Abstract
Background
Hearing loss (HL) is common in older adults and is associated with substantial functional decline, yet community‐based evidence on its determinants remains limited in China, particularly across different methods of hearing assessment.
Objectives
To investigate associations and predictors of HL among older adults in southern China using audiometric and self‐reported assessments, and to compare patterns across two community‐based cohorts.
Methods
Data were analyzed from 2664 adults aged ≥ 60 years in Shenzhen and 30,518 adults aged ≥ 50 years from the Guangzhou Biobank Cohort Study (GBCS). Moderate‐to‐severe HL was defined as a pure‐tone average (PTA) ≥ 35 dB hearing level in the better‐hearing ear, calculated from air‐conduction thresholds at 500–8000 Hz. HL was assessed using pure‐tone audiometry in Shenzhen and a validated self‐reported measure in GBCS. Multivariable logistic regression estimated adjusted odds ratios (aORs) with 95% confidence intervals (CIs). Extreme gradient boosting with SHAP values was used to assess predictor importance.
Results
Older age, male sex, and lower household income were consistently associated with higher odds of HL in both cohorts. In Shenzhen, metabolic disease (aOR = 1.28, 95% CI: 1.05–1.57) and otitis media (aOR = 2.65, 95% CI: 1.63–4.33) were positively associated with HL, whereas thyroid disease showed an inverse association. In GBCS, alcohol consumption (aOR = 1.28, 95% CI: 1.15–1.43), arthritis (aOR = 1.37, 95% CI: 1.23–1.52), and stroke (aOR = 1.64, 95% CI: 1.06–2.45) were positively associated, while overweight status and nonmanual occupation were inversely associated. Machine‐learning analyses consistently identified age, sex, education, income, and chronic diseases as key predictors.
Conclusions
HL in older adults shows both shared and cohort‐specific associations across assessment methods, highlighting sociodemographic and health‐related disparities. Targeted community‐based screening and prevention strategies are warranted.
Keywords: community-based study, hearing loss, older adults
1. Introduction
Hearing loss (HL) is a major and growing global public health concern as populations age. The World Health Organization estimates that more than 25% of adults aged ≥ 60 years have disabling HL, with prevalence continuing to rise worldwide [1]. Beyond impairing communication and reducing quality of life [2], HL imposes substantial social and economic burdens, with global costs exceeding USD 981 billion in 2019 [3]. Accumulating evidence further links HL to cognitive decline [4–6], depression [7], and frailty [8], compounding multimorbidity in older adults. Despite the availability of effective interventions such as hearing aids, their uptake remains low. In a study, only 29.2% of individuals with HL used hearing aids, with even lower utilization among Black, Hispanic, and low‐income populations [9]. This substantial treatment gap highlights the urgency of improving early identification and prevention strategies in community settings.
While randomized controlled trials (RCTs) and clinic‐based studies have established the efficacy of hearing interventions, their reliance on selected populations or administrative data limits the generalizability of findings to the broader aging community [10–12]. Community‐based, population‐representative data remain scarce in China, and variation in diagnostic approaches further reduces comparability across studies [13, 14]. These gaps highlight the need for large community samples with standardized data collection to better characterize HL and its determinants in real‐world populations.
To address the evidence gap and improve generalizability beyond clinical samples, we examined HL and its associated factors using cross‐sectional data from two independent cohorts of community‐dwelling older adults in southern China. Hearing was assessed using pure‐tone audiometry in Shenzhen and a standardized self‐reported measure in the Guangzhou Biobank Cohort Study (GBCS). By integrating machine‐learning–based prediction with traditional epidemiological analyses, this study aimed to identify robust and potentially modifiable factors of HL.
2. Methods
2.1. Study Design and Population
This study used data from two independent sources of community‐dwelling older adults in southern China: (1) a routine health screening program for adults aged ≥ 60 years in Longgang District, Shenzhen and (2) the baseline survey of the GBCS. Given differences in hearing assessment methods (audiometry vs. self‐reported measures) and covariate definitions between the two studies, the two study populations were analyzed separately. Study findings are presented comparatively to identify consistent risk factors across populations and measurement approaches.
2.2. Shenzhen Sample
Residents aged ≥ 60 years in Longgang District are provided annual health check‐ups at community health centers. Participants were enrolled using consecutive sampling between January and December 2024. Individuals were eligible if they completed pure‐tone audiometry and the risk factor questionnaire. Participants with missing audiometric or PSQI data were excluded. Written informed consent was obtained from all participants. Details of the study have been reported elsewhere [15].
2.3. GBCS Sample
Details of GBCS studies have been reported previously [16–18]. Briefly, baseline examinations were conducted from September 2003 to January 2008, including 30,518 participants without self‐reported cardiovascular disease at baseline. GBCS is a collaboration among Guangzhou Twelfth People’s Hospital and the Universities of Hong Kong and Birmingham. Participants were recruited from the Guangzhou Health and Happiness Association for the Respectable Elders (GHHARE), a large community organization for local residents aged ≥ 50 years. Membership requires a nominal monthly fee of 4 RMB and covered approximately 7% of this age group in Guangzhou across all districts.
2.4. Outcome Variable
In the Shenzhen sample, trained examiners conducted pure‐tone audiometry in sound‐attenuated environments. Air‐conduction thresholds were measured at 500, 1000, 2000, 4000, and 8000 Hz, consistent with the diagnostic standard pure‐tone audiometry frequency range (0.125–8 kHz) recommended by the ASHA Guidelines for Manual Pure‐Tone Threshold Audiometry [19]. The inclusion of 8000 Hz is further supported by evidence that cochlear damage at frequencies above 8000 Hz precedes threshold shifts at conventional frequencies (250–8000 Hz), rendering high‐frequency measurement particularly sensitive for early detection of age‐related HL [20]. The pure‐tone average (PTA, dB HL) of the better ear was calculated as the mean threshold across these frequencies. Moderate‐to‐severe hearing impairment was defined as PTA ≥ 35 dB HL, while PTA < 35 dB HL was classified as normal/mild impairment. This dichotomous grouping (< 35 vs. ≥ 35 dB HL) served as the primary outcome.
In GBCS, hearing status was self‐assessed at baseline with five response options: excellent, good, fair, poor, and unable to hear. We classified poor and unable to hear as HL (coded as 1), and excellent/good/fair as no HL (coded as 0). Although self‐reported hearing has potential for misclassification, research shows that in older adults it exhibits reasonable validity for detecting HL [21].
2.5. Risk Factor Data Collection
In the Shenzhen sample, a structured questionnaire and clinical examination were used to collect information on potential HL risk factors. Demographic characteristics included age, sex, educational attainment (primary school or less, middle school, college, or above), annual household income (≤ ¥9999; ¥10,000–49,999; ≥ ¥50,000), and lifelong occupation (manual, nonmanual, and other). Lifestyle factors included smoking status (never vs. ever) and alcohol consumption (never vs. current or former). Medical history included physician‐diagnosed cardiovascular, cerebrovascular, and musculoskeletal diseases, as well as metabolic diseases, defined as one or more of the following physician‐diagnosed conditions: diabetes mellitus or hyperlipidemia. Histories of chronic otitis media and otitis externa were specifically collected due to their relevance to auditory function. Occupational or environmental noise exposure (yes/no) was defined as regular exposure to loud noise without proper hearing protection.
