Understanding headache classification coding within the veterans health administration using ICD-9-CM and ICD-10-CM in fiscal years 2014–2017

Objectives Understand the continuity and changes in headache not-otherwise-specified (NOS), migraine, and post-traumatic headache (PTH) diagnoses after the transition from ICD-9-CM to ICD-10-CM in the Veterans Health Administration (VHA). Background Headache is one of the most commonly diagnosed chronic conditions managed within primary and specialty care clinics. The VHA transitioned from ICD-9-CM to ICD-10-CM on October-1-2015. The effect transitioning on coding of specific headache diagnoses is unknown. Accuracy of headache diagnosis is important since different headache types respond to different treatments. Methods We mapped headache diagnoses from ICD-9-CM (FY 2014/2015) onto ICD-10-CM (FY 2016/2017) and computed coding proportions two years before/after the transition in VHA. We used queries to determine the change in transition pathways. We report the odds of ICD-10-CM coding associated with ICD-9-CM controlling for provider type, and patient age, sex, and race/ethnicity. Results Only 37%, 58% and 34% of patients with ICD-9-CM coding of NOS, migraine, and PTH respectively had an ICD-10-CM headache diagnosis. Of those with an ICD-10-CM diagnosis, 73–79% had a single headache diagnosis. The odds ratios for receiving the same code in both ICD-9-CM and ICD-10-CM after adjustment for ICD-9-CM and ICD-10-CM headache comorbidities and sociodemographic factors were high (range 6–26) and statistically significant. Specifically, 75% of patients with headache NOS had received one headache diagnoses (Adjusted headache NOS-ICD-9-CM OR for headache NOS-ICD-10-CM = 6.1, 95% CI 5.89–6.32. 79% of migraineurs had one headache diagnoses, mostly migraine (Adjusted migraine-ICD-9-CM OR for migraine-ICD-10-CM = 26.43, 95% CI 25.51–27.38). The same held true for PTH (Adjusted PTH-ICD-9-CM OR for PTH-ICD-10-CM = 22.92, 95% CI: 18.97–27.68). These strong associations remained after adjustment for specialist care in ICD-10-CM follow-up period. Discussion The majority of people with ICD-9-CM headache diagnoses did not have an ICD-10-CM headache diagnosis. However, a given diagnosis in ICD-9-CM by a primary care provider (PCP) was significantly predictive of its assignment in ICD-10-CM as was seeing either a neurologist or physiatrist (compared to a generalist) for an ICD-10-CM headache diagnosis. Conclusion When a veteran had a specific diagnosis in ICD-9-CM, the odds of being coded with the same diagnosis in ICD-10-CM were significantly higher. Specialist visit during the ICD-10-CM period was independently associated with all three ICD-10-CM headaches.


Background
Headache is one of the most commonly diagnosed chronic conditions managed within primary and specialty care clinics. The VHA transitioned from ICD-9-CM to ICD-10-CM on October-1-2015. The effect transitioning on coding of specific headache diagnoses is unknown. Accuracy of headache diagnosis is important since different headache types respond to different treatments.

Methods
We mapped headache diagnoses from ICD-9-CM (FY 2014/2015) onto ICD-10-CM (FY 2016/2017) and computed coding proportions two years before/after the transition in VHA. We used queries to determine the change in transition pathways. We report the odds of

