Aging and mortality among people with diagnosed HIV in Italy: recent trends and projections

Introduction

Since the advent of effective antiretroviral therapies, the life expectancy of people diagnosed with HIV (PWDH) has increased significantly. In developed countries, it is now common for PWDH to live up to 70 years and beyond.^1-3 While we stress this is a positive development, it nonetheless poses new challenges for health systems. PWDH present age-related chronic conditions more frequently and at a younger age compared to the general population; contributing factors include the specific effects of HIV infection in older adults, the long-term effects of the infection, and prolonged use of antiretroviral therapy (ART).^1,4-6 Therefore, HIV prevention and care interventions must include more and more the management of health complications associated with HIV in the elderly, in long-term infections, and in prolonged ART use among PWDH.
To this end, effective methods for accurately predicting age-related chronic conditions among PWDH are fundamental for resource allocation, long-term strategy planning, and informing possible interventions. Having reliable methods to clearly quantify and predict the future age structure of the PWDH population is a necessary step toward this goal. However, this is not a trivial task. As mortality trends in the PWDH population differ from those of the general population and the data often noisy, many commonly techniques are unsuitable for PWDH populations.^2,3,7-11  
These efforts are further complicated by the nature of HIV surveillance data, typically structured in discrete categories. While such practice is useful in some contexts, significant variability can occur within age bands (for example, disease susceptibility and/or mortality risk). Discretely age-structured data is also ill-suited for forecasting – while mechanistic age-time models can be easily employed for, e.g., 1-year age resolutions, this is generally not possible for 5- or 10-year discretely-binned data, even for short-term extrapolation. As such, HIV surveillance data alone is often insufficient to address questions in which evolving, age-dependent dynamics play an important role. Developing robust models of PWDH demographic evolution is thus not only a challenge of forecasting; in general, the continuous population age-structure must be reconstructed by assimilating discrete age-structured surveillance data within an age-continuous framework. 
To this end, in Viguerie et al.,³ the authors have recently developed a multi-stage computational pipeline that combines ensemble demographic reconstruction based on partial differential equations (PDEs), data assimilation, and scientific machine learning techniques to provide reconstructions and forecasts for the PWDH population in the United States at an age-continuous level. In the present work, a modified version of this computational pipeline was applied to the Italian PWDH population. 
However, differences in the surveillance practices make the Italian case substantially more difficult. While the United States NHSS reports deaths among all PWDH, deaths among PWDH in Italy, as in many European countries, are only recorded among PWDH with a previous AIDS diagnosis. This group accounts for only approximately 22%-27% of Italian PWDH.^12-14
Given these challenges, in addition to the straightforward utility provided by the demographic forecasts, the present study is also, to the author’s knowledge, the first attempt to rigorously estimate the age-structured diagnosed HIV prevalence in Italy: that is, the total number of PWDH at each given age in Italy. Previous efforts in this direction have significant limitations: survey studies by Camoni et al. 2015¹⁴ e Camoni et al. 2018¹⁵ do not provide detailed information on the population age structure, and various modelling studies either do not consider PWDH deaths at all13 or consider only deaths among PWDH with an AIDS diagnosis¹⁶, limiting their utility. The present study seeks to overcome these challenges and provide complete, age-structured estimates of the Italian PWDH population in recent years, as well as forecast its evolution into the next decade.

Methods

Study population

In this study, aggregated data from 01.01.2012 and 31.12.2024 was used, extracted from two independently surveillance systems: the National HIV Surveillance System and the Italian National AIDS Register (RAIDS). The first system collects data on new HIV infections diagnosed in Italy annually at the individual (but anonymized) level. This system has achieved national coverage since 2012. The second system collects data on new AIDS diagnoses since 1982, with non-anonymized individual data is collected using name and surname. While the surveillance systems do not directly update vital status, for AIDS diagnoses, the vital status is determined annually by a procedure of record linkage between RAIDS and the Italian mortality database. As the individual data in the National HIV Surveillance System are anonymized, a similar procedure cannot be applied for PWDH without an AIDS diagnosis.