Sleep quality was assessed using the Pittsburgh Sleep Quality Index (PSQI). Seven components (subjective sleep quality, sleep latency, sleep duration, sleep efficiency, sleep disturbances, hypnotic use, and daytime dysfunction) were scored from 0 to 3 and summed to yield a global score (0–21), with higher values indicating poorer sleep quality. Following established recommendations, PSQI > 7 indicated a sleep disorder (coded as 1), while ≤ 7 indicated normal sleep (coded as 0). Component scoring followed the official PSQI guidelines [22].
In GBCS, baseline data were collected through standardized interviews and physical examinations conducted by trained personnel. Demographics included age, sex, and socioeconomic status, defined by education (primary or below, middle school, college, or above), occupation (manual, nonmanual, and others), and annual household income (≤ ¥30,000; > ¥30,000; unknown; 1 USD ≈ 7 RMB). Medical history was based on self‐report and physician diagnosis and included metabolic diseases (diabetes mellitus, hypertension, and hyperlipidemia), cardiovascular diseases (coronary heart disease, stroke, angina, myocardial infarction, or peripheral vascular disease), and other conditions (thyroid disease, fracture history, and arthritis).
Trained personnel measured height and weight to compute body mass index (BMI, kg/m2). Blood pressure was measured three times after ≥ 5 min seated rest using an Omron 705CP sphygmomanometer (Omron Corporation, Kyoto, Japan), and the mean of the last two readings was recorded. Hypertension was defined as systolic blood pressure ≥ 140 mmHg or diastolic blood pressure ≥ 90 mmHg. Sleep‐related information included insomnia and hypnotic use > 1 time/week (coded as 1).
2.6. Statistical Analysis
Data processing and statistical analysis were performed using Stata 18 and R Version 4.3. Multiple imputation by chained equations (MICE) was used to handle missing data by generating five imputed datasets. Continuous variables were summarized as mean ± standard deviation (SD) and compared using t‐tests, while categorical variables were expressed as frequency (percentage) and compared using chi‐square tests.
Multivariable logistic regression was used to determine factors independently associated with HL. Multicollinearity was assessed using generalized variance inflation factor (GVIF), with all values < 2.
Additionally, an Extreme Gradient Boosting (XGBoost) model was developed in each sample to assess predictor importance using SHapley Additive exPlanations (SHAP) values. The dataset was split into training (70%) and test (30%) sets using stratified random partitioning. Five‐fold cross‐validation with early stopping (20 rounds) was applied to prevent overfitting. Key hyperparameters included: maximum tree depth = 4, learning rate (eta) = 0.05, gamma = 1, subsample = 0.8, colsample_bytree = 0.8, minimum child weight = 3, L2 regularization (lambda) = 1, and L1 regularization (alpha) = 0.5. To address potential class imbalance in the GBCS sample (negative‐to‐positive ratio approximately 16:1), models with varying scale_pos_weight values (1, 5, 10, and 16.3) were evaluated; as cross‐validation AUC was consistent across all values (range: 0.681–0.682), the unweighted model was retained for its superior training stability. For the GBCS sample, the optimal classification threshold was determined using the Youden index. The primary purpose of the XGBoost analysis was variable importance ranking rather than clinical prediction; performance metrics are reported for methodological transparency.
3. Results
3.1. Baseline Characteristics
A total of 2664 participants from the Longgang ENT Hospital sample and 30,518 from the GBCS sample were included in the analysis. Table 1 shows that, in the Shenzhen sample, 39.5% (1053 individuals) of participants had moderate or greater HL (PTA ≥ 35 dB). Those with HL were older (mean age 69.4 years, SD 6.4 years) than those without (mean age 66.5 years, SD 5.2 years, p < 0.001) and more likely to be male (p < 0.001). HL was associated with lower educational level (p < 0.001), lower household income (p = 0.002), smoking (p < 0.001), metabolic disease (p = 0.042), and otitis media (p < 0.001).
TABLE 1.
Baseline characteristics of study participants from Shenzhen.
| Total | No or mild hearing loss (PTA < 35 dB) N = 1611 | Moderate or greater hearing loss (PTA ≥ 35 dB) N = 1053 | p value | |
|---|---|---|---|---|
| Age, years, mean (SD) | 2664 | 66.49 (5.16) | 69.41 (6.43) | < 0.001 |
| Sex, N (%) | < 0.001 | |||
| Women | 1398 (52.48) | 925 (57.42) | 473 (44.92) | |
| Men | 1266 (47.52) | 686 (42.58) | 580 (55.08) | |
| Obesity status, N (%) | 0.287 | |||
| Underweight | 28 (1.05) | 12 (0.74) | 16 (1.52) | |
| Normal weight | 2213 (83.07) | 1344 (83.43) | 869 (82.53) | |
| Overweight | 373 (14.00) | 224 (13.90) | 149 (14.15) | |
| Obesity | 50 (1.88) | 31 (1.92) | 19 (1.80) | |
| Education, N (%) | < 0.001 | |||
| Primary or below | 1231 (46.21) | 684 (42.46) | 547 (51.95) | |
| Middle school | 1309 (49.14) | 840 (52.14) | 469 (44.54) | |
| College or above | 124 (4.65) | 87 (5.40) | 37 (3.51) | |
| Occupation, N (%) | 0.290 | |||
| Manual | 2214 (83.11) | 1324 (82.18) | 890 (84.52) | |
| Nonmanual | 165 (6.19) | 105 (6.52) | 60 (5.70) | |
| Others | 285 (10.70) | 182 (11.30) | 103 (9.78) | |
| Household income, Yuan/year, N (%) | 0.002 | |||
| < 10,000 | 1274 (47.82) | 726 (45.07) | 548 (52.04) | |
| 10,000–49,999 | 821 (30.82) | 516 (32.03) | 305 (28.96) | |
| ≥ 50,000 | 569 (21.36) | 369 (22.91) | 200 (18.99) | |
| Smoking status, N (%) | < 0.001 | |||
| Never | 2190 (82.21) | 1360 (84.42) | 830 (78.82) | |
| Current/ex‐smoker | 474 (17.79) | 251 (15.58) | 223 (21.18) | |
| Alcohol use, N (%) | 0.088 | |||
| Never | 2350 (88.21) | 1435 (89.08) | 915 (86.89) | |
| Current/ex‐drinker | 314 (11.79) | 176 (10.92) | 138 (13.11) | |
| Metabolic disease, N (%) | 0.042 | |||
| No | 2085 (78.27) | 1282 (79.58) | 803 (76.26) | |
| Yes | 579 (21.73) | 329 (20.42) | 250 (23.74) | |
| Noise exposure, N (%) | 0.519 | |||
| No | 2633 (98.84) | 1594 (98.94) | 1039 (98.67) | |
| Yes | 31 (1.16) | 17 (1.06) | 14 (1.33) | |
| Cerebrovascular disease, N (%) | 0.208 | |||
| No | 2621 (98.39) | 1589 (98.63) | 1032 (98.01) | |
| Yes | 43 (1.61) | 22 (1.37) | 21 (1.99) | |
| Thyroid disease, N (%) | 0.013 | |||
| No | 2642 (99.17) | 1592 (98.82) | 1050 (99.72) | |
| Yes | 22 (0.83) | 19 (1.18) | 3 (0.28) | |
| Musculoskeletal disease, N (%) | 0.684 | |||
| No | 2580 (96.85) | 1562 (96.96) | 1018 (96.68) | |
| Yes | 84 (3.15) | 49 (3.04) | 35 (3.32) | |
| Otitis media, N (%) | < 0.001 | |||
| No | 2584 (97.00) | 1578 (97.95) | 1006 (95.54) | |
| Yes | 80 (3.00) | 33 (2.05) | 47 (4.46) | |
| Otitis externa, N (%) | > 0.999 | |||
| No | 2653 (99.59) | 1604 (99.57) | 1049 (99.62) | |
| Yes | 11 (0.41) | 7 (0.43) | 4 (0.38) | |
| Cardiovascular disease, N (%) | 0.434 | |||
| No | 1530 (57.43) | 935 (58.04) | 595 (56.51) | |
| Yes | 1134 (42.57) | 676 (41.96) | 458 (43.49) | |
| Sleep disorder, N (%) | 0.389 | |||
| No | 2107 (79.09) | 1283 (79.64) | 824 (78.25) | |
| Yes | 557 (20.91) | 328 (20.36) | 229 (21.75) |
Abbreviation: PTA = pure‐tone average (dB).
