Epidemiology and factors associated with Talaromyces marneffei detection in the mainland of China, 2022–2024: a cross-sectional study of hospitalized patients

Abstract

Background

Epidemiological studies of Talaromyces marneffei (T. marneffei) infection are largely confined to HIV-positive populations and specific geographic regions, whereas its population-scale characteristics and environmental factors associated with infection in China are lacking. This study investigates spatiotemporal patterns, host susceptibility, and environmental correlates of T. marneffei detection in Chinese mainland.

Methods

Targeted next-generation sequencing (tNGS) data from patients hospitalized for acute respiratory tract infections (ARTIs) from 2022 to 2024 were used for epidemiological, spatiotemporal, and co-detection analyses. Geographical detector models with the q-statistic were employed to quantify the associations of meteorological, host distribution, and social factors on detection risk, where the q-statistic measures the proportion of spatial variance explained by each factor.

Results

Among 2,316 reported cases, we identified significant spatial clustering, with bimodal seasonal peaks in detection rate. Males and individuals aged 41–50 showed the highest susceptibility. Pneumocystis jirovecii was the predominant co-detected pathogen; 5 pathogens were positively and 16 were negatively correlated with T. marneffei. Univariate analysis using the q-statistic revealed that dew point temperature had the strongest explanatory power for T. marneffei detection at 48.55%. Bivariate interaction analysis demonstrated that paired factor combinations exhibited enhanced explanatory power for disease prevalence compared with single factors. Average air pressure, which alone explained only 1.4% of the spatial variance, showed markedly higher explanatory power when paired with other variables.

Conclusions

This study revealed spatiotemporal heterogeneity, population susceptibility, and environmental factors associated with T. marneffei detection in Chinese mainland, which can guide disease monitoring and control through enhanced surveillance and tailored interventions in epidemic hotspots and among high-risk groups.

Supplementary Information

The online version contains supplementary material available at https://doi.org/10.1186/s12889-026-28001-z.

Introduction

Talaromycosis is an invasive fungal infection caused by the thermally dimorphic pathogen Talaromyces marneffei (T. marneffei), which is endemic mainly to tropical and subtropical regions of Southeast Asia [1]. It is a pathogenic species within TalaromycesTalaromyces capable of causing systemic infections in both humans and animals [2]. This disease primarily affects the monocyte-macrophage reticuloendothelial system [3], resulting in diverse and severe clinical manifestations. The main clinical features involve the respiratory system and are accompanied by a systemic inflammatory response [4]. However, the disease course is often complicated by multiple comorbidities, posing a significant challenge for clinical management, frequently resulting in severe treatment-related adverse reactions, rapid progression, and high mortality rates [5]. Notably, T. marneffei was ranked as the second most concerning fungal pathogen globally in 2018, underscoring its significant public health threat [6].

Historically, talaromycosis has been a leading opportunistic cause of mortality among acquired immunodeficiency syndrome (AIDS) patients in endemic areas [7, 8]. Although traditionally linked to human immunodeficiency virus (HIV)-infected individuals, talaromycosis is increasingly reported in HIV-negative populations, who often experience more severe symptoms, persistent infections, and diagnostic delays [9]. Moreover, the widespread adoption of highly active antiretroviral therapy (HAART) and enhanced HIV prevention efforts have contributed to a decline in HIV-associated cases [10]. In contrast, the number of cases of talaromycosis is increasing among HIV-negative individuals, potentially because of factors such as organ transplantation, autoimmune diseases, and the use of biological agents [11]. Despite this epidemiological shift, research on non-HIV-associated T. marneffei infections remains limited.

Since 1988, T. marneffei infections have been reported among patients with advanced HIV, and the disease incidence has gradually increased in areas where the incidence of HIV has also continued to increase [12]. The distribution of T. marneffei infections clearly shows regional characteristics, and it is historically concentrated in Asian countries such as Thailand, Vietnam, Malaysia, India, and southern regions of China [13–16]. However, its geographical distribution now extends beyond traditional epidemic areas, with sporadic cases recently reported in nonepidemic areas [17, 18]. This expansion may be driven by factors such as climate change, improved diagnostic capabilities, and population mobility [19]. Despite this evolving epidemiology, much of the existing research, surveillance data, and public health focus remains disproportionately concentrated on long-established high-incidence hotspots, potentially overlooking emerging trends and vulnerable populations in new areas.

