Identifying multiple sclerosis in linked administrative health data in Aotearoa New Zealand

Multiple sclerosis (MS) is a chronic, inflammatory demyelinating disease of the central nervous system.¹ MS causes a range of neurological symptoms, including spasticity, optic neuritis, sensory disturbances, weakness, motor coordination impairment and cognitive dysfunction, as well as profound fatigue.¹

The New Zealand national MS prevalence study (NZMSPS) identified 2,917 people with definite MS who were resident in Aotearoa on census day 2006, yielding an estimated prevalence of 72.4 per 100,000.^2,3 This was an estimated 96.7% of diagnosed cases.^2,3 Other studies of people with MS in Aotearoa have used online cohorts,⁴ or older regional cohorts.⁵

For the NZMSPS, various organisations holding data about MS, including consultant neurologists, Multiple Sclerosis New Zealand (MSNZ), the Ministry of Health – Manatū Hauora and district health boards, contacted people living with MS and invited them to participate on behalf of researchers.³ The study was also advertised publicly.³ Consenting participants completed questionnaires and shared their medical records, which were assessed by a study neurologist.³ Once established, this cohort has been used to understand mortality,⁶ disability profiles,⁷ the relationship between receipt of disease-modifying therapies and disability⁸ and clinical features and geographic distributions of Māori and non-Māori living with MS.⁹

The Integrated Data Infrastructure (IDI) contains de-identified linked-administrative, census and survey data for an “ever resident” population of Aotearoa, and is accessible for research in the public interest.¹⁰ Information is available across many domains of life, including (but not limited to) health, education, work and family composition.¹⁰ The IDI allows for a wide range of research, including cross-sectional and longitudinal analyses, research about small populations, intervention analyses and examination of the social determinants of health.¹⁰

Previous research has characterised a range of medical conditions in the IDI, including autism and mental health conditions,¹¹ attention-deficit/hyperactivity disorder,¹² sudden unexpected deaths among infants¹³ and chronic conditions.¹⁴ To our knowledge, there has been little research characterising chronic autoimmune conditions like MS in the IDI. Identifying such conditions may be challenging, as the IDI does not contain diagnostic codes from primary care or outpatient settings,¹⁵ and it is likely that such disorders are commonly diagnosed in these settings.

Research aims

This study aims to:

1.       Use health service data in the IDI to identify a cohort of individuals likely to have been diagnosed with MS.

2.       Compare the characteristics of the identified cohort to previous studies, including the NZMSPS.

Methods

The Integrated Data Infrastructure (IDI)

The IDI is a large research database developed and managed by Stats NZ containing de-identified linked-administrative, census and survey data.¹⁰ Records in the IDI are linked probabilistically to a central “spine” of people who have been ever-resident in Aotearoa.¹⁰ In accordance with Stats NZ confidentiality requirements, all counts in this paper have been random rounded to base 3, and some small categories have been combined or could not be presented.

Data sources for identifying MS cases

Individuals likely to have been diagnosed with MS were identified based on service use in the following areas:

• publicly funded hospitalisations (January 1988–June 2022)
• privately funded hospitalisations (January 2001–December 2020)
• interRAI needs assessments of older adults (July 2014–June 2022)
• disability assessments through national disability support services (Socrates) (January 1998–June 2022)
• pharmaceutical dispensings (January 2006–June 2022).

Hospital discharges with the ICD-9 code 340 or the ICD-10 code G35 recorded as the primary or secondary diagnosis were used to identify people with MS from both publicly funded and privately funded hospitalisations. The disability support service and interRAI data contain specific flags for MS. Pharmaceutical dispensings of the eight funded disease-modifying therapies (DMTs)—interferon beta-1-alpha, interferon beta-1-beta, glatiramer acetate, teriflunomide, dimethyl fumarate, fingolimod, natalizumab and ocrelizumab—were taken as evidence of MS.¹⁶

Cohort characteristics

The following characteristics were compared between the IDI MS cohort and prior studies: gender, self-identified ethnic distribution, Māori descent (ancestry), regional distribution and functional limitations.^2,5,7,9,17 We also report on age of first record of MS and distribution by area deprivation using the 2018 New Zealand Index of Deprivation (NZDep2018).¹⁸

Demographic data were sourced from the personal details table, which contains information on gender, birth year and ethnicity for all individuals in the IDI.¹⁹ Ethnic group information is source-ranked, with more reliable data sources prioritised.¹⁹ Everyone was counted in every level 1 ethnic group with which they identify—Māori, Pacific peoples, Asian, Middle Eastern/Latin American/African (MELAA) and European.