In the GBCS sample, 5.8% (1763 individuals) of participants reported poor or unable hearing. HL was associated with older age, male sex, lower education, lower household income, underweight status, smoking, alcohol use, arthritis, thyroid disease, hypertension, coronary heart disease, stroke, myocardial infarction, and insomnia (all p < 0.05) (Table 2).
TABLE 2.
Baseline characteristics of study participants in the GBCS cohort.
| Total | Poor or unable to hear N = 1763 | Normal/mild hearing N = 28755 | p value | |
|---|---|---|---|---|
| Age, years, mean (SD) | 30518 | 61.76 (7.00) | 66.51 (7.57) | < 0.001 |
| Sex, N (%) | < 0.001 | |||
| Women | 22075 (72.33) | 21042 (73.18) | 1033 (58.59) | |
| Men | 8443 (27.67) | 7713 (26.82) | 730 (41.41) | |
| Obesity status, N (%) | 0.019 | |||
| Underweight | 1388 (4.55) | 1301 (4.52) | 87 (4.93) | |
| Normal weight | 14946 (48.97) | 14026 (48.78) | 920 (52.18) | |
| Overweight | 10998 (36.04) | 10417 (36.23) | 581 (32.96) | |
| Obesity | 3186 (10.44) | 3011 (10.47) | 175 (9.93) | |
| Education, N (%) | < 0.001 | |||
| Primary or below | 13100 (42.93) | 12181 (42.36) | 919 (52.13) | |
| Middle school | 14684 (48.12) | 13989 (48.65) | 695 (39.42) | |
| College or above | 2734 (8.96) | 2585 (8.99) | 149 (8.45) | |
| Occupation, N (%) | 0.146 | |||
| Manual | 18634 (61.06) | 17519 (60.93) | 1115 (63.24) | |
| Nonmanual | 7278 (23.85) | 6885 (23.94) | 393 (22.29) | |
| Others | 4606 (15.09) | 4351 (15.13) | 255 (14.46) | |
| Household income, Yuan/year, N (%) | < 0.001 | |||
| < 5000 | 792 (2.60) | 707 (2.46) | 85 (4.82) | |
| 5000‐ | 1846 (6.05) | 1716 (5.97) | 130 (7.37) | |
| 10000‐ | 5503 (18.03) | 5120 (17.81) | 383 (21.72) | |
| 20000‐ | 7804 (25.57) | 7344 (25.54) | 460 (26.09) | |
| 30000‐ | 8237 (26.99) | 7820 (27.20) | 417 (23.65) | |
| ≥ 50000 | 6336 (20.76) | 6048 (21.03) | 288 (16.34) | |
| Smoking status, N (%) | < 0.001 | |||
| No | 24641 (80.74) | 23376 (81.29) | 1265 (71.75) | |
| Yes | 5877 (19.26) | 5379 (18.71) | 498 (28.25) | |
| Alcohol use, N (%) | < 0.001 | |||
| No | 20926 (68.57) | 19827 (68.95) | 1099 (62.34) | |
| Yes | 9592 (31.43) | 8928 (31.05) | 664 (37.66) | |
| Frequent hypnotic use (> 1 time/week), N (%) | 0.089 | |||
| No | 29645 (97.14) | 27944 (97.18) | 1701 (96.48) | |
| Yes | 873 (2.86) | 811 (2.82) | 62 (3.52) | |
| Daytime napping, N (%) | 0.093 | |||
| No | 10587 (34.69) | 10008 (34.80) | 579 (32.84) | |
| Yes | 19931 (65.31) | 18747 (65.20) | 1184 (67.16) | |
| Arthritis, N (%) | < 0.001 | |||
| No | 21087 (69.10) | 19954 (69.39) | 1133 (64.27) | |
| Yes | 9431 (30.90) | 8801 (30.61) | 630 (35.73) | |
| Thyroid disease, N (%) | 0.002 | |||
| No | 27885 (91.37) | 26238 (91.25) | 1647 (93.42) | |
| Yes | 2633 (8.63) | 2517 (8.75) | 116 (6.58) | |
| Fracture history, N (%) | 0.428 | |||
| No | 25974 (85.11) | 24485 (85.15) | 1489 (84.46) | |
| Yes | 4544 (14.89) | 4270 (14.85) | 274 (15.54) | |
| Diabetes mellitus, N (%) | 0.633 | |||
| No | 28099 (92.07) | 26481 (92.09) | 1618 (91.78) | |
| Yes | 2419 (7.93) | 2274 (7.91) | 145 (8.22) | |
| Hypertension, N (%) | < 0.001 | |||
| No | 20889 (68.45) | 19775 (68.77) | 1114 (63.19) | |
| Yes | 9629 (31.55) | 8980 (31.23) | 649 (36.81) | |
| Hyperlipidemia, N (%) | 0.277 | |||
| No | 27323 (89.53) | 25731 (89.48) | 1592 (90.30) | |
| Yes | 3195 (10.47) | 3024 (10.52) | 171 (9.70) | |
| Cardiovascular disease, N (%) | 0.160 | |||
| No | 17885 (58.60) | 16880 (58.70) | 1005 (57.01) | |
| Yes | 12633 (41.40) | 11875 (41.30) | 758 (42.99) | |
| Coronary heart disease, N (%) | 0.004 | |||
| No | 29472 (96.57) | 27791 (96.65) | 1681 (95.35) | |
| Yes | 1046 (3.43) | 964 (3.35) | 82 (4.65) | |
| Stroke, N (%) | < 0.001 | |||
| No | 30292 (99.26) | 28557 (99.31) | 1735 (98.41) | |
| Yes | 226 (0.74) | 198 (0.69) | 28 (1.59) | |
| Angina pectoris, N (%) | 0.062 | |||
| No | 30244 (99.10) | 28504 (99.13) | 1740 (98.70) | |
| Yes | 274 (0.90) | 251 (0.87) | 23 (1.30) | |
| Myocardial infarction, N (%) | 0.023 | |||
| No | 30407 (99.64) | 28656 (99.66) | 1751 (99.32) | |
| Yes | 111 (0.36) | 99 (0.34) | 12 (0.68) | |
| Peripheral vascular disease, N (%) | 0.101 | |||
| No | 30475 (99.86) | 28717 (99.87) | 1758 (99.72) | |
| Yes | 43 (0.14) | 38 (0.13) | 5 (0.28) | |
| Insomnia, N (%) | 0.012 | |||
| No | 25300 (82.90) | 23877 (83.04) | 1423 (80.71) | |
| Yes | 5218 (17.10) | 4878 (16.96) | 340 (19.29) |
3.2. Association Between Participant Characteristics and HL
Table 3 shows that in the Shenzhen sample, after mutual adjustment for all variables in the model, older age (aOR = 1.09, 95% CI: 1.07–1.11), male sex (aOR = 1.84, 95% CI: 1.52–2.24), metabolic disease (aOR = 1.28, 95% CI: 1.05–1.57), and otitis media (aOR = 2.65, 95% CI: 1.63–4.33) were independently and positively associated with HL. Higher educational level (aOR ranging from 0.58 to 0.75, p < 0.03), higher household income (aOR ranging from 0.72 to 0.82, p < 0.05), and thyroid disease (aOR = 0.26, 95% CI: 0.06–0.81) were inversely associated with HL.
