Prevalence of multiple sclerosis in the Lazio region, Italy: use of an algorithm based on health information systems

Multiple sclerosis (MS) is the most common disabling neurological disorder of young adults around the world. It is typically diagnosed between the ages of 20 and 40, thus affecting individuals in their most productive years. The management of MS is complex and involves a comprehensive team as well as specific diagnostic, therapeutic and rehabilitative services representing a considerable burden for both patients and healthcare systems [1].

Italy is considered a high prevalence country for MS with about 70,000 people affected and an estimated prevalence of approximately 110 per 100,000 population [2, 3]. A number of studies have been conducted in different Italian cities and provinces, mainly based on data of patients recruited in neurological/rehabilitation clinics or MS centers, showing an uneven distribution of MS within the country [4]. It is well known that Sardinia and Sicily (the two largest Italian islands) represent very high-risk areas [5, 6] although a prevalence higher than 100/100,000 has been estimated in some North Italian cities such as Verona, Padova, Ferrara and Genoa [7‐10]. While the Northern part of the country as well as Sardinia and Sicily have been particularly well studied, relatively limited epidemiological data are available for Central and Southern regions [11‐15]. Moreover, most studies have been carried out at municipal/provincial level while to date prevalence estimates at larger geographic areas (i.e. region) are not common in Italy.

The availability of MS prevalence estimates is considered a valuable information to assess resource utilization and costs associated with the disease as well as for health care planning and monitoring purposes; furthermore, comparing prevalence estimates across different geographical areas can help to understand the contribution of genetic and environmental factors in MS etiology [16].

Different methodological approaches and data sources are used for estimating the prevalence of MS, including case ascertainment from hospitals and clinic records, neurologists, general practitioners (GP) and other physicians, disease registries, and health administrative databases. In recent years, there has been a growing interest in developing methods based on health administrative databases to estimate MS prevalence and different algorithms have been proposed depending on data sources available [17‐21].

The Lazio Region is part of the Italian National Health System (NHS) which provides universal health insurance for its residents, including coverage for GP service, hospital services and drug prescriptions. In the Lazio region health information systems (HIS) are comprehensive and contain high-quality information; consequently, health administrative data are largely used to measure the occurrence of acute and chronic diseases [22, 23].

The principal aims of this study were to estimate the prevalence of MS in the Lazio region using population-based health administrative databases and to describe the geographical distribution of MS across Local Health Units (LHUs) and districts within the region. Additionally, we evaluated the validity of the case-finding algorithm based on administrative data using a cohort of patients with definite diagnosis enrolled at MS treatment centers as the gold standard.

This study contributes to the knowledge on the epidemiology of MS in Southern Europe. In particular, it provides new insights on MS prevalence in Central Italy, an area that has been considerably less studied than Sicily and Sardinia and the Northern part of the country. We used health information systems to produce a prevalence estimate of MS at the regional level and to describe its geographic distribution across different areas within the region. The large population enabled us to provide a robust estimate of MS prevalence which ranks Lazio between the areas at high risk in Italy, although with an uneven distribution across LHUs and districts.

A huge number of studies assessed MS prevalence in European countries, revealing spatial heterogeneity in the distribution of disease [2, 26]. Italy has been particularly well studied, although most of studies were carried out in the Northern areas of the country and in the two largest islands. Considering prevalence estimates from 2000 onwards, they ranged from 94/100,000 to 149/100,000 in the North [27, 28], and from 127/100,000 to 224/100,000 in the islands [6, 29]. Moreover, some studies revealed an heterogeneous distribution of the disease even within small areas [5, 27], calling for specific analytical studies to detect possible MS risk factors.

So far only few data were available on MS prevalence in Central Italy coming from studies conducted in the district of L’Aquila and in some areas of the Lazio region [11‐13]. In a 2007 prevalence study conducted in the province of Frosinone the overall crude prevalence was 95.0 per 100,000 (94.4 when standardized on the European population) [12]. Applying capture–recapture techniques a period prevalence of 41.3/100,000 was estimated in the city of Rome between 1980 and 1990 [13]. Both in Frosinone and Rome we observed higher prevalence estimates than those previously reported. Although comparing results across studies presents a number of challenges, this finding is consistent with results of several studies showing a progressive increase in MS prevalence in different Italian cities and provinces [8, 28] as well in other countries [30].