Data on Māori descent, region and usual residence were sourced from the administrative population census (APC), with geographical data measured at 30 June 2022.²⁰ Māori descent data in the APC is sourced from birth records.²⁰ Area deprivation was classified by linking small area information (meshblocks) with NZDep2018 scores. Functional limitations were measured using the Washington Group Short Set in the 2018 census.²¹

Comparisons were made for three different cohorts, i) the overall cohort of people classified with MS at any point, whether or not they subsequently moved overseas or died (“overall cohort”), ii) those resident and classified with MS by 30 June 2022 (“resident in June 2022”), and iii) those resident and classified with MS by 30 June 2018 (“resident in June 2018”; functional limitations only).

There was no missing data for age, gender, ethnicity, region or area deprivation for the relevant MS cohorts. Māori descent was unknown for 19% (1,515/7,890) of the overall MS cohort. Of those resident in June 2018, 12–13% (516–549/4,278) were missing data for individual functional limitation domains.

Crude prevalence

Crude prevalence rates for the total resident population (N=5,029,095) and population sub-groups were calculated using estimates from the June 2022 APC.²⁰ No age restrictions were applied as children can be (rarely) affected by MS.

Age standardisation

Direct standardisation to the total resident population in June 2022²⁰ was used to allow for comparisons across groups with different age structures. Age-adjusted rates are presented by gender and ethnic group, for those of Māori descent and by region. Age-specific rates were calculated for the following age groups: <40, 40–49, 50–59 and 60+ years. Due to small counts, those aged 40–59 were combined to calculate the age-adjusted rate for Pacific peoples, so this rate may not be directly comparable.

Area deprivation trends

Binary logistic regression was used to estimate associations between NZDep2018 decile and MS records for the resident population in June 2022. Analyses were stratified by ethnic group and Māori descent. Models controlled for age and gender. Under 1% (38,313/5,024,235) of people without a record of MS were excluded due to missing NZDep2018, having implausible ages (>110), or not being classified as male or female.

Software

SAS Enterprise Guide 8.3 was used for data management. Tabulations were conducted in Microsoft Excel. R was used to conduct regression analyses and create figures.

Results

Identification of MS cases

Counts of people with MS classified from each data source for the overall cohort and for those resident in June 2022 are summarised in Table 1. Overall, 7,890 people were identified with service use with MS recorded or indicated, of which 4,860 were classified as resident in June 2022. Some people with MS were identified from more than one data source.

Most people were identified from a publicly funded hospitalisation event, with 87% (6,897/7,890) of people in the overall MS cohort identified this way. This was followed by receipt of DMTs at 33% (2,616/7,890), receipt of disability support at 25% (1,965/7,890), interRAI assessments at 12% (918/7,890) and privately funded hospitalisations at 1% (177/7,890). The distribution was broadly similar for those resident in June 2022, excepting that 49% (2,373/4,860) had received DMTs for MS.

View Table 1–3, Figure 1–3.

There was a considerable degree of overlap between data sets. However, for the overall cohort, 47% (3,216/6,897) of those identified from publicly funded hospitalisations were only identified from this source. This compares to 28% (258/918) for interRAI assessments, 19% (495/2,616) for pharmaceutical records, 15% (27/177) for privately funded hospitalisations and 9% (168/1,965) for disability support services.

National prevalence

The estimated national crude prevalence of MS from this study is 96.6 per 100,000 for the resident population in June 2022.

Demographic characteristics

Table 2 presents demographic characteristics. Of the total cohort, 73% (5,766/7,890) were women, 95% (7,509/7,890) identified as European and 5% (399/7,890) identified as Māori.

Age-adjusted prevalence in June 2022 was estimated at 145.1 per 100,000 for women compared to 48.9 per 100,000 for men. Estimated age-adjusted prevalence was highest for Europeans (124.7 per 100,000), followed by MELAA (85.5), Māori (41.8), Asian (16.8) and Pacific peoples (11.1) ethnic groups. The age-adjusted rate of MS was slightly higher for those of Māori descent (48.6) than those with Māori ethnicity.

Age of first record

The mean (SD) age of first record was 47.4 (15.4) years for the overall cohort. Mean age of first record was similar for men at 48.6 (15.8) years and women at 46.9 (15.3) years. Across ethnic groups, mean age of first record was oldest for Europeans at 47.7 (15.3) years, followed by MELAA, Pacific peoples, Māori and Asian ethnic groups at 43.9 (16.0) years, 42.4 (18.5) years, 40.7 (14.4) years and 37.1 (14.7) years, respectively. Mean age of first record was 40.4 (14.5) years for people of Māori descent.