TABLE 3.
Univariable and multivariable logistic regression analysis of hearing loss among participants from Shenzhen.
| Variable | Crude OR (95% CI) | Crude p value | Adjusted OR (95% CI) † | Adjusted p value |
|---|---|---|---|---|
| Age, years | 1.09 (1.08–1.11) | < 0.001 | 1.09 (1.07–1.11) | < 0.001 |
| Sex | ||||
| Women | Reference | Reference | ||
| Men | 1.65 (1.41–1.93) | < 0.001 | 1.84 (1.52–2.24) | < 0.001 |
| Obesity status | ||||
| Underweight | 2.06 (0.98–4.48) | 0.060 | 1.72 (0.78–3.87) | 0.180 |
| Normal weight | Reference | Reference | ||
| Overweight | 1.03 (0.82–1.29) | 0.804 | 1.05 (0.83–1.34) | 0.672 |
| Obesity | 0.95 (0.52–1.67) | 0.856 | 1.17 (0.63–2.13) | 0.611 |
| Education | ||||
| Primary or below | Reference | Reference | ||
| Middle school | 0.70 (0.60–0.82) | < 0.001 | 0.75 (0.62–0.89) | 0.001 |
| College or above | 0.53 (0.35–0.79) | 0.002 | 0.58 (0.36–0.93) | 0.027 |
| Occupation | ||||
| Manual | Reference | Reference | ||
| Nonmanual | 0.85 (0.61–1.18) | 0.332 | 0.88 (0.60–1.31) | 0.542 |
| Others | 0.84 (0.65–1.09) | 0.188 | 0.86 (0.63–1.16) | 0.330 |
| Household income, Yuan/year | ||||
| < 10,000 | Reference | Reference | ||
| 10,000–49,999 | 0.78 (0.65–0.94) | 0.008 | 0.82 (0.67–0.99) | 0.044 |
| ≥ 50,000 | 0.72 (0.58–0.88) | 0.002 | 0.72 (0.56–0.92) | 0.009 |
| Smoking status | ||||
| Never | Reference | Reference | ||
| Current/ex‐smoker | 1.46 (1.19–1.78) | < 0.001 | 1.09 (0.85–1.40) | 0.482 |
| Alcohol use | ||||
| Never | Reference | Reference | ||
| Current/ex‐drinker | 1.23 (0.97–1.56) | 0.088 | 1.05 (0.80–1.38) | 0.728 |
| Metabolic disease | ||||
| No | Reference | Reference | ||
| Yes | 1.21 (1.01–1.46) | 0.042 | 1.28 (1.05–1.57) | 0.015 |
| Noise exposure | ||||
| No | Reference | Reference | ||
| Yes | 1.26 (0.61–2.57) | 0.520 | 1.27 (0.59–2.71) | 0.540 |
| Cerebrovascular disease | ||||
| No | Reference | Reference | ||
| Yes | 1.47 (0.80–2.69) | 0.211 | 1.17 (0.61–2.22) | 0.632 |
| Thyroid disease | ||||
| No | Reference | Reference | ||
| Yes | 0.24 (0.06–0.70) | 0.022 | 0.26 (0.06–0.81) | 0.039 |
| Musculoskeletal disease | ||||
| No | Reference | Reference | ||
| Yes | 1.10 (0.70–1.70) | 0.684 | 1.24 (0.76–1.99) | 0.384 |
| Otitis media | ||||
| No | Reference | Reference | ||
| Yes | 2.23 (1.43–3.54) | < 0.001 | 2.65 (1.63–4.33) | < 0.001 |
| Otitis externa | ||||
| No | Reference | Reference | ||
| Yes | 0.87 (0.23–2.90) | 0.830 | 0.78 (0.19–2.79) | 0.715 |
| Cardiovascular disease | ||||
| No | Reference | Reference | ||
| Yes | 1.06 (0.91–1.25) | 0.434 | 0.95 (0.80–1.13) | 0.560 |
| Sleep disorder | ||||
| No | Reference | Reference | ||
| Yes | 1.09 (0.90–1.31) | 0.389 | 1.04 (0.85–1.28) | 0.679 |
†Adjusted for age, sex, educational level, occupation, household income, smoking, alcohol consumption, obesity status, comorbidities (metabolic disease, cardiovascular disease, thyroid disease, and musculoskeletal disease), noise exposure, otitis media, otitis externa, and sleep quality.
In the GBCS sample, after mutual adjustment for all variables in the model, older age (aOR = 1.09, 95% CI: 1.08–1.10) and male sex (aOR = 1.64, 95% CI: 1.43–1.88) were strongly and positively associated with HL. Alcohol consumption (aOR = 1.28, 95% CI: 1.15–1.43), arthritis (aOR = 1.37, 95% CI: 1.23–1.52), and stroke (aOR = 1.64, 95% CI: 1.06–2.45) also showed significant positive associations. In contrast, nonmanual occupation (aOR = 0.77, 95% CI: 0.66–0.88), overweight status (aOR = 0.86, 95% CI: 0.77–0.96), and higher household income (aOR ranging from 0.63 to 0.70, p < 0.01) were inversely associated with HL (Table 4).
TABLE 4.
Univariable and multivariable logistic regression analysis of hearing loss in the GBCS cohort.
