?Mathematical formulae have been encoded as MathML and are displayed in this HTML version using MathJax in order to improve their display. Uncheck the box to turn MathJax off. This feature requires Javascript. Click on a formula to zoom.Abstract
Objective
We utilized national claims-based data to identify the change in odds of diagnosis of ALS following possible-ALS-symptoms-and whether the change varies in urban/rural areas.
Methods
Insurance claims were obtained from the Merative MarketScan databases, 2001–2021 in the United States. Individuals with incident ALS were identified and matched on age, sex, and enrollment period to individuals without ALS. For all individuals, claims for 8 possible-ALS-symptoms in the time before any ALS diagnosis were identified. We then used conditional logistic regression to estimate the odds of being diagnosed with ALS following these symptoms and whether the association varied by urban/rural location.
Results
19,226 individuals with ALS were matched to 96,126 controls. Patients with ALS were more likely to live in an urban area (87.0% vs 84.5%). Of those with ALS 84% had 1+ of our 8 possible-ALS-symptom compared to 51% of controls. After adjustment for confounders, having possible-ALS-symptoms increased the odds of a future ALS diagnosis by nearly 5-fold. A dose-response pattern was present with increasing odds as the number of symptoms increased. In all models, urban areas were associated with increased odds of diagnosis with ALS while the effect of having a symptom was smaller in urban places. Urban cases of ALS are diagnosed at younger ages.
Conclusions
These results suggest symptoms may appear and be noted years before the diagnosis of ALS. Additionally, rural patients are diagnosed at later ages with a greater dependence on symptoms than urban patients. These results highlight potential improvements for screening for ALS.
Keywords:
Introduction
Improving quality of life and survival among people with amyotrophic lateral sclerosis (ALS) depends on early diagnosis. Yet symptoms later attributed to ALS may appear a year before diagnosis (Citation1). Understanding the timing and magnitude of possible-ALS-symptoms may inform prodromal and diagnostic red flags. Despite symptoms being present, an ALS diagnosis may be missed when the symptoms are overlooked or ambigious. The potential for missed symptoms may be especially high for patients with limited access or long referral times to specialists, such as those living in rural areas. Neurologists tend to practice inurban centers (Citation2–4) resulting in greater travel distances for rural patients (Citation5). These barriers to care may cause rural patients to be more likely to have their symptoms missed.
In this study, we quantify the association between having possible-ALS-symptoms and later having a diagnosis of ALS while stratifying metropolitan or non-metropolitan county. We hypothesized that the risk of ALS will be increased in people following possible-ALS-symptoms and further than the increase will vary according to whether the patient lives in an metropolitan or non-metropolitan county.
Materials and methods
Data source
Data from the Merative Marketscan Commercial Claims and Encounters and the Medicare Coordination of Benefits databases from 2001 to 2021 were used for this study. The Commercial Claims and Encounters database is private insurance of a mostly 64 population while the Medicare Coordination of Benefits database is private Medicare supplemental health plans. Most people under 65 have a private health insurance while Medicare supplemental plans are purchased by about 20% of Americans over the age of 65. Unlike analysis of electronic health records, this population is insured and may differ from the uninsured or those in Medicare Part C.
These datasets include claims-based data from inpatient, outpatient, and pharmacy claims with longitudinal follow-up during the enrollment period (Citation6). The University of Iowa IRB has determined analysis of Marketscan is exempt from review (analysis of deidentified secondary data). No information prior to 2001 was included in this study.
Case definition
We identified patients with ALS using ICD-9-CM (335.20) and ICD-10-CM (G12.21) diagnosis codes on any claim. Patients were also included if they had taken the ALS-specific medications edaravone or riluzole. Sodium phenylbutyrate-taurursodiol was not included as release occurred in September 2022. Medication dispensing events were identified using National Drug Code numbers. The date of diagnosis was defined as either the earliest claim with an ALS diagnosis or dispensing of ALS medication. We required at least one year of enrollment in Marketscan prior to the diagnosis date to ensure we are identifying new cases. These codes were selected to focus as narrowly as possible on ALS and not motor neuron diseases more broadly.
We identified healthy controls – people without ALS – that were the same sex, age, and enrolled in Marketscan during the same period as the case. For instance, a male enrollee with ALS born in 1933, who first appears in the data in 2003, and is diagnosed with ALS in 2007 would be matched to a male born in 1933, enrolled from 2003 to 2007, but without an ALS diagnosis or prescription at any point during their enrollment period. While we have information during the entire enrollment period, we do not have information about diagnoses before or after the individual was enrolled and contributing data to Marketscan. Up to five controls were identified for each case.
