Time intervals to key diagnostic milestones were significantly longer in PSP/CBD compared to age- and sex-matched PD at each successive stage of the evolving diagnostic process from index symptom onset. In addition to the greater delay in identifying parkinsonism (which influences treatments and narrows differential diagnoses), it is notable that PSP/CBD appears not to be readily considered, even once the parkinsonian syndrome has been identified. The complexity of the diagnostic journey for individuals with PSP/CBD was also greater than for those with PD, with more differential diagnoses, primary diagnoses and changes to diagnosis overall. Individuals with PSP/CBD had more emergency admissions prior to diagnosis and a greater number of specialties involved in their diagnostic process with longer delays to a movement disorder specialist review, in particular, which was a key step in making a diagnosis.

Some previous studies in PSP/CBD report similar time delays [10, 11], some longer [12], and some shorter [5, 13]. None, however, are directly comparable due to differences in the precision and method of symptom onset definition, for example self-report versus first documentation in the medical record, or unclear definitions of symptom onset. In PD, the time to diagnosis is often rapid (9–15 months) [14–16], consistent with our study findings, though longer time intervals are reported [17–19]. Extending comparisons to other neurodegenerative diseases, the time intervals from symptom onset to diagnosis in PSP/CBD are also longer than those reported for MND [20–29], which range from 9.3 [20] to 16.2 months [29], and are most similar to diagnostic intervals reported in dementia, especially frontotemporal dementia [30–33]. Fewer studies have evaluated the frequency of changes to diagnosis in PSP and CBD. Of 4141 cognitively impaired patients visiting a dementia research centre approximately a year apart, 19 PSP patients received their diagnosis on both visits, four on their first visit only and eight their second visit only. In CBD, 45 received the diagnosis on both visits, 14 the first visit only, and 15 the second visit only [34].

In general, there has been little research into the determinants of diagnostic delay in PSP/CBD. In one previous study, there was a significant correlation between diagnostic latency and age of onset, with older patients having shorter diagnostic latency overall (r =  − 0.23, p < 0.001). The average age of our cohort was 78 years and so the time intervals reported may therefore be even worse in younger patients. In PD proposed determinants include sex (longer for men to present to primary care) [16], motor phenotype (longer in non-tremor dominant presentations) [16, 35], and greater motor and non-motor severity [35], though the association of greater disease severity with longer diagnostic delay would seem somewhat counterintuitive.

An initial appraisal of presenting symptoms/signs is a critical first step to systematically evaluate the greater diagnostic uncertainty and delay in PSP/CBD. The high prevalence of postural instability at symptom onset in PSP/CBD suggests it is underappreciated as a core feature of atypical parkinsonism. If realised earlier, this might narrow differential diagnoses and expedite movement disorder review. This is reflected in more recent changes to diagnostic criteria. Although postural instability, for example, is included in the UK PD Brain Bank criteria, it is not included in the 2015 Movement Disorder Society PD criteria as its presence in early disease should prompt consideration of alternative diagnoses, including PSP. Key identified categorical misdiagnoses, particularly vascular or primary psychiatric disorders, could be reduced by targeted history and examination. The higher frequency of TIA/stroke diagnoses in PSP/CBD (p = 0.012), for example, suggests that focal symptoms or signs in PSP/CBD are being identified but misattributed, perhaps due to a clinical reasoning process favouring common diseases. Careful assessment of the speed of symptom onset and symptom progression should broadly differentiate vascular and neurodegenerative pathologies. The greater number of neuropsychiatric diagnoses in PSP/CBD (p < 0.01), suggests training psychiatrists to perform targeted motor examinations in those with predominantly cognitive or behavioural presentations may reduce misdiagnosis in PSP/CBD or trigger movement disorder specialist referral, an approach previously suggested to differentiate DLB and AD [36]. Clinicians also need to be aware of the tendency to over-diagnose PD once parkinsonism is recognised, which was the case in over 50% of our PSP/CBD cases, similar to the extent of overdiagnosis in other studies [2, 37]. Atypical parkinsonian disorders were rarely included amongst differential diagnoses for identified parkinsonism, suggesting available red flags such as early falls are not being recognised. The high prevalence of tremor in PD may partially explain the shorter time to final diagnosis, being particularly amenable to clinician diagnostic pattern recognition. Tremor pattern recognition probably also shortens patient appraisal delay, removing potential barriers to presentation as patients may delay attending their GP with non-specific symptoms [38].