| Variable | Crude OR (95% CI) | Crude p value | Adjusted OR (95% CI) † | Adjusted p value |
|---|---|---|---|---|
| Age, years | 1.09 (1.09–1.10) | < 0.001 | 1.09 (1.08–1.10) | < 0.001 |
| Sex | ||||
| Women | Reference | Reference | ||
| Men | 1.93 (1.75–2.13) | < 0.001 | 1.64 (1.43–1.88) | < 0.001 |
| Obesity status | ||||
| Underweight | 1.02 (0.81–1.27) | 0.868 | 0.86 (0.68–1.08) | 0.207 |
| Normal weight | Reference | Reference | ||
| Overweight | 0.85 (0.76–0.95) | 0.003 | 0.86 (0.77–0.96) | 0.009 |
| Obesity | 0.89 (0.75–1.04) | 0.154 | 0.92 (0.77–1.09) | 0.335 |
| Education | ||||
| Primary or below | Reference | Reference | ||
| Middle school | 0.66 (0.59–0.73) | < 0.001 | 0.95 (0.84–1.07) | 0.403 |
| College or above | 0.76 (0.64–0.91) | 0.003 | 0.84 (0.68–1.04) | 0.108 |
| Occupation | ||||
| Manual | Reference | Reference | ||
| Nonmanual | 0.90 (0.80–1.01) | 0.071 | 0.77 (0.66–0.88) | < 0.001 |
| Others | 0.92 (0.80–1.06) | 0.248 | 1.01 (0.87–1.17) | 0.918 |
| Household income, Yuan/year | ||||
| < 5000 | Reference | Reference | ||
| 5000‐ | 0.63 (0.47–0.84) | 0.002 | 0.66 (0.50–0.89) | 0.006 |
| 10000‐ | 0.62 (0.49–0.80) | < 0.001 | 0.70 (0.55–0.91) | 0.006 |
| 20000‐ | 0.52 (0.41–0.67) | < 0.001 | 0.67 (0.52–0.87) | 0.002 |
| 30000‐ | 0.44 (0.35–0.57) | < 0.001 | 0.68 (0.53–0.89) | 0.004 |
| ≥ 50000 | 0.40 (0.31–0.51) | < 0.001 | 0.63 (0.48–0.83) | < 0.001 |
| Smoking status | ||||
| No | Reference | Reference | ||
| Yes | 1.71 (1.53–1.90) | < 0.001 | 0.94 (0.82–1.08) | 0.403 |
| Alcohol use | ||||
| No | Reference | Reference | ||
| Yes | 1.34 (1.21–1.48) | < 0.001 | 1.28 (1.15–1.43) | < 0.001 |
| Frequent hypnotic use (> 1 time/week) | ||||
| No | Reference | Reference | ||
| Yes | 1.26 (0.96–1.62) | 0.089 | 1.15 (0.86–1.52) | 0.328 |
| Daytime napping | ||||
| No | Reference | Reference | ||
| Yes | 1.09 (0.99–1.21) | 0.093 | 0.98 (0.88–1.09) | 0.715 |
| Arthritis | ||||
| No | Reference | Reference | ||
| Yes | 1.26 (1.14–1.39) | < 0.001 | 1.37 (1.23–1.52) | < 0.001 |
| Thyroid disease | ||||
| No | Reference | Reference | ||
| Yes | 0.73 (0.60–0.89) | 0.002 | 0.88 (0.72–1.06) | 0.190 |
| Fracture history | ||||
| No | Reference | Reference | ||
| Yes | 1.06 (0.92–1.20) | 0.428 | 1.01 (0.88–1.16) | 0.871 |
| Diabetes mellitus | ||||
| No | Reference | Reference | ||
| Yes | 1.04 (0.87–1.24) | 0.633 | 0.97 (0.81–1.16) | 0.738 |
| Hypertension | ||||
| No | Reference | Reference | ||
| Yes | 1.28 (1.16–1.42) | < 0.001 | 0.97 (0.87–1.08) | 0.538 |
| Hyperlipidemia | ||||
| No | Reference | Reference | ||
| Yes | 0.91 (0.77–1.07) | 0.277 | 0.95 (0.79–1.13) | 0.552 |
| Cardiovascular disease | ||||
| No | Reference | Reference | ||
| Yes | 1.07 (0.97–1.18) | 0.160 | 0.95 (0.84–1.07) | 0.399 |
| Coronary heart disease | ||||
| No | Reference | Reference | ||
| Yes | 1.41 (1.11–1.76) | 0.004 | 1.06 (0.83–1.35) | 0.629 |
| Stroke | ||||
| No | Reference | Reference | ||
| Yes | 2.33 (1.53–3.41) | < 0.001 | 1.64 (1.06–2.45) | 0.020 |
| Angina pectoris | ||||
| No | Reference | Reference | ||
| Yes | 1.50 (0.95–2.25) | 0.064 | 1.24 (0.77–1.90) | 0.346 |
| Myocardial infarction | ||||
| No | Reference | Reference | ||
| Yes | 1.98 (1.03–3.47) | 0.026 | 1.28 (0.66–2.29) | 0.434 |
| Peripheral vascular disease | ||||
| No | Reference | Reference | ||
| Yes | 2.15 (0.74–4.98) | 0.108 | 1.52 (0.51–3.68) | 0.402 |
| Insomnia | ||||
| No | Reference | Reference | ||
| Yes | 1.17 (1.03–1.32) | 0.012 | 1.12 (0.98–1.27) | 0.105 |
†Adjusted for age, sex, education, occupation, household income, physical exposure, insomnia, hypnotic use, daytime napping, arthritis, thyroid disease, fracture history, diabetes mellitus, hypertension, hyperlipidemia, cardiovascular disease, coronary heart disease, stroke, angina, myocardial infarction, peripheral vascular disease, smoking, alcohol use, and BMI, as appropriate.
3.3. XGBoost Model and Feature Importance
In the Shenzhen sample, the XGBoost model achieved an AUC of 0.694 on the held‐out test set (training AUC: 0.710; accuracy: 67.2%; sensitivity: 40.3%; specificity: 84.7%). In the GBCS sample, the model achieved an AUC of 0.695 (training AUC: 0.708; accuracy: 66.6%; sensitivity: 65.3%; specificity: 66.7%), with an optimal classification threshold of 0.108. As the primary aim was variable importance ranking, these metrics are reported for transparency rather than predictive utility. ROC curves for both models are presented in Supporting Figures 1 and 2.
SHAP value analysis identified age, sex, education, household income, and smoking status as the most influential predictors across both samples. In the GBCS sample, age showed the highest mean |SHAP| value (0.191), followed by sex (0.042), education (0.025), arthritis (0.022), alcohol use (0.015), household income (0.014), and smoking status (0.014). In the Shenzhen sample, age, sex, and education were also leading predictors, while metabolic disease contributed more prominently compared with the GBCS sample (Figure 1). Multicollinearity testing showed that all predictors had adjusted GVIFs below 2, indicating no significant multicollinearity (Supporting Tables 1 and 2).
FIGURE 1.
These figures depict top predictor importance based on SHAP values in the XGBoost model. Subpanels (a) and (b) correspond to the Shenzhen cohort and the GBCS cohort, respectively. The horizontal axis shows the mean absolute SHAP values, and the vertical axis lists the predictors included in the model. Bars represent the relative contribution of each predictor to the model’s output as estimated by SHAP.
3.4. Sensitivity Analysis
The sensitivity analysis restricted to GBCS participants aged ≥ 60 years (n = 17,120; HL prevalence: 8.0%) yielded a similar pattern of predictor importance in the XGBoost model. Age remained the dominant predictor (mean |SHAP| = 0.082), followed by sex (0.037), alcohol use (0.011), and household income (0.006), broadly consistent with the findings from the full GBCS sample (Supporting Figure 3).
4. Discussion
In this study, we analyzed two population‐based samples to examine factors associated with HL in older adults. In the Shenzhen sample, metabolic disease and otitis media showed stronger positive associations with HL, whereas in the GBCS, alcohol consumption and arthritis were more prominently associated. In contrast, thyroid disease in the Shenzhen sample and overweight status and nonmanual occupation in the GBCS sample were inversely associated with HL. Despite these cohort‐specific patterns, several associations showed broadly similar directional trends across the two cohorts. Although HL was assessed using different approaches—pure‐tone audiometry in the Shenzhen sample and self‐reported hearing status in the GBCS sample—self‐reported measures may be influenced by educational attainment, cognitive status, health awareness, and sociocultural perceptions of hearing difficulties, potentially introducing reporting bias and limiting direct comparability between cohorts. Nevertheless, associations for older age, male sex, lower household income, and smoking remained consistent across both cohorts despite differences in hearing ascertainment, suggesting that these findings may not be entirely attributable to measurement‐specific bias and may instead reflect broader population‐level patterns of HL in older adults.