Regional distribution

Estimated crude and age-adjusted prevalence rates for each region are shown in Figure 1, which demonstrates a strong latitudinal gradient. Due to small counts, Gisborne and Hawke’s Bay, and Nelson and Marlborough, were combined to estimate age-adjusted prevalence rates. Estimated crude and age-adjusted prevalence rates were generally similar. The highest crude prevalence rate was in Southland at 209.8 per 100,000, while the lowest crude prevalence was in Auckland at 59.5 per 100,000.

Functional limitations

Figure 2 compares the distributions of functional limitations across six domains (mobility, self-care, cognition, vision, communication and hearing) between those with a record of MS by June 2018 and the total resident population. People with MS had greater levels of functional limitations compared to people of the same age for all age groups across all domains except for hearing. Functional limitations were particularly pronounced for mobility and self-care.

For the hearing domain, the <40 and 40–49 age categories were combined, and 9% (120/1,302) of those with MS reported some difficulty with hearing, while under 1% (6/1,302) reported a lot of difficulty or being unable to hear.

Area deprivation

Figure 3 shows the crude prevalence of MS per 100,000 population across area deprivation quintiles for those classified with MS and resident in June 2022 stratified by ethnic groups (excluding Pacific peoples group due to small counts). This shows a reduced prevalence of MS for the overall cohort, and for people with Māori ethnicity and Māori descent, among the most deprived quintiles. This pattern was not evident for the European ethnic group, for whom those living in the least deprived quintile had the lowest estimated prevalence.

Table 3 shows the results of logistic regression models estimating the association between area deprivation and MS records, controlling for age and sex. For the total population, the odds of having a record of MS were 0.97 (0.96, 0.98) times lower with each one decile increase in NZDep2018. Stronger associations were apparent between area deprivation and MS records for the Māori, Pacific peoples and Māori descent groups, with estimated odds ratios of 0.90 (0.87, 0.93) for both the Māori ethnicity and Māori descent groups, and 0.87 (0.80, 0.95) for the Pacific peoples ethnic group. For the European ethnic group, the odds of having a record of MS were 1.01 (1.00, 1.02) times greater given a one decile increase in NZDep2018. There was no evidence of an association between area deprivation and MS records for the Asian or MELAA ethnic groups.

Discussion

This study identified 7,890 people with MS in the IDI within the period from January 1988 to June 2022. Most people were identified from publicly funded hospital records. MS was more common in women, European and MELAA ethnic groups, and people living in more Southern regions. Estimated prevalence of MS was lower among those living in the most deprived quintiles for the overall resident population, Māori, Pacific peoples and those of Māori descent. People with MS reported greater levels of functional impairment, especially for mobility and self-care.

Comparisons with the prior studies

The estimated prevalence rate of 96.6 per 100,000 is considerably higher than the estimate of 72.4 per 100,000 from the NZMSPS in 2006.^2,3 Prior research demonstrated an increase in the prevalence of MS in Aotearoa over the period 1968–2006⁵ and it is possible that this represents a continuation of this trend. Increasing rates of MS have been observed in other countries.²²

Demographic characteristics of the cohort from this study are broadly similar to previous research. The cohort was predominantly women and European, with lower prevalence rates for Māori, Pacific peoples and Asian ethnic groups, consistent with the NZMSPS.^2,9,17 While prevalence for MELAA has not previously been reported for Aotearoa, the estimated prevalence for this group was 72.6 per 100,000, higher than most other ethnic groups.

However, prevalence rates among those of Māori ethnicity and Māori descent (crude rates 33.1 per 100,000 and 36.1 per 100,000, respectively) were considerably higher than that reported in the NZMPS (15.9 per 100,000 and 20.6 per 100,000), consistent with research suggesting the prevalence of diagnosed MS may be increasing among Māori.^2,9,17

Average age of onset in the NZMSPS was 35 years.⁷ A study of MS incidence in Aotearoa in 2014 found that the average age of onset was 37.8 (26.1–49.5) years, while the average age of diagnosis was 42.4 (29.7–55.1) years (In an email from DF Mason, October 2024), indicating that a substantial delay between onset of symptoms and diagnosis is typical. The greater average age of first record in the current study (47.4 years) likely reflects delays between initial diagnosis and having an event suggestive of MS captured in the IDI. Consequently, people in the MS IDI cohort will often have lived with symptoms of MS for some time prior to their first record of MS in the IDI.

