Neurological Conditions: estimating the prevalence in Scotland of selected conditions using GP and Hospital Admissions datasets

3 Findings

3.1 All Conditions

Table 1 shows the data for each condition from the two datasets. Annex C includes notes on why data is not available for some conditions.

It is important to emphasise again that the general practice dataset does not cover the whole of Scotland: the general practices whose data are included represented 72.7% of registered patients at the time of the data extract. If the general practice dataset covered 100% of registered patients, and prevalence was the same for the missing practices, the true general practice recorded diagnosis figures could be almost 40% higher.

Table 1: Number of people with selected neurological conditions recorded in GP practice dataset*, and estimated from hospital admissions dataset

Neurological Condition	Number of People
	General practice: recorded diagnoses*	Hospital admissions-derived estimates
Acquired Brain Injury
Brain Injury - Infection	24,121	n/a
Brain Injury - Other	1,803	3,185
Brain Injury - Trauma	229,255	18,749
Brain Injury - Vascular	79,830	89,198
Ataxia	2,677	3,944
Brain Tumour	7,841	7,299
Cerebral Palsy	6,416	5,936
Cluster Headache	7,312	852
Corticobasal Degeneration	32	n/a
Degenerative Spinal Disease	17,490	4,996
Dystonia	8,405	1,452
Epilepsy	57,588	40,352
Essential Tremor	10,457	1,300
Functional Neurological Disorder	8,040	1,756
Huntington's Disease	519	387
Limbic Encephalitis	n/a	n/a
ME / Chronic Fatigue Syndrome	19,496	2,820
Migraine	273,598	35,052
MOG Antibody Disease	n/a	n/a
Motor Neurone Disease	523	682
Multiple Sclerosis	11,961	10,720
Multiple System Atrophy	86	191
Muscle Disorders – Dermatomyositis	482	279
Muscle Disorders - Inclusion Body Myositis	n/a	n/a
Muscle Disorders - Muscular Dystrophies	1,249	2,151
Muscle Disorders - Polymyositis	375	285
Myasthenia Gravis	1,868	1,502
Neuromyelitis Optica Spectrum Disorder	115	143
NMDAr Encephalitis	n/a	n/a
Parkinson's Disease	7,571	7,681
Peripheral Neuropathy
Peripheral Neuropathy - due to diabetes	5,331	3,092
Peripheral Neuropathy - due to disease other than diabetes	146	485
Peripheral Neuropathy - due to Hereditary/Idiopathic	2,437	1,219
Peripheral Neuropathy - due to infection	1,712	1,322
Peripheral Neuropathy - due to mechanical causes	87,166	60,550
Peripheral Neuropathy - due to other causes	31,307	13,520
Peripheral Neuropathy - due to toxins	280	549
Post Polio Syndrome (PPS)	12	260
Progressive Supranuclear Palsy (PSP)	120	123
Spina Bifida / Hydrocephalus	7,136	8,017
Tourette's (Tic Disorders)	2,932	467

* The general practice dataset does not cover the whole of Scotland: the general practices whose data are included represented 72.7% of registered patients at the time of the data extract. If the general practice dataset covered 100% of registered patients, and prevalence was the same for the missing practices, the true general practice recorded diagnosis figures could be almost 40% higher.

Individuals are counted against each condition for which they have a diagnosis recorded. In both datasets, some individuals could have had a diagnosis for more than one condition. Therefore, the condition sub-totals should not be added up as their total will not represent the number of unique individuals with a diagnosis.

It is important to note that it is not known the extent to which the datasets for each condition overlap: this is likely to vary across conditions.

Differences between the general practice and hospital admissions data for specific conditions are likely to be attributable to where in the healthcare system people obtain a diagnosis and access care. For example, people with migraine or ME/CFS are more likely to access general practice services than to be admitted to hospital. In addition, the hospital admissions dataset does not include data from outpatient clinics, where many neurological conditions are diagnosed.

For some conditions shown above, comparisons with findings from other sources are presented below. The focus is on conditions where sources of comparative estimates are relatively robust, and where prevalence is generally used as the measure of the condition. (For some other conditions the measure generally used is incidence, or the number of new cases in a specified time period.)

Some of the comparative estimates differ from the general practice and/or hospital admissions data. The possible reasons highlight the strengths and limitations of using electronic health records as sources for prevalence estimates. For example, some differences could be attributable to different methods of case ascertainment. Other differences could reflect the complexity of clinical coding for some conditions, which makes it difficult to identify those patients in such records.

Table 1 is included in a supporting document which accompanies this publication. A full list of supporting documents accompanying this publication is shown below, and they are discussed in Annex D.

General Practice and Hospital Admissions estimates (includes Table 1 above).

General Practice recorded diagnoses.

Hospital Admissions estimates.

Primary Care Informatics: list of general practice Read codes.