Mathematical modelling

The basic model employed herein is an age-structured partial differential equation (PDE) model describing the age-structured HIV prevalence in Italy over time: its output is a function u(a,t) giving the total number of PWDH aged a at time t. At a basic level, the model works as follows.  If, in addition to u(a,t), one also has the age-dependent mortality hazard μ(a,t), giving the probability of death at age a at time t (assuming one is alive at time t) and λ(a,t), the number of new HIV diagnoses among persons aged a at time t, one can promptly advance to the time t+1 following the relation:

u(a,t+1) = (1-μ(a-1,t))u(a-1,t)+λ(a-1,t)

Note, however, that not suitable representations of μ(a,t) and λ(a,t) are directly available from data: their determination is the core methodological challenge of the present work.
Discretely age-structured deaths among PWDH with a previous AIDS diagnosis (2012-22) and new HIV diagnoses (2012-24) were used, provided by the Italian National Health Institute.¹³ These data are age-structured into discrete 5-year age bands for the ages 15 to 85, plus two additional age bands comprising all PWDH under 15 and over 85, for a total of sixteen total age groups.
While PWDH prevalence data were not available for any year, this model requires these data for its initial point – thereafter future states can be determined by tracking deaths and diagnoses across ages over time. To this end, several sources were triangulated to obtain suitable initial conditions. Specifically, the survey data were used,^14,15,17 together with surveillance data in subsequent years, to estimate age-specific HIV prevalence in 2011. More specifically, the data in Camoni et al. 2015¹⁴ e Camoni et al. 2018¹⁵ provide discrete age-structured snapshots of Italian PWDH in 2012 and 2014 over coarse age structures (under 25, 25-49, 50-59, and over 60 years). Together with the mortality and diagnosis data, the shift in the population age structure over the two-year period allows to estimate plausible age distributions. 
Then, the population ensemble was updated using the provided dignosis and mortality data, following the computational procedure described in Viguerie et al.³ Broadly, this pipeline includes ensemble Kalman inversion to estimate age-specific mortality rates over time, ensemble population sampling for uncertainty quantification, and non-negative dynamic mode decomposition (nnDMD) to predict the evolution of mortality rates and HIV diagnoses, employed with minimal modification; further details are provided in the online Supplementary Materials.
As the age-structured mortality data contains only deaths among PWDH with an AIDS diagnosis, the reconstructed mortality hazards correspond only to this specific group, estimated as 23.1% of the PWDH population following Regine et al.¹³ To estimate mortality among PWDH without an AIDS diagnosis, IRR data from the ICONA cohort were used^2,3,18 together with the mortality hazards reconstructed in this paper for the AIDS-diagnosis group. A more detailed mathematical explanation is provided in the online Supplementary Materials.
Given the inherent uncertainty, an ensemble of 1,000 age distributions was generated, each one individually consistent with initial data. New HIV diagnoses were also handled in this manner, with each individual ensemble member updated at each time step with an independently sampled age distribution of new HIV diagnoses, each consistent with the discrete age structure reported in the surveillance data. The ensemble elements simulated independently over the entire period; the point estimates ported herein are the ensemble means, with obtained by adding and subtracting two sample standard deviations from the ensemble mean.

Results

The results of this study show that the PWDH population in Italy has aged significantly in the last decade. An immediate effect is the increase in the PWDH population in Italy, which has grown from ~112,000 in 2012 to ~130,000 in 2024. However, this growth has clearly slowed in recent years, with beginning around 2018-2019. While predict continued (albeit slow) growth in the immediate short term, peaking at approximately 133,000 in 2032, the forecasts here presented show a slow decline beginning in 2033 and continuing through 2035 (Figure 1, Table 1).

The predicted growth through 2032 is driven by continued improvements in PWDH life expectancy and not through increases in new diagnoses, which are forecast to decline over the same period (Table 2). Furthermore, the projected population decline beginning in 2033 is entirely driven by decreases in new diagnoses and increase in the overall PWDH population age; increases in PWDH mortality on a per-person basis were not predicted (Figure 2, Tables 1-3). 

