This study is a prospective observational analysis conducted within the framework of the Report-AGE project (ClinicalTrials.gov identifier: NCT01397682), an ongoing multicenter epidemiological cohort aimed at investigating clinical and biological determinants of outcomes in hospitalized older adults. The Report-AGE project has been previously described in detail [15]. Participants were consecutively recruited among patients aged 65 years or older admitted to acute care wards, including geriatric medicine, cardiology, urology, surgery, and neurology, at three Italian National Institute on Health and Aging (Istituto Nazionale di Ricovero e Cura dell’Anziano, Istituto a carattere scientifico, IRCCS INRCA) hospitals between September 2011 and October 2021. During the overall enrollment period, more than 5000 patients were included in the parent cohort, with written informed consent obtained from over 90% of eligible individuals.

Eligibility criteria for inclusion in the Report-AGE cohort comprised age ≥ 65 years, hospitalization lasting at least 24 h, and the ability to provide informed consent. Patients with hematological malignancies or active SARS-CoV-2 infection were excluded. For the purposes of the present analysis, patients who died during the index hospitalization were excluded to minimize potential bias related to acute terminal conditions and to focus on post-discharge mortality.

The present study was conducted as a predefined substudy of the Report-AGE cohort and included patients discharged alive for whom biological samples were available and suitable for LTL assessment. Blood samples were collected at hospital admission and processed and stored according to standardized protocols. The resulting analytical sample consisted of 872 patients and was broadly representative of the overall cohort with respect to age distribution, sex, and long-term vital status. Vital status and mortality outcomes were ascertained through hospital discharge records and linkage with national population registries, ensuring complete follow-up without the need for active participant contact. The study was conducted in accordance with the principles of the Declaration of Helsinki and was approved by the Ethics Committees of the IRCCS INRCA network (CE INRCA 20031, 04/02/2021). All participants provided written informed consent prior to enrollment; consequently, individuals with severe cognitive impairment or impaired consciousness may be underrepresented in the study population.

High molecular weight genomic DNA was extracted from peripheral blood leukocytes using a commercial kit (Qiagen, Milan, Italy), following the manufacturer’s instructions. DNA purity and concentration were assessed using a NanoDrop spectrophotometer (Thermo Fisher Scientific, Waltham, MA, USA) by measuring absorbance at 260 and 280 nm. LTL was quantified as the telomere-to-single copy gene (T/S) ratio by quantitative real-time polymerase chain reaction (PCR), according to the method originally described by Cawthon [16], with minor modifications.

Briefly, each reaction was performed using 20 ng of DNA in a 5 µL volume combined with 10 µL of master mix. A reference DNA sample was serially diluted (1.68-fold) to generate a standard curve consisting of five concentrations ranging from 30 to 2 ng per reaction. To minimize inter-assay variability, telomere and single-copy gene (36B4) amplifications were performed on the same plate. The 36B4 gene, encoding the acidic ribosomal phosphoprotein P0, served as the reference gene in the conventional qPCR assay for telomere length measurement. Primer sequences and concentrations were as previously reported [16].

The thermal cycling conditions consisted of an initial denaturation step at 95 °C for 10 s, followed by 30 cycles of 95 °C for 5 s, 57 °C for 15 s, and 72 °C for 20 s. All samples were analyzed in duplicate, and results were expressed as T/S ratios normalized to a calibrator sample (Roche, Milan, Italy), allowing comparison across different runs. Quantitative PCR reactions were performed using a Rotor-Gene real-time PCR system (Qiagen).

All samples were analyzed in duplicate, and measurements were accepted only when the difference between duplicate Ct values was below 0.5 cycles. Samples exceeding this threshold were reanalyzed. To further reduce technical variability, telomere and single-copy gene amplifications were performed on the same plate using the same calibrator sample across runs whenever possible.

