Summary of evidence
Overall, this systematic review showed that the analyzed studies do not represent a valid source of epidemiological data on the prevalence of dementia in Italy. We found that, until now, no epidemiological data for 10 out of 20 Italian regions were available and that the majority of epidemiological studies were performed at the level of municipalities, with most studies conducted in Northern Italy. Some regions were more affected by dementia than others.
Irrespective of geographical distribution, the prevalence rates of dementia reported in the Italian studies vary widely which may be due to important differences in methodological approaches and population age ranges. In the 16 studies analyzed, we found five different age ranges of study samples, a discrepancy that makes it difficult to compare the results of these studies and suggests a lack of methodological consensus. Furthermore, it is important to note that the only nationwide survey on the prevalence of dementia in Italy with a sample size >3000 subjects excluded those who were older than 84 years, an age range associated with a high rate of dementia.
From a diagnostic methodological perspective, the majority of Italian studies on the prevalence of dementia and AD included in our analysis adopted a two-phase design, but not all of them used sampling of screen negatives and none of them adopted the weighting back method. Furthermore, the informant interview was performed in only a minority of selected publications.
Overall, the Italian studies included in this review had lower ADI quality scores than those of European studies and, unlike the finding reported in the meta-analysis of Prince et al. , quality showed only a slight tendency to improve over time.
The finding of the lack of robust recent epidemiological data is in accordance with the global data reported in a meta-analysis, which showed that the number of epidemiological studies on the prevalence of dementia in high-income countries peaked in the 1990s and subsequently dropped off sharply . Even if the prevalence of dementia and AD has not changed significantly over time , the paucity of epidemiological data on the prevalence of dementia in Italy over the last ten years is regrettable and has important implications from economic and social points of view. Indeed, annual updates of the actual number of patients with dementia residing in a country should be the first step in creating a policy supporting patients and their families. It is also noteworthy that the geographic distribution of territorial Alzheimer Evaluation Units in Italy is not homogeneous, with the majority located in the north of the country . This geographic distribution might explain, at least in part, why most epidemiological studies on dementia and AD in Italy have been performed in northern regions.
Another issue regarding studies on the prevalence of age-related diseases like dementia and AD is the timing of publication in relation to the time of survey. At a national level, health policy strategy is dependent on accurate and current estimates of the size of the problem [1, 4]. The gap between the dates of the surveys and their dates of publication, together with the scarcity of recent data, suggests that the available publications on the prevalence of dementia in Italy may not represent an up-to-date source of information for health economic policy planning regarding patients with dementia.
Epidemiological studies on prevalence of dementia in Italy show low methodological quality. As Prince et al.  reported, multiphase methods in general tend to underestimate the prevalence of dementia and overestimate the precision. In accordance with other epidemiological studies , our analysis confirmed that many studies omitted the informant interview. Furthermore, prevalence estimates may reflect the diagnostic criteria adopted by each study. For example, a study that evaluated the prevalence of dementia using different systems of classification found that the proportion of subjects with dementia varied from 3 % when International Classification of Diseases (ICD)-10 criteria were used to 29 % when DSM-III criteria were applied . Similarly, the variability observed in European epidemiological studies has been attributed precisely to the clinical criteria adopted [1, 6]. In our analysis, 31 % of the studies used neither the DSM-IV criteria nor the NINCDS-ADRDA criteria. This finding represents a major methodological issue considering that only the latter diagnostic criteria have been validated with post-mortem data .
The weak ADI quality scores of the Italian studies, along with evidence that quality showed only a slight tendency to improve over time, has important implications at the national healthcare system level. Since no national survey commissioned by the Italian government has been performed in Italy, we suggest that the Italian healthcare system should urgently institute nationally representative surveys using the highest quality epidemiological methods, as defined in the ADI 2009 report, and repeat them at regular intervals to track any changes in the prevalence of dementia or AD [4, 6].
Based on the findings of our systematic review, we believe that the development of a national plan might be an appropriate strategy to obtain epidemiological estimates on dementia using the current healthcare system and, at the same time, we encourage researchers to undertake national surveys. A national plan might help overcome differences between Italian regions, whilst the detailed estimates obtained in this way might be useful for policymaking, planning, and allocation of health and welfare resources.
This review has several limitations First, our selected studies included surveys that were not specifically dedicated to the prevalence of dementia [10, 24, 25], which may have resulted in a bias in the types of publications included in the review. Second, although we reported that 75 % of studies were published before the year 2000, this finding might be due to our search methodology as PubMed/Medline and Embase were the only databases searched. However, this bias is unlikely to be substantial since all studies included in our analysis (Table 2) were also included in the most relevant meta-analysis published in this field  Third, the quality of studies included in this review was low. Fourth, The mean gap of four years between the year of survey and the year of publication should also be taken into account. Fifth, the review was not listed on an international prospective register of systematic reviews such as PROSPERO . Sixth, this review has the intrinsic methodological limitation that the prevalence rates derived from all the analyzed studies have not been standardized or compared with those of a reference population, e.g. one chosen for age and sex. Finally, it should also been taken into consideration that although the quality of the studies only slightly improved over time, our literature search for the studies on prevalence began in 1980 and 75 % of the selected studies were published prior to 2000. Therefore, many of the included studies were unlikely to have been conducted in conformity with current requirements for epidemiological studies .