Повышенное использование медицинских услуг two За несколько десятилетий до первого демиелионирующего события, которое отличается по возрасту и полу.

Introduction

Prodromal phases have been recognized and described across several immune‐mediated and neurodegenerative conditions, including rheumatoid arthritis, type 1 diabetes, and Parkinson's disease (PD) and may last months to years before classical manifestation of disease.
1
Until relatively recently, multiple sclerosis (MS), an immune‐mediated neurologic condition, was not thought to have a prodromal phase.
2
However, compared to controls, people with MS have higher rates of physician, hospital, and pharmacy‐related visits before classical disease manifestations, suggesting a measurable prodromal phase.
3
Biological evidence of preclinical disease includes higher serum neurofilament levels a median 6 years pre‐MS onset.
4
Nonetheless, much remains unclear, including when elevations in healthcare use begin relative to classical MS onset.
1
Linked health administrative and registry‐related data provide opportunity to examine patterns of healthcare use years before classical manifestation of disease. This approach has been critical in narrowing the etiologic time‐window in other diseases,
5
and informing the design of risk factor studies, and intervention studies aimed at preventing disease initiation. In MS, the few studies that examined healthcare use beyond 5 years pre‐onset were limited in scope to primary care, examining visits to general practitioners (GPs) only, or assessed the period before diagnosis rather than before the earlier onset of classical MS symptoms.
6
,
7
Although MS is highly heterogeneous, with sex and age influencing clinical presentation and outcomes including healthcare use, their association with the pre‐onset or prodromal phase were rarely delineated.
8
Finally, most prior studies have focused on one region or country, potentially limiting generalizability.

We examined patterns of healthcare use, measured as annual rates of hospital and physician visits, including primary and secondary (i.e., specialist) care and prescriptions filled, in the 2–3 decades before onset of classical MS manifestations by accessing two international cohorts of persons with MS (PwMS) and matched controls in Ontario, Canada and Sweden. We also examined the effect on healthcare use by participants’ sociodemographic characteristics, including age, sex, socioeconomic status (SES), and urban/rural residency.
8
,
9

Methods

Data sources

We used population‐based health administrative and, or registry data from Canada's most populous province, Ontario (14.6 million) and Sweden (10.3 million [2019]). Regions were selected due to their sizable populations and accessible data from their publicly funded, universal healthcare systems, including all medically necessary hospital and physician services. Administrative data included dates of all hospital visits (in and outpatients) and physician visits (all specialties, with primary care/GP also available in Ontario) via the Discharge Abstract Database/Ontario's Health Insurance Plan and Sweden's National Patient Register, Table  S1 . Also accessed were emergency department (ED) visit dates in Ontario through the National Ambulatory Care Reporting System and prescription dispensation dates in community/outpatient pharmacies via Sweden's National Prescription Register. Sociodemographic and residency information (sex, birth/death dates, SES, and urban/rural residency) were accessed via Ontario's Registered Persons Database and Sweden's Total Population Register and Longitudinal Integration Database. SES was expressed as quintiles, estimated by linking each person's 6‐digit residential postal code to census‐derived neighborhood‐level average income (Canada) and using each person's annual disposable income, or highest earning parent if <18 years‐of‐age (Sweden). Urban/rural residency was based on the postal code's second character (Canada), and the median county‐level population density/sqKM (Sweden).
10
Neurologist‐recorded MS symptom onset date was extracted from the Swedish MS Registry (which captures ≈80% of Sweden's MS population).
11
Within each region, data were linked deterministically using each person's unique encoded healthcare number and analyzed at ICES (Ontario) or released to researchers (Sweden).

Data were available/accessed: April 1, 1991–March 1, 2020 in Ontario, except for ED visits (available July 1, 2000–March 1, 2020) and January 1, 2001–December 31, 2019 in Sweden, except for prescriptions (January 1, 2005–December 31, 2019). Study end dates were chosen partly to avoid SARS‐CoV‐2/COVID‐19 pandemic‐related healthcare disruptions.

Study populations

Our primary approach included two large ‘administrative’ cohorts, selected within each region by accessing diagnostic information captured from the physician and hospital visits. PwMS were identified using a validated algorithm, requiring ≥3 hospital or physician MS diagnostic codes in any combination (recorded as International Classification of Diseases [ICD]‐9/10: 340/G35).
12
,
13
The earliest MS/demyelinating disease‐related code (Table  S2 ) defined MS onset (index date). At least 5 years of residency in the region (≥90% of days annually, or from birth if <5‐years old) with no MS/demyelinating disease‐related codes pre‐index date were required.

A smaller “clinical cohort” was created with neurologist‐diagnosed PwMS identified from the Swedish MS Registry (overlap with the administrative cohort was permitted). Here, the index date was the earlier of MS symptom onset date or first MS/demyelinating code.

For each of the three cohorts, PwMS were matched by sex, birth year, region (3‐digit postal code or county) at their index year, and residency time in the region, with up to five controls from the general population without MS/demyelinating disease‐related codes, and not in the Swedish MS Registry. Both PwMS and controls had to remain resident (≥90% of days/year) until the algorithm was fulfilled (the third MS code [administrative cohorts]), or MS diagnosis date [clinical cohort], Figure  S0 .

Outcome: Healthcare use patterns

Healthcare use pre‐index date was measured as yearly counts of physician, hospital, ED visits (Ontario) and prescriptions filled (Sweden; by drug class, Level 2, Anatomical Therapeutic Chemical Classification System). All hospital visits were examined, regardless of length, with inpatients (>1 day) reported separately. Up to 29 years of data were available in Ontario (19 years for ED visits) and 19 years in Sweden (15 years for prescriptions).