Introduction
Headache is one of the most commonly diagnosed chronic conditions managed within primary and specialty care clinics [1,2]. Headache has a lifetime prevalence of 66% [2][3][4][5][6][7][8][9]. Half of people with a headache history actively experience headache attacks [8,10]. Understanding the public health impact of headache and the treatment needs of people with headache requires accurate data, especially when reporting disease prevalence. The Veterans Health Administration (VHA) is the largest integrated healthcare system within the United States (U.S.) and provides electronic health record (EHR) data over time for veteran enrollees [11]. The VHA transitioned from the International Classification of Diseases and Related Health Problems, 9 th revision, Clinical Modification (ICD-9-CM) to the 10 th revision on October 1, 2015. From this point onward, each ICD-9-CM diagnosis required manual recoding as an ICD-10-CM diagnosis in the EHR by healthcare provides managing headache.
The new ICD-10-CM codes were expected to increase diagnostic specificity, hence aiding healthcare providers, payers, and policymakers in understanding disease prevalence, establishing appropriate reimbursement rates, and improving care quality and delivery. While resources exist to map ICD-9-CM conditions onto ICD-10-CM, [12,13] the effect of this transition on coding of specific headache diagnoses is unknown. Accuracy of headache diagnosis is important since different headache types respond to different treatments, and newer, more expensive migraine-specific treatments, are emerging rapidly and may not be available to patients unless they are accurately diagnosed and coded as having migraine.
While coding of healthcare conditions such is guided by the ICD, diagnosis of headache conditions is guided by the International Classification of Headache Disorders (ICHD). This classification schema initially divides headache diseases into primary (e.g., migraine) and secondary (e.g., post-traumatic headache [PTH] headache. Primary headaches are those for which there is no known underlying etiology, whereas secondary headaches are attributable to an underlying condition or conditions thought to cause or have close temporal relationship to when a headache condition began [14,15]. The first edition of the ICHD criteria was published in 1988 whereas its fourth edition is currently being developed [15,16]. Headache classification and criteria have evolved over time, incorporating medicine's growing understanding of headache conditions and the utility of previously used criteria in guiding clinicians in making distinct headache diagnoses [15,17]. For example, when considering patients with migraine with typical aura, the ICHD-3 beta diagnostic criteria had a false positive rate of 16.9%, whereas its predecessor, the ICHD-3, had a false positive rate of 10.5%, as compared to the diagnosis made by treating physicians [17]. It should also be noted that, while there is no diagnostic category in ICHD for "headache," ICD does contain a "headache" symptom code. Clinicians were using the ICHD-3 beta criteria to diagnose headache diseases during the time when the VHA transitioned from ICD-9-CM to ICD-10-CM, clinicians were using the ICHD-3 beta criteria, which was introduced in 2013 and supplanted by ICHD-3 in 2018. Previous work mapping any headache diagnosis with the transition from ICD-9-CM to ICD-10-CM noted a 0.75 odds ratio of being coded with a headache disorder in ICD-10-CM if an ICD-9-CM code existed for any type of headache [3,18]. Little if any work has examined diagnostic trajectories of specific headache conditions with the transition from ICD-9-CM to ICD-10-CM. After mapping ICD-9-CM headache diagnoses onto the ICD-10-CM system, we sought to understand the continuity and specificity of coding specific headache subtypes. We also explored whether provider type (primary care versus specialty care) and/or patient characteristics (age, sex, race/ethnicity) influenced ICD-10-CM coding.

Study design
We measured headache coding proportions two years before and two years after the transition to ICD-10-CM to examine changes in headache classifications unadjusted by patient or characteristics. All patients included in the analyses had at least one VHA visit during the first two years of ICD-10-CM utilization (FY2016/2017) to allow for an opportunity for a headache condition to be coded. We utilized eight categories of headache present in both ICD-9-CM and ICD-10-CM coding systems: migraine, tension-type headache, trigeminal-autonomiccephalalgias (TAC), other primary headache disorders, PTH, post-whiplash headache, other secondary headache disorders, and headache NOS.

Mapping headache diagnoses from ICD-9-CM onto ICD-10-CM
The ICD-10-CM is a system used by healthcare providers to classify and code all diagnoses, symptoms and procedures recorded in conjunction with health care in the US. An ICD-10-CM is a seven-character, alphanumeric code. Each code begins with a letter, and that letter is followed by two numbers. The first three characters of ICD-10-CM are the "category." The category describes the general type of the injury or disease. The category is followed by a decimal point and the subcategory. For example, G43 is "migraine," whereas G43.1XX is "migraine with aura." To develop corresponding ICD-10-CM diagnoses for the headache conditions above, we cross-walked the original ICD-9-CM codes for each condition coded in in FY 2014-2017 to ICD-10-CM codes using general equivalency mapping (GEMs). As such, the period in this study occurred exclusively during a time after ICHD-3 beta was published and before ICHD-3 was published (i.e., 2013 to 2018). Publicly available codes from the Centers for Disease Control and the Centers for Medicare and Medicaid Services [12] were used and crossreferenced with guidance from the American Academy of Neurology (AAN) [19] and the American Headache Society (AHS) [20]. The code mappings in Table 1 were independently reviewed by four neurologists specializing in headache care-two United Council for Neurologic Subspecialties (UCNS) Headache Medicine Certified and two UCNS Headache Medicine eligible neurologists. This crosswalk was applied to the Women Veterans Cohort Study (WVCS) of 1.14 million OEF/OIF/OND veterans with and without headache in order to update the WVCS cohort with the full complement of headache diagnoses to be examined in the VHA Headache Centers of Excellence (HCoE) cohort [11].