A strong latitudinal gradient was observed, with greater estimated prevalences of MS further south, as has been previously reported.^2,5,17 The latitudinal gradient may be related to greater vitamin D exposure in northern regions, which is thought to be potentially protective against MS, as well as differences in ethnic composition across Aotearoa.^2,9

However, a lower prevalence of MS was found for Auckland compared to the surrounding regions, Northland and Waikato, opposite to the pattern observed in the NZMSPS.² Auckland is the largest region in Aotearoa and has a greater share of the population born overseas (42.9% at the 2023 census).²³ Evidence from Denmark, also a high-risk country for MS, shows that first generation migrants arriving from low-risk countries have a lower MS risk than those born in Denmark, but a higher risk than in their country of birth.²⁴ The risk of developing MS was greater for those who migrated earlier in life, while the risk for second-generation migrants was greater than for ethnic Danes.²⁴ This suggests a strong role for environmental factors, which may include exposure to vitamin D, tobacco, obesity and age of primary Epstein–Barr infection.²⁴ Differences in characteristics between those who migrate and those who do not (e.g., the “healthy-migrant effect”) may also play a role.²⁴ Investigating whether migration influences the MS prevalence in Auckland warrants further study.

We are not aware of previous research using the Washington Group Short Set to characterise functional capacity for people living with MS. Prior research in Aotearoa using the Expanded Disability Status Scale (EDSS), a MS-specific measure of disability particularly related to mobility, found higher levels of disability with increasing age,⁷ consistent with mobility-related disability in this study.

Differences in estimated prevalence by area deprivation

This study found lower prevalence rates of recorded MS among those living in the most deprived areas for the total population, for Māori and Pacific peoples and for people of Māori descent. This may represent diagnostic barriers in more deprived areas.

Evidence for an association between socio-economic position and MS risk, as well as the direction of such relationships, is conflicting.^24,25 Studies conducted in more unequal countries tend to find lower risk of MS among those with low socio-economic position, possibly reflecting underdiagnosis due to difficulties accessing and navigating healthcare.²⁵ Notably, some potential risk factors for MS (e.g., smoking) are socio-economically patterned.²⁵

Differences in estimated prevalence across ethnic groups

Internationally, there is a perception that MS may be less common among minority ethnic groups.²⁵ This may create barriers to testing and diagnosis, thereby reinforcing this belief.²⁵ In Aotearoa, lower rates of MS among Māori have been ascribed to lower frequency of high-risk genetic alleles relating to the human leukocyte antigens system.²⁶ However, using ethnic categories to understand genetic disease risk can be highly problematic, especially given their imprecise and often overlapping nature.²⁷ Furthermore, there are well-known access issues among Māori and Pacific peoples in the healthcare system.^12,28,29 People identifying with Māori and Pacific peoples ethnic groups are more likely to live in deprived areas,²⁹ and less likely to have a record of MS if living in more deprived areas. Ethnic-specific and socio-economic barriers may intersect to reduce access to MS testing and diagnosis for these groups.

Coverage errors

Twenty percent of patients assessed by two specialty clinics in California had previously been misdiagnosed with MS.³⁰ Given people were classified with MS based on any instance of MS recorded/indicated in the IDI, it is likely that some people re-assessed as not having MS at a later point were included in the MS cohort. By contrast, NZMPS participants who had not seen a neurologist in the previous 12 months were re-assessed by a study neurologist.³

It is also likely that people with MS are missing from this cohort. People with established diagnoses of MS may be missing if they have not received DMTs, disability support or interRAI needs assessments, and have not been hospitalised with MS listed as a discharge diagnosis. It is likely there are people living with MS who are undiagnosed.

Reassuringly, the characteristics of the MS IDI cohort are broadly consistent with previous research, and the number of people identified with MS is plausible. This suggests that the cohort is useful for understanding patterns, although it should not be considered an accurate count of everyone in Aotearoa with MS.

Strengths and limitations

Classifying MS in the IDI allows researchers to examine a wide range of outcomes across multiple life domains.

Several important limitations warrant mention. Firstly, it is not possible to identify the subtype of MS, even among those hospitalised, as the ICD data in the IDI does not include this information. Importantly, disease-related outcomes and DMT eligibility varies by MS subtype.^1,7,17 Other important clinical information is unavailable, such as age of onset and diagnosis, and EDSS scores.

While the IDI is useful for understanding patterns, the cohort of people with MS will include some people who do not have MS and exclude some people who do. Without access to detailed clinical information, it is difficult to determine the extent to which this is the case. We also did not include formal validation against high quality incidence data.

Conclusion

The study identified 4,860 resident people in June 2022 with health records suggestive of MS between 1988 and 2022, resulting in an estimated crude prevalence of 96.6 per 100,000. The MS cohort was broadly similar to previously research in Aotearoa, except that the estimated prevalence among Māori was greater, and the estimated prevalence for Auckland was lower than surrounding regions. Overall, this study demonstrates that the IDI appears to be a viable tool to measure and monitor MS prevalence, demographic patterns and treatment rates in Aotearoa.