Despite overall progress in epidemiological studies of T. marneffei, several important issues remain overlooked, such as regionally imbalanced data, unclear population characteristics, and a lack of early diagnostic techniques. In particular, the high-risk factors for this disease have not been clearly identified, and quantitative predictive models for climate‒host‒pathogen interactions are still lacking. This cross-sectional study, which is based on detection data for T. marneffei in hospitalized patients with acute respiratory tract infections (ARTIs) collected from 2022 to 2024 across the Chinese mainland, systematically analyzes its spatiotemporal distribution patterns and epidemiological characteristics. By constructing a model that incorporates both environmental factors and temporal trends in terms of the detection rate, the present study aims to uncover key factors associated with T. marneffei detection and to provide a scientific basis for the development of region-specific control measures and rational allocation of medical resources. Although they are derived from a defined regional context, our findings on the environmental correlates of transmission provide a model for understanding and mitigating the threat of talaromycosis and similar climate-sensitive fungal pathogens under global change.

Methods

Data source

We conducted a cross-sectional analysis of targeted next-generation sequencing (tNGS) data from hospitalized patients with ARTIs across the Chinese mainland. Data were obtained from a standardized, nationwide repository established and maintained by KingMed (Guangzhou, Guangdong, China). Respiratory specimens were collected from a network of collaborating hospitals and processed centrally in KingMed’s laboratories using uniform protocols.

The study cohort comprised inpatients clinically diagnosed with ARTIs whose respiratory samples were submitted for tNGS testing—typically in cases of severe, atypical, or unresolved infection. This reflects a real-world, clinically selected population rather than a systematic sample of all ARTI admissions. tNGS has been clinically validated as a reliable diagnostic method for respiratory pathogen detection [20]. A T. marneffei detection was defined as a tNGS result meeting a validated bioinformatics threshold of ≥ 10 reads per 100,000 sequencing reads (RPhK). This threshold effectively distinguishes true positives from background noise, and cases meeting this criterion were classified as positive for T. marneffei identification. All such positive results are hereafter referred to as “detection” in this epidemiological analysis.

For spatial analyses, each case was assigned to the city of the submitting hospital. The data spanned from January 2022 to December 2024 and included 4,758 collaborating hospitals distributed across 314 cities in all 31 provincial-level regions of the Chinese mainland. The hospital network encompasses tertiary, secondary, and primary care institutions in both urban and rural settings, ensuring broad geographic representation.

Meteorological data were sourced from the National Oceanic and Atmospheric Administration (NOAA, https://www.ngdc.noaa.gov/) and included average temperature (TEMP), minimum temperature (MIN_TEMP), maximum temperature (MAX_TEMP), dew point temperature (DEWP), precipitation (PRCP), average wind speed (WDSP), maximum sustained wind speed (MXSPD), and average air pressure (STP). Additional data, including the enhanced vegetation index (EVI) for 2023, normalized difference vegetation index (NDVI) for 2023, gross domestic product (GDP) for 2020, and population density (POP) for 2020, were obtained from the Resource and Environmental Science Data Platform (RESDP, https://www.resdc.cn/). The distribution of bamboo rats was obtained from the Global Biodiversity Information Facility (GBIF, https://www.gbif.org/) and included three species: Rhizomys pruinosus, Rhizomys sinensis, and Rhizomys sumatrensis. The three bamboo rat species were selected based on their documented roles as natural reservoirs of T. marneffei. Rhizomys sinensis was the species from which T. marneffei was first isolated in 1956 [21]. Subsequent studies have confirmed the presence of T. marneffei in all three species, with high prevalence rates reported across their respective habitats [22–24]. The selection of these three species ensures comprehensive coverage of the primary reservoir hosts of T. marneffei in the Chinese mainland. For geographical detector analysis, all environmental variables were converted to raster format with a unified spatial resolution of 20 km × 20 km. Bamboo rat occurrence records were processed into a density surface, while meteorological and socioeconomic variables were spatially interpolated to match the same grid resolution and coordinate system before being sampled for analysis.

Sample inclusion/exclusion criteria

Hospitalized patients diagnosed with ARTIs were eligible for inclusion if they met the following criteria: (1) availability of a respiratory specimen subjected to tNGS; and (2) complete clinical and demographic metadata, including sex, name, date of birth, sampling date, geographic location (e.g., Guangdong, Guangzhou), sampling site (e.g., nasopharyngeal), and type of detection. Samples with missing or incomplete metadata were excluded from subsequent analysis.