PLOS ONE
Understanding headache classification coding using ICD-9-CM and ICD-10-CM Patients were identified as (1) no headache coded at all (e.g., headache NOS in ICD-9-CM and no headache diagnosis in ICD-10-CM); (2) the same headache disorder coded (e.g., headache NOS in both ICD-9-CM and ICD-10-CM); or (3) a new headache diagnosis or more than one new headache diagnoses coded instead of the original headache type (e.g., headache NOS in ICD-9-CM followed by migraine in ICD-10-CM). Second, we investigated changes in ICD-9-CM/ICD-10-CM coding relative to age, sex, and race/ethnicity. Third, we examined ICD-10-CM coding by different types of providers, comparing utilization by generalists (i.e., primary care, emergency department and urgent care) and specialists (i.e., neurologists and physiatrists). Provider type was based on stop codes that indicate clinic type as shown in S1 Table. Multivariate modeling of ICD-10-CM headache coding Multivariate modeling was used to quantify the likelihood of receiving a given diagnosis of headache NOS, migraine, or PTH in ICD-10-CM when the original ICD-9-CM diagnosis of headache NOS, migraine, or PTH was made by a PCP, controlling for: ICD-9-CM headache diagnoses, ICD-10-CM headache diagnoses, and sociodemographic factors. Persons diagnosed in ICD-9-CM by other provider types were not included in these analyses, given that: (1) primary care providers manage a majority of patients with headache, and; (2) we sought to understand whether headache coding changed when a specialist was involved after the transition to ICD-10-CM. Adjusted odds ratios (aORs) and associated 95% Wald confidence intervals (CIs) were generated using logistic regression models containing the following variables: headache NOS, migraine, and PTH in ICD-9-CM and ICD-10-CM, followed by patient age, sex, and race/ethnicity. Logistic regression was used to determine the odds of a specific headache diagnosis in ICD-10-CM being associated with a specific ICD-9-CM headache diagnosis made by a primary care provider, controlling for other headache morbidity, ICD-10-CM diagnoses made by either a generalist or specialist, age, sex, and race/ethnicity. A three-level provider type variable for the ICD-10-CM visit(s) was created. Generalists consisted of visits to either primary care, urgent care, or emergency care whereas specialists were classified as either neurologists or physiatrists. Diagnostic codes for neurology visits (6.74%) were coded first, regardless of whoever else saw the veteran. Physiatry clinic visits (3.51%) were coded next and were assigned if the patient was not seen in a neurology clinic and regardless if they were also seen by a generalist. The physician of record would be a generalist only if the patient was not seen by either neurology or physiatry services (89.75%). SQL server version 2014 was used to run queries and Python version 3.7.4 were used to generate Figures. Logistic regression was performed using SAS, version 9.4 (SAS Institute, Cary, NC).

Results
We report herein on the presence/absence of headache diagnoses as well as specific diagnostic patterns in ICD-10-CM.

Headache diagnoses & specifics in ICD-10-CM (FY2016-FY2017)
Headache NOS. In considering the pathways of headache from FY2014/2015 into FY 2016/2017 for patients with an ICD-9-CM diagnosis of headache NOS (Fig 1, Panel A), 63.2% (29,424/46,549) of patients did not receive any ICD-10-CM headache diagnostic code. Of the headache NOS patients (N = 17,307) who were coded with an ICD-10-CM headache diagnosis in FY2016 and FY2017, 75% (N = 12,933/17,307) had one headache diagnosis within any of the eight headache categories; two-thirds of these patients (66%, 8556/12,993) were recoded as headache NOS in ICD-10-CM, while the remaining patients received a new, different
Changes in ICD-9-CM and ICD-10-CM headache coding across age, sex, and race/ethnicity. In Table 4 we present the changes in headache coding for the different Table 3. Probability modeled is ICD-10-CM headache NOS, migraine, and post-traumatic headache, with and without modeling specialist/generalist.