Possible-ALS-symptom identification
ALS symptoms were based on codes previously reported in an analysis of Medicare data (Citation7). Additionally, in our criteria we added falls as a symptom. The considered symptoms are classified into two categories – bulbar or motor symptoms. We retained subdivisions within each as described previously (Citation7). We cross-walked ICD-9-CM codes to ICD-10-CM using the Center for Medicare and Medicaid Services General Equivalence Crosswalk. The codes used are described in . We included only diagnoses from before the ALS diagnosis date. As these diagnoses occurred before the diagnosis of ALS and the exact cause of the symptoms is uncertain, we use the terms “possible-ALS-symptoms” to describe symptoms consistent with the clinical presentation of ALS but where ALS pathology is not the only possible cause.
Table 1. Diagnosis codes used to define ALS symptoms by origin and sub-classification.
Urban/rural categorization
We defined all counties included in a metropolitan statistical area (MSA) as an urban area. MSAs are metros with populations of at least 50,000 defined by the Office of Management and Budget. If a person was ever reported as having lived in an MSA in Marketscan, we coded them as metropolitan. Meanwhile, enrollees who do not have a code for residing in an MSA were defined as rural. Roughly 85% of Americans and a similar percentage of enrollees in Marketscan live in one of these MSAs. Of the metropolitan residents in our data, 93.5% exclusively lived in a metropolitan area.
Primary analysis
Our primary outcome was whether the person is diagnosed with or treated for ALS. Our primary exposures are 1) whether they previously had any possible-ALS-symptoms, 2) whether they resided in a metropolitan or non-metropolitan area, and 3) an interaction between these two exposures. We included adjustment for healthcare utilization during the pre-ALS period, complexity, and the Elixhauser comorbidities (Citation8–10). Estimation was performed using conditional logistic regression and included the sampling strata to account for matching.
Secondary analyses
We conducted secondary analyses to better understand the relationship between past possible-ALS-symptoms, urban dwelling, and ALS. First, we estimated “system-specific” models to determine whether the association differed whether the symptoms were motor or bulbar origin. Second, we included the number of the eight defined symptoms that were observed to look for a dose-response. Third, since the different symptoms may have different relationships with the odds of developing ALS, we estimated eight symptom-specific models.
A provider noting symptoms and ordering a referral to neurology is an example of idealized care – this is not a delay. However, this pattern would result in finding an association between the symptoms and the future diagnosis. To address this possible concern, we fit a series of regressions where we excluded people who were first diagnosed with the possible-ALS-symptom within a short period of time before they were diagnosed with ALS. If our result is due entirely due to unavoidable delays associated with a referral and testing, then we would expect the OR to decrease rapidly to a relatively constant estimate after a short exclusion period (∼90 days). On the other hand, if this association persists with long exclusion periods (∼1 year), then it is likely that these diagnoses of symptoms may be overlooked symptoms of ALS. We vary the exclusion period from 30 days to 2 years so that we can identify the point at which the association starts increasing. The exclusion period where the association starts to increase is a plausible estimate of the earliest time when symptoms are manifest, noted by the medical provider, but do not result in a diagnosis of ALS. Note that the date of diagnosis of ALS is not the same as the date of symptom onset or when ALS pathology is present. Next, given the lack of previously established urban/rural differences in incidence, we theorized any observed differences result from different rates of detection/diagnosis due to greater access to care. If ALS incidence in urban and rural areas is the same but people residing in urban areas are diagnosed earlier due to greater access to care, we would expect ALS cases in urban areas to be diagnosed at younger ages. We performed a survival analysis to quantify differences in the age at diagnosis. Specifically, we estimated the Kaplan-Meier survival function and Cox regression stratified by 1) living in a metropolitan or non-metropolitan country, 2) having or not having ALS symptoms, and 3) the interaction of these two variables.
Finally, as urban patients, as described in the introduction, face lower barriers to obtaining all types of medical care, we may expect all diagnoses to be elevated compared to rural residents. To evaluate this potential threat, we used a quasi-Poisson regression to estimate the relationship between the number of days with health care encounters and living in an urban or rural area among the control enrollees.