Our findings align with previous reports. Results are consistent with previous studies identifying age, male sex [23, 24], and lower income as major risk factors for age‐related HL [25, 26]. The association between metabolic disease and HL observed in the Shenzhen sample also accords with findings from a systematic review by Rim et al. [27] and a cross‐sectional study in the Democratic Republic of Congo [28]. Furthermore, smoking emerged as a significant factor associated with HL in both samples, consistent with previous epidemiological studies suggesting that tobacco exposure may contribute to age‐related auditory decline and the burden of hearing impairment [29–32]. In addition, alcohol consumption—representing an adverse lifestyle factor—was significantly associated with HL, in accordance with evidence from an Australian cohort study [33], a cross‐sectional study in England [26], and a Japanese cross‐sectional study [34]. Taken together, these findings further support the multifactorial nature of age‐related HL and highlight the potential contribution of metabolic conditions and adverse lifestyle factors, particularly smoking and alcohol consumption, to auditory decline in older adults.
A notable and clinically relevant finding in our data is the independent association between arthritis and HL in the GBCS sample. Emerging evidence suggests an increased prevalence of hearing impairment in patients with inflammatory arthritis, particularly rheumatoid arthritis [25, 35]. Several mechanisms could explain this association. First, systemic inflammation characteristic of inflammatory arthritis may contribute to cochlear microvascular dysfunction, immune‐mediated injury, and neuroinflammation in the inner ear, with cytokines such as TNF‐α and IL‐1β implicated in cochlear degeneration [36–38]. Autoimmune inner‐ear processes (disease‐specific autoimmunity or cross‐reactive antibodies) may also precipitate sensorineural injury in susceptible individuals [39–41]. Additionally, comorbid vascular disease or shared cardiometabolic risk factors common in people with arthritis may induce ischemic injury to the cochlea via microvascular compromise [42–46]. Prolonged exposure to analgesics and anti‐inflammatory agents, such as regular NSAID or high‐dose salicylate use, has been associated with increased risk of hearing impairment and may exacerbate cochlear vulnerability [47–49]. These pathways—chronic low‐grade inflammation, immune dysregulation, microvascular insufficiency, and drug‐related ototoxicity—overlap with biological processes implicated in age‐related auditory decline, including inflammaging [50] and cochlear microvascular deterioration [51, 52], and may therefore accelerate HL in older adults [53, 54].
This study has several strengths. First, both samples were derived from population‐based cohorts, enabling the examination of factors associated with HL in populations that more accurately reflect real‐world aging communities and helping address the current evidence base largely shaped by clinic‐based studies. Second, analysis of two independent community samples enhanced the robustness and generalizability of the findings. Third, the application of both traditional multivariable logistic regression and machine‐learning approaches produced broadly consistent results, strengthening confidence in the observed associations.
This study also has limitations. First, the analysis was based on cross‐sectional data from two cities in southern China, which limits generalizability and precludes causal inference. Second, we acknowledge that inclusion of the 8000 Hz threshold may yield higher prevalence estimates compared with studies using the conventional four‐frequency PTA (0.5, 1, 2, and 4 kHz). However, high‐frequency hearing assessment is considered more sensitive for detecting early cochlear damage and may facilitate earlier identification of age‐related HL. Third, the two cohorts differed in age eligibility criteria (≥ 60 years in Shenzhen vs. ≥ 50 years in GBCS), and this difference may partially account for the substantially lower HL prevalence observed in GBCS (5.8% vs. 39.5%), which should be considered when interpreting cross‐cohort comparisons of prevalence estimates. However, the sensitivity analysis suggests that the observed cross‐cohort consistency in key determinants is robust to differences in age structure. At last, several factors potentially associated with HL were not available in the present datasets, including hearing aid use, ototoxic medication exposure, detailed occupational noise exposure history, and cognitive status. These factors may influence both hearing perception and hearing‐related behaviors and could therefore contribute to residual confounding or biased effect estimates. Further studies with more comprehensive assessments are needed to clarify these associations.
5. Conclusion
In two large community‐based cohorts of older adults in southern China, we identified both shared and cohort‐specific associations with HL. Older age, male sex, and lower socioeconomic status showed broadly consistent associations across cohorts, while metabolic conditions and adverse lifestyle factors, particularly smoking and alcohol consumption, may also contribute to HL in aging populations. These findings highlight persistent disparities in hearing health and support the need for improved early identification and targeted prevention strategies in community settings.
Funding
No funding was received for this manuscript.
Conflicts of Interest
The authors declare no conflicts of interest.
Supporting Information
Additional supporting information can be found online in the Supporting Information section.
Supporting information
Supporting Information Supporting information is available online and includes additional tables and figures supporting the findings of this study. Specifically, the supporting section provides multicollinearity diagnostics (Supporting Tables 1–2), XGBoost model performance metrics and sensitivity analyses (Supporting Tables 3–4), receiver operating characteristic curves for the Shenzhen and GBCS XGBoost models (Supporting Figures 1–2), and SHAP importance rankings for the GBCS subsample aged ≥ 60 years (Supporting Figure 3).
Zhu, Dian , Zhong, Xutong , Li, Ruiqiang , Zeng, Xianhai , Xu, Lin , Li, Juanjuan , Hearing Loss in Older Adults: Consistent Determinants Across Two Community‐Based Cohorts in Southern China, Journal of Aging Research, 2026, 4809368, 13 pages, 2026. 10.1155/jare/4809368
Lin Xu and Juanjuan Li share joint senior authorship.
Dian Zhu and Xutong Zhong contributed to the work equally and should be considered co‐first authors.
Academic Editor: Fenise Selin Karalı
Contributor Information
Lin Xu, Email: l.xu.5@bham.ac.uk.
Juanjuan Li, Email: ljjent@163.com.
Fenise Selin Karalı, Email: skarali@biruni.edu.tr.
Data Availability Statement
The data that support the findings of this study are available on request from the corresponding author. The data are not publicly available due to privacy or ethical restrictions.