Headache
Headache NOS Migraine PTH

Without Specialty With Specialty Without Specialty With Specialty Without Specialty With Specialty
Characteristic aOR (95% CI)
Changes in ICD-9-CM and ICD-10-CM coding when specialists involved in headache diagnosis. In the first two years of ICD-10-CM implementation, relatively few people who had seen a PCP in the last two years of ICD-9-CM saw a specialist (10.25%). When considering the involvement of specialists in recording ICD-10-CM headache diagnoses, for predicting headache NOS in ICD-10-CM, all headache morbidity was significantly and positively associated with the probability of having headache NOS diagnosed. The largest odds ratio was for headache NOS in ICD-9-CM (aOR = 5.71, 95% CI: 5.50-5.92; Table 3). The type of provider seen in the ICD-10-CM two-year window was also significantly independently associated with the probability of a headache NOS diagnosis ( . In all three outcome models, the addition of the threelevel specialist variable reduced the odds ratios for the other two headache comorbidities while remaining independently associated with the probability of diagnosis of each headache outcome. The addition of the specialist variables to the ICD-10-CM PTH model resulted in Hispanic veterans having significantly reduced odds of getting a PTH diagnosis.

Discussion
Using the last two years of ICD-9-CM coding and first two years of ICD-10-CM coding in VHA, we mapped the same patients with an ICD-9-CM headache diagnosis onto the ICD-10-CM system. We also sought to understand the contribution of age, sex, race/ethnicity, and provider type in influencing headache coding during the earliest part of the transition to ICD-10-CM. In considering headache NOS, migraine, and PTH, several findings are noteworthy. First, while patients received an ICD-9-CM diagnosis headache diagnosis, a majority of patients did not receive any ICD-10-CM code for headache despite having at least one office visit with a provider who could have coded an ICD-10-CM diagnosis within the EHR. Reasons for the lack of coding headache diagnoses in ICD-10-CM include: the headache condition may have resolved, the headache was stable and hence not an "active issue," or may not have been the focus of the office visit. Not coding patients with headache diagnoses with the transition to ICD-10-CM has similarly been reported with VHA data, such that patients with any headache diagnosis prior to the transition had a 0.75 odds of receiving any headache diagnosis in the same ICD-10-CM time period examined here [21].
Second, when patients did receive an ICD-10-CM code, most patients continued receiving a single code which corresponded to their ICD-9-CM headache code. Among ICD-9-CM headache NOS patients, nearly half continued with a headache NOS during ICD-10-CM; 20% were coded as migraine in ICD-10-CM. When an additional diagnosis was given to ICD-10-CM headache NOS, it also tended to be migraine. As such, most patients with ICD-9-CM headache NOS diagnosis retained the least specific headache diagnosis possible or were either recoded as having a more specific headache diagnosis. More specific ICD-10-CM headache diagnosis may have resulted from a more complete headache evaluation during or after the transition of coding schemas, a "clean slate" of the patient's health concerns ("problem list") may have compelled the re-evaluation of headache history and diagnosis, and the new coding schema provided more diagnostic choices. Also, providers had additional time using ICHD-3 beta headache criteria in the first two years of the transition compared to the two prior years, which may also account for increased use of migraine diagnostic codes. Additional reasons for the recoding of ICD-9-CM headache NOS to ICD-10-CM migraine may be more intrinsic to the differences of these two-coding schema. For example, ICD-9-CM code of 346.20 ("variants of migraine, not elsewhere classified, without mention of intractable migraine without mention of status migrainosus") has five corresponding ICD-10-CM codes, including G43.809 ("other migraine, not intractable, without status migrainosus") and G43.D0 ("ophthalmoplegic migraine, not intractable"). However, for our analysis, we grouped all migraine ICD codes together; hence, the improved specificity of ICD-10-CM may be more applicable to individual types of migraine rather than migraine as a whole. Generalists and specialists both continued to use a headache NOS diagnosis, despite there being clear guidance in the International Classification of Headache Disorders 3 rd edition (ICHD-3) regarding the criteria for each headache type [9]. When ICD-9-CM diagnosed migraine patients received an ICD-10-CM code, it was overwhelmingly a code for migraine, either alone or in conjunction with a second code (most commonly headache NOS). Migraine may be the most familiar as well as the best understood of the headache disorders, given the high rates in which patients present to Primary and Specialty Care clinics for headache care [3,7]. Another reason for the potential persistence of migraine diagnosis coding is that migraine is the only headache type allowed as a service-connected disability within VHA. ICD-9-CM diagnosed PTH showed the most changes in coding during ICD-10-CM with a greater percentage of these patients being coded with ICD-10-CM headache NOS and migraine or migraine alone. The migraine phenotype of PTH is the most commonly encountered, and may explain why these patients received an ICD-10-CM migraine diagnosis [22,23].