Statistical analyses

Concordant with data access and privacy agreements, analyses were performed within each region using a common plan. Results for the three cohorts were reported separately, preserving regional differences (e.g., primary care and ED visits available in Ontario but not Sweden which had prescription data). Yearly rates of each healthcare outcome pre‐index date were compared between PwMS and matched controls using negative binomial regression models fitted via generalized estimating equations (GEE) with an exchangeable working correlation matrix. The logarithm of person time was included as a model offset. Findings were reported as rate ratios (RR, 95% confidence intervals [CI]), adjusted for sex, and, at the index date: age (continuous), SES quintiles, urban/rural area, and calendar year (continuous). Sociodemographic effects (sex, age, SES, and urban/rural area) were examined using interaction terms with age at 20, 30, 40, and 50 years‐old reported via coefficient estimates. Analyses were conducted using SAS_V9.4 (SAS Institute, Cary, USA) and R_V4.0.5 (R Foundation, Vienna, Austria).

Results

Cohorts

The administrative cohorts included 35,018 PwMS and 136,007 matched controls in Ontario, and 10,269 PwMS and 51,297 matched controls in Sweden, Table  1 . Sweden's clinical cohort included 7604 PwMS and 37,974 matched controls. Across all cohorts, females predominated and most lived in urban areas. Within each region, the MS and matched cohorts were demographically very similar. For example, at the index date, the mean ages (and standard deviations) for the MS and matched control cohorts were as follows: in Ontario, the MS cohort was 42.7 (14.1) years and matched cohort was 42.9 (13.6) years. In Sweden, the health administrative MS and matched cohort were both 41.7 (14.1) years, and the clinical MS cohort was 36.5 (11.8) years and matched cohort was 36.6 (11.8) years. Sweden's clinical cohort was younger on average (37 years) than the administrative cohorts at their respective index dates (42–43 years). For PwMS in Sweden's clinical cohort with both index dates ( n  = 5835), the mean time between dates was 1.1 years (SD = 1.74). The average follow‐up time pre‐index (MS onset) exceeded 15 years in Ontario and 11 years in Sweden.

Table 1 Characteristics of the people with multiple sclerosis (PwMS) and matched controls at MS onset (the index date) used to examine healthcare use in the years leading up to MS onset in Canada (1991–2020) and Sweden (2001–2019). Ontario, Canada Sweden Health administrative cohort PwMS‐control
Health administrative cohort PwMS‐control
Clinical cohort PwMS‐control
Characteristic, at the index date (MS onset): PwMS ( N  = 35018) Control ( N  = 136007) PwMS ( N  = 10269) Control ( N  = 51297) PwMS ( N  = 7604) Control ( N  = 37974) Sex Female 24239 (69.2%) 94671 (69.6%) 7044 (68.6%) 35197 (68.6%) 5208 (68.5%) 26013 (68.5%) Male 10779 (30.8%) 41336 (30.4%) 3225 (31.4%) 16100 (31.4%) 2396 (31.5%) 11961 (31.5%) Age at Index,
a
years: Mean (SD) 42.7 (14.1) 42.9 (13.6) 41.7 (14.1) 41.7 (14.1) 36.5 (11.8) 36.6 (11.8) Categories <18 years 451 (1.3%) 1420 (1.0%) 206 (2.0%) 1011 (2.0%) 233 (3.1%) 1151 (3.0%) 18 to <30 5917 (16.9%) 21394 (15.7%) 2230 (21.7%) 11135 (21.7%) 2323 (30.5%) 11592 (30.5%) 30 to <50 18583 (53.1%) 73608 (54.1%) 4979 (48.5%) 24886 (48.5%) 3967 (52.2%) 19827 (52.2%) ≥50 10067 (28.7%) 39585 (29.1%) 2854 (27.8%) 14265 (27.8%) 1081 (14.2%) 5404 (14.2%) Age at diagnosis, years: Mean (SD) 43.9 (14.2) 44.1 (13.7) 43.0 (14.4) 37.8 (12.0) 37.8 (12.0) 43.0 (14.4) Index year, categories 1991–2000 8984 (25.7%) 37190 (27.3%) — — — — 2001–2010 14727 (42.1%) 57200 (42.1%) 4229 (41.2%) 21135 (41.2%) 2902 (38.2%) 14504 (38.2%) 2011–2019/2020 11307 (32.3%) 41617 (30.6%) 6040 (58.8%) 30162 (58.8%) 4702 (61.8%) 23470 (61.8%) SES SES
1 (lowest income quintile, least affluent) 6361 (18.2%) 23483 (17.3%) 1463 (14.2%) 6412 (12.5%) 1145 (15.1%) 5514 (14.5%) 2 6658 (19.0%) 26099 (19.2%) 1905 (18.6%) 8417 (16.4%) 1243 (16.3%) 5925 (15.6%) 3 7287 (20.8%) 28204 (20.7%) 2616 (25.5%) 12240 (23.9%) 1949 (25.6%) 9060 (23.9%) 4 7387 (21.1%) 29348 (21.6%) 2337 (22.8%) 12634 (24.6%) 1810 (23.8%) 9242 (24.3%) 5 (highest income quintile, most affluent) 7325 (20.9%) 28873 (21.2%) 1948 (19.0%) 11594 (22.6%) 1457 (19.2%) 8233 (21.7%) Area of residence
b