The transition from primary to secondary care is another key milestone in the diagnostic process. While there are no national referral recommendations for PSP/CBD in the UK, in PD national guidelines explicitly encourage urgent referral to movement disorder specialists to expedite diagnosis, determine prognosis, prevent distress arising from diagnostic delay, and allow early multidisciplinary involvement [39–41], a rationale which is clearly also applicable to PSP/CBD. PSP/CBD patients, however, take approximately 8 months from index symptom to be referred to secondary care, longer than those with PD. As a similar proportion (66.7%) of index symptoms in both PSP/CBD and PD are identified in primary care, the longer transition from primary to secondary care identified in PSP/CBD may, in part, reflect a greater reliance on “test of time” diagnostic strategies [42] to evaluate non-specific symptoms, as well as a greater difficulty in attributing non-specific symptoms/signs to PSP/CBD, or indeed parkinsonism. The differential diagnoses formulated to explain such symptoms at this stage are critical however as this influences the location, as well as timing, of subsequent referrals. No PSP/CBD secondary care referral, for example, queried a specific diagnosis of PSP/CBD, likely due to its rarity within primary care, in contrast to 63% of PD referrals, while parkinsonism was proposed in only a third of PSP/CBD patient referrals compared to 70% in the PD group. The greater tendency for GPs in PSP/CBD to misattribute identified symptoms to non-parkinsonian conditions is also likely an important determinant of subsequent diagnostic delay as this frequently resulted in a breadth of referrals to medical and surgical specialities. Our findings are in keeping with a study of referral patterns in pathologically confirmed cases of PSP and PD from the Queen Square Brain Bank [37]. The Step-Back PSP study, while primarily evaluating symptoms in the pre-diagnostic period (the period at least 1 year prior to diagnosis), also noted that PSP patients had more consultations to ENT (p = 0.028) and ophthalmology (p = 0.016) compared to age- and sex-matched PD [12].

For nearly a third of PSP/CBD patients, predominantly women, their first secondary care contact after symptom onset is not due to a GP referral but an emergency presentation, three times higher than in their age-sex matched PD counterparts. Given the high fracture rate in those presenting with falls, the reported female preponderance in this group may reflect a higher fracture risk due to lower bone mineral density in women. Initial emergency presentations in PSP/CBD, unlike PD, also do not reliably trigger subsequent outpatient referrals to relevant specialities, resulting in a higher frequency of recurrent emergency attendances prior to outpatient referral and review. Such emergency attendances (predominantly due to falls) result in a high rate of hospitalisation (over 90% of all attendances in both groups). The longer duration of hospitalisation in those with PSP/CBD compared to PD, suggests that the rehabilitation of PSP/CBD patients, prior to diagnosis, may already be slower than their age-sex matched PD counterparts. In addition, unscheduled admissions rarely resulted in inpatient diagnoses, a missed diagnostic opportunity. While patients with atypical parkinsonism have been previously identified as having a higher fracture risk in the 2 years preceding diagnosis [43], the current analysis suggests that, in addition to falls and fractures, recurrent emergency presentations associated with a lengthy rehabilitative process, may serve as additional diagnostic red flags for PSP/CBD.

PSP/CBD is also difficult to diagnose even once individuals have reached secondary care unless there is movement disorder expertise. PSP/CBD was diagnosed in only 11.1% of initial outpatient reviews in the PSP/CBD group compared to 60.0% diagnosed with PD in the PD group, while parkinsonism was diagnosed in 40% of PSP/CBD cases and two-thirds of PD cases after an initial outpatient neurology or COTE review. Although PSP/CBD patients first reached secondary care by approximately 8 months on average, it took individuals with PSP/CBD nearly 2 years from symptom onset to receive specialist movement disorder input, approximately three times longer than PD patients. This is shorter than the time reported in a series of 16 pathologically confirmed PSP cases where the time to first movement disorder clinic visit was 2.9 years [11]. The necessity of such input, particularly in PSP/CBD, is evident in that only one individual received a diagnosis of PSP/CBD prior to movement disorder review, in contrast to 24 (80.0%) PD cases who received their diagnosis prior to this input. In PD in contrast a previous review has shown that the accuracy of diagnosis in PD is not significantly lower in non-experts (73.8%) [44]. In addition, in a more recent retrospective survey of 1775 patients with PD, there was no difference in diagnostic certainty between a general neurologist or movement disorder neurologist [45].

Having reached a movement disorder specialist, approximately 50% of those with PSP/CBD were diagnosed at their first movement disorder assessment, reaching approximately 75% by their final review. Several studies have previously highlighted the improved diagnostic accuracy at final compared to initial review [2, 5, 37]. Whilst movement disorder specialists may be particularly adept at pattern recognition and have a greater familiarity with atypical features or rarer phenotypes [46], 25% of PSP/CBD did not receive their diagnosis in life, serving to highlight the persistent difficulties in diagnosis, even after movement disorder specialist input.