Third, age, sex and race/ethnicity were independently associated with ICD-10-CM diagnoses adjusting for the three select headache diagnoses in both ICD-9-CM and ICD-10-CM. The decrease in all three headache types with increasing age and marked gender differences have been seen previously. The associations of race/ethnicity and headache coding across the three select headache diagnoses are interesting in the context of veterans being seen in the integrated healthcare system as well as adjusting for specialist visits. Black and Hispanic veterans were significantly less likely to receive an ICD-10-CM PTH diagnosis after adjusting for their ICD-9-CM diagnoses and ICD-10-CM comorbidity as well as age, sex, and specialist visits. These findings match what has been reported in the literature regarding disparities in headache care and may be explained by social constructs such as race and socioeconomic status [24].
Fourth, among those that did receive an ICD-10-CM code, only 10.25% saw a specialist for headache. In community-dwelling U.S. patients with at least one office-based visit, 10.93% of those with a self-reported history of headache saw a neurologist [25]. There is a marked shortage of neurologists and headache medicine providers to see patients with headache, especially within high-income countries [26,27]. A similar shortage of neurologists, and particularly, headache neurologists, exists within the VHA. This has contributed to Congressional and VHA commitment to expanding access to headache medicine providers through the established of a national VHA Headache Centers of Excellence program.
Lastly, when patients saw providers with additional training and expertise in headache diagnosis and management, their odds of being diagnosed and coded with any of the three headache types of interest increased, compared to when only a generalist was involved. These findings highlight the importance of headache specialists in diagnosing and coding headache conditions; however, we did observe that even among specialists, patients continue to be coded with the least specific headache diagnosis. Adjustment in the models for specialist visits did not appreciably change the aORs for sociodemographic factors, suggesting these did not confound their observed associations with ICD-10-CM diagnoses.
Limitations to our work are worth noting. Although our study is a mapping study, it is considered a "transitioning study" due to its longitudinal study design (ICD-9-CM at FY 2014/ 2015 and ICD-10-CM at FY 2016/2017) and not cross-sectional study where both coding systems are used simultaneously. Therefore, the results are interpreted differently. Medicallydiagnosed headache disorders and have not been validated by chart review or structured patient interviews. Next, it is possible that ICD-9-CM headache diagnoses were updated after the two-year period of observation that we selected. However, a longer period after the beginning of the transition could have allowed additional time for headaches to either change (e.g., to no longer become an "active issue") or for patients to develop a new type of headache with or without an interceding event. Among patients who received two headache diagnoses in ICD-10-CM, we cannot comment if they developed a new headache disorder or if it was the same headache condition being coded differently. Again, keeping the examination period short hopefully minimized the appearance of new headache types in the interim between the ICD-9-CM and ICD-10-CM diagnoses. Next, coding may not be synonymous with diagnosis. Providers may have documented in the EHR a diagnosis of migraine and coded them as headache not otherwise specified. Formal chart review to determine the agreement between ICD-10-CM headache codes and unstructured clinic note data would be required to understand if this is a concern. Also, we considered a diagnosis by a specialist as superseding that of a generalist. While this hypothesis is supported by the additional headache training specialists receive, we are not aware studies comparing diagnostic and coding accuracy across provider types. Finally, whether greater degrees of headache severity, disability, or frequency may have impacted diagnoses which were coded was not explored, as these are not routinely available in administrative data.

Conclusion
In the first two years where the largest integrated healthcare system in the U.S. transitioned to a coding schema intended to promote diagnostic specificity, we demonstrate both promise and opportunities for improvement and future inquiry. While headache NOS continued to be used commonly by generalists and specialists, there was a high rate of reclassification from headache NOS to another more specific headache diagnosis, chief among them being migraine, even after controlling for age, gender, and race/ethnicity. It is puzzling that patients with ICD-9-CM migraine or PTH diagnoses sometimes received a less specific diagnosis of headache NOS; this finding deserves further exploration to determine if there are any diagnostic challenges or systematic biases in the coding of this important headache diagnosis. Future work will also explore the impact of the ICD-9-CM/ICD-10-CM transition on other headache diagnoses as well as examined whether receipt of a more specific headache diagnosis or diagnoses changes healthcare utilization and headache care quality and delivery.
Supporting information S1 Table. Clinical visits stop codes assigned to primary care providers, neurologists, and physiatrists.