References
- 1. WHO, Deafness and Hearing Loss (Fact Sheet), 2025, World Health Organization, https://www.who.int/news-room/fact-sheets/detail/deafness-and-hearing-loss. [Google Scholar]
- 2. Ciorba A., Bianchini C., Pelucchi S., and Pastore A., The Impact of Hearing Loss on the Quality of Life of Elderly Adults, Clinical Interventions in Aging. (2012) 7, 159–163, 10.2147/cia.s26059. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 3. McDaid D., Park A. L., and Chadha S., Estimating the Global Costs of Hearing Loss, International Journal of Audiology. (2021) 60, no. 3, 162–170, 10.1080/14992027.2021.1883197. [DOI] [PubMed] [Google Scholar]
- 4. Conceição Santos de Oliveira D., Gomes-Filho I. S., Araújo E. M. et al., Association Between Hearing Loss and Cognitive Decline in the Elderly: a Systematic Review with meta-analysis Study, PLoS One. (2023) 18, no. 11, 10.1371/journal.pone.0288099. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 5. Dillard L. K., Pinto A., Mueller K. D. et al., Associations of Hearing Loss and Hearing Aid Use with Cognition, Health-Related Quality of Life, and Depressive Symptoms, Journal of Aging and Health. (2022) 35, no. 7-8, 455–465, 10.1177/08982643221138162. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 6. Yuan J., Sun Y., Sang S., Pham J. H., and Kong W. J., The Risk of Cognitive Impairment Associated with Hearing Function in Older Adults: a Pooled Analysis of Data from Eleven Studies, Scientific Reports. (2018) 8, no. 1, 10.1038/s41598-018-20496-w. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 7. Li F., Jin M., Ma T., and Cui C., Association Between age-related Hearing Loss and Depression: a Systematic Review and meta-analysis, PLoS One. (2025) 20, no. 1, 10.1371/journal.pone.0298495. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 8. Assi S., Garcia Morales E. E., Windham B. G. et al., Hearing Loss and Frailty Among Older Adults: the Atherosclerosis Risk in Communities Study, Journal of the American Medical Directors Association. (2023) 24, no. 11, 1683–1689.e5, 10.1016/j.jamda.2023.08.023. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 9. Reed N. S., Garcia-Morales E. E., Myers C. et al., Prevalence of Hearing Loss and Hearing Aid Use Among US Medicare Beneficiaries Aged 71 Years and Older, JAMA Network Open. (2023) 6, no. 7, 10.1001/jamanetworkopen.2023.26320. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 10. Huang A. R., Morales E. G., Arnold M. L. et al., A Hearing Intervention and Health-Related Quality of Life in Older Adults: a Secondary Analysis of the ACHIEVE Randomized Clinical Trial, JAMA Network Open. (2024) 7, no. 11, 10.1001/jamanetworkopen.2024.46591. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 11. Tonelli M., Wiebe N., Lunney M. et al., Associations Between Hearing Loss and Clinical Outcomes: Population-Based Cohort Study, EClinicalMedicine. (2023) 61, 10.1016/j.eclinm.2023.102068. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 12. Patterson R., Schuh M., Bush M. L., Nieman C. L., Robler S. K., and Emmett S. D., Expanding Clinical Trials Designs to Extend Equitable Hearing Care, Ear and Hearing. (2022) 43, no. S1, 23S–32S, 10.1097/aud.0000000000001165. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 13. Liang Z., Li A., Xu Y., Qian X., and Gao X., Hearing Loss and Dementia: a Meta-Analysis of Prospective Cohort Studies, Frontiers in Aging Neuroscience. (2021) 13, 10.3389/fnagi.2021.695117. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 14. Collaborators G. U. H. L. et al., Hearing Loss Prevalence, Years Lived with Disability, and Hearing Aid Use in the United States from 1990 to 2019: Findings from the Global Burden of Disease Study, Ear and Hearing. (2024) 45, no. 1, 257–267. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 15. Li J. J., Gao W. X., Wang J. et al., A Composite Risk Score to Identify Older Adults at High Risk of Hearing Loss in a Community Screening Program, BMC Geriatrics. (2025) 25, no. 1, 10.1186/s12877-025-06208-w. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 16. Xu L., Jiang C. Q., Lam T. H., Thomas G. N., Zhang W. S., and Cheng K. K., Aortic Arch Calcification and Vascular Disease: the Guangzhou Biobank Cohort Study, Cardiology. (2010) 117, no. 4, 260–264, 10.1159/000323285. [DOI] [PubMed] [Google Scholar]
- 17. Xu L., Jiang C. Q., Lam T. H., Thomas G. N., Zhang W. S., and Cheng K. K., Passive Smoking and Aortic Arch Calcification in Older Chinese Never Smokers: the Guangzhou Biobank Cohort Study, International Journal of Cardiology. (2011) 148, no. 2, 189–193, 10.1016/j.ijcard.2009.11.001. [DOI] [PubMed] [Google Scholar]
- 18. Jiang C., Thomas G. N., Lam T. H. et al., Cohort Profile: the Guangzhou Biobank Cohort Study, a Guangzhou-Hong Kong-Birmingham Collaboration, International Journal of Epidemiology. (2006) 35, no. 4, 844–852, 10.1093/ije/dyl131. [DOI] [PubMed] [Google Scholar]
- 19. Guidelines for Manual pure-tone Threshold Audiometry, Asha. (1978) 20, no. 4, 297–301. [PubMed] [Google Scholar]
- 20. Škerková M., Kovalová M., and Mrázková E., High-Frequency Audiometry for Early Detection of Hearing Loss: a Narrative Review, International Journal of Environmental Research and Public Health. (2021) 18, no. 9, 10.3390/ijerph18094702. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 21. Oosterloo B. C., Homans N. C., Baatenburg de Jong R. J., Ikram M. A., Nagtegaal A. P., and Goedegebure A., Assessing Hearing Loss in Older Adults with a Single Question and Person Characteristics; Comparison with Pure Tone Audiometry in the Rotterdam Study, PLoS One. (2020) 15, no. 1, 10.1371/journal.pone.0228349. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 22. Buysse D. J., Reynolds C. F.III, Monk T. H., Berman S. R., and Kupfer D. J., The Pittsburgh Sleep Quality Index: a New Instrument for Psychiatric Practice and Research, Psychiatry Research. (1989) 28, no. 2, 193–213, 10.1016/0165-1781(89)90047-4. [DOI] [PubMed] [Google Scholar]
- 23. Dillard L. K., Matthews L. J., and Dubno J. R., Sex- and Race-specific Prevalence of Hearing Loss Across the Adult Lifespan and Associated Factors, JAMA Otolaryngology–Head & Neck Surgery. (2025) 151, no. 6, 576–583, 10.1001/jamaoto.2025.0534. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 24. Dillard L. K., Matthews L. J., and Dubno J. R., Race- and Sex-specific Differences in the Risk of Incident Hearing Loss and Associated Factors, Scientific Reports. (2025) 15, no. 1, 10.1038/s41598-025-96937-0. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 25. Li F. F., Fu Z. Y., Han K. et al., Trends and Driving Factors of age-related Hearing Loss and Severity over 30 years: a cross-sectional Study, BMC Geriatrics. (2025) 25, no. 1, 10.1186/s12877-025-06066-6. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 26. Tsimpida D., Kontopantelis E., Ashcroft D., and Panagioti M., Socioeconomic and Lifestyle Factors Associated with Hearing Loss in Older Adults: a cross-sectional Study of the English Longitudinal Study of Ageing (ELSA), BMJ Open. (2019) 9, no. 9, 10.1136/bmjopen-2019-031030. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 27. Rim H. S., Kim M. G., Park D. C. et al., Association of Metabolic Syndrome with Sensorineural Hearing Loss, Journal of Clinical Medicine. (2021) 10, no. 21, 10.3390/jcm10214866. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 28. Sokolo J. G., Nkodila A. N., Lema G. M., Matanda R. N., Kabututu P. Z., and Longo-Mbenza B., Influence of Metabolic Syndrome in the Occurrence of Hearing Disorders in the Democratic Republic of Congo, Endocrine and Metabolic Science. (2025) 17, 10.1016/j.endmts.2025.100227. [DOI] [Google Scholar]