To avoid overcounting due to repeated tNGS testing of the same patient, a structured deduplication algorithm was applied. Specifically, records with identical sex, name, date of birth, and hospital location were considered operational duplicates—i.e., multiple test records originating from the same patient. For each set of duplicates, only the earliest positive sample was retained for further analysis. This strategy ensured that each patient contributed only once to the epidemiological analysis, thereby preventing bias in case counts and co‑detection estimates that could arise from repeated testing.

Statistical analysis

The frequency of T. marneffei detection was analyzed. Detection rates were calculated as the number of patients with a positive tNGS result divided by the total number of patients in the tNGS-tested ARTI cohort. A descriptive epidemiology approach was used to characterize T. marneffei detection. Detection rates and counts were calculated across provinces, cities, months, weeks, genders, and age groups. Gender differences were assessed using the chi-square test and variations among age groups were examined using the chi-square test of independence. The null hypothesis assumed equal detection rates across age groups, with expected frequencies proportional to the number of tested patients in each age group.

Spatial autocorrelation analysis, including global and local autocorrelations, was employed to analyze spatial patterns of T. marneffei detection on the Chinese mainland from 2022 to 2024. Moran’s I index and the Getis-Ord General G index were used to estimate global autocorrelations. Anselin’s local Moran’s I index and the Getis-Ord Gi* statistic were applied in local autocorrelation analyses.

Co-detection patterns were described by calculating the frequency and proportion of other pathogens identified in samples positive for T. marneffei. To characterize the co-detection profile, we performed an exploratory pairwise association analysis by calculating odds ratios (ORs) and 95% confidence intervals for 25 common ARTI-related pathogens. An OR > 1 with P < 0.05 was interpreted as a positive association, whereas an OR < 1 with P < 0.05 indicated a negative association. Given the exploratory nature of this analysis and the non-independence among the 25 comparisons (as pathogen detections may share common host factors, seasonality, and testing practices), no correction for multiple comparisons was applied, as such correction would be overly conservative and could increase the risk of false-negative findings in this context [25]. Therefore, the results should be interpreted cautiously based on ORs, confidence intervals, and nominal p-values, and are not intended to infer direct causality or definitive biological interactions.

A geographical detector model was used to quantify the associations of environmental factors with the spatial heterogeneity of detection rates. Prior to analysis, all the environmental variables were processed to the same spatial resolution and temporally matched to the case data period where applicable. All geographical detector analyses were implemented through Python scripts (Python 3.9), following the computational framework of the q-statistic [26]. The T. marneffei detection rate for each spatial unit was calculated at the city level. For continuous environmental variables, discretization was optimized using the Optimal Multi-scale Geographical Detector (OMGD) approach [27], which automatically searches over a predefined number of strata (5–10 layers) and selects the discretization scheme that maximizes the q-statistic. This data-driven procedure ensures that the stratification reflects the intrinsic spatial variance of each factor. The factor detector in this model calculates the q-statistic to measure the explanatory power of each factor on the spatial distribution of the T. marneffei detection rate. The value of q ranges from 0 to 1, where a larger q indicates that the factor explains a greater proportion of the spatial heterogeneity. Bivariate interaction analysis was performed following the standard framework of the geographical detector. This analysis compares the individual q-values of two factors [q(X₁) and q(X₂)] with the q-value of their overlay [q(X₁∩X₂)]. The interaction is classified into one of five types (nonlinear-weaken, univariate‑weaken, bivariate‑enhance, independent, or nonlinear‑enhance) according to the predefined decision rules. No formal significance tests were performed for the interaction q-values, because the classification of interaction types is determined by relative comparisons among the q-values of the individual and combined factors, rather than by hypothesis testing against a null distribution.

Results

Regional distribution characteristics

From 2022 to 2024, a total of 2,316 hospitalized patients with T. marneffei detected by tNGS were identified in the nationwide cohort of hospitalized patients with ARTIs, representing a detection proportion of 0.27% among all tNGS-tested ARTI patients in this study network. T. marneffei detection was distributed in 157 cities across 25 provinces (Fig. 1A-B). Significant spatial autocorrelation was observed each year (Moran’s I: 0.055384 in 2022, 0.168443 in 2023, 0.143621 in 2024, overall 0.198556, all P < 0.05), indicating persistent clustering of detections (Table S1). Global clustering was further confirmed by Getis-Ord General G analysis (P < 0.05, Table S2). Local spatial autocorrelation analysis revealed significant regional differences and clustering patterns for T. marneffei detection. A persistent high-high cluster was observed in southern China, whereas a low-low cluster predominated in northern China. The outliers included high-low types, mainly in the north, and low-high types, in the south. Minor interannual variations in cluster intensity and extent were noted (Fig. 1C). Hotspots were consistently identified in southern China, particularly in the eastern coastal and densely populated regions. Compared with 2022, cold spots in the north were predominant and expanded in 2023 and 2024 (Fig. 1D).