Going into more detail, in line with Camoni et al. 2015,¹⁴ people aged 65 or older were estimated to represent slightly under 4% of PWDH in 2012 and 5.0% in 2014. By 2020, this share is estimated to increase to over 10%; the prediction is that nearly 40% of Italian PWDH will be over 65 years by 2035. The effect is even more marked in the over-75-year age group: less than 1% of PWDH were estimated to be over 75 in 2012, nearly 4% by 2024, and should exceed 10% by 2032, approaching 15% by 2035 (Figure 2, Table 1).

This change is determined by a shift in PWDH mortality trends. Further declines in mortality among PWDH were predicted, especially at younger ages, in continuity with long-term trends. Consequently, the percentage of PWDH deaths occurring in the oldest age groups is predicted to increase significantly. In 2012, persons over 65 years were estimated to comprise 10% of total PWDH deaths and persons over 75 years 3%. By 2024, these percentages had increased to 36% and 15%, respectively. The projections of this study show a continuation of these trends, with persons over 65 comprising over 50% of PWDH deaths by 2030 and person over 75 comprising over 25% of PWDH deaths by 2031. By 2035, forecasts show that these age groups will comprise 65% and 40% of total PWDH deaths, respectively (Table 2).

Regarding new HIV diagnoses, continued, steady decrease in the future was projected, with annual new diagnoses dropping to 2000 by 2035. While total declines across all age groups was forecast, also a slight shift in the relative age distribution of new diagnoses is forecast, with the average age of new HIV diagnoses increasing over time. This gradual shift is a clear, long-term trend: while persons over 45 years were fewer than 30% of new HIV diagnoses in 2012, this percentage steadily has steadily increased over time, exceeding 35% in 2018 and 40% in 2020. The forecasts of this study suggest this trend should continue through the early 2030s, reaching approximately 44% of new diagnoses in 2029 and plateauing thereafter.

While average age of new diagnoses is increasing, it is nonetheless increasing fare more slowly compared to the average age of the overall PWDH population. These dynamics lead to a skewed age distribution by the end of the forecasting period, with a large concentration of long-surviving cohorts at older ages (likely those diagnosed in the 1990s-2010s, when transmission levels were higher) and a smoother, more extended distribution at younger ages, particularly visible by the end of the forecast period (Figure 2, Table 1). Similar demographic profiles are also predicted for PWDH in comparable high-income countries.^2,3