Rural 4189 (12.0%) 17434 (12.8%) 2495 (24.3%) 12443 (24.3%) 1767 (23.2%) 8808 (23.2%) Urban 30829 (88.0%) 118573 (87.2%) 7774 (75.7%) 38854 (75.7%) 5837 (76.8%) 29166 (76.8%) Follow‐up pre‐index date: mean (SD), years 15.2 (SD:6.9) 15.8 (SD:6.9) 11.0 (SD:3.7) 11.2 (SD:3.8) 11.3 (SD:3.8) 11.5 (SD:3.8)
Note : PwMS‐control Ontario: 3127 cases had one control; 4162 had two controls; 4651 had three controls; 4787 had four controls; 18,291 had five controls. Sweden: in the health administrative cohort, four patients had one control, five had two, four had three, nine had four, and 10,247 had five. In the clinical cohort, four patients had one control, four had two, six had three, six had four, and 7584 had five controls. SES Canada: A minority of the cohort (0.3%; 103 PwMS; 423 controls), with missing socioeconomic status were assigned to category 3. SES Sweden: For individuals under 18 years at the index date, the socioeconomic status of the highest‐earning parent was used. A minority of the entire cohort, regardless of age (health administrative cohort: 38 (1 PwMS, 37 controls), clinical cohort: 29 (0 PwMS, 29 controls)) with missing socioeconomic status was assigned to category 3. N‐Miss, number of missing; Q, quartile; SD, standard deviation; SES, socioeconomic status.
a
Index date in the clinical cohort was MS symptom onset, as recorded by the neurologist for the majority (6430/7604; 85%), while for the remainder, it was the first MS/demyelinating‐related code.
b
Ontario: A minority (45 PwMS and 100 controls) with missing area of residence were assigned to urban status. Sweden: For individuals under 18 years at the index date and without a recorded residential area, the residential area used was that of the mother (health administrative cohort: 16/206; 7%, clinical cohort: 7/233: 3%).

Characteristics of the people with multiple sclerosis (PwMS) and matched controls at MS onset (the index date) used to examine healthcare use in the years leading up to MS onset in Canada (1991–2020) and Sweden (2001–2019).

Note : PwMS‐control Ontario: 3127 cases had one control; 4162 had two controls; 4651 had three controls; 4787 had four controls; 18,291 had five controls. Sweden: in the health administrative cohort, four patients had one control, five had two, four had three, nine had four, and 10,247 had five. In the clinical cohort, four patients had one control, four had two, six had three, six had four, and 7584 had five controls. SES Canada: A minority of the cohort (0.3%; 103 PwMS; 423 controls), with missing socioeconomic status were assigned to category 3. SES Sweden: For individuals under 18 years at the index date, the socioeconomic status of the highest‐earning parent was used. A minority of the entire cohort, regardless of age (health administrative cohort: 38 (1 PwMS, 37 controls), clinical cohort: 29 (0 PwMS, 29 controls)) with missing socioeconomic status was assigned to category 3.

N‐Miss, number of missing; Q, quartile; SD, standard deviation; SES, socioeconomic status.

Index date in the clinical cohort was MS symptom onset, as recorded by the neurologist for the majority (6430/7604; 85%), while for the remainder, it was the first MS/demyelinating‐related code.

Ontario: A minority (45 PwMS and 100 controls) with missing area of residence were assigned to urban status. Sweden: For individuals under 18 years at the index date and without a recorded residential area, the residential area used was that of the mother (health administrative cohort: 16/206; 7%, clinical cohort: 7/233: 3%).

Healthcare use patterns before MS onset: The health administrative cohorts

Physician visits and hospitalizations

In Ontario, yearly rates of physician visits were consistently higher for PwMS versus controls by 20% or more in the 20 years pre‐index date, with some evidence of elevated rates in up to 28 (of 29) years pre‐index, Figure  1 . In Sweden, yearly physician visit rates, which did not include primary care, were elevated up to 15 (of 19) years pre‐index; the rates were significantly higher in PwMS than controls in the 10 years pre‐index, exceeding 10% (RR > 1.1) in the 6 years pre‐index. In both regions, the sharpest increase occurred in the 2 years pre‐index date, where RRs for PwMS versus controls exceeded 1.5 (50% higher) in Ontario and 1.3 (30%) in Sweden. In the year pre‐index, RRs exceeded 2, being 2.25 (2.21–2.28) in Ontario and 2.51 (2.44–2.59) in Sweden. For Sweden, this represented the highest RR across all outcomes.

Associations between multiple sclerosis and physician, hospital, and emergency department visits before a first demyelinating event in Ontario, Canada (1991–2020) and Sweden (2001–2019). Top: physician visits: Ontario, Canada (left), Sweden (right). Middle: all hospital visits: Ontario, Canada (left), Sweden (right). Bottom: ED visits: Ontario, Canada (2000‐2020) (left). Total number of events for cases and controls (right). The x ‐axis shows the rate ratios (RR) and corresponding 95% confidence intervals (CI); a RR > 1 indicates higher healthcare use in the multiple sclerosis cohort relative to the matched controls. The y ‐axis represents the years before the index date, e.g., “−1” indicates the year before the index date, “−2” the second year prior and so on.

This pattern remained largely consistent when all hospital visits were examined, especially in Ontario where rates were consistently ≥20% higher for PwMS versus controls in each of the 20 years pre‐index date, exceeding 30% higher in the 7 years prior, peaking at 1.90 (1.84–1.96) in the year pre‐index, Figure  1 . In Sweden, the elevated RRs were primarily confined to the 2 years pre‐index, although some RRs reached statistical significance up to 16 years pre‐index (years −12, −15, and −16). In the second‐year pre‐index, the RR was 1.16 (1.07–1.25), rising sharply to 2.41 (2.27–2.56) in the year pre‐index.

For inpatient hospitalizations, the relative rate pre‐index remained remarkably high, especially in the year prior, peaking at a RR of 2.63 (2.53–2.74) in Ontario and 2.36 (2.23–2.50) in Sweden, Figure  S1 . Moreover, in Ontario, RRs consistently exceeded 1.2 (20% higher rates) for PwMS versus controls for 17 (of 29) years pre‐index.