Fig. 1.

Regional distribution of T. marneffei detection from 2022 to 2024. A Distribution of T. marneffei detection across the Chinese mainland in various provinces from 2022 to 2024. The color gradient represents the rate of T. marneffei detection among all the ARTIs. Bubble size in each region indicates the number of T. marneffei detection cases. B Distribution of T. marneffei detection across the Chinese mainland in various cities from 2022 to 2024. The color gradient illustrates the T. marneffei detection rate among the total reported ARTIs. Bubble size reflects the number of T. marneffei detection cases. C Local spatial autocorrelation analysis of T. marneffei detection from 2022 to 2024. D Distributions of cold spot and hotspot regions of T. marneffei detected from 2022 to 2024

Changes in T. marneffei distribution over time

From 2022 to 2024, the detection rate of T. marneffei tended to increase, with epidemic characteristics exhibiting clear seasonal fluctuations in distribution (Fig. 2). The annual detection rates also consistently exhibited bimodal distributions, peaking around April and August, with these two months accounting for 9.15% and 10.10% of the total annual detections, respectively. The actual monthly detection rates were shown in Table S3 and ranged from 0.15% to 0.60%.

Fig. 2.

Temporal distribution of T. marneffei detection from 2022 to 2024. A Weekly temporal distribution of T. marneffei detection cases and corresponding rates during the study period. The primary y-axis represents the cumulative case counts per observation week, while the secondary y-axis depicts the weekly rate. B Monthly temporal distribution of T. marneffei detection cases and corresponding rates during the study period. The primary y-axis represents the cumulative case counts per observation month, while the secondary y-axis depicts the monthly rate. C The figure presents the seasonal decomposition results for monthly T. marneffei detection cases. The figure displays the original observed series alongside its decomposed seasonal, trend, and residual components, revealing underlying patterns

Demographic characteristics of the T. marneffei-detected population

The prevalence of T. marneffei was significantly greater in males than in females across the Chinese mainland from 2022 to 2024, with an overall male-to-female ratio of 3.5:1 (P < 0.05), and the monthly ratios ranged from 2.66 to 4.43 throughout the study period (Fig. 3A). Examination of the age distribution of T. marneffei detection revealed that the rate was lower in younger age groups and progressively increased with age, peaking in individuals aged > 40 and ≤ 50 years and declining in older age groups (Table S4). Notably, the rate of T. marneffei detection among all the ARTI cases was consistent with the distribution of T. marneffei case counts (Fig. 3B). Chi-square test of independence revealed significant differences in the detection rate of T. marneffei among the different age groups (P < 0.05).

Fig. 3.

Population distribution of T. marneffei detection from 2022 to 2024. A Gender distribution of T. marneffei detected from 2022 to 2024. Stacked bars represent T. marneffei detection cases. Lines represent rates of T. marneffei detection. B Age distribution of T. marneffei detected from 2022 to 2024. Columns represent T. marneffei detection cases. Lines show T. marneffei detection rate

Statistical analysis of co-detection among T. marneffei

Among individuals in whom T. marneffei was detected, a total of 25 ARTI pathogens were reported as co-detections, 10 of which each affected more than 5% of the T. marneffei‑positive individuals. Pneumocystis jirovecii was the predominant pathogen, accounting for 27.68% of the T. marneffei co-detections (Fig. 4A). Risk analysis of these 25 ARTI pathogens revealed that five were significantly positively correlated with T. marneffei, namely, Pneumocystis jirovecii, Aspergillus, human coronavirus, Cryptococcus, and human parvovirus B19. Conversely, 16 pathogens, namely, Haemophilus influenzae, Klebsiella pneumoniae, Staphylococcus aureus, Streptococcus pneumoniae, Acinetobacter baumannii, Pseudomonas aeruginosa, Influenza virus, Mycoplasma pneumoniae, human parainfluenza virus, human adenovirus, Moraxella catarrhalis, human respiratory syncytial virus, human metapneumovirus, Bordetella pertussis, human bocavirus and Chlamydia psittaci were significantly negatively correlated (Fig. 4B) (Table S5).

Fig. 4.