All analysis were completed in R 4.2.2 (Citation11). The conditional logistic regression, Kaplan-Meier, and Cox proportional hazards models were estimated with the survival package (Citation12). For all regression models, standard errors were clustered by the sampling strata. In the Cox models, we used robust standard errors.
Results
We identified 19,226 people newly diagnosed with ALS and 96,126 people without ALS matched based on age, sex, and enrollment period. Patients with ALS are slightly more likely to live in urban areas (87.0% of people with an ALS diagnosis live in an urban area compared to 84.5% of the matched non-ALS control enrollees) and are much more likely to have one or more of the possible-ALS-symptoms (83.7% vs 50.7%), . Differences were larger for higher number of symptoms (e.g., 25.5% of people who will develop ALS had 4 symptoms compared to only 4.7% of controls) and specific symptoms (e.g., 35.1% of people who will develop ALS had muscle weakness compared to 5.2% of controls).
Table 2. Summary of Study Cohort. All values are reported as number (percent) when discrete or mean (inner quartile range) for continuous values. Age, sex, and enrollment time was including in the matching scheme and noted here (all with standardized mean differences below 0.001).
Both having one or more possible-ALS-symptoms (OR = 4.92; 95% CI: 4.36, 5.56) and living in an urban area (OR = 1.44; 95% CI: 1.28, 1.61) were independent risk factors a diagnosis of ALS, . The interaction between having a possible-ALS-symptom and living in an urban area is less than 1 (OR = 0.83; 95% CI: 0.73, 0.94). In other words, having a symptom in a rural dweller increases the odds of an ALS diagnosis by a factor of 4.92 while the same symptom has a smaller increase when the person lives in an urban area (0.83 × 4.92 = 4.08) – having a possible-ALS-symptom reduces the urban bias of ALS. Considering only motor or only bulbar symptoms had similar increases in the odds of getting an ALS diagnosis (OR = 3.75 and 3.80). The OR for living in an urban area and the interaction were similar between all three models. Full details are included in Supplemental Tables S1–S3.
Table 3. Adjusted odds ratios for acquiring a future diagnosis of ALS after having any of the study symptoms, only bulbar, or only motor symptoms, living in an urban area, and the interaction of having the symptom and living in an urban area.
The individual symptoms had ORs above 1, , and all except falls and swallowing problems were statistically significant. The main effect for living in a metropolitan county was similar at an OR of approximately 1.3 across all symptom-specific models. The interactive effect was significant and negative for four (speech problems, muscle weakness, pain, and other symptoms), non-significant and negative for two (gait problems, involuntary movements) and non-significant and positive for two (swallowing problems, falls). Coefficients are reported in Tables S4-S11.
Figure 1. Odds Ratio for a Future Diagnosis of ALS by Specific Symptom. All symptoms were associated with increased odds of a future ALS diagnosis; however, the effect was variable by the specific symptom. More common and less ALS-specific symptoms had lower odds ratios. Points are the estimated OR, lines denote the 95% CI.
Table 4. Full model estimates for acquiring an ALS diagnosis by number of symptoms. Note the interaction between living in an urban area and having 8 symptoms could not be estimated (there were not individuals with 8 symptoms in both rural and urban settings).
There was a dose-response relationship between the number of symptoms and the odds of developing ALS, . While any number of symptoms was associated with statistically significant increased odds, the greatest increase occurring at 5 symptoms (OR = 33.20). The interaction effect between number of symptoms and living in an urban area remained consistent at an OR near 0.7.
The odds of an ALS diagnosis following possible-ALS-symptoms are persistently high even if we impose exclusion periods of over a year, . Full coefficients are reported in Table S12 for selected delays. Starting 1 year before the diagnosis date, the relationship between having possible-ALS-symptoms and a future ALS diagnosis starts to increase rapidly. This increase in the OR indicates an increase in possible-ALS-symptoms starting 1year prior to ALS diagnosis.
Figure 2. Odds Ratio for a Future Diagnosis of ALS by Exclusion Period Length. For each exclusion period, we estimate a model where we exclude anyone who had their first possible-ALS-symptoms within the exclusion period before their ALS diagnosis. For example, if the exclusion period was 90 days, we would remove any cases of ALS where the first time they were diagnosed with the possible-ALS-symptom was less than 90 days before they were diagnosed with ALS. Each point is the estimate from a model incorporating that exclusion period and the lines denote 95% CIs. The OR associated with possible-ALS-symptoms starts to increase rapidly starting approximately 1 year before the person is diagnosed with ALS. This suggests ALS symptoms may start appearing and being noted clinically at least 1 year before the diagnosis of ALS is made.
