People living with undiagnosed HIV infection and a low CD4 count: estimates from surveillance data, Italy, 2012 to 2014

Background and aims Late HIV diagnosis is associated with onward HIV transmission, higher morbidity, mortality and healthcare costs. In Italy, more than half of people living with HIV were diagnosed late during the last decade, with a CD4 count < 350 cells/mm3 at diagnosis. We aimed to determine the number and characteristics of people living with undiagnosed HIV infection and low CD4 counts in Italy. Methods Data on newly reported HIV diagnoses from 2012 –2014 were obtained from the national HIV surveillance system. We used the European Centre for Disease Prevention and Control HIV modelling tool to calculate the undiagnosed prevalence and yearly diagnosed fraction (YDF) in people with low CD4 count. Results The estimated annual number undiagnosed HIV infections with low CD4 count was on average 6,028 (95% confidence interval (CI): 4,954–8,043) from 2012–2014. In 2014, most of the undiagnosed people with low CD4 count were men (82.8%), a third acquired HIV through sex between men (MSM) (35.0%), and heterosexual transmission (33.4%), respectively. The prevalence of undiagnosed HIV infection was 11.3 (95% CI: 9.3–14.9) per 100,000 residents ranging from 0.7 to 20.8 between Italian regions. Nationally the prevalence rate was 280.4 (95% CI: 173.3–450.2) per 100,000 MSM, 8.3 (95% CI: 4.9–13.6) per 100,000 heterosexual men, and 3.0 (95% CI: 1.4–5.6) per 100,000 women. The YDF was highest among heterosexual women (27.1%; 95% CI: 16.9–45.2%). Conclusions These findings highlight the importance of improving efforts to identify undiagnosed HIV infections primarily among men, both MSM and heterosexual men.


Introduction
Late diagnosis of HIV remains a major public health concern worldwide [1 -3]. In 2015, close to half (48%) of people newly diagnosed with HIV in European countries were late presenters, with CD4 counts below 350 cells/mm 3 at diagnosis, including 28% with advanced HIV infection (CD4 < 200 cells/mm 3 ) [1]. In Italy, despite HIV testing and healthcare being free of cost for the individual, more than half of the ca 4,000 people diagnosed with HIV annually are diagnosed with a CD4 count < 350 cells/mm 3 , and ca 40% are at the symptomatic stage of infection when diagnosed [4,5].
The late diagnosis of HIV infection has negative consequences, both at the individual and population levels. People presenting late respond insufficiently to antiretroviral therapy (ART) and treating them is often complex and costly. Individuals who are in an advanced stage of immunosuppression due to HIV are at high risk of clinical events and death [6,7]. At population level, those diagnosed late are a potential source of HIV transmission for a considerable period [8]. Low CD4 count and high viral load contribute significantly to the risk of sexual transmission [9].
Given the prevalence of late diagnoses, it is likely that a considerable number of people with low CD4 counts remain undiagnosed in Italy. Mammone et al. estimated that there are 12,000-18,000 undiagnosed people living with HIV in Italy [10], although no estimate of those undiagnosed with a low CD4 count was calculated. Knowing the numbers of people living with undiagnosed HIV and with a low CD4 count could be useful in predicting the prevalence of late HIV presentation and its consequences of poor prognosis and onward transmission. Being aware of the overall numbers of those who are undiagnosed and of the respective numbers in relevant subpopulations can support the monitoring of national and local HIV prevention strategies, the  c For regions (one in the Central area and one in the North area) that did not collect data on the clinical stage and CD4 count, the number of undiagnosed was estimated assuming the distribution of clinical stage and CD4 count observed at national level ( Table 2).
revision of health policies and the allocation of economic resources to prevention and control efforts [11]. The aim of this study was to estimate the number of people living with undiagnosed HIV and a low CD4 count in Italy, to analyse the characteristics of this population and to evaluate the prevalence of undiagnosed HIV infection in the general population.

Methods
To estimate the number of people living with undiagnosed HIV and a low CD4 count, we applied the model proposed by Lodwick et al. [12]. This model is part of the European Centre for Disease Prevention and Control (ECDC) HIV modelling tool. The tool uses routine surveillance data to calculate estimates of the number of people living with HIV as well as of those not yet diagnosed. It does not depend on historical data, i.e. it can be used even with limited years of available data [13]. Of the two models included we choose the London model as it applies to the way HIV data was collected through the Italian HIV surveillance system effective as of 2012.