ED visits (Ontario) and prescriptions filled (Sweden)

Due to data availability (19 years for ED visits and 15 for prescriptions filled pre‐index), the number of individuals contributing to these analyses was smaller: 19,598 PwMS and 73,166 controls for ED visits and 6,501 PwMS and 32,467 controls for prescriptions filled, Table  S3 .

The RRs for ED visits pre‐index were among the highest of all the healthcare use examined, peaking at 3.04 (2.94–3.13) in the year prior, Figure  1 . While this dropped sharply in the second year pre‐index (RR = 1.65), RRs still exceeded 1.2 in each of the 18 (of 19) years pre‐index date, P  < 0.0001.

In contrast, prescriptions filled were among the most modest estimates observed, with the RR primarily being elevated in the year pre‐index (1.24, 1.13–1.37), Figure  S1 .

Sex and healthcare use

Sex differences in the RRs between PwMS and controls were observed. Male PwMS exhibited a higher degree of elevation in healthcare use versus females, particularly in the 1–5 years pre‐index date, Figures  2 and 3 , Figure  S2 . Males had higher physician visit RRs than females in Ontario in each of the 5 years pre‐index date, peaking in the year pre‐index (males: RR = 2.83 [2.75–2.92] and females: 1.98[1.95–2.01]). Similar sex differences were evident up to 5 years pre‐index for inpatient hospitalizations, and in the 1–2 years pre‐index for all hospital and ED visits.

Associations between multiple sclerosis and physician and hospital visits before a first demyelinating event in Ontario, Canada (1991–2020) and Sweden (2001–2019) by sex at the index date (within each panel, left: females; right: males). Top: physician visits, Ontario, Canada. Middle: physician visits, Sweden. Bottom: all hospital visits, Ontario, Canada. For a description of the x ‐ and y ‐axes, see caption to Figure  1 .

Associations between multiple sclerosis and hospital and emergency department (ED) visits before a first demyelinating event in Ontario, Canada (2000–2020) and Sweden (2001–2019) by sex at the index date (within each panel, left: females; right: males). Top: all hospital visits, Sweden. Bottom: ED visits, Ontario, Canada. For a description of the x ‐ and y ‐axes, see caption to Figure  1 .

Sweden's rate ratios of hospital and physician visits also differed by sex mainly in the year pre‐index date. For hospital visits, RRs for males peaked at 3.02 (2.66–3.43) and 2.19 (2.05–2.34) for females. For physician visits, RRs were 2.97 (2.80–3.15) for males versus 2.32 (2.24–2.40) for females. No sex difference was observed for prescriptions filled.

In contrast to these sex differences in the magnitude of elevations, the duration of elevated healthcare use was similar for males and females. For example, across a total of 29 years of follow‐up, Ontario's physician visit rate ratios were significantly higher for both sexes in each of 28 years pre‐index date, and up to 26–28 years pre‐index for hospital visits (all such visits and for inpatients only).

Age and healthcare use

PwMS older (vs younger) at the index date exhibited higher visit RRs, particularly in the year pre‐index date, Figures  4 , 5 , 6 , 7 , 8 , Figure  S3 . For example, in Ontario, the RRs for physician visits showed an age gradient in the year pre‐index, 2.31 (2.27–2.34) for 50‐year‐olds and 1.66 (1.63–1.70) for 20‐year‐olds. The age‐gradient was particularly high for inpatient visits, with RRs in the year pre‐index being 2.76 (2.65–2.89) for 50‐year‐olds versus 1.64 (1.57–1.71) for 20‐year‐olds. For ED visits, RRs in the year pre‐index were 3.09 (2.98–3.19) for 50‐year‐olds and 1.98 (1.88–2.09) for 20‐year‐olds. Similar patterns were observed in Sweden; for example, RRs for physician and hospital visits in the year pre‐index decreased steadily by age being 2.5–2.6 for 50‐year‐olds to 1.7 for 20‐year‐olds.

Associations between multiple sclerosis and physician visits before a first demyelinating event in Ontario, Canada (1991–2020), by age at the index date (from left to right: ages 20, 30, 40, and 50 years). For a description of the x ‐ and y ‐axes, see caption to Figure  1 .

Associations between multiple sclerosis and physician visits before a first demyelinating event in Sweden (2001–2019) by age at the index date (from left to right: ages 20, 30, 40, and 50 years). For a description of the x ‐ and y ‐axes, see caption to Figure  1 .

Associations between multiple sclerosis and hospital visits before a first demyelinating event in Ontario, Canada (1991–2020), by age at the index date (from left to right: ages 20, 30, 40, and 50 years). For a description of the x ‐ and y ‐axes, see caption to Figure  1 .

Associations between multiple sclerosis and hospital visits before a first demyelinating event in Sweden (2001–2019) by age at the index date (from left to right: ages 20, 30, 40, and 50 years). For a description of the x ‐ and y ‐axes, see caption to Figure  1 .

Associations between multiple sclerosis and emergency department (ED) visits before a first demyelinating event in Ontario, Canada (2000–2020), by age at the index date (from left to right: ages 20, 30, 40, and 50 years). For a description of the x ‐ and y ‐axes, see caption to Figure  1 .

These age‐related differences in RRs largely disappeared by the second‐ or third‐year pre‐index date in both regions. In Ontario, where the longest follow‐up pre‐index date was available, RRs remained elevated for most of the follow‐up, irrespective of the patients’ age at the index date. For example, for those aged 20 years, both hospital (all and inpatients) and physician visits were higher for PwMS versus controls in each year, and ED visits were higher in each of the 19 years. For the 50‐year‐olds, hospital, physician, and ED visits were each elevated for almost all available years; for example, for 28 (of 29) years for hospital visits, all 28 years for physician visits and 18 (of 19) years for ED visits. In Sweden, this extended period of high healthcare use was most evident for younger individuals. For example, physician visit RRs were elevated for those aged 20, 30, or 40 years at the index date for almost every year pre‐index until year 18 (of 19), and hospital visits through until year 17 (of 19) pre‐index.