Statistical analysis of pathogens co-detected with T. marneffei. A Statistical analysis of cases in which ARTI and T. marneffei were co-detected from 2022 to 2024. Columns show T. marneffei co-detection case numbers. Scatter diagram representing the percentage of co-detection cases involving each T. marneffei. B Risk analysis of co-detection in T. marneffei detection cases from 2022 to 2024. Pathogens with a negative correlation are shown in blue, those with a positive correlation are shown in red, and those with no significant correlation are shown in gray

Analysis of environmental factors associated with T. marneffei detection

Univariate analysis revealed that dew point temperature had the strongest explanatory power for T. marneffei detection (48.55%), followed by average temperature (46.24%), annual minimum temperature (37.51%) and bamboo rat distribution at the available data resolution (36.06%) (Table S6). Bivariate analysis further indicated that pairwise factor interactions substantially increased the overall explanatory power of T. marneffei detection. Notably, factors combined with dew point temperature generally exhibited high explanatory power, while average air pressure had lower explanatory power (1.4%) in univariate analysis, but its explanatory power was significantly enhanced when combined with other variables (Fig. 5).

Fig. 5.

Environmental factors associated with T. marneffei detection from 2022 to 2024. A Percentage of independent variance in T. marneffei detection explained by individual environmental factors. The vertical axis represents environmental impact factors. The horizontal axis represents explanatory power. B Percentage of joint variance in T. marneffei detection explained by the combined effect of two factors. The numbers in the matrix represent the percentage of joint variance explained by the corresponding paired factors

Discussion

Unlike previous studies predominantly focused on HIV-positive cohorts or localized geographical areas, this study provides a novel, nationwide epidemiological profile of T. marneffei detection by analyzing tNGS data from hospitalized patients with ARTIs across the Chinese mainland (2022–2024). By integrating multidimensional environmental datasets and employing spatial statistics and geographical detector models, we characterize the spatiotemporal distribution, identify high-risk demographics, and quantify the key interactive environmental factors associated with T. marneffei detection, particularly dew point temperature. These findings, derived from a real-world clinical cohort, contribute to an evidence base for understanding its epidemiology and informing targeted control strategies.

The pronounced differences in detection rates observed between the northern and southern regions are consistent with the established understanding of T. marneffei biology and underscore the critical role of climatic factors [28]. Specifically, the warmer annual temperatures, higher precipitation, and greater humidity in southern regions favor the environmental survival and transmission of this dimorphic fungus [29]. Spatial analyses revealed sporadic high-low outlier clusters within the generally low-risk northern region, indicating the presence of localized high-rate areas that warrant closer attention in basic research and public health policies aimed at prevention and control measures. In contrast, we identified low-high outlier clusters in the southern endemic zone, highlighting the considerable intraregional heterogeneity of T. marneffei detection. These localized low-risk areas surrounded by high-detection regions face an elevated risk of pathogen introduction from adjacent high-risk zones, suggesting that enhanced monitoring may be needed to prevent an increase in detection rates in these local populations. Furthermore, high-rate areas persisting in eastern coastal and densely populated zones share geographical continuity with previously reported high-rate regions in Southeast Asia [30]. This spatial relationship among established hotspots in the northern and southern regions suggests that current interventions show limited efficacy in reducing risk or preventing local transmission, enabling northward expansion of the pathogen’s endemic range.

Temporal analysis revealed a bimodal seasonal distribution of T. marneffei detection rates, a pattern that is closely related to environmental changes. This trend was similar to that observed in Ho Chi Minh City, where AIDS-associated T. marneffei cases increased by 30% during the May–November rainy season [31]. Seasonality is driven mainly by elevated temperature and humidity levels, which likely facilitate infection by promoting fungal sporulation and airborne survival and providing favorable conditions for T. marneffei survival, reproduction, and spore dissemination [32–34]. Given the critical influence of climatic factors, especially humidity, on transmission dynamics, medical institutions should consider enhanced screening for respiratory infections during high-humidity periods and heightened vigilance for T. marneffei detection in cases of unexplained pneumonia, particularly in immunocompromised individuals, to mitigate these seasonal risks. The study period spanned the transition of China’s COVID‑19 control policies, during which healthcare utilization, hospitalization patterns, and tNGS submission practices may have changed, particularly before and after the substantial lifting of zero‑COVID measures in late 2022. These changes could influence the observed temporal trends. To mitigate this concern, we conducted annual stratified analyses for all main epidemiological outcomes. The results demonstrated that the overall spatial pattern—high‑risk clustering in southern provinces and low‑risk clustering in northern provinces—remained consistent across 2022, 2023, and 2024. However, changes in case counts over time should be interpreted cautiously and cannot be directly equated with year‑to‑year changes in the true population incidence, as they may also reflect variations in healthcare‑seeking behavior and testing practices.