Data source
We obtained data from the Italian National HIV Surveillance system (INHS) on people aged > 15 years who were diagnosed with HIV between 2012 and 2014 and reported to the INHS by June 2015 [5].
In Italy, the reporting of new HIV diagnoses is mandatory since July 2008 for all clinicians [14]. There are 173 Infectious Diseases Centers (IDC) in Italy that offer free monitoring and health management to all HIV-positive persons, including non-nationals and undocumented migrants [15]. Most people who test HIV-positive are diagnosed at IDCs directly and those who are tested in other health facilities are further referred to an IDC for confirmatory testing and diagnosis.
Data on new diagnoses are collected by regional surveillance systems and sent annually to the INHS coordinated by the Italian national institute of health in Rome. While the surveillance did not cover all regions previously, since 2012 there was 100% geographical coverage with all regions reporting data to the INHS [5].
The INHS collects the following data on an annual basis: (i) demographic data (age, sex, nationality, geographical area of diagnosis, and geographical area of residence), (ii) clinical information (clinical stage, CD4 counts, and viral load) and (iii) HIV exposure group data (people who inject drugs (PWID), heterosexual

Adjustment for missing values and reporting delay
The estimates obtained from the model were adjusted for reporting delays or underreporting of HIV diagnoses with HIV-related symptoms as proposed in the original publication [12]. As information on clinical stage and CD4 count were missing in around 30% of the INHS records, the missing values were adjusted under the assumption that the distribution of clinical stage and CD4 count among diagnosed cases with missing data was similar to that of diagnosed cases with available data [17]. In brief, the estimates obtained from the London method were divided by the proportion of all diagnoses with symptoms, where the CD4 count was known.
The following steps were used for the adjustment: first, the proportion of missing data was calculated relative to the clinical stage and CD4 count for each characteristic of the new diagnoses. Second, the estimates of undiagnosed people living with HIV were multiplied by the inverse of the missing proportion according to each characteristic. Last, the reporting delay to the INHS was considered, with the adjusting of the annual estimates by a reporting delay of 5%, introduced for each year of the 3 years, given that all new HIV diagnoses were notified to the surveillance system within 3 years after diagnosis [5]. In other words, it was assumed that in 2015 the INHS data were all complete for 2012 (100%), and almost complete for 2013 (95%) and 2014 (90%). Therefore, for each characteristic we adjusted the estimates multiplying them by the following: Table 1 shows the proportion of missing data (of CD4 count and/or clinical stage) with the respective 'adjustment factors' applied to undiagnosed people living with HIV estimates by main characteristics.  Table 1).

Characteristics of people undiagnosed and newly diagnosed with HIV and with low CD4 count in 2014
The characteristics of both people undiagnosed and newly diagnosed with HIV and a low CD4 count were described for the year 2014 to compare characteristics of those undiagnosed with low CD4 count with new HIV diagnoses with a low CD4 count.
The yearly diagnosed fraction (YDF) in people with a low CD4 count (CD4 < 350 cells/mm 3 or CD4 < 200 cells/ mm 3 ) was calculated according to main characteristics for the year 2014. YDF has been recently proposed by Sasse et al. [18] to evaluate the ratio of new diagnoses among people living with HIV who can be diagnosed in a given year. In our study, YDF was calculated among people with HIV and with a low CD4 count according to the following formula: Adjustment factor = (1proportion of missing − annual reporting delay) 1 .

Prevalence of undiagnosed HIV infection with low CD4 count in 2014
To evaluate the prevalence of undiagnosed people living with HIV and a low CD4 count for the year 2014, the rate expressed was calculated as follows: As a denominator, the population aged > 15 years up to 75 years estimated by the ISTAT was used [16].The described undiagnosed prevalence of HIV infection was also calculated by region of residence and by HIV exposure group. As denominator, the female population for heterosexual women and the male population for men was used. For MSM, a proportion of 3% of the adult male population was assumed, given that published data reveals estimates of MSM ranging from 2% to 4% of the male population in Italy [19][20][21]. Thus, for heterosexual men the remaining 97% of male population was used.  (Table 2). Clinical stage at HIV diagnosis was reported for 70% of people, 39% of them were diagnosed at clinical advanced stage (clinical stage B or C). Table 2, shows the distribution of new HIV diagnoses by main characteristics; these were similar during the 3 years: the majority were men, more than half aged between 25 and 44 years, and more than one third were MSM. More than half were diagnosed late, namely with CD4 count < 350 cells/mm 3 .  (Table 3). Also for the undiagnosed people with CD4 count < 200 cells/mm 3 , the highest estimates were in men, those living in the North and in MSM.