SES and healthcare use

In both regions, the durations over which the different types of healthcare use were elevated for PwMS versus controls were similar regardless of SES, Figure  S4 . However, in the year pre‐index date, a gradient across socioeconomic quintiles was observed, with individuals in the most affluent quintiles showing modestly higher RRs for the ED (Ontario) and hospital (Sweden). RRs for ED visits ranged between 3.2–3.4 and 2.7 (quintiles 4.5 to 1.2). Similarly, RRs for hospital visits (Sweden) ranged from >3 to 1.8–2.1 (quintiles 4.5 to 1.2).

Urban versus rural residency and healthcare use

Across both regions, healthcare use patterns (duration and magnitude) for PwMS versus controls were largely similar regardless of a person's rural or urban residency, Figure  S5 , with one exception. For Ontario's ED visits, RRs were higher in the year pre‐index date for urban (RR = 3.14, 3.05–3.25) versus rural dwellers (RR = 2.30, 2.08–2.56).

Healthcare use patterns before MS onset: Sweden's clinical cohort

Rates of health service use, especially physician visits, were higher for PwMS versus controls up to 5 years before MS symptom onset. Similar to the administrative cohorts, the highest RRs were in the year pre‐index date, Figure  S6 . While largely similar patterns were observed by age, sex, Figures  S7 and S8 , and SES (not shown), estimates were modest. For example, as before, an age‐gradient was evident in the year pre‐index date, with the RRs for hospital visits being 1.5/1.4/1.3/1.2 for those aged 50/40/30/20 years at MS onset.

Discussion

In this large international matched cohort study, we observed an extended period of higher healthcare use before classical onset of MS. Yearly rates of healthcare use rose steadily for over two decades for PwMS versus matched population controls as MS onset (the index date) approached, escalating particularly 6–7 years pre‐MS onset. In Ontario, Canada, where the largest cohort and longest follow‐up were available, annual physician and hospital visit rates were consistently higher by ≥20% for 20 (of 29) years, rising to ≥30% for hospital visits in the 7 years pre‐index. A similarly extended period was observed for Sweden, with higher yearly physician visit rates up to 15 (of 19) years, exceeding 10% in each of the 6 years pre‐index date. Moreover, these higher rates were evident across a diverse range of healthcare settings, peaking 1–2 years pre‐index date in both regions. Among the highest were for Ontario's ED visits, where RRs were >3 times greater for PwMS versus controls in the year pre‐index, followed by inpatient hospitalizations and physician visits, which were more than twice as high in Canada and Sweden. Combined, findings suggest that MS may start decades before classical onset, with PwMS healthcare use trajectories deviating from controls in early life or early adulthood.

These results have several implications. First, findings place MS within the life course model of chronic disease,
14
complementing, but greatly extending the MS disease continuum concept.
15
This has public health implications, indicating the need for a life course approach to prevent MS. Second, findings inform the most likely period representing the MS prodromal phase. Third, healthcare disparities were evident; sociodemographics, particularly age and sex, affected healthcare use even during this early period. Fourth, findings are critical to inform the etiologically relevant time period when searching for risk factors for MS onset. Together, we propose a life course disease framework for MS. Briefly, early life adverse exposures, from in utero (e.g., maternal low sunlight/serum vitamin D,
16
smoking,
17
and diabetes
18
) to early adulthood (e.g., low sunlight/serum vitamin D, smoking, obesity, and maltreatment)
19
,
20
may place genetically susceptible individuals at heightened risk of disease. This is consistent with our observed higher healthcare use stemming back to early life. Biologically, MS initiates after accrual of a critical combination of exposures, including the necessary agent (Epstein Barr Virus). The prodromal phase is likely represented by the final escalation in healthcare use 6–7 years pre‐MS onset, consistent with serum neurofilament light level elevations observed over a similar period.
4
Our study, combined with recent advances in (multi‐panel) biomarker discovery,
21
suggest that there is tangible opportunity to identify such individuals earlier in their disease course. This is critical since delays in MS diagnosis are associated with greater disability accrual at diagnosis, lower brain reserve,
22
,
23
and delayed disease‐modifying drug initiation.
24

Sociodemographics affected findings, primarily in the final 1–2 years pre‐onset (“prodromal phase”), with males and those older at MS onset exhibiting higher increases in visit rates across healthcare settings in both regions. Also, disproportionately higher ED (Ontario) and hospital (Sweden) visit increases were observed among the most affluent participants (highest SES quintiles). Urban (vs. rural) dwellers also exhibited higher ED visit increases (Ontario). Sociodemographic factors influence healthcare use and outcomes after an MS diagnoses.
9
,
25
,
26
For example, males and older persons (who may also be more vulnerable due to lower brain reserve
27
) typically experience more rapidly progressing disease.
28
We demonstrate that these factors influence healthcare use even before classical manifestations of MS. We found few other comparable studies, although similar age and sex‐related patterns were observed, albeit in a smaller study, conducted in one Canadian province (British Columbia) over a shorter time period, in which all 5 years pre‐onset were combined.
8

Interestingly, and despite availability of universal healthcare, our observed effect of SES on ED and hospital visits in the year pre‐index date is suggestive of a very early indicator of health inequity. This is of potential concern; lower SES at MS onset is associated with poorer outcomes, including increased risk of disability accrual and mortality.
26
,
29
Social determinants of health affect healthcare use and access in complex ways, and can reflect individual‐level (wealth, education‐level) and structural factors.
25
While the disproportionally higher ED visit rates among urban dwellers with MS in the year pre‐index date may reflect, in part, the access to EDs and other healthcare services in such areas, combined, findings underscore the need to access diverse populations, including by SES and geography. Diversity is particularly relevant when incorporating data derived from routine healthcare visits into risk prediction tools aimed at identifying PwMS early in their disease. Such efforts are paramount to prevent bias and perpetuating existing healthcare disparities.
30