Using the Kaplan-Meier estimator, we find a younger age at diagnosis of ALS for those who live in urban areas compared to those who do not, . Having any possible-ALS-symptoms, , and having more symptoms, , also increases the hazard of developing ALS. When considered jointly, , both having symptoms and living in a metropolitan area increase the hazard of being diagnosed with ALS; however, the difference between the urban and rural curves is smaller in those with symptoms than without. Cox proportional hazards regression shows that these shifts are statistically significant and persist after adjustment for confounders, .
Figure 3. Kaplan-Meier Survival Curves for Age at ALS Diagnosis. Panel A shows a younger age-at-diagnosis among people living in urban vs non-urban locations. Panel B shows a younger age-at-diagnosis for people with vs without possible-ALS-symptoms. Panel C shows younger age-at-diagnosis with more possible-ALS-symptoms. Panel D shows an interaction between having symptoms and living in an urban area – both with and without symptoms, people in urban areas are diagnosed at younger ages than their non-urban peers; however, the difference between urban and non-urban patients with a symptom history is much smaller.
Table 5. Cox proportional hazards regression on age at diagnosis with ALS. Hazard ratios > 1 indicate younger age at diagnosis.
We find an 8% increase in the overall number of days with health care utilization among urban control enrollees compared to rural control enrollees (IRR = 1.08, 95% CI: 1.06, 1.10). The median number of days with a health care claim per year of enrollment was 7.84 (IQR: 3.58–15.5) among urban enrollees and 7.49 (IQR: 3.50–14.6) among rural enrollees.
Discussion
We tested the hypotheses that 1) symptoms of ALS appear in administrative record before the ALS diagnosis, 2) there are urban/rural differences in the rate of diagnosis, and 3) there are interactions between these exposures. We found having any of eight possible-ALS-symptoms is associated with a nearly 5-fold increase in the odds of being diagnosed with ALS in the future. Across all models, urban dwellers had greater odds of being diagnosed with ALS; however, the effect of having a symptom was smaller among urban residents compared to rural residents.
We believe the urban/rural differences in the rate of diagnosis do not represent a difference in ALS risk or ALS presentation in rural areas versus urban ones but rather that ALS diagnosis was more common in urban areas due to greater access to care. While diagnoses of all types were more common among urban control enrollees compared to rural control enrollees, the increase was approximately 8% - far less than the 40% seen for ALS. This suggests the barriers to care for complex conditions, like ALS, exceed the barriers to more typical types of health care services. Additionally, our survival analysis found that having possible-ALS-symptoms or living in an urban area was associated with a younger age at ALS diagnosis. If the age-dependent incidence of ALS is similar between urban and rural places (Citation13,Citation14), this finding would suggest people with ALS are being diagnosed earlier in their disease course and with presumably lower symptom burden when they live in urban areas. Indeed, ALS may be more common in rural settings because of greater pesticide exposures (Citation13,Citation15). If this is the case, our result would be stronger as we found a greater incidence of diagnosis among urban dwellers.
There appeared to be no difference in whether these were chiefly motor or bulbar symptoms; both were associated with a nearly 4-fold increase in the odds of developing an ALS diagnosis. Certain symptoms had greater impacts on the odds of acquiring an ALS diagnosis. For instance, muscle weakness had an OR of nearly 10 while other, less-specific symptoms (e.g., falls) had lower ORs. There was a dose-response pattern between the number of symptoms and the odds of being diagnosed with ALS which peaked at five symptoms (OR = 33.20 compared to no symptoms). The decline in ORs at very high number of symptoms likely stems from ALS having been considered and ruled out.
Our finding that possible-ALS-symptoms predate the diagnosis of ALS is supported by prior literature. A chart review of people diagnosed with ALS found the earliest symptoms included limb and mobility problems, bulbar symptoms, and respiratory problems (Citation16). Another study of 202 people with limb-onset ALS found that half of patients first presented to a non-neurologist, which had implications for time to diagnosis – patients who presented to orthopedists had a longer time to diagnosis (Citation17). Our work directly builds and expands upon the prior analysis performed using Medicare claims finding increased rates of the 8 symptoms considered here (Citation7). Our findings and the broader literature are consistent with current paradigms for prodromal and other early ALS symptoms.