Estimates of people living with undiagnosed HIV and a low CD4 count
Among the undiagnosed people living with HIV and with low CD4 count, the proportion of those with CD4 count < 200 cells/mm 3

Characteristics of people undiagnosed and newly diagnosed with HIV and with low CD4 count in 2014
In Table 4, for the year 2014, main characteristics of people undiagnosed and newly diagnosed with HIV and with a low CD4 count are compared. The characteristics of those newly diagnosed and with a low CD4 count were similar to those of people with undiagnosed HIV and with a low CD4 count.
Many people undiagnosed and with CD4 count < 350 cells/mm 3 were men and older than 35 years, while a third were MSM and, another third were heterosexual men. About a third were born abroad and nearly half resided in the North of Italy. Similarly, men (MSM and heterosexual men), people older than 35 years, and those living in the North were among those most represented among undiagnosed people with HIV and with CD4 count < 200 cells/mm 3 (Table 4).
In Table 4 Figure 1 shows the prevalence of people with undiagnosed HIV and with CD4 < 350 cells/mm 3 , calculated as a rate per 100,000 adult residents. Overall, this rate was 11.3 (95% CI: 9.3-14.9) per 100,000 residents older than 15 years. The prevalence of people with undiagnosed HIV varied between the different Italian regions from 0.7 per 100,000 (Calabria) to 20.8 per 100,000 adults (Liguria); North and Central areas showed higher rates of those undiagnosed with a low CD4 count ( Figure 1A ). -450.2) per 100,000 MSM, for heterosexual men it was 8.3 (95% CI: 4.9-13.6) per 100,000 heterosexual men. The prevalence rates among MSM ranged from 6.2 (Basilicata, Molise, and Valle d'Aosta) to 450.6 (Liguria); almost all regions showed rates higher than 16 per 100, 000 ( Figure 1C), in particular five regions showed a regional rate higher than 300.0 per 100,000 MSM (Umbria, Sicilia, Toscana, Lombardia, and Liguria) (results not shown in the figure). The regional rates varied for heterosexual men from 0.4 (Basilicata, and Friuli) to 16.1 (Valle d'Aosta) per 100,000 heterosexual men; almost half of the Italian regions (nine regions) had an estimated prevalence rate higher than 8.0 per 100,000 heterosexual men ( Figure 1D).

Prevalence of people living with undiagnosed HIV and with low CD4 cell count in 2014
The annual rate of new diagnoses in Italy was 6.1 per 100,000 adult residents in 2014, ranging from 2.0 (Calabria region) to 11.1 (Lazio region) [5] (data not shown).  stage. e In the Central area, the proportion of missing values were concentrated mainly in one region (100% in one region and less than 5% in the remaining three regions) that does not routinely collect data on the clinical stage and CD4 count.