Primary Care Informatics: background to list of general practice Read codes.

3.2 Specific Conditions

Epilepsy

GP-recorded diagnoses dataset: 57,588

Hospital admissions dataset: 40,352

Comparative data. In 2015 SIGN (the Scottish Intercollegiate Guidelines Network) published a guideline on the diagnosis and management of epilepsy in adults. This stated that there were 54,000 people with active epilepsy in Scotland. This figure derived from a 2011 publication by the Joint Epilepsy Council of the UK and Ireland on epilepsy prevalence and incidence. This statistics publication reviewed a variety of data sources to estimate, for each UK nation and for Ireland, the number of people with a diagnosis of epilepsy and a prescription of anti-epileptic drugs.

The higher GP figure could reflect that people are less likely to be admitted to hospital if their seizures are mild or their condition is well-managed.

Huntington's Disease

GP-recorded diagnoses dataset: 519

Hospital admissions dataset: 387

Comparative data. A 2013 study explored the prevalence of adult Huntington's disease in the UK based on diagnoses recorded in general practice records. It reported, for the period 1990-2010, an average prevalence rate for Scotland of 16.1 cases per 100,000 population aged 21 and over. If this rate was applied to the current population estimate (aged 21 and over, using the Mid-2021 Population Estimates for Scotland), it would give a figure of 689 people.

Migraine

GP-recorded diagnoses dataset: 273,598

Hospital admissions dataset: 35,052

Comparative data. In 2020 the Migraine Trust commissioned an independent rapid research review about who is living with migraine in the UK. The review cites a number of UK and international studies, including the Global Burden of Disease Study in 2016, which surveyed people from several countries including the UK. It stated that, based on the information available, around one in five adults in the UK may be living with migraine or migraine-like symptoms (23%). If this 23% rate was applied to the current adult population estimate (aged 18 and over, using the Mid-2021 Population Estimates for Scotland), it would give a figure of 1,024,631 people.

This variation in numbers is likely to reflect the different methods of case ascertainment, and where people are most likely to access healthcare. A population survey will identify people who report migraine but are not seeking healthcare. Many people with migraine will not seek GP care; and most will never require hospital treatment.

Motor Neurone Disease (MND)

GP-recorded diagnoses dataset: 523

Hospital admissions dataset: 682

Comparative data. A national register for MND integrates clinical care, audit, research and evaluation to provide ongoing comprehensive monitoring of every person living with MND in Scotland. A study on the Changing epidemiology of motor neurone disease in Scotland, drawing on this register and other data sources, found that, in 2017, there were 422 people living with MND in Scotland. In 2021, MND Scotland evidence to the Scottish Parliament Health & Sport Committee's Inquiry into Technology and Innovation in the NHS stated that MND Specialist Nurses were supporting over 450 people with MND in Scotland.

Further work would be needed to explore why the GP and hospital admissions figures are higher than those derived from the national register, which research shows has a high coverage of the population with MND.

Multiple Sclerosis (MS)

GP-recorded diagnoses dataset: 11,961

Hospital admissions dataset: 10,720

Comparative data. There is a national register for MS in Scotland whose aim is to improve the care of people with MS through systematic and comprehensive audit of their diagnosis and early management. The register collects data on MS incidence (the number of new cases). The latest national report from this register states that, since the register was established in 2010, data have been collected on 5,878 people with a confirmed diagnosis of MS, 578 of whom received a diagnosis in 2021. (Patients who choose not to be contacted by an MS Specialist Nurse and paediatric patients are excluded from these data.) A 2019 study drawing on findings from this register suggested that, in the light of the high incidence rates it shows, previous estimates of prevalence in Scotland were likely to be under-estimates.

The MS register does not set out to measure prevalence. Further work would be needed to explore the relationship between its data and the higher GP and hospital admissions figures.

Myalgic Encephalomyelitis/Chronic Fatigue Syndrome (ME/CFS)

GP-recorded diagnoses dataset: 19,496

Hospital admissions dataset: 2,820

Comparative data. AScottish Good Practice Statement on ME-CFS published in 2010 stated that epidemiological evidence for Scotland is lacking but that a population prevalence of at least 0.2‐0.4% is widely accepted, and that over 20,000 people in Scotland may be affected by ME/CFS. This figure would represent a prevalence rate of approximately 0.4%. The 2021 NICE guidance on ME and CFS cited data for England and Wales which would also represent a population prevalence rate of 0.4%. If this 0.4% rate was applied to the current population estimate (all ages, using Mid-2021 Population Estimates for Scotland), it would give a figure of 21,920 people.

Myasthenia Gravis

GP-recorded diagnoses dataset: 1,868

Hospital admissions dataset: 1,502

Comparative data. A study of the prevalence and incidence of neuromuscular conditions in the UK between 2000 and 2019 using primary care (GP) data showed, for 2019, a rate of 33.7 cases per 100,000 population. If this rate was applied to the current population estimate (all ages, using the Mid-2021 Population Estimates for Scotland), it would give a figure of 1,847 people.