Demographic analysis revealed that the rate of T. marneffei detection is consistently higher among males, with an overall male-to-female ratio of 3.5:1. This stable sex disparity is likely attributable to differences in occupational exposure, as males dominate high-risk outdoor occupations such as construction, agriculture, and other fields involving direct soil contact [35]. Previous studies have shown that patients with recent occupational or other soil contact history, especially during the rainy season, have a higher risk of T. marneffei infection [36]. Although specific anti-cytokine autoantibody (ACAAs) syndromes, notably adult-onset immunodeficiency due to anti-interferon-gamma autoantibodies, are well documented in females [37], their population-level signal appears to be outweighed by occupational exposure in this cohort. Our study did not collect individual‑level occupational data; nevertheless, the observed male predominance and the peak detection in the 41–50 year age group, an age range when males in China are actively engaged in outdoor and agricultural labor, indirectly support the hypothesis that occupational exposure contributes to detection risk. In addition to this occupational pattern, the unimodal peak among individuals aged 41–50 years may also reflect age-related changes in immune function that reduce host defense against fungal infection [38]. Future prospective studies incorporating detailed occupational histories are warranted to directly quantify this association. Consequently, enhanced health education, proactive screening initiatives, and heightened clinical vigilance are imperative for the early detection and management of talaromycosis in this high-risk age group.

Systematic analysis of ARTI-related co-detection agents of T. marneffei revealed numerous significant interactions among T. marneffei and other infectious agents. Pairwise correlation analysis identified five pathogens significantly positively correlated with T. marneffei detection, suggesting that these agents may share similar infection conditions or immune escape mechanisms. For Aspergillus species, both T. marneffei and Aspergillus are opportunistic fungi that primarily infect immunocompromised hosts [39], and co‑infections with T. marneffei and Aspergillus have been documented in HIV‑positive patients [40]. Cryptococcus is another common opportunistic pathogen in AIDS patients, and concurrent T. marneffei and Cryptococcus infections have been reported in both HIV‑positive and HIV‑negative individuals [41, 42]. Human parvovirus B19 infection in immunocompromised hosts often leads to chronic viremia and anemia due to impaired viral clearance [43, 44], and such sustained immune dysregulation may create a permissive environment for opportunistic fungal invasion. Human coronavirus may directly damage the respiratory epithelium or suppress immune cell function [45], thereby attenuating host resistance to fungal infection and facilitating T. marneffei colonization and invasion. The most frequent co-detection was Pneumocystis jirovecii, which has been confirmed in immunodeficient patients by multiple diagnostic methods [46]. The critical role of cell-mediated immunity in preventing Pneumocystis jirovecii infection is well established. Individuals with impaired cell-mediated immunity, particularly those with reduced CD4⁺ T‑cell counts or function, are at markedly increased risk for Pneumocystis jirovecii pneumonia [47]. Beyond HIV‑positive patients, Pneumocystis jirovecii pneumonia is also a significant opportunistic infection in non‑HIV immunocompromised populations, including solid organ transplant recipients, autoimmune conditions and those with primary immunodeficiency disorders [48]. Supporting this shared vulnerability, concurrent T. marneffei and Pneumocystis jirovecii infection has been documented in a child with a known STAT1 mutation—a genetic disorder that disrupts immune signaling pathways—further substantiating that both opportunistic pathogens can co‑occur in hosts with underlying cell‑mediated immune defects [49]. This comorbidity likely reflects a shared susceptibility in hosts with impaired cell-mediated immunity. Alternatively, 16 pathogens were negatively correlated, which may be related to differences in host immune status or specific characteristics of pathogen biology. For example, Streptococcus pneumoniae and influenza virus are well-established causes of community-acquired infections [50–52] that can trigger host inflammatory and immune responses [53, 54], potentially indirectly suppressing the risk of T. marneffei infection by activating innate host immunity.

Univariate and bivariate analyses identified key climatic and host-related factors associated with T. marneffei detection. Dew point temperature was the primary factor, which is consistent with its previously established role in enhancing pathogen survival and transmission [55]. A high dew point environment can significantly slow the evaporation rate of respiratory droplets, allowing droplets containing pathogens to maintain their hydrated liquid form for a longer period of time [56]. While rainfall is a recognized seasonal trigger for various infections [57], our model highlights the superior explanatory power of the dew point. Unlike relative humidity, the dew point directly reflects the absolute moisture content in the air, which is critical for maintaining fungal hydration and viability. This distinction suggests that dew point temperature may serve as a more consistent and generalizable predictor of environmental risk across different regions, offering a precise metric for climate-based surveillance. Although our findings are derived from a specific regional context, the identified dominance of dew point temperature elucidates a climate-sensitive association that provides a transferable model for understanding the dynamics of other environmentally persistent, climate-sensitive fungal pathogens under global change.