Discussion
Estimating the number of people living with undiagnosed HIV and with a low CD4 count enables the identification of determinants for a delayed access to care. We estimated the number of people with HIV and with a low CD4 count in Italy who are not yet diagnosed using an easy, reproducible, and validated model [13]. The strength of this study is that it provided estimates of demographic characteristics of undiagnosed people with HIV. The average yearly number of people living with undiagnosed HIV infection and CD4 < 350 cells/ mm 3 was 6,000 over the period 2012 to 2014, with a similar pattern across the years. The estimate of people with low CD4 count corresponded to 40% of the total number of people (including asymptomatic) with undiagnosed HIV infection in Italy (i.e. 15,000) [10]. The same proportion of people (40%) with CD4 count < 350 cells/mm 3 was found in France for the estimated undiagnosed people with HIV in 2010 [22]. Our numbers indicate there are a substantial number of people with undiagnosed HIV in Italy who need to be treated immediately. Failure to diagnose these individuals will result in greater morbidity and mortality for them, risk  of onward transmission and greater costs accrued for the health system.
Focusing on the most recent year analysed, the prevalence of undiagnosed HIV infections was 11.3 per 100,000 adults in the resident population in 2014, ranging from 0.7 to 20.8 for different regions. Differences in regional prevalence could be attributed to factors, such as (i) different spread of HIV infection [5,15], (ii) different levels of HIV risk awareness [23,24], and (iii) the risk groups prevalent in each region.
This is in line with a cross-sectional study [20] that indicated a higher prevalence of people diagnosed and linked to care in northern Italy. Moreover, despite IDCs being well distributed throughout Italy, surveillance data indicates higher numbers of new diagnoses of HIV infection and AIDS, as well as of HIV-positive people under treatment, in the North [5,15].
Different levels of HIV risk awareness were confirmed in a respective study, which showed that people living in the North were less aware of HIV risk factors compared with those in the Centre and South of Italy [23]. Furthermore, a study showed that regional differences of HIV risk awareness seem to be correlated with different socioeconomic factors and lifestyles existent in North and South Italy (unpublished data). Differences in regional prevalences of undiagnosed HIV infection were very similar to those observed among HIV-positive people diagnosed that were linked to care [15] as well as to differences observed among new HIV diagnoses across the Italian regions [5]. These findings confirm that, at least in Italy, regions with high rates of new diagnoses also encompass a high proportion of both diagnosed and undiagnosed people [5,15]. This highlights the importance of the regional differences in the spread of HIV infection that can be observed at a wider level across European countries as well as within the United States [25,26].
In addition, the highest prevalence of undiagnosed HIV infection was observed among MSM in whom it was 280 per 100,000 MSM, whereas among heterosexual men it was 8 per 100,000 heterosexual men, and among heterosexual women it was 3 per 100,000 female residents, with large differences across the Italian regions. Even though MSM have been reported to have high HIV testing rates compared with other key populations in high-income countries [24,[27][28][29], as well as, the highest perception of the risk of HIV infection [23,24], the study findings show that they account for the highest number (2,115) and the highest proportion (35%) of undiagnosed people with a low CD4 count in 2014. MSM in Italy are also the subgroup most represented (nearly 50%) among the total population of undiagnosed HIV people (including asymptomatic), as estimated by Mammone et al. [10]. This could be attributed to a high rate of new infections in this group during the most recent years [1,[30][31][32] combined with a large number of undiagnosed people who contribute to ongoing transmission [10,22,33]. In addition, a high HIV prevalence and a high proportion of MSM with undiagnosed HIV could be attributed to high levels of sexual activity and to some risk behaviours for sexual transmission of HIV [34]. Therefore, test-seeking behaviour should be encouraged and voluntary counselling and testing made more accessible in Italy, a country where the stigma against HIV and homosexuality may still be prevalent [20,35].
Focusing on the most recent year in our analysis, a high proportion of undiagnosed people with low CD4 count was reported among heterosexual men (33.4%), whereas in other Italian studies this population accounted for a quarter of the total undiagnosed (including those asymptomatic [10]), and a quarter of new diagnoses reported to the INHS [5]. The higher proportion of heterosexual men among undiagnosed with a low CD4 count could depend, partly, on the fact that heterosexuals were more likely to have a longer undiagnosed interval (time lag from infection to HIV diagnosis) as shown in other studies worldwide [8,[36][37][38]. In Italy, Mammone et al. [37] estimated that heterosexuals had a far longer lag from infection to HIV diagnosis compared with MSM (7.7 vs 3.7 years).
We found a YDF of 20.6% which was similar to that reported recently by Sasse [18] on the total HIV population in the European countries. Among heterosexual women the YDF was the highest (27.1%) compared to the other groups, suggesting a more frequent access to HIV testing, likely facilitated from routine screening during pregnancy in this population [5,39]. This result may mean a certain degree of success with regard to testing in this group. The highest YDF (26.6%) among people living in the North compared to the other areas may be an indicator of the wider availability of IDCs and HIV testing services in this area [40]. A higher detection could represent a more efficient and therefore better surveillance system.
This study has some limitations. First, we assumed that people with HIV who develop AIDS, or other HIV-related symptoms, will almost certainly present for care, and as consequence, will be all diagnosed with HIV and notified to the surveillance system (assumption of London method) [12,13]. However, the assumption of the London method can be considered acceptable for our study, as HIV testing and access to care are free in all IDCs and the proportion of people living with HIV who do not attend the IDCs should be reasonably low.
Another limitation was the assumption that CD4 counts in those where the information was not available was the same as in those with available information. This assumption was supported by other studies conducted on the Italian HIV Surveillance data [10]. The missing CD4 count information, In the Italian national HIV surveillance data, mainly in the Central regions, may make the estimates less robust. However, in the remaining areas the proportion of missing data were lower than 10%.
In terms of the reporting delay we assumed a constant decrease over the 3 years. This had a small impact on the estimates as it was sufficiently low. Other limitations which may have a considerable impact on the eventual estimates include the effect of new testing strategies, the changes over time in the reporting of data, and the different quality of data in the surveillance systems of all the regions.

Conclusions
About 6,000 HIV-positive people with low CD4 counts, remained annually undiagnosed between 2012 and 2014 in Italy. This indicates that ca 40% of the 15,000 total undiagnosed people living with HIV in Italy were in immediate need of diagnosis, linkage to care and antiretroviral treatment in order to avert high HIVrelated morbidity, mortality and healthcare costs.
The majority of those with undiagnosed HIV and with low CD4 counts were MSM and heterosexual men, and there were large differences in prevalence of undiagnosed HIV infections with low CD4 across the Italian regions. These findings highlight the importance of improving HIV testing availability, with a focus on men, in order to diagnose and provide treatment to those living with undiagnosed HIV in Italy.