Parkinson's Disease

GP-recorded diagnoses dataset: 7,571

Hospital admissions dataset: 7,681

Comparative data. A 2017 study of theprevalence and incidence of Parkinson's Disease using primary care (GP) data was undertaken by Parkinson's UK with advice from clinical experts. It projected that, in 2018, 12,184 people in Scotland aged 20 and over would have Parkinson's Disease. The study assessed that its prevalence and incidence estimates were consistent with findings from other studies which it reviewed.

It is not known to what extent the GP and hospital admissions data shown above overlap; and whether both figures represent an under-estimate of prevalence in the population.

Progressive Supranuclear Palsy (PSP)

GP-recorded diagnoses dataset: 120

Hospital admissions dataset: 123

Comparative data. A 2022 systematic review of prevalence studies of Progressive Supranuclear Palsy and Corticobasal Syndrome found that the best three prevalence studies (two undertaken in the UK) gave a pooled rate of 7.1 people per 100,000 population per year. If this rate was applied to the current population estimate (all ages, using the Mid-2021 Population Estimates for Scotland), it would give a figure of 389 people.

3.3 Strengths and Limitations

General Practice Dataset

• Innovation. This is the first general practice extract of this kind for neurological conditions in Scotland. It offers a starting point for potential future work to analyse variation between NHS Boards, regional inequalities and comparisons with other countries.
• Incomplete dataset. The general practice data presented in this report does not include data for the whole of Scotland. Two out of the 14 NHS Boards did not participate in the data extract, and data from 21 general practices from the other 12 NHS Boards were removed following a PHS quality assurance process. Data from 653 general practices were extracted and retained for analysis. These general practices had a total population of 4,260,960 registered patients, representing 72.7% of the 5,858,622 registered patients at the time of the data extract. Annex D shows details of excluded data. While it would be possible to obtain crude estimates of the true national figures based on the proportion of registered patients covered by the general practice dataset, PHS analysts advise against doing so. Further work with clinical and statistical input would be required to produce robust estimates.
• Clinical coding. Whether general practice datasets of this kind are comprehensive depends on the extent to which the conditions are diagnosed and recorded using the appropriate Read codes. This Primary Care Informatics guidance for general practices in Scotland highlights the benefits of accuracy, completeness and consistency in the coding of patient records.

Hospital Admissions Dataset

• Mapping of Read codes to ICD codes. SBoD used an in-house mapping file to map Read codes used by general practices across to ICD10 codes used for hospital admissions. In addition, the PHS Terminology Services team provided a full list of ICD10 codes for each specific condition, which were used to supplement the codes mapped from the Read codes. ICD codes are less granular and in larger groups than Read Codes. As a result, for some neurological conditions, the hospital admissions dataset may be capturing patients with a broader diagnosis.
• The potential of hospital records to under-estimate the prevalence of disease. Estimates drawing on hospital inpatient/daycase data only could under-estimate the prevalence of disease in the population. This will vary according to the condition. For some conditions, people diagnosed are more likely to be supported by GPs, or in outpatient clinics from which detailed data are not available. They might never require hospital admission, or only very late in the progression of the condition.
• Timeframe. Using a 20-year lookback will not capture those discharged from hospital with one of the listed neurological conditions before the year 2000 and who subsequently did not have a hospital admission until the end of the observation period (31 December 2019). This will result in an under-estimate of prevalent cases. Creating a longer lookback, such as 30 or 40 years, is possible. However, this would require additional mapping work to be undertaken between ICD10 and ICD 9 (the previous ICD version), which is beyond the scope of the present work.

Recovery from Neurological Conditions

General practice electronic records include data from as far back as the 1940s. For this data extract it was decided to set no date limits for when a relevant Read code was entered into the clinical system. Therefore, a patient would be identified in the search if their record included one of the relevant Read codes, regardless of when the Read code had been entered in the system. This approach was taken in order to identify patients who could have been diagnosed with progressive or lifelong conditions a long time ago.

However, for some neurological conditions, such as brain injury, migraine and epilepsy, recovery from or management of the condition is possible. Therefore, the presence of a relevant code in either of the general practice or hospital admissions datasets might not indicate a current or chronic condition. As such the estimates could overstate current prevalence and future healthcare need.

Re-diagnosis of Neurological Conditions

This work could not take account of re-diagnosis: for example people initially diagnosed with one condition (such as Parkinson's Disease) who are subsequently re-diagnosed with a different one (such as Progressive Supranuclear Palsy (PSP). As noted earlier, in both datasets individuals could appear more than once if they have a diagnosis for more than one condition.

See also Annex F which describes differences between the two datasets on issues such as timeframe and mortality.