We examined the years before MS symptom onset in Sweden. Although determined retrospectively, this approach could provide valuable insights into the true prodromal phase.
1
,
3
Despite a smaller cohort, we still found evidence of higher healthcare use up to 5 years pre‐MS symptom onset. The ability to recognize MS earlier may depend on understanding both (naturally overlapping) time periods—before MS symptom onset and before a first demyelinating‐disease code. The latter is captured in real time, representing the patient's current movement though the health system and is likely the most amenable to “real‐time” risk prediction tool. It is also plausible that some aspects captured before a first demyelinating‐disease code in our study represent part of the diagnostic workup of MS and/or may be affected by referral wait times which could vary regionally and over time. While published literature are rather limited, wait times in both regions appear relatively short. According to a 2021/2022 Canada‐wide survey of healthcare professionals and neurologists working at MS clinics, the majority of survey respondents indicated that wait times were 3 months or less.
31
In Sweden, the “national care guarantee” in the Patient Care Act (2015) mandates that all referrals to a neurologist must result in an appointment within 90 days.
32
While we were unable to find published results indicating actual time to a neurologist appointment, we were able to access information from the Swedish MS Registry for all those who received a diagnosis of MS. From 2001 to 2024, the median time from referral to diagnosis was relatively short, ranging from 0.06 to 0.34 years (i.e., 22–124 days).
33
This, combined with findings from the Swedish clinical cohort support the interpretation that the increased healthcare use is likely related to a prodrome and not exclusively due to barriers to healthcare access for typical MS symptoms.

Strengths, weaknesses, and future opportunities

Study strengths included access to two international administrative cohorts in North America and Europe, maximizing generalizability of findings across regions. Moreover, this enabled diverse healthcare encounters to be examined over an extensive period. Selection bias was avoided by accessing population‐based data. Recall bias was virtually eliminated as data were recorded in “real‐time,” as the patient accessed healthcare. The MS and comparator cohorts without MS were also carefully matched by sex, birth year, and region (3‐digit postal code or county) at their index year, and residency time in the region, resulting in demographically very similar cohorts. We purposely examined the period before MS onset, defined as a first demyelinating event or MS symptom onset, the latter of which could arguably be least affected by delays in diagnosis or changes to the MS diagnostic criteria over the study period. However, it is possible that long delays in diagnosis could affect recall and the accuracy of the date of symptom onset in the clinical cohort. Despite our robust approach, residual confounding due to unmeasured variables may still occur. Limitations included inability to access complementary factors, such as lifestyle (diet and smoking), although these are difficult to capture on an entire population, and before disease onset. Both regions had universal healthcare; however, we could not capture issues that did not prompt healthcare use. We examined heterogeneity in healthcare use by age, sex, SES, and for urban/rural dwellers. Other social determinants of health may be of value to examine (cultural identity and education‐level/health literacy). Our findings are critical in helping inform the design of such future work. Studies are also warranted to phenotype healthcare encounters over a similar extended period before MS onset (beyond the typical 5 years examined previously).
34
Such phenotyping could include the types of medical specialists consulted (e.g., neurologists vs other specialities) in the years leading up to clinical MS onset.
34
Such information could potentially be used to assess if raising awareness among certain medical specialties that are consulted more often by people developing MS could help in the prompt identification of MS, without causing unintended harm.
35
Ethical, appropriate early identification of MS could, in the future, provide opportunity for enrollment in clinical trials of putative neuroprotective agents. A similar approach has shown promise in radiologically isolated syndrome (RIS).
36
,
37
,
38
However, while an important group of individuals, RIS is rare,
38
leading to significant challenges in clinical trial enrolment, exemplified by the early termination of a recent trial due to slow recruitment.
39
Moreover, inequity in MRI access can limit such opportunities.
40
,
41
In other neurologic diseases, prodromal symptomology combined with risk markers, demographics (e.g., age and sex), family history of disease, and imaging biomarkers are used to identify at‐risk individuals for neuroprotective trials.
42
,
43
For example, the Movement Disorder's criteria can identify prodromal Parkinson's with high specificity (99%), modest sensitivity (67%), and a positive predictive value of 40% over 3 years.
42
Another future possibility for MS could be a stepwise approach, whereby high‐risk individuals are flagged by the non‐specialist for additional screening using low‐cost biomarkers (e.g., serum neurofilament light) before additional testing is warranted.
1

Conclusion

Findings suggest an extended period of high healthcare use, evident up to two decades before MS onset, escalating 6–7 years before MS onset, and peaking in the year prior. Specific levels of healthcare were disproportionality used among PwMS versus controls, such as the ED. Healthcare use also differed across sociodemographic groups in the final 1–2 years leading up to classical MS onset, with the most (vs least) affluent patients disproportionately accessing the ED and hospital. A higher healthcare burden (visits) was also evident for males and persons older at onset. Efforts to identify PwMS in the earliest phase of disease need to include sociodemographically diverse populations and encompass all levels of healthcare, including primary and secondary/specialist/ED care. Such initiatives are required to maximize success in the early identification and management of MS globally, and are critical to avoid introducing bias into risk‐prediction tools.
21
Finally, findings inform a life course approach to understanding MS risk and is critical to public health efforts and disease prevention.

