting more information such as an impression or diagnosis from the provider. For vignette #1, while physicians assigned things like “post-traumatic headache,” “closed head injury,” or “head injury with secondary concussion,” MCs remained mostly within the bounds of language used in the narrative. In the absence of a physician-assigned diagnosis, MCs primarily assigned codes including headache and fall, or at most, nonspecific head injury (see Table 2). MCs noted during the interviews that there was little for them to use in their process:

I can’t give a diagnosis code…. I can code the fall part. I can code the symptoms that she came with. But definitely need more information to finish coding this one….. we need the doctor’s impression.

Similarly, vignette #2 demonstrated most ED physicians identifying “concussion,” (n=25 of 26) while the majority of MCs again coded “headache” in addition to “photosensitivity,” “fall,” and “dizziness.” Three-quarters of the MCs indicated that more information is needed (see Table 3). None of the physicians identified the fall as something that would potentially be part of the diagnosis or ICD-10-CM code.

The vignettes demonstrate that MCs code for information that is directly presented. They are neither trained nor encouraged to speculate regarding diagnostics, and require specific terminology and qualifiers (i.e., concussion without LOC) in order to accurately identify ICD-10-CM codes for encounters. Although some discussion about overreach, as mentioned previously, arose during the interviews, the vignettes demonstrate only minimal overreach, in a context where they were prompted and given an opportunity to do so.

MCs had suggestions to improve ICD coding for head injuries in the ED setting. These fell into three categories: (1) facilitating better documentation, (2) training and education, and (3) understanding the importance of accurate ICD-10-CM coding (See Table 4). Participants shared their awareness of the importance of specificity in their work and their reliance on documentation. One respondent described that, “the better documentation, the better it is for us to pick up the code that signifies the greatest specificity” (MC05). Suggestions included engaging provider champions to promote understanding of the importance of detailed documentation and better awareness of the type of information MCs need for head injury coding (e.g., LOC, location on head, trauma). Further suggestions included templates in EMRs that prompt healthcare providers to include complete documentation.

The amount of training that MCs receive and have access to varies widely. One participant shared that there are limited training opportunities available, especially on specific injuries. Additionally, MCs were aware that the data that they input are used for public health surveillance and research. They felt that better awareness among healthcare providers of the importance of this could help improve the documentation that they provide.

This evaluation has several potential limitations. First, participants were recruited through a convenience sample, and results may not represent the larger population of MCs in the United States. Specifically, diversity in career level may not have been adequately represented. Second, interviews could be subject to response bias, where the questions asked could have affected the study participants’ responses. Finally, assessment of ICD-10-CM coding by MCs was conducted with hypothetical vignettes that may not reflect the type of information that MCs receive in the course of their work.

In the absence of detailed information and due to time pressures to get claims processed, many MCs in this evaluation said they may default to an “unspecified” code. Some also noted that they may unintentionally use the wrong coding path (non-trauma versus trauma) or may attempt to interpret or make assumptions about a diagnosis based on available patient characteristics to inform their coding decisions. Inadequate access to training was indicated as one key contributing factor for unreliable or inconsistent coding practices.^25–27 Currently, training opportunities for MCs are limited or require resources (e.g., cost to participate), and when available, may be insufficient.^28,29 To our knowledge, there is currently no training available to MCs specific to coding head injuries. To this end, development of no or low-cost education and training opportunities for MCs is warranted to address current information gaps and improve coding practices in the ED for head injuries.