Discussion

The present study is subject to several important limitations that should be considered when interpreting these results. In Italy, deaths among PWDH are recorded only among PWDH with a previous AIDS diagnosis, a group that has represented approximately 25% of new diagnoses in recent years,¹³ reflecting long-standing patterns of testing and diagnosis. To estimate mortality among the broader PWDH population, data-driven estimates were combined with IRR and mean age at death data cohort studies comparing Italian PWDH populations with and without AIDS diagnoses. Given the lack of alternative surveillance data, such an approach is necessary; however, both the reconstructions of past demographic data, as well as future projections, assume the ICONA data generalizes appropriately to the overall PWDH population. 
The methods here used cannot specifically distinguish between PWDH born in Italy and those born abroad. The latter group is an important, growing portion of the Italian PWDH population; the present study may not entirely account for demographic effects owing to the particularities of this group (for example, repatriation leading to substantial decreases in the overall PWDH population). While this limitation was acknowledged, non-Italian PWDH tend to be younger compared to domestic PWDH.¹³ Therefore, while outmigration may affect the overall prevalence estimates and forecasts shown herein, it is unlikely to substantially change the overall demographic dynamics.
These projections of future HIV diagnoses and PWDH mortality trends are based on a fully data-driven approach, nonnegative DMD (for details, see references and online Supplementary Materials). While this strategy avoids reliance on explicit structural assumptions, it assumes continuity in the underlying processes shaping recent trends. In this regard, causal drivers of diagnoses and mortality are incorporated into future projections insofar as they are reflected in the pre-existing data. Rapid and/or substantial changes in HIV prevention, testing, or care in the coming years, however, could affect these dynamics. Possible such disruptions may include, but are certainly not limited to, large-scale changes in PrEP uptake, shifts in HIV testing strategies, improvements in early diagnosis, developments in ART efficacy, or shifts in long-term ART adherence (for example, due to increased use of long-acting injectables)¹⁹. Should such changes occur, future outcomes may differ substantially from the forecasts provided in the present, which assume continuation of the present (2012-2024) trends. Furthermore, the lack of an underlying mechanistic, parameterized structure limits the interpretability of nnDMD and related methods. While useful for producing forecasts, interpreting the outputs in terms of specific, attributable causes or phenomena is not straightforward and limits the utility of such methods for intervention planning or scenario analysis. 
As PWDH prevalence is not specifically recorded by Italian HIV surveillance systems, direct validation with data is not possible. Furthermore, the available literature on this topic does not yield any work suitable for direct quantitative comparison. However, the results of this study are consistent with such work. Specifically, the modelling analysis¹⁶ considered only deaths among PWDH with an AIDS diagnosis; the model used in the present study produced nearly identical results when considering only those deaths. The results are also broadly consistent with those reported in Viguerie et al.²⁰; while the estimated mortality risk is somewhat lower in the present, the discrepancies are consistent with differences in VLS rates in the respective populations.
While acknowledging these limitations, the findings of this study provide a clear and consistent picture of the ongoing demographic transformation of the PWDH population in Italy. Although new transmissions have decreased steadily, improvements in life expectancy have led to substantial growth in the Italian PWDH population over the past decade. These results show this growth was dramatic from 2012-2018, though has slowed somewhat in more recent years. Although PWDH life expectancy forecast will continue to improve going forward, the increasing age of the Italian PWDH population will cause the population to peak in the early 2030s, with a slow decline beginning shortly thereafter. 
However, in absolute terms, the declining number of PWDH will be driven by fewer PWDH at younger ages. The number of older PWDH is already rapidly increasing and the forecasts show that this growth will accelerate dramatically in the coming years, with no signs of slowing by 2035. This demographic shift has important implications for public health planning and HIV care delivery. An ageing PWDH population is likely to experience a higher burden of chronic comorbidities, polypharmacy, and functional decline, requiring closer integration among HIV services, multidisciplinary health approaches, geriatric care, and management of non-communicable diseases. Surveillance systems and care models that were originally designed for a younger population of PWDH may need to be adapted to reflect the growing importance of long-term care, multimorbidity, and age-related health needs.^2,12 Anticipating these changes through demographic forecasting is therefore essential for informing resource allocation, workforce training, and the design of future HIV prevention and care strategies in Italy.

Conflicts of interest: none declared.

Publication history: a version of this work was posted on medRxiv: https://doi.org/10.64898/2026.02.02.26345395

Authorship: all authors contributed to study design and conception. Surveillance data were provided by B. Suligoi, V. Regine, and L. Pugliese. Data analysis, software implementation, and tabulation were performed by A. Viguerie. Original draft was written by A. Viguerie. All authors contributed to revision/editing of final manuscript.

