Prevalence and mortality of patients with multiple sclerosis in France in 2012: a study based on French health insurance data

This was a cross-sectional study using data from the French National Health Insurance database (SNIIRAM) linked with the National hospital discharge database (PMSI) [12]. The SNIIRAM database contains individualized, anonymous and comprehensive data on all reimbursements of patients’ health expenditure. Data are derived from the various French Health Insurance schemes, including the General scheme and the various special insurance schemes. Altogether, these schemes cover nearly the entire population residing in France (65.5 million inhabitants in 2012). Besides data on health expenditure reimbursements, demographic data are available, including the year of birth, gender, area of residence, geographical deprivation [13], and date of death. The SNIIRAM also contains information on patient eligibility for 100% health insurance coverage for LTD encoded in the International Classification of Diseases, 10th Revision (ICD-10). Similar diagnostic coding is used for the payment of disability pensions for patients of the general scheme. The SNIIRAM database is linked by a unique patient identifier to the PMSI database which provides medical information about all private and public hospital stays in France, including hospitalisation dates and diagnoses coded according to ICD-10.

We defined four selection criteria to identify a prevalent MS case: reimbursement of a disease-modifying treatment, LTD for MS, disability pensions for MS, and hospitalisation for MS. Since MS is a chronic disease with low or no relapse activity for several years, our selection period was not limited to 2012. To maximize the chances of identifying MS cases, we extended the identification period back in time according to the maximum data storage period available for this study in the databases, i.e., 2011–2012 for SNIIRAM and 2008–2012 for the PMSI.

The selection criteria were: (1°) at least one reimbursement for a disease-modifying treatment (interferon beta, glatiramer acetate, fingolimod, or natalizumab); (2°) LTD for MS with at least one reimbursement related or not related to MS; (3°) at least one disability pension payment for MS (general scheme patients only); and (4°) at least one hospitalisation with a discharge diagnosis of MS (principal, related or associated diagnoses). The ICD-10 code used for these three latter criteria was “G35”. An individual was identified as a prevalent MS case if (1°) he/she met at least one of the four selection criteria and (2°) was alive on 31 December 2012 and (3°) had at least one reimbursement in 2012. This latter criterion was added to ensure that MS patients had received reimbursed care on the national territory in 2012.

On 31 December 2012, 99,123 patients with MS in France were identified from the SNIIRAM/PMSI databases and 68.4% of these patients were identified by at least two selection criteria. LTD status for MS allowed the identification of 83.3% of MS cases, followed by hospitalisation (69.6%), reimbursement of a disease-modifying treatment (40.1%) and payment of a disability pension for MS (16.3%). 13.5% of the identified MS cases were captured only through hospitalisation records. The sex ratio (female/male) was 2.5. The mean age of the population identified on 31 December 2012 was 50.3 ± 14.2 years and one half of MS cases were between the ages of 40.0 and 60.0 years. 78% of MS cases in metropolitan France lived in urban areas. The distribution of MS cases according to geographical deprivation index was similar to that of the general population.

On 31 December 2012, the national crude prevalence of multiple sclerosis in France was 151.2 per 100,000 inhabitants [95% confidence interval (CI) 150.3–152.2] 210.0 per 100,000 in women (95% CI 208.4–211.5) and 88.7 per 100,000 in men (95% CI 87.6–89.7). The national prevalence rate based on the information provided by LTD status only was 125.9 per 100,000 inhabitants. The crude prevalence rates by gender and age-group are shown in Table 1. The peak prevalence of MS occurred between the ages of 50 and 54 years in women and 55 and 59 years in men. Prevalence rates after standardization on the European and World populations are presented in Table 2 to allow international comparisons.

The standardized prevalence in each region of France is shown in Table 3 and represented graphically in Fig. 1 for each department. Higher standardized prevalence rates were observed in North-Eastern regions of France (e.g., Lorraine, Picardie, or Alsace, with close to 200 MS cases per 100,000 inhabitants) than in South-Western regions (Languedoc-Roussillon, Corse, and Poitou–Charentes with approximately 130 MS cases per 100,000 inhabitants). The standardized prevalence rates in French overseas departments were considerably lower, ranging from 22.9 in Reunion Island (Indian Ocean) to 49.4 MS cases per 100,000 inhabitants in Martinique (Caribbean).

Among the 78,805 general scheme beneficiaries identified as MS cases, 1080 died in 2013. The mean age at death was 66.6 ± 13.9 years (68.2 ± 13.9 in women and 64.1 ± 13.5 in men). The crude mortality rate in 2013 for people identified as having MS on 31 December 2012 was 13.7 per 1000 MS cases (11.4 in women and 20.3 in men). Age-specific and gender-specific crude mortality rates are presented in Fig. 2. The SMR, as compared with the general population, was 2.56 (95% CI 2.41–2.72). The SMR was 2.55 (95% CI 2.35–2.75) in women and 2.58 (95% CI 2.34–2.83) in men.

The French MS prevalence was estimated at 151.2 per 100,000 inhabitants (prevalence rate of 155.6/100,000 standardized to the European population) in 2012 based on data from the French healthcare databases including ICD-10 G35 diagnoses from LTD, hospitalisations, disability pensions, and reimbursements of MS disease-modifying treatment. Patients with MS had a 2.5-fold increased risk of all-cause mortality compared with the general population.