Unexpectedly, a novel finding was that there was little difference in the overall survival from carefully documented symptom onset to death in those with PSP/CBD and their age- and sex- matched PD counterparts. In PD, older age and male sex have been identified as independent worse prognostic factors. Due to 1:1 sex-matching in PD to match the PSP/CBS sex ratio, survival in PD in our sample has likely been extended as relatively more women were included in the PD group. However, conversely, in that the median age in PSP (79.9 years) was also older than the entire incident PD population in PINE (73.8 years), age matching may have decreased survival in our matched PD population. At the time of the study, the NINDS-PSP criteria were used for diagnosis which did not permit the diagnosis of non-Richardson’s syndrome subtypes of PSP. Given the PINE study was an incident cohort of parkinsonism, cortical PSP subtypes (which may have greater diagnostic delay and a poorer prognosis) are likely under-represented potentially extending survival in our PSP cohort. However, the proportion of cognitive subtypes are relatively small [47, 48] and unless cases had a purely cognitive presentation (therefore likely excluding the corticobasal presentations included within the cognitive subtype), they are likely to have been identified within the PINE study. Finally, unlike the lead-time bias seen in association with screening programmes, whereby an earlier diagnosis due to screening makes it appear as if individuals are surviving longer, the contribution of diagnostic delay should also be considered. Ultimately our findings relating to survival from index symptom require replication in a bigger incidence cohort (requiring multi-centre involvement), with access to primary care records to identify index symptoms.

A particular strength of this study is the inclusion of the entire recruited PSP/CBD population of a community based, prospective incident study of parkinsonism in the UK, meaning PSP/CBD cases are likely to be representative of the motor predominant PSP/CBD population. Additional strengths include the detailed systematic review of initial symptoms and signs from primary and secondary care records, the availability of detailed longitudinal data from research and clinical records, and the comparative analysis in PSP/CBD and PD matching on variables identified as possible determinants of diagnostic delay.

The study is not without limitations. PSP and CBD are distinct diseases with differentiating clinical features but were considered together for analysis due to small numbers (but akin to many studies in PSP/CBD due to their rarity). There is significant overlap between PSP and CBD clinically, but nonetheless with two CBD cases, our results mostly apply to those with PSP. The PINE study is an incidence study of parkinsonism so, while representative of the motor predominant PSP/CBD population, may under-represent those with a predominantly cognitive initial presentation. Primary and secondary care referrals of patients to a neurologist with a specialist interest in movement disorders were also actively sought as part of the PINE study, which may have reduced the time to movement disorder specialist and increased diagnostic accuracy compared to real-world practice in the UK. Unless explicitly documented, it is impossible to determine how frequently differential diagnoses or changes to primary diagnosis were communicated to patients, and the extent to which this impacted their perception and experience of their diagnostic journey. Pathological verification of all clinical diagnoses was not achieved (although our 30% post-mortem confirmation is good for a clinical study). This is never possible outside brain bank studies which, as detailed previously, are highly selected and susceptible to bias. During life clinicians rely on clinical diagnosis to guide management, and so even without pathological confirmation of all cases, studies on best clinical diagnosis are important. In addition, final clinical diagnostic accuracy in our study is also likely to be high due to, at minimum, annual reviews from diagnosis to death and a full case note review after death. While, without post-mortem data in all cases, we may have missed some with atypical PSP, the effect of this potential bias would tend to make “true” delays in all cases of PSP/CBD even longer without better diagnostic tests. The PINE study also preceded the updated MDS diagnostic criteria for PSP [49]. While this is unlikely to have altered diagnosis in most cases and there is some evidence that clinical expert diagnosis is better than rigid research criteria [50], it has prevented subtyping of PSP cases in our cohort which may have masked potential differences in diagnostic delay and survival between subtypes. Implementation of relatively complex diagnostic criteria however takes time and so our study will likely reflect the day-to-day clinical diagnostic process and experience of PSP/CBD patients in the UK.

In conclusion, individuals with PSP/CBD had a longer and more complex diagnostic pathway than individuals with age-sex matched PD. There are several ways this could be improved including greater recognition that balance issues may be parkinsonian, raised awareness of red flags for PSP including emergency admissions preceding diagnosis for unexplained falls/fractures and earlier access to movement disorder specialists. We found little difference in overall survival from symptom onset to death in PSP/CBD and age-sex matched PD. If replicated, this has important implications for our understanding of prognosis in PSP/CBD.