Most degenerative cervical myelopathy remains undiagnosed, particularly amongst the elderly: modelling the prevalence of degenerative cervical myelopathy in the United Kingdom

In this study, we provided the first estimates of the age-stratified prevalence of DCM using two data sources—UK hospital admissions (HES) data, and published spinal cord compression (SCC) prevalence data. There was a mismatch in estimates using these two measures, with estimates from HES data lower than those from SCC data in all age groups other than 50–59 years (Table 2, Fig. 3). This mismatch is particularly pronounced in the oldest age group, > 79 years.

A mismatch exists between prevalence estimates

A number of explanations could account for the generalised mismatch between prevalence estimates.

First, SCC-derived DCM prevalence may represent an overestimation of true prevalence. This could occur if the conversion rate from SCC to DCM used in this study is greater than the rate in the UK population. However, this conversion rate is derived from a high-quality meta-analysis [10], which includes data from ten different countries across three continents, making the conversion rate an unlikely source of inaccuracy. Overestimation could also occur if the SCC prevalence used in this study, measured in a Japanese population, is higher than the SCC prevalence in the UK population on which the hospital data are based. However, the same meta-analysis found that the prevalence of SCC was in fact higher in American/European populations than Asian populations [10], suggesting that the SCC-derived prevalence may in fact be an underestimate. Hence, overestimation of the SCC-derived prevalence is unlikely to explain the observed mismatch.

Second, hospital-derived prevalence could represent an underestimation of true prevalence. This might have several causes. For instance, ICD-10 coding has a poor sensitivity for DCM [27]—hence, a number of DCM cases are likely to have been missed. However, the ICD-10 codes used in this study are considered to provide the best possible balance of sensitivity against specificity [15]. Inaccuracy in the hospital-derived prevalence estimate could also arise from the inherent assumptions of steady-state populations and average duration of disease when extrapolating from incidence to prevalence, although this is standard methodology for calculating prevalence [22, 28]. Hospital-derived prevalence may actually be an overestimate, since the calculations are based on FCEs rather than admissions, and a single admission may have more than one FCE.

Another potential cause for underestimation in the hospital-derived prevalence is case ascertainment bias. These prevalence estimates are based solely on hospital admissions primarily due to DCM, for instance, for surgical management or in those with severe and emergent symptoms. This is likely a specific but poorly sensitive metric to use in estimating population prevalence, although it may be of benefit to those involved in the planning and delivery of surgical services. By contrast, prevalence derived from studies which recruit healthy volunteers for an MRI scan to detect SCC and neurological exam to detect DCM [12] yielded, as expected, higher estimates of DCM prevalence.

One interpretation of this mismatch is therefore that this latter metric is capturing DCM patients with a milder phenotype, or who are perhaps not undergoing or being offered surgical treatment of their DCM and hence would only be captured using community-based metrics [29]. If our estimate of prevalence using HES data is primarily reflecting surgical management of DCM, then this may also explain the differing trends of prevalence by age (Fig. 3) as rates of surgery are higher in those who are still of working age [30], despite older DCM patients having more severe disease [14].

However, case ascertainment bias due to our inpatient-only analysis is unlikely to fully explain the mismatch—preliminary analysis of outpatient data showed that the mean incidence of outpatient attendances for DCM was just over a tenth the incidence of inpatient FCEs. Hence, even adding outpatient attendances, the mismatch remains. Previous estimation of DCM prevalence using primary care data has also yielded an overall prevalence of 0.04% [29], a fifth of our estimate. Although the HES database does not include primary care data, the diagnosis of DCM requires secondary care assessment [5], so this is unlikely to be a substantial source of case ascertainment bias.

Having explored the above explanations, we are left with the possibility that the prevalence mismatch represents underdiagnosis across the population. Awareness of DCM amongst both patients and healthcare professionals is also poor [31], which may lead to delayed diagnosis [6]. This interpretation is also supported by our finding of a widening diagnostic gap with age, discussed below.

Hospitalised DCM prevalence declines in older age groups

In addition to the generalised mismatch between prevalence estimates, we found differing trends in prevalence in older age groups. SCC-derived estimates showed increasing prevalence in the oldest age groups, whilst HES-derived estimates showed decreasing prevalence (Table 2, Fig. 3). This potentially separate phenomenon also has several possible explanations.

First, older patients may be less likely to be hospitalised for DCM, and instead may be living with DCM and thus absent from our estimates using these metrics. Hospital admissions for a different problem, where the patient also has DCM, would not be included in the observed prevalence values, since these are based on primary hospital diagnosis. Reduced hospital admissions for DCM in older age groups could represent a health inequality—older DCM patients may be considered ‘not suitable for surgery’ due to the presence of comorbidities or frailty. However, such an inequality would be against management guidance for DCM [7], and hence is unlikely to provide a full explanation.

Second, low hospital prevalence in older age groups may reflect a cohort effect, i.e., an effect of year of birth that is independent of age. Hypothetically, those age > 75 at the time of these HES data may have had limited access to MRI investigation as younger adults, since MRI scans were not in routine use in the NHS until the 1980s [32], meaning that potentially fewer cases were picked up. To investigate this hypothesis, we generated cohort graphs using the available data (Fig. 4). Different birth cohorts have different observed prevalence over the same age ranges, but the narrow range of years for which HES data is available and the potential impact of case ascertainment bias already described means that this hypothesis can be neither confirmed nor refuted. Data from a wider time period and more sensitive measures of case ascertainment would be required to fully explore this hypothesis. However, on an empirical level, a cohort effect is unlikely to fully account for the differing trends—as a progressive disease, even cases which were not picked up initially should have been detected later.

Third, the conversion rate from SCC to DCM may be negatively affected by age. SCC-derived prevalence is based on a non-stratified conversion rate, since age-stratified conversion rates from SCC to DCM are not available in the literature. Thus, the discrepancy between our two metrics could reflect a false assumption that this rate of conversion or diagnosis is static across age bands. Whilst possible, this explanation seems unlikely—if anything, the conversion rate is likely to be positively affected by age, given the increased vulnerability of the spinal cord [33] and greater severity of degenerative changes [14] with age.

Finally, underdiagnosis in older age groups may explain the differing trends in prevalence. The signs and symptoms of DCM are non-specific [34] and may be mistaken for ‘getting older’ [1]. For instance, a case–control study of patients presenting with hip fracture found that 18% had previously undiagnosed DCM [35]. Physical findings may also be masked by comorbidities such as diabetic neuropathy [36], which become more common with age. Thus, underdiagnosis may act as a unifying explanation for both the decline in hospitalised DCM prevalence, and the generalised mismatch between estimates.

Access to timely diagnosis and surgery is critical in DCM

If underdiagnosis is even partially responsible for both the generalised prevalence mismatch and the apparent decline in older age groups, this has potentially serious implications. Delayed diagnosis leads to greater disability [6], so timely access to diagnosis is critically important. Furthermore, our results suggest that underdiagnosis may be more likely in older age groups. Older patients still gain meaningful benefit from surgery [14], which should therefore be performed in cases of moderate and severe DCM [7] regardless of age. Thus, the potential impact of underdiagnosis on both health system utilisation and patient quality of life is highly relevant.

留言 (0)

沒有登入
gif