Our study has several limitations. First, inherent to all retrospective database studies, we are limited by what is recorded in the administrative record. We observe symptoms only through diagnoses and only observe diagnoses that are recorded. This is necessarily a less rich data source than contained in a chart or clinical evaluation. However, health insurance claims data have the advantage of being prospectively collected and sheer number of patients. Indeed, our sample of nearly 20,000 new cases of ALS represents nearly 20% of all cases likely newly diagnosed between 2001–2021 in the United States (Citation18). Such large numbers of cases would be impossible with chart review or clinical evaluation. Additionally, we are bound by our years of data and have no information about enrollees before or after their enrollment. Second, the presentation of ALS may overlap with other neurological diseases and up to 10% of people with ALS may be undiagnosed (Citation19,Citation20). While the ALS diagnostic codes are likely very specific (>99%) and sensitive (78.9%-91.6%) these codes may follow a series of delayed or inaccurate non-ALS diagnoses(Citation21–23). It is possible that our non-ALS controls may develop ALS after the end of their enrollment period. However, if we misclassify cases as controls this would bias our estimates toward no association. Third, we defined urban residents as anyone who ever resided in a metropolitan county. This may misclassify residents of rural parts of counties as urban. Some rural dwellers are near specialist care while others may face very long distances to obtain care. While access to care correlates with rurality, there exists significant heterogeneity in terms of access to specialist health care within that case. Fourth, we used a limited set of a priori defined possible-ALS-symptoms. While we found a potential relationship between these symptoms, urban/rural settings, and future hazard of ALS, the relationship for other symptoms may occur even earlier, have a stronger association, or both. Discovering a broader set of “pre-ALS” symptoms should be explored in future work.
Retrospective analyses find symptoms precede diagnosis of ALS by a year or more. There is an urgent, unmet need to accelerate the detection-to-diagnosis timeline for ALS. Our results suggest that rural populations, which are disproportionately elderly and face limited access to specialist care, are more dependent on major symptoms to arrive at a diagnosis compared to their urban peers. With later, more severe disease at diagnosis, these individuals will have spent more time living with unmitigated symptoms of ALS including higher risk of falls, injuries, and other preventable morbidity. Outreach and telehealth have the potential to reduce some of these disparities (Citation24,Citation25); however, any effort to improve the detection-to-diagnosis process will need to consider the particular needs of rural populations. A failure of improvement to consider the barriers and challenges of the rural elderly will exacerbate the existing disparities in access to care, diagnosis, and outcomes.
Declaration of interest
The authors report there are no competing interests to declare.
Data availability and deposition
The data used in this analysis was used under a license agreement with Merative Marketscan. The authors cannot redistribute this data; however, interested users may contact Merative for licensing terms.
Supplemental Material
Download MS Word (64.1 KB)Code availability
The data extraction, processing, and analysis code is shared at https://github.com/iacobus42/als-sx.
Correction Statement
This article was originally published with errors, which have now been corrected in the online version. Please see Correction (http://dx.doi.org/10.1080/21678421.2024.2326288).
Additional information
Funding
References
- Richards D, Morren JA, Pioro EP. Time to Diagnosis and Factors Affecting Diagnostic Delay in Amyotrophic Lateral Sclerosis. In: Araki T, editor. Amyotrophic Lateral Sclerosis. Brisbane (AU): Exon Publications Copyright: The Authors.; 2021.
- Lin CC, Callaghan BC, Burke JF, Skolarus LE, Hill CE, Magliocco B, et al. Geographic variation in neurologist density and neurologic care in the United States. Neurology 2021;96:e309–e21.
- Hammond G, Luke AA, Elson L, Towfighi A, Joynt Maddox KE. Urban-rural inequities in acute stroke care and in-hospital mortality. Stroke 2020;51:2131–8.
- Horton DK, Graham S, Punjani R, Wilt G, Kaye W, Maginnis K, et al. A spatial analysis of amyotrophic lateral sclerosis (ALS) cases in the United States and their proximity to multidisciplinary ALS clinics, 2013. Amyotroph Lateral Scler Frontotemporal Degener. 2018;19:126–33.
- Armstrong MD, Hansen G, Schellenberg KL. Rural residence and diagnostic delay for amyotrophic lateral sclerosis in Saskatchewan. Can J Neurol Sci. 2020;47:538–42.
- Kulaylat AS, Schaefer EW, Messaris E, Hollenbeak CS. Truven health analytics marketscan databases for clinical research in colon and rectal surgery. Clin Colon Rectal Surg. 2019;32:54–60.