Author Contributions

HT: Conceptualization, funding acquisition, investigation, methodology, project administration, resources, supervision, and writing—original draft. FZ: Formal analysis, funding acquisition, investigation, methodology, visualization, and writing—review and editing. KE: Data curation, formal analysis, and writing—review and editing. AA: Data curation, formal analysis, and writing—review and editing. AM: Data curation, formal analysis, investigation, methodology, and writing—review and editing. PL: Data curation, formal analysis, and writing—review and editing. KMcK: Conceptualization, data curation, funding acquisition, investigation, methodology, and writing—review and editing. JH: Data curation, funding acquisition, project administration, resources, supervision, and writing—review and editing. YZ: Conceptualization, formal analysis, funding acquisition, investigation, methodology, supervision, and writing—review and editing. CM: Conceptualization, funding acquisition, investigation, methodology, project administration, supervision, validation, and writing—review and editing. RAM: Conceptualization, funding acquisition, investigation, methodology, visualization, and writing—review and editing.

Funding information

The study was funded in part by: the National MS Society and MS Canada (RG5063A4/1/RFA‐2103‐37392; EGID:P002/903124; PI:Tremlett). The study funders had no role in the methods, analyses, writing of the manuscript, or the decision to submit it for publication.

Conflict of Interest

Feng Zhu, Karl Everett, Ayesha Asaf, Ali Manouchehrinia, Ping Li and Yinshan Zhao: Have nothing to disclose. Kyla A McKay is funded by the Swedish Research Council for Health, Working Life and Welfare and has received speaker honoraria from Biogen (2022) and Sanofi‐Aventis (2023). Jan Hillert received honoraria for serving on advisory boards for Biogen, Bristol‐Myers‐Squibb, Janssen, Merck KGaA, Novartis, Sandoz, and Sanofi‐Genzyme and speaker's fees from Biogen, Janssen, Novartis, Merck, Teva, Sandoz, and Sanofi‐Genzyme. He has served as P.I. for projects sponsored by, or received unrestricted research support from, Biogen, Bristol‐Myers‐Squibb, Janssen, Merck KGaA, Novartis, Roche, and Sanofi‐Genzyme. His MS research is funded by the Swedish Research Council and the Swedish Brain foundation. Colleen Maxwell receives research funding from a University of Waterloo Research Chair, CIHR, Multiple Sclerosis Society of Canada, National Multiple Sclerosis Society, CMSC and the Public Health Agency of Canada. Ruth Ann Marrie receives research funding from: CIHR, MS Canada, Crohn's and Colitis Canada, National Multiple Sclerosis Society, CMSC, The Arthritis Society, US Department of Defense, Pfizer Foundation. Helen Tremlett has, in the last 5 years, received research support from the Canada Research Chair Program, the National Multiple Sclerosis Society, the Canadian Institutes of Health Research, Multiple Sclerosis Canada, the Multiple Sclerosis Scientific Research Foundation and the EDMUS Foundation (“Fondation EDMUS contre la sclérose en plaques”). In addition, in the last 5 years, has had travel expenses or registration fees prepaid or reimbursed to present at CME conferences or attend meetings (as a member of the International Advisory Committee on Clinical Trials in Multiple Sclerosis) from the Consortium of MS Centres (2023), the Canadian Neurological Sciences Federation (2023), National MS Society (2022, 2023, 2024), ECTRIMS/ ACTRIMS (2017–2024), and American Academy of Neurology (2019). Speaker honoraria are either declined or donated to an MS charity or to an unrestricted grant for use by HT's research group.

Supporting information

Figure S0 .Overview of the health administrative and clinic cohorts, the index dates, and study period.

Figure S1 . The associations between multiple sclerosis (MS) and inpatient hospital visits and prescriptions filled before a first demyelinating event in Ontario, Canada (1991–2020) and Sweden (2001–2019).

Figure S2 . The associations between multiple sclerosis (MS) and inpatient hospital visits and prescriptions filled before a first demyelinating event in Ontario, Canada (1991–2020) and Sweden (2001–2019) by sex.

Figure S3 . The associations between multiple sclerosis (MS) and inpatient hospital visits and prescriptions filled before a first demyelinating event in Ontario, Canada (1991–2020) and Sweden (2001–2019) by age.

Figure S4 . (Panels A–H). The associations between multiple sclerosis (MS) and physician, hospital (all and in‐patients separately), emergency department visits and prescriptions filled before a first demyelinating event in Ontario, Canada (1991–2020) and Sweden (2001–2019) by socioeconomic status.

Figure S5 . (Panels A–H). The associations between multiple sclerosis (MS) and physician, hospital (all and inpatients separately), emergency department visits and prescriptions filled before a first demyelinating event in Ontario, Canada (1991–2020) and Sweden (2001–2019) by rurality of residence.

Figure S6 . (Panels A–C). The associations between multiple sclerosis (MS) and physician, hospital visits before MS symptom onset in Sweden (2001–2019).

Figure S7 . (Panels A–D). The associations between multiple sclerosis (MS) and physician, hospital and emergency department visits before MS symptom onset in Sweden (2001–2019) by sex at the index date.

Figure S8 . (Panels A–D). The associations between multiple sclerosis (MS) and physician, hospital visits before MS symptom onset in Sweden (2001–2019) by age at the index date.

Table S1 . Summary of health administrative and clinical data and its availability over the study period by region.

Table S2 . Summary of the International Classification of Diseases (ICD‐9/10) CNS demyelinating disease‐related diagnostic codes used to identify the index date (multiple sclerosis onset) in the health administrative cohorts.

Table S3 . Characteristics of the people with multiple sclerosis (PwMS) and matched controls at MS onset (the index date) used to examine healthcare use in the years leading up to MS onset in Canada for ED visits (2000/1–2020) and Sweden for prescriptions filled (2005–2019).