References

1. Trickey A, Sabin CA, Burkholder G, et al. Life expectancy after 2015 of adults with HIV on long-term antiretroviral therapy in Europe and North America: a collaborative analysis of cohort studies. Lancet HIV 2023;10(5):e295-307. doi:10.1016/S2352-3018(23)00028-0
2. O’Connor SM, Song R, Viguerie A, et al. Projected Increases in Older People with HIV in the United States Through 2040. Open Forum Infect Dis 2025;12(11):ofaf656. doi:10.1093/ofid/ofaf656
3. Viguerie A, Iacomini E. Ageing and mortality of persons with HIV: a novel data-driven approach. Eur J Appl Math 2025:1-27. doi:10.1017/S0956792525100247
4. Collins LF, Armstrong WS. What It Means to Age with HIV Infection. JAMA Netw Open 2020;3(6):e208023. doi:10.1001/jamanetworkopen.2020.8023
5. Bingham A, Shrestha RK, Khurana N, Jacobson EU, Farnham PG. Estimated Lifetime HIV-Related Medical Costs in the United States. Sex Transm Dis 2021;48(4):299-304. doi:10.1097/OLQ.0000000000001366
6. Rodés B, Cadiñanos J, Esteban-Cantos A, Rodríguez-Centeno J, Arribas JR. Ageing with HIV: Challenges and biomarkers. EBioMedicine 2022;77:103896. doi:10.1016/j.ebiom.2022.103896
7. Taborelli M, Suligoi B, Serraino D, et al. Increased kidney disease mortality among people with AIDS versus the general population: a population-based cohort study in Italy, 2006-2018. BMJ Open 2022;12(12):e064970. doi: 10.1136/bmjopen-2022-064970
8. Grippo F, Grande E, Zucchetto A, et al. Mortality of people with AIDS in Italy: comparison of AIDS surveillance and multiple cause-of-death registries. Ann Ist Super Sanita 2022;58(2):139-45. doi: 10.4415/ANN_22_02_10
9. Taborelli M, Suligoi B, Toffolutti F, et al. Excess liver-related mortality among people with AIDS compared to the general population: an Italian nationwide cohort study using multiple causes of death. HIV Med 2020;21(10):642-49. doi: 10.1111/hiv.12937
10. Suligoi B, Virdone S, Taborelli M, et al. Excess mortality related to circulatory system diseases and diabetes mellitus among Italian AIDS patients vs. non-AIDS population: a population-based cohort study using the multiple causes-of-death approach. BMC Infect Dis 2018;18(1):428. doi: 10.1186/s12879-018-3336-x
11. Grande E, Zucchetto A, Suligoi B, et al. Multiple cause-of-death data among people with AIDS in Italy: a nationwide cross-sectional study. Popul Health Metr 2017;15(1):19. doi: 10.1186/s12963-017-0135-3
12. Pugliese L, Regine V, Suligoi B. Le sorveglianze HIV e AIDS in Italia dal 2013 al 2022: verso l’implementazione della sorveglianza unificata. Boll Epidemiol Naz 2023;4(3):20-26.
13. Regine V, Pugliese L, Ferri M, Santaquilani M, Suligoi B. Aggiornamento delle nuove diagnosi di infezione da HIV e dei casi di AIDS in Italia al 31 dicembre 2024. Not Ist Sup Sanita 2025;38:3-59.
14. Camoni L, Raimondo M, Dorrucci M, et al. Estimating minimum adult HIV Prevalence: A cross-sectional study to assess the characteristics of people living with HIV in Italy. AIDS Res Hum Retroviruses 2015;31(3):282-87. doi: 10.1089/aid.2014.0154
15. Camoni L, Raimondo M, Urciuoli R, et al. People diagnosed with HIV and in care in Italy in 2014: results from the second national survey. AIDS Care 2018;30(6):760-64. doi: 10.1080/09540121.2017.1400639
16. Navarra A, Piselli P, Tavelli A, et al. OC5 Progresses in the continuum of care of people living with HIV, using surveillance and cohort data. Italy, 2012-2023. In: Oral Communication. BMJ 2025:A4-A5. doi: 10.1136/sextrans-ICAR-2025.4
17. Raimondo M, Camoni L, Suligoi B, Pezzotti P, CARPHA Study Group. HIV-Positive Individuals on Antiretroviral Therapy and with Viral Load Suppressed in 12 Infectious Diseases Clinics in Italy: Successes and Disparities in the HIV Continuum of Care. AIDS Res Hum Retroviruses 2017;33(6):575-82. doi: 10.1089/aid.2016.0256
18. Giacomelli A, Lanini S, De Vito A, et al. All-cause and Cause-specific Mortality in People with HIV in Italy in 1997-2022: Data From the Icona Cohort. Open Forum Infect Dis 2025;12(8):ofaf455. doi: 10.1093/ofid/ofaf455
19. Viguerie A, O’Shea J, Johnston M, et al. Impact of increased uptake of long-acting injectable antiretroviral therapy on HIV incidence and viral suppression in the United States under 2021 FDA guidelines. AIDS 2025;39(8):1024-31. doi: 10.1097/QAD.0000000000004144
20. Viguerie A, Iacomini E. Aging and mortality of persons with HIV: a novel data-driven approach. Eur J Appl Math 2025:1-27. doi:10.1017/S0956792525100247