Of the 1,468 patients visiting our tertiary headache center for the first time, 69.5% had previously received a headache diagnosis at the PSLC. At our center, 99.5% were diagnosed with at least one headache diagnosis during their initial visit. Migraine was underdiagnosed at the PSLC, while TTH was overdiagnosed. The coefficient of determination (R²) for migraine, TTH, and CH at the PSLC fell below the acceptable 80%, suggesting insufficient adherence to ICHD-3 guidelines in the PSLC setting. This highlights the need for improving diagnostic accuracy in primary and secondary headache care.

The global initiative against Headache, ‘Lifting The Burden,’ introduced the current framework of three levels of headache care [21]. The majority of people with headaches are diagnosed and treated at the primary and secondary level of care and only a fraction of those patients require tertiary headache care in a 90:9:1% split [22]. Therefore, it is essential that headaches are correctly recognized, diagnosed, and managed in the PSLC [23]. Accurate diagnosis ensures that the patient receives the appropriate treatment, such as pharmacological therapy or lifestyle modifications, which can significantly reduce the frequency and severity of headaches over time [24, 25]. Additionally, misdiagnosis or delayed diagnosis may lead to ineffective treatments, unnecessary investigations, and prolonged patient suffering [24, 25]. Therefore, diagnostic accuracy is essential for long-term management, outcomes, and prognosis.

In this study, 67% of patients presenting at a tertiary headache center received the same diagnosis they had previously been given at lower levels of care. While this indicates a substantial level of agreement, it also reveals that approximately one-third of these patients either received a new diagnosis or were diagnosed for the first time solely at the tertiary care level. This points to discrepancies or missed diagnoses in the primary and secondary care settings [26,27,28], emphasizing the need for improved diagnostic accuracy.

Migraine was the predominant diagnosis in our cohort at both the PSLC and TLC, but a notable discrepancy existed between the two levels. At PSLC, the likelihood of a migraine diagnosis when meeting ICHD-3 criteria was substantially lower compared to TLC, indicating underdiagnosis at PSLC. In addition, more than 60% of patients without a prior PSLC diagnosis met all ICHD-3 criteria for migraine, with about half reporting aura, which was not correctly identified or considered when making a diagnosis.

The issue of underdiagnosis of migraine is well-documented in previous studies. A large multi-center study across seven countries found correct migraine diagnoses by general practitioners in only 28% of cases, with similar trends observed at secondary care levels [4]. Studies from Russia, Italy, and China reported accurate prior diagnosis rates from 12 to 27% [29,30,31].

Comparing our findings with those previous reports, it seems that German physicians in the PSLC demonstrate above-average accuracy in diagnosing migraine. In line with this observation, a 2012 German study reported that 63% of individuals meeting the ICHD-2 criteria for migraine and consulting a physician were correctly diagnosed with migraine [17]. However, significant room for improvement remains, with minimal progress since 2012. Enhanced diagnostic strategies and awareness campaigns are needed, with some authors suggesting that most patients with disabling, episodic headaches should be diagnosed with migraine by default at the primary level [32].

Regarding specific migraine symptoms, a comprehensive review identified nausea, pain exacerbation by physical activity, and photophobia as the most sensitive clinical features for diagnosing migraine [33]. Rai et al. identified non-throbbing pain, non-temporal pain, and the first physician not being a neurologist as predictor factors for inappropriate migraine diagnosis [34]. Radtke and Neuhauser associated nausea/vomiting, photophobia/phonophobia, unilateral pain, and severe headache with recognition of migraine at PSLC in Germany [17]. Our findings align partially, indicating that moderate to severe pain, photophobia, nausea, vomiting, or aura increased correct diagnosis likelihood, while unilateral pain, throbbing pain, and phonophobia did not. While not all diagnostic features are mandatory for diagnosing migraine [10, 35], awareness of all associated symptoms may be beneficial for cases that do not meet al.l diagnostic criteria.

TTH appears to be often misdiagnosed in the PSLC. Among patients initially diagnosed with TTH in the PSLC, 13% should have been diagnosed with migraine instead and 69% with migraine alone according to the ICHD-3 criteria. Previous studies highlight this misdiagnosis tendency. The Spectrum study found 32% of TTH diagnoses were actually migraine [36]. In the TEDDI study, only 2.4% of patients diagnosed with TTH in the emergency department met all criteria and 7.6% met all but one criteria for TTH [37]. These findings underscore the importance of thorough evaluation and adherence to standardized diagnostic criteria to prevent misdiagnosis and ensure appropriate management of headache disorders, particularly in distinguishing between tension-type headache and migraine.