- Williams JR, Fitzhenry D, Grant L, Martyn D, Kerr DA. Diagnosis pathway for patients with amyotrophic lateral sclerosis: retrospective analysis of the US Medicare longitudinal claims database. BMC Neurol. 2013;13:160.
- Cai R, Zhang Y, Simmering JE, Schultz JL, Li Y, Fernandez-Carasa I, et al. Enhancing glycolysis attenuates Parkinson’s disease progression in models and clinical databases. J Clin Invest. 2019;129:4539–49.
- Zhang Q, Schultz JL, Aldridge GM, Simmering JE, Kim Y, Ogilvie AC, et al. COVID-19 case fatality and Alzheimer’s disease. J Alzheimers Dis. 2021;84:1447–52.
- Simmering JE, Welsh MJ, Schultz J, Narayanan NS. Use of glycolysis-enhancing drugs and risk of Parkinson’s disease. Mov Disord. 2022;37:2210–6.
- Team. RC. R: A Language and Environment for Statistical Analysis 2021 [4.2.2.]. Available from: https://www.R-project.org/.
- TM. T, T. L, A. E, C. C. survival: Survival Analysis 2023 [Available from: https://cran.r-project.org/web/packages/survival/.
- Breckenridge CB, Berry C, Chang ET, Sielken RL, Jr., Mandel JS. Association between Parkinson’s disease and cigarette smoking, rural living, well-water consumption, farming and pesticide use: systematic review and meta-analysis. PLoS One. 2016;11:e0151841.
- Ayton D, Ayton S, Barker AL, Bush AI, Warren N. Parkinson’s disease prevalence and the association with rurality and agricultural determinants. Parkinsonism Relat Disord. 2019;61:198–202.
- Korner S, Kammeyer J, Zapf A, Kuzma-Kozakiewicz M, Piotrkiewicz M, Kuraszkiewicz B, et al. Influence of environment and lifestyle on incidence and progress of amyotrophic lateral sclerosis in A German ALS population. Aging Dis. 2019;10:205–16.
- Lee JW, Kang S-W, Choi WA. Clinical course of amyotrophic lateral sclerosis according to initial symptoms: an analysis of 500 cases. Yonsei Med J. 2021;62:338–43.
- Kano O, Iwamoto K, Ito H, Kawase Y, Cridebring D, Ikeda K, et al. Limb-onset amyotrophic lateral sclerosis patients visiting orthopedist show a longer time-to-diagnosis since symptom onset. BMC Neurol. 2013;13:19.
- Mehta P, Raymond J, Punjani R, Larson T, Han M, Bove F, et al. Incidence of amyotrophic lateral sclerosis in the United States, 2014–2016. Amyotroph Lateral Scler Frontotemporal Degener. 2022;23:378–82.
- Kwan J, Vullaganti M. Amyotrophic lateral sclerosis mimics. Muscle Nerve. 2022;66:240–52.
- Ghasemi M. Amyotrophic lateral sclerosis mimic syndromes. Iran J Neurol. 2016;15:85–91.
- Pisa FE, Verriello L, Deroma L, Drigo D, Bergonzi P, Gigli GL, et al. The accuracy of discharge diagnosis coding for amyotrophic lateral sclerosis in a large teaching hospital. Eur J Epidemiol. 2009;24:635–40.
- Beghi E, Logroscino G, Micheli A, Millul A, Perini M, Riva R, et al. Validity of hospital discharge diagnoses for the assessment of the prevalence and incidence of amyotrophic lateral sclerosis. Amyotroph Lateral Scler Other Motor Neuron Disord 2001;2:99–104.
- Chiò A, Ciccone G, Calvo A, Vercellino M, Di Vito N, Ghiglione P, et al. Validity of hospital morbidity records for amyotrophic lateral sclerosis. A population-based study. J Clin Epidemiol. 2002;55:723–7.
- Fabbri M, Caldas AC, Ramos JB, Sanchez-Ferro Á, Antonini A, Růžička E, et al. Moving towards home-based community-centred integrated care in Parkinson’s disease. Parkinsonism Relat Disord. 2020;78:21–6.
- Wechsler LR, Tsao JW, Levine SR, Swain-Eng RJ, Adams RJ, Demaerschalk BM, et al. Teleneurology applications: report of the telemedicine Work Group of the American Academy of Neurology. Neurology 2013;80:670–6.