Different methods for calculating MS prevalence based on health administrative data have been recently developed [17‐21] demonstrating the growing interest in the use of HIS as a source of information to estimate and monitor the burden of the disease. Administrative databases are similar among studies although algorithms may vary according to different context or specific objectives [19].

Given the good quality and completeness of health administrative databases in the Lazio region, they have been increasingly used to estimate the prevalence of chronic diseases and to measure the performance of health care organizations and the quality of services offered [22, 23, 31]. To identify cases of MS we used hospital discharge diagnoses, prescriptions of medications used exclusively for the treatment of MS and disease-specific ticket exemption code. The recently proposed algorithms for identifying MS cases from administrative databases mainly rely on drug prescriptions, in particular in those countries, like Italy, with a universal health coverage [20, 21]. Considering the high cost of drugs for the treatment of MS, the self-funding of these medications is likely to be negligible. It is known that some patients with MS do not take drugs or use drugs other than those included in the algorithm and therefore they cannot be captured through medication claims. The use of both hospital discharge and exemption registries allowed us to recover part of this untreated or differently treated population. In fact, although there are large overlaps between the populations registered in the three databases, each source of data provides a specific and unique contribution to the identification of MS cases. It is clear that the algorithm based on administrative records only allows to identify patients who have contacts with the Regional Health Service that, considering the characteristics of the disease and its chronic nature, is supposed to be a very likely event. Nonetheless, the overall prevalence of MS may be underestimated by the approach we used. As part of this study, data from medical charts of patients entering five MS centers in the period 2006–2011 were collected. By record linkage procedures, we estimated that 85 % of patients attending the centers were included in the population identified by the algorithm based on health administrative data. While this figure is not far from sensitivity estimates observed in validation studies in other settings [22], it is likely that patients in the first steps of the disease with a clinical examination in an ambulatory specialist setting and/or untreated are not detected through our methodology. Another reason for the lack of complete identification of clinical records in our administrative datasets may be in the low quality in reported personal demographic data in the medical charts with consequent failed record linkage procedures. On the other hand, the population of MS patients identified showed a male to female ratio and an age distribution consistent with those reported by other studies carried out both in Italy and elsewhere supporting the reliability of our results [5, 6, 9, 11].

The study also showed an uneven geographical distribution, with a gradient in prevalence decreasing from the eastern mountainous part of the region to the western areas closer to the coast. This is in line with other studies conducted at different latitudes, showing lower MS incidence in the coastal area compared to the inland area [32, 33]. Diet and lifestyle factors have been hypothesized to modulate the risk of MS across regions at the similar latitude. Lifestyles of people in coastal areas, generally involving more time outdoors, match evidence showing that exposure to sunlight is associated with a lower risk of developing MS [34‐36]. As sun exposure is the determining factor for vitamin D status in most populations, it has been suggested that sun exposure to some extent mediates its effect on MS risk through vitamin D [37].

This study has both strengths and limitations. The standardized methodology and the contribution of medical records in validating the case identification algorithm are the main strengths of the study. Our procedure based on population-based administrative data has different advantages: it is time efficient, allows to easily provide updated MS estimates, and enables to avoid problems related to small sample size. We chose to ascertain an MS case from at least one claim in one of the three administrative databases due to MS clinical practices and the reliable coding process utilized within the Lazio region.

The major limitation is the probable underestimation of MS prevalence. Although the use of a combination of multiple sources of data (hospital discharge, ticket exemption, and prescription registries) contributes to produce more reliable estimates, only cases diagnosed and recorded in administrative databases could be captured by the algorithm. The availability of a cohort of patients with definite diagnosis of MS allowed us to calculate the proportion of cases not identified through HIS and therefore to estimate the extent of the actual population prevalence underestimation. Secondly, the use of administrative data instead of clinical information to identify the population affected by MS limits its use for more analytical purposes.

In conclusion, this study produced an estimate of MS prevalence in the Lazio region using population-based health administrative databases and described the geographical distribution of MS within the region. Although some limitations must be considered including possible prevalence underestimation, administrative databases represent an attractive source of information to measure the burden of MS allowing for periodic updates of prevalence estimates, useful for monitoring prevalence trends at population level and to ensure appropriate healthcare resources allocation. Moreover, considering that we used administrative databases including information available in other Italian regions, the proposed algorithm could be tested to obtain population-based prevalence estimates of MS in different areas of the country and to analyze geographic differences as in the case of other chronic diseases [38].