Acknowledgments

This study was in part supported by ICES (formerly the “Institute for Clinical Evaluative Sciences”), which is funded by an annual grant from the Ontario Ministry of Health (MOH) and the Ministry of Long‐Term Care (MLTC). ICES is an independent, nonprofit research institute whose legal status under Ontario's health information privacy law allows it to collect and analyze healthcare and demographic data, without consent, for health system evaluation and improvement. We acknowledge the members of the team who facilitated gaining funding for the wider program of work related to the MS prodrome (Dr. M. Ehsan Karim, Dr. José Wijnands, and Fardowsa L.A. Yusuf, MSc). JW also contributed to designing the technical plan while a research associate at the University of British Columbia, Canada.

Funding Statement

This work was funded by National Multiple Sclerosis Society
grant RG5063A4/1/RFA‐2103‐37392; Multiple Sclerosis Society of Canada
grant EGID:P002/903124.

Data Availability Statement

Data sharing agreements prohibit public availability. Access maybe granted via ICES (for details, see acknowledgments) and upon reasonable request for Sweden’s de‐identified data, as permitted by current legislation protecting personal data. This document used data adapted from the Statistics Canada Postal Code OM Conversion File, which is based on data licensed from Canada Post Corporation, and/or data adapted from the Ontario Ministry of Health Postal Code Conversion File, which contains data copied under license from ©Canada Post Corporation and Statistics Canada. The dataset from this study is held securely in coded form at ICES. While legal data sharing agreements between ICES and data providers (e.g., healthcare organizations and government) prohibit ICES from making the dataset publicly available, access may be granted to those who meet prespecified criteria for confidential access, available at www.ices.on.ca/DAS (email: das@ices.on.ca ). The full dataset creation plan and underlying analytic code are available from the authors upon request, understanding that the computer programs may rely upon coding templates or macros that are unique to ICES and are therefore either inaccessible or may require modification. Parts of this material are based on data and/or information compiled and provided by the Ontario Ministry of Health and the Canadian Institute for Health Information (CIHI). The analyses, conclusions, opinions, and statements expressed herein are solely those of the authors and do not reflect those of the funding or data sources; no endorsement is intended or should be inferred.

References

Associated Data

Supplementary Materials

Figure S0 .Overview of the health administrative and clinic cohorts, the index dates, and study period.

Figure S1 . The associations between multiple sclerosis (MS) and inpatient hospital visits and prescriptions filled before a first demyelinating event in Ontario, Canada (1991–2020) and Sweden (2001–2019).

Figure S2 . The associations between multiple sclerosis (MS) and inpatient hospital visits and prescriptions filled before a first demyelinating event in Ontario, Canada (1991–2020) and Sweden (2001–2019) by sex.

Figure S3 . The associations between multiple sclerosis (MS) and inpatient hospital visits and prescriptions filled before a first demyelinating event in Ontario, Canada (1991–2020) and Sweden (2001–2019) by age.

Figure S4 . (Panels A–H). The associations between multiple sclerosis (MS) and physician, hospital (all and in‐patients separately), emergency department visits and prescriptions filled before a first demyelinating event in Ontario, Canada (1991–2020) and Sweden (2001–2019) by socioeconomic status.

Figure S5 . (Panels A–H). The associations between multiple sclerosis (MS) and physician, hospital (all and inpatients separately), emergency department visits and prescriptions filled before a first demyelinating event in Ontario, Canada (1991–2020) and Sweden (2001–2019) by rurality of residence.

Figure S6 . (Panels A–C). The associations between multiple sclerosis (MS) and physician, hospital visits before MS symptom onset in Sweden (2001–2019).

Figure S7 . (Panels A–D). The associations between multiple sclerosis (MS) and physician, hospital and emergency department visits before MS symptom onset in Sweden (2001–2019) by sex at the index date.

Figure S8 . (Panels A–D). The associations between multiple sclerosis (MS) and physician, hospital visits before MS symptom onset in Sweden (2001–2019) by age at the index date.

Table S1 . Summary of health administrative and clinical data and its availability over the study period by region.

Table S2 . Summary of the International Classification of Diseases (ICD‐9/10) CNS demyelinating disease‐related diagnostic codes used to identify the index date (multiple sclerosis onset) in the health administrative cohorts.

Table S3 . Characteristics of the people with multiple sclerosis (PwMS) and matched controls at MS onset (the index date) used to examine healthcare use in the years leading up to MS onset in Canada for ED visits (2000/1–2020) and Sweden for prescriptions filled (2005–2019).

Data Availability Statement

Data sharing agreements prohibit public availability. Access maybe granted via ICES (for details, see acknowledgments) and upon reasonable request for Sweden’s de‐identified data, as permitted by current legislation protecting personal data. This document used data adapted from the Statistics Canada Postal Code OM Conversion File, which is based on data licensed from Canada Post Corporation, and/or data adapted from the Ontario Ministry of Health Postal Code Conversion File, which contains data copied under license from ©Canada Post Corporation and Statistics Canada. The dataset from this study is held securely in coded form at ICES. While legal data sharing agreements between ICES and data providers (e.g., healthcare organizations and government) prohibit ICES from making the dataset publicly available, access may be granted to those who meet prespecified criteria for confidential access, available at www.ices.on.ca/DAS (email: das@ices.on.ca ). The full dataset creation plan and underlying analytic code are available from the authors upon request, understanding that the computer programs may rely upon coding templates or macros that are unique to ICES and are therefore either inaccessible or may require modification. Parts of this material are based on data and/or information compiled and provided by the Ontario Ministry of Health and the Canadian Institute for Health Information (CIHI). The analyses, conclusions, opinions, and statements expressed herein are solely those of the authors and do not reflect those of the funding or data sources; no endorsement is intended or should be inferred.