The results of bivariate analysis underscore the central role of the dew point temperature, showing that its interaction with other factors explained a substantial proportion of the detection risk. The average and annual minimum temperatures also exhibited high explanatory power, likely because of the temperature dependence of fungal growth and virulence [58]. Notably, while the documented distribution of bamboo rats as a biological factor exhibited lower individual explanatory power than meteorological factors, its inclusion in bivariate combinations significantly enhanced the joint explanatory power. The three bamboo rat species exhibit distinct ecological characteristics and habitat distributions in Chinese mainland. Rhizomys pruinosus has the broadest potential distribution, with a high-suitability area covering Yunnan, Guizhou, Guangxi, Guangdong, Hunan, Jiangxi, Fujian, and Hainan. Rhizomys sinensis is more geographically concentrated, with its core high-suitability region in Sichuan and Guizhou. Rhizomys sumatrensis has the most limited distribution, primarily in southwestern Yunnan. The overlapping and distinct distributions of these reservoir species may contribute to the spatial heterogeneity of T. marneffei detection risk across different regions of China [59]. This synergy suggests that areas environmentally suitable for T. marneffei transmission often overlap with areas where bamboo rats are known to exist. This finding is consistent with the established role of bamboo rats as a key reservoir host, with studies frequently detecting T. marneffei in these animals [60]. Critically, epidemiological links have been reported, with clusters of T. marneffei infection observed in immunocompetent individuals with a history of hunting wild bamboo rats [23]. However, the bamboo rat distribution data used in this study were derived from GBIF presence-only occurrence records, which are subject to sampling bias and do not reflect true population density of bamboo rats or the intensity of human-animal contact. Therefore, the observed spatial overlap should be interpreted as an ecological association at the species distribution level rather than a quantitative measure of reservoir exposure risk. Nevertheless, this observed spatial overlap reinforces the established ecological link between the pathogen and its reservoir host at the population level. Beyond these climatic and host factors, the environmental microbiome may serve as a crucial selective pressure driving fungal adaptation and virulence evolution. Recent evidence has demonstrated that amoebal predation can induce significant phenotypic changes in environmental fungi, ultimately increasing their pathogenic potential in infection models [61]. Although our study did not directly examine interactions between T. marneffei and amoebae, their coexistence in humid soils and similar microenvironments is plausible. This perspective integrates microbial predator‒prey dynamics into the environmental factors of fungal epidemiology and merits future investigation.

The number of T. marneffei detections documented in our nationwide analysis substantially exceeds historical reports of non-HIV-associated cases in China. This divergence likely reflects major advancements in diagnostic methodology rather than a sudden epidemic shift. The increased ascertainment can be attributed to the superior sensitivity of tNGS compared with traditional culture-based methods [62], our systematic active surveillance framework within a large hospital network, and the focus on all hospitalized ARTI patients—a population that captures the growing non-HIV demographic often underrepresented in earlier studies. Consequently, these data establish a recalibrated baseline for the detection of T. marneffei in the modern diagnostic era.

Beyond refining the burden estimate, our study provides new epidemiological insights. We delineate previously unreported fine-scale spatial heterogeneity, identifying specific high-risk outlier clusters in northern regions and localized low-risk areas within southern endemic zones. Furthermore, we define the demographic profile of the dominant hospitalized patients at a national scale, characterized by a detection peak among middle-aged adults, which contrasts with the younger age distribution typical of HIV-associated cohorts. The systematic documentation of novel co-detection patterns, including both positive and negative associations with specific respiratory pathogens, suggests complex within-host ecological interactions that warrant further investigation.

These findings offer a clear foundation for targeted public health action. The quantification of dew point temperature as the principal environmental factor supports its integration into regional climate-based early warning systems, particularly in southern endemic provinces, to prompt heightened clinical suspicion during high-risk periods. Given the pronounced bimodal seasonal peak, health care facilities in hotspots could implement seasonal screening protocols or preemptive diagnostic guidelines for at-risk patients presenting with respiratory symptoms. The marked spatial heterogeneity argues for implementing a tiered surveillance framework that prioritizes intensive laboratory screening and public health education in persistent southern hotspots, while establishing sentinel monitoring in northern regions with outlier clusters to track potential range expansion. Moreover, the distinct demographic profile—particularly the elevated detection rate among males aged 41–50—can inform tailored risk communication and occupational health measures for middle-aged male populations in endemic areas, who may have higher environmental exposure risks. Collectively, these data-driven insights can optimize resource allocation, refine clinical guidelines, and support evidence-based policy development for talaromycosis containment. Notably, these implications are derived from a hospitalized patient cohort, and extrapolation to the general community requires appropriate caution.