- 29. Cruickshanks K. J., Klein R., Klein B. E. K., Wiley T. L., Nondahl D. M., and Tweed T. S., Cigarette Smoking and Hearing Lossthe Epidemiology of Hearing Loss Study, JAMA. (1998) 279, no. 21, 1715–1719, 10.1001/jama.279.21.1715. [DOI] [PubMed] [Google Scholar]
- 30. Dawes P., Cruickshanks K. J., Moore D. R. et al., Cigarette Smoking, Passive Smoking, Alcohol Consumption, and Hearing Loss, Journal of the Association for Research in Otolaryngology. (2014) 15, no. 4, 663–674, 10.1007/s10162-014-0461-0. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 31. Nondahl D. M., Cruickshanks K. J., Dalton D. S. et al., Serum Cotinine Level and Incident Hearing Loss: a Case-Control Study, Archives of Otolaryngology-Head and Neck Surgery. (2004) 130, no. 11, 1260–1264, 10.1001/archotol.130.11.1260. [DOI] [PubMed] [Google Scholar]
- 32. Uraguchi K. et al., Impact of Smoking on age-related Deterioration in Hearing: a Cohort Study Using Health Checkups, Archives of Gerontology and Geriatrics. (2025) . [DOI] [PubMed] [Google Scholar]
- 33. Gopinath B., Tran Y., Tang D., Burlutsky G., McMahon C., and Mitchell P., Association Between Birthweight and Hearing Loss in Older Adults, Maturitas. (2022) 157, 57–61, 10.1016/j.maturitas.2021.11.008. [DOI] [PubMed] [Google Scholar]
- 34. Takahashi H., Suzuki J., Motoike I. N. et al., Relationship Between age-related Hearing Loss and Alcohol Consumption in a Japanese Population, Scientific Reports. (2025) 16, no. 1, 10.1038/s41598-025-29634-7. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 35. Chaitidis N., Theocharis P., Festas C., and Aritzi I., Association of Rheumatoid Arthritis with Hearing Loss: a Systematic Review and meta-analysis, Rheumatology International. (2020) 40, no. 11, 1771–1779, 10.1007/s00296-020-04609-1. [DOI] [PubMed] [Google Scholar]
- 36. Fujioka M., Okano H., and Ogawa K., Inflammatory and Immune Responses in the Cochlea: Potential Therapeutic Targets for Sensorineural Hearing Loss, Frontiers in Pharmacology. (2014) 5, 10.3389/fphar.2014.00287. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 37. Fuentes-Santamaria V. et al., The Role of Glia in the Peripheral and Central Auditory System Following Noise Overexposure: Contribution of TNF-Alpha and IL-1beta to the Pathogenesis of Hearing Loss, Frontiers in Neuroanatomy. (2017) 11. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 38. do Amaral J. B., Peron K. A., Soeiro T. L. T. et al., The Inflammatory and Metabolic Status of Patients with sudden-onset Sensorineural Hearing Loss, Frontiers in Neurology. (2024) 15, 10.3389/fneur.2024.1382096. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 39. Mijovic T., Zeitouni A., and Colmegna I., Autoimmune Sensorineural Hearing Loss: the otology-rheumatology Interface, Rheumatology. (2013) 52, no. 5, 780–789, 10.1093/rheumatology/ket009. [DOI] [PubMed] [Google Scholar]
- 40. Ciorba A., Corazzi V., Bianchini C. et al., Autoimmune Inner Ear Disease (AIED): a Diagnostic Challenge, International Journal of Immunopathology and Pharmacology. (2018) 32, 10.1177/2058738418808680. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 41. Athanasopoulos M., Samara P., and Athanasopoulos I., Decoding the Impact of autoinflammatory/autoimmune Diseases on Inner Ear Harmony and Hearing Loss, Exploration of Immunology. (2024) 73–89, 10.37349/10.37349/ei.2024.00129. [DOI] [Google Scholar]
- 42. Khosravipour M. and Rajati F., Sensorineural Hearing Loss and Risk of Stroke: a Systematic Review and meta-analysis, Scientific Reports. (2021) 11, no. 1, 10.1038/s41598-021-89695-2. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 43. Guo Q., Li Z., Luo C. et al., Metabolic Shifts in Murine Inner Ears Mimicking Erythrocytes and Plasma Reveal Diagnostics and Early Predictors for age-related Hearing Loss, iScience. (2025) 28, no. 9, 10.1016/j.isci.2025.113285. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 44. Dillard L. K., Ghosh S., Dubno J. R. et al., The 25-Year Incidence and Progression of Hearing Loss in the Framingham Offspring Study, JAMA Network Open. (2025) 8, no. 10, 10.1001/jamanetworkopen.2025.39371. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 45. Kim S. J., Reed N., Betz J. F. et al., Association Between Microvascular Retinal Signs and Age-Related Hearing Loss in the Atherosclerosis Risk in Communities Neurocognitive Study (ARIC-NCS), JAMA Otolaryngology–Head & Neck Surgery. (2020) 146, no. 2, 152–159, 10.1001/jamaoto.2019.3987. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 46. Xu R., Qi X., Li Y. et al., Association Between Retinal Vascular Fractal Dimension and Hearing Loss: a cross-sectional Study, Scientific Reports. (2025) 15, no. 1, 10.1038/s41598-025-16451-1. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 47. Curhan S. G., Eavey R., Shargorodsky J., and Curhan G. C., Analgesic Use and the Risk of Hearing Loss in Men, Americas Journal of Medicine. (2010) 123, no. 3, 231–237, 10.1016/j.amjmed.2009.08.006. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 48. Lin B. M., Curhan S. G., Wang M., Eavey R., Stankovic K. M., and Curhan G. C., Duration of Analgesic Use and Risk of Hearing Loss in Women, American Journal of Epidemiology. (2017) 185, no. 1, 40–47, 10.1093/aje/kww154. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 49. Clark D. P. Q., Zhou Z., Hussain S. M. et al., Low-Dose Aspirin and Progression of Age-Related Hearing Loss: a Secondary Analysis of the ASPREE Randomized Clinical Trial, JAMA Network Open. (2024) 7, no. 7, 10.1001/jamanetworkopen.2024.24373. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 50. Kociszewska D. and Vlajkovic S., Age-Related Hearing Loss: the Link Between Inflammaging, Immunosenescence, and Gut Dysbiosis, International Journal of Molecular Sciences. (2022) 23, no. 13, 10.3390/ijms23137348. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 51. Shi X., Physiopathology of the Cochlear Microcirculation, Hearing Research. (2011) 282, no. 1-2, 10–24, 10.1016/j.heares.2011.08.006. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 52. Gratton M. A., Schmiedt R. A., and Schulte B. A., Age-Related Decreases in Endocochlear Potential Are Associated with Vascular Abnormalities in the Stria Vascularis, Hearing Research. (1996) 94, no. 1-2, 116–124, 10.1016/0378-5955(96)00011-1. [DOI] [PubMed] [Google Scholar]
- 53. Watson N., Ding B., Zhu X., and Frisina R. D., Chronic Inflammation-Inflammaging-in the Ageing Cochlea: a Novel Target for Future Presbycusis Therapy, Ageing Research Reviews. (2017) 40, 142–148, 10.1016/j.arr.2017.10.002. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 54. Kalinec G. M., Lomberk G., Urrutia R. A., and Kalinec F., Resolution of Cochlear Inflammation: Novel Target for Preventing or Ameliorating Drug-Noise- and age-related Hearing Loss, Frontiers in Cellular Neuroscience. (2017) 11, 10.3389/fncel.2017.00192. [DOI] [PMC free article] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.
Supplementary Materials
Supporting Information Supporting information is available online and includes additional tables and figures supporting the findings of this study. Specifically, the supporting section provides multicollinearity diagnostics (Supporting Tables 1–2), XGBoost model performance metrics and sensitivity analyses (Supporting Tables 3–4), receiver operating characteristic curves for the Shenzhen and GBCS XGBoost models (Supporting Figures 1–2), and SHAP importance rankings for the GBCS subsample aged ≥ 60 years (Supporting Figure 3).
Data Availability Statement
The data that support the findings of this study are available on request from the corresponding author. The data are not publicly available due to privacy or ethical restrictions.