Age-adjusted telomere length residuals were derived using an external reference population of apparently healthy individuals aged 40–90 years. This population comprised 593 subjects (median age of 58 years (IQR 52–67), 38.4% males) drawn from a previously published cohort of community-dwelling individuals recruited in central Italy within prevention-oriented studies [17]. Participants underwent standardized clinical evaluation, including detailed medical history, physical examination, and laboratory testing. Individuals with major chronic conditions were excluded, including diabetes (based on clinical history or fasting glucose criteria), cardiovascular disease (myocardial infarction, stroke, or peripheral vascular disease), and other relevant comorbidities, and were not receiving medications known to interfere with metabolic or inflammatory pathways. All participants were Caucasian, sampled under fasting conditions, and derived from a geographically and lifestyle-homogeneous setting, broadly adhering to a Mediterranean dietary pattern. This reference cohort represents a geographically comparable reference to the Report-AGE study population [18].

Linear regression models were fitted with LTL as the dependent variable and chronological age as the independent variable. For each participant in the hospitalized geriatric cohort, telomere length residuals were calculated as the difference between observed telomere length and the value predicted by the age-specific regression line derived from the healthy reference population. Positive residuals indicate longer-than-expected telomeres for age, whereas negative residuals indicate shorter-than-expected telomeres. The same age-adjusted telomere residuals derived from the healthy reference population were used across all survival analyses, including exploratory analyses focused on 1-year, 5-year, and 10-year mortality.

Comorbidities were recorded at admission and coded according to the International Classification of Diseases, Ninth Revision (ICD-9). Multimorbidity was assessed using the Charlson comorbidity index (CCI). Frailty was quantified as previously described [19] using a frailty index (FI) based on the deficit accumulation model [20, 21]. Frailty was defined as a FI ≥ 0.20 according to established thresholds [22]. Laboratory variables were assessed using standard commercial methods. The glomerular filtration rate (GFR) was estimated by the creatinine-based Berlin Initiative Study-1 (BIS1) equation specifinally used to estimate GFR in older adults [23]. All covariates were assessed at hospital admission.

Continuous variables are described as medians with interquartile ranges (IQRs), while categorical variables are reported as absolute numbers and percentages. The distribution of continuous variables was assessed using the Shapiro–Wilk test. Comparisons between survivors and non-survivors were performed using the Wilcoxon rank-sum test for continuous variables and the chi-square test for categorical variables, as appropriate.

LTL was analyzed as an age-adjusted measure. Age-adjusted telomere length residuals were obtained as described above and analyzed both as a continuous variable (per standard deviation decrease) and as a categorical variable based on a data-driven cut-point derived within the study population using maximally selected rank statistics, as implemented in the surv_cutpoint function of the survminer R package. This approach identifies the threshold that maximizes separation in survival between groups while enforcing a minimum proportion of observations in each group (minprop = 0.10). This cut-point was used to dichotomize participants into groups with shorter versus longer telomere length relative to chronological age.

The association between age-adjusted telomere length residuals and all-cause mortality was investigated using Cox proportional hazards regression models. Univariable Cox models were first fitted to evaluate the crude association between telomere length residuals and mortality. Multivariable Cox models were then constructed to adjust for potential confounders, including demographic characteristics, comorbidity burden, polypharmacy, and laboratory parameters. Covariates were selected based on their clinical relevance and on significant differences observed between survivors and non-survivors at baseline. Continuous covariates were modeled per clinically meaningful increments where appropriate. Results of Cox regression analyses are presented as hazard ratios (HRs) with corresponding 95% confidence intervals (CIs). The proportional hazards assumption was assessed using Schoenfeld residuals and was not violated.

To investigate whether the association between telomere length residuals and mortality differed according to frailty status, interaction terms between telomere length residuals and frailty strata were included in multivariable Cox models. Stratified analyses were subsequently performed to explore the association between telomere length and mortality within different frailty strata.

All statistical analyses were performed using R software (version 4.5.2). A two-sided p value < 0.05 was considered statistically significant.