Table 4 summarizes the methodologies and results of the previous studies carried out in France to estimate the MS prevalence at a national or local level, presenting their respective strengths and limits. The estimations (per 100,000 inhabitants) based on the capture-recapture method in Haute-Garonne (138–149) and in Lorraine (188.2) are close to our estimations (141.8 and 200.2, respectively).

Despite many studies on the prevalence of multiple sclerosis in European countries, published data are not sufficient to allow reliable comparison between European countries. This is essentially due to a lack of standardized estimates as published rates are rarely age-and-sex-adjusted to the European population. Furthermore, prevalence estimates are often provided for regions or cities in a given country and some European regions are underrepresented. The methodological quality of epidemiological studies, therefore, is highly variable but has improved in the more recent studies due to the use of administrative databases [2, 14]. When searching for recently published estimates of MS prevalence in neighbouring countries of France, very few studies were found to provide MS prevalence rates standardized to the European population [15‐19]. Although no direct comparison can, therefore, be established, these recent data suggest that our standardized MS prevalence rate of 155.6 per 100,000 for the 2013 European population is in an intermediate position between those reported in studies from northern (203.4 and 175 per 100,000 in UK and Germany) and southern (96.0 and 70.6 per 100,000 in Italy and Spain) neighbouring countries.

France is located in the middle of Western Europe. We observed a prevalence gradient with a higher prevalence (190–200 per 100,000) in North-Eastern France and a lower prevalence in the Southern (126 per 100,000 in Corsica-Mediterranean island) and South-Western France (130–140 per 100,000). Age-and-sex-standardized rates determined by an identical method of measurement at the same date from the claims database of a homogeneous national health system support our geographical comparisons according to region. The prevalence gradient observed in our study corresponding to about seven additional cases per 100,000 inhabitants by degree of latitude is in line with that observed by Simpson et al. for Western Europe (8.1 cases per 100,000 inhabitants by degree of latitude) [20]. The standardized prevalence gradient appears to be slightly more marked than in 2004 [3], although classification of regions according to MS prevalence remains globally similar. The present study is unable to determine the respective roles of possible ethnic or environmental factors that might explain this gradient.

We found an excess mortality in MS patients compared to the general population with an SMR of 2.56. This result is consistent with the international literature [8] in which SMR ranges between 2.47 and 2.89, but differs from the result (SMR 1.48) published by Leray et al. [11]. This difference can be explained by the mean follow-up period which was only 15 ± 10 in this study, compared with the longer follow-up period reported in other international studies. In our study, the MS population was followed for only 1 year, but all of the prevalent MS populations were followed, therefore, including young, recent diagnosed patients as well as patients with a longstanding diagnosis.

While crude mortality rates were higher in men than in women, SMRs were similar in both sexes, suggesting that sex-and-age-related differences in MS mortality simply reflect sex-and-age-related differences in the general population, unrelated to the disease.

To our knowledge, this is the first study to provide such an accurate estimate of French MS prevalence and one of the largest studies of national MS prevalence. One of the main strengths of this study is to provide regional estimates of MS prevalence using identical method of measurement. It facilitates regional comparisons and can be helpful for health decision makers to better plan health policies according to MS prevalence.

The prevalent MS population was identified using health insurance reimbursement data from two databases (SNIIRAM and PMSI) with completely independent data collection. Controls insuring the quality of data exist for each of the selection criteria. Information on reimbursement of drugs is collected by entering the specific bar code and transmitted thanks to the patient’s personal smart card. The high cost of treatments for MS excludes the hypothesis of non-reimbursed health care. LTD status for MS is requested by the patient’s GP and validated by the health insurance medical adviser who codes the diagnosis according to ICD-10. A similar process applies to disability pension. Similarly, hospitalisation data are systematically recorded by hospital physicians after hospitalisation to determine the cost of each hospital stay. Physicians specialised in these codes validate and transmit this information to paying bodies. Controls are performed to ensure the validity of the ICD—10 codes used. However, the SNIIRAM-PMSI databases have been created for payment management purposes. Just like other claims and hospitalisation databases, its main limitation is the lack of case validation. The anonymization of the data makes it difficult to match clinical cohort data with SNIIRAM-PMSI data to perform such case validation. As explained by Moulis et al., in the PMSI, only events accompanying the diagnosis or complicating the disease are encoded. Moreover, LTD attributions are not exhaustive for a given disease. If a patient has already obtained an LTD for a certain disease, he may not systematically request a new LTD for another disease. Yet, in our study, the number of information sources observed for each case identified (an average of almost two) and the closeness of our results and the estimates provided by the regional studies with case validations in Haute-Garonne and Lorraine are convincing elements for the quality of the information. With a 5-year pre-index period to detect hospitalisation diagnoses and a 2-year period to detect other criteria, our MS prevalence estimates are probably underestimated by at least 15% as MS cases characterized by a single relapse or benign MS may have been missed. Missing benign MS cases may slightly impact the mortality estimates. Benign MS cases may be less likely to die prematurely compared to other MS cases. Therefore, this bias could have led to a slight overestimation of the mortality of MS patients in our study.

MS incidence was not estimated in this study as identification of incident cases from claims data is a challenging process requiring a long patient history of diagnosis and treatment. Previous studies have reported that a 4-year history is insufficient to accurately classify MS patients as incident on the basis of claims data [21].