Regarding CH, studies in Italy, East Europe, and Spain reported high rates of misdiagnosis at initial consultations, with only a minority of CH cases correctly identified [38,39,40]. Interestingly, German physicians in PSLC seem to exhibit higher diagnostic accuracy, with a smaller proportion of CH cases misdiagnosed in our cohort. Nonetheless, there remains a need for improvement, as one out of five CH patients was still missed in PSLC.

The Aids to Management of Headache Disorders in Primary Care proposed that CH should be diagnosed in PSLC and managed in TLC [41]. Despite the recommendation that CH should “never be missed” in PSLC, it appears that the goal of a correct diagnosis at PSLC has not yet been fully achieved. Thus, there is still work to be done in terms of improving both physician education in PSLC and raising awareness among the general population [4, 39].

Though to a much lesser extent, misdiagnosis of CH has also been previously described, particularly among men with migraine [42]. In our cohort, ipsilateral autonomic symptoms were linked to CH misdiagnosis. While these symptoms are typical of CH, they can also appear in migraine and other headache disorders. For example, Karsan et al. reported cranial autonomic symptoms in 74% of migraine patients [43], and Togha et al. found a prevalence of 61% [44].

Another factor contributing to the under-recognition of primary headache disorders in PSLC could stem from the design of the International Classification of Headache Disorders (ICHD), which prioritizes specificity over sensitivity, potentially leading to under-diagnosis [24]. Consequently, the guidelines should be viewed as guidance rather than strict rules. It is important to acknowledge that some patients may not fully meet al.l criteria for a specific headache disorder, or criteria fulfillment may be unclear or vary between attacks [45].

Our findings may not be generalizable to all patients seen at PSLC, as this study focuses on those with potentially more complex diagnostic and treatment challenges. Furthermore, these results should not be viewed as a criticism of the diagnostic accuracy of PSLC physicians. Rather, the intent is to identify areas for potential improvement in diagnostic practices, with the aim of enhancing overall care quality and reducing unnecessary diagnostic delays.

The novelty of this study lies in its systematic evaluation of diagnostic practices for three major headache disorders across different levels of care in a large German population. In addition, we were able to identify key factors contributing to potential diagnostic errors. Our findings contribute to the broader scientific understanding by confirming and expanding upon results from similar studies conducted in other countries or with different designs. These insights have the potential to enhance the accuracy of headache diagnoses, improve treatment outcomes, optimize healthcare resource allocation, and promote a more standardized approach to headache care across all levels of the healthcare system.

To enhance diagnostic consistency and accuracy in primary and secondary care settings, ongoing training and education are essential and have been shown to be effective [11]. This could involve expanding headache education during medical training, offering e-learning modules, and providing regular headache training programs for physicians of the PSLC [25]. The ICHD-3 diagnostic criteria can be challenging for non-neurologists, and the large number of distinct diagnoses may be overwhelming for daily use in PSLC settings. Therefore, developing a user-friendly format for clinical use might be beneficial and could help physicians to be more confident in making accurate diagnoses and reduce unnecessary referrals [33]. Diagnostic tools such as validated and standardized questionnaires or AI-assisted diagnostic support might be supportive in this. Additionally, it remains crucial to monitor diagnostic consistency across care levels and share the results with providers as part of a continuous improvement strategy. Research should also focus on identifying areas of disagreement in diagnoses between care levels to address these issues systematically [5].

A key strength of this study is its large sample size, allowing robust conclusions even for less common headache disorders. We thoroughly cross-referenced cases with discrepancies between TLC diagnoses and ICHD-3 criteria to mitigate potential errors.

However, some limitations exist. Reliance on patient-reported diagnoses from PSLC may introduce bias and under-documentation, potentially overestimating our results. The lack of detailed data regarding the reasons for patient referrals to our tertiary center may have introduced selection bias by disproportionately including cases with greater diagnostic complexity or uncertainty. However, identifying why those patients failed to get diagnosed in the PSLC was also part of this study’s aim. Doctor’s letters may also introduce reporting bias, as not all symptoms or their absence are consistently documented, despite a standardized template. This makes it challenging to determine whether symptoms were omitted during assessment or genuinely absent. A similar concern applies to the possibility of prior visits to another TLC. However, given that the nearest alternative center is ≈ 200 km away, such instances are very rare and unlikely to have a significant impact on our results. Additionally, the occurrence of multiple headache types described by patients could bias our prediction models. We minimized this potential bias by including only patients meeting none or only one of the ICHD-3 diagnoses of migraine, TTH, and CH.