Limitations of the study

While this study provides systematic information on the epidemiological characteristics and influencing factors of T. marneffei detection across the Chinese mainland, it has several limitations. First, our case definition relied on tNGS detection in a clinically defined ARTI cohort. tNGS cannot definitively distinguish invasive disease from colonization, and the cohort was selected based on clinical needs, potentially introducing selection bias toward more severe or complex cases. Second, spatial analysis assigned cases by hospital location rather than actual exposure site, meaning the identified hotspots may be influenced by healthcare access factors such as hospital distribution, referral patterns, and testing practices. Relatedly, our deduplication process could not reliably identify or exclude imported cases. Nevertheless, the strong spatial correlation with environmental data supports the ecological validity of the identified hotspots. Third, the absence of occupational data precluded direct assessment of occupation-related risks, and our study only included hospitalized ARTI patients, thus not reflecting the geographic distribution of T. marneffei in the general population (including asymptomatic carriers or mild outpatient cases). While travel-related detections are likely negligible, we acknowledge our inability to definitively exclude such cases. Additionally, the bamboo rat distribution data from GBIF are presence-only records subject to sampling bias and do not represent true population density or human-animal contact intensity; therefore, the observed spatial overlap should be interpreted as an ecological association at the species distribution level rather than a quantitative measure of reservoir exposure risk. Finally, and most importantly, the lack of systematic data on HIV status and other immunocompromising conditions precluded stratified analyses to determine how risk factors differ between immunocompetent and immunocompromised subgroups. Given that talaromycosis is a well-recognized AIDS-defining illness in East and Southeast Asia, this absence represents a major limitation that may have biased the observed demographic, seasonal, and spatial patterns. Future studies should prioritize collecting HIV-related data to enable subgroup analyses. Although we performed annual stratified analyses to assess the stability of spatial patterns, we cannot completely exclude residual confounding arising from the COVID-19 policy shift and associated changes in testing and healthcare-seeking behaviors.

Conclusions

This nationwide study explores the spatiotemporal epidemiology and multifactorial environmental and host factors associated with T. marneffei detections across the Chinese mainland. We identified significant geographical clustering with pronounced seasonal fluctuations (bimodal peaks in April and August), reflecting synergistic environmental and host correlates. Males and individuals aged 41–50 years exhibited heightened susceptibility. Critically, dew point temperature emerged as the predominant environmental predictor. Further, frequent polymicrobial co-detections implicated diverse respiratory pathogens, underscoring clinical complexities in disease management. Nevertheless, because HIV status and other immunocompromising conditions were not available in this dataset, the observed epidemiological patterns may differ between HIV-positive and HIV-negative subgroups. Therefore, our findings primarily reflect the characteristics of T. marneffei detection in the overall hospitalized ARTI population without stratification by immune status. The findings underscore the necessity for enhanced pathogen surveillance and respiratory infection control measures in endemic regions during peak transmission seasons, and highlight the urgent need for future studies incorporating HIV and other immune status data to enable stratified analyses.

Authors’ contributions

JGX, RS and YMS conceptualized the study. XCL, PL and JLY contributed the original data for analysis. YMS and RS verified the underlying data. XCL and MW performed the formal analysis, while YMS, JGX and RHJ validated the data. Data analysis was co-led by YMS and JGX. JLY and MW were responsible for data visualization. Project administration was managed by YMS and RS. The study was supervised by JGX and RHJ. YMS and XCL drafted the original manuscript, and XW and RHJ reviewed and edited it. All the authors reviewed and approved the final manuscript. Each author had final responsibility for the decision to submit for publication.

Funding

This work was supported by the Prevention and Control of Emerging and Major Infectious Diseases-National Science and Technology Major Project (grant number 2025ZD01900100, 2025ZD01900114 and 2025ZD01900116, Tengfei initiative (grant number 2025NITFID502 and 2025NITFID513) by the National Key Laboratory of Intelligent Tracking and Forecasting for infectious Diseases, and 2025 High Innovation Plan· Young Top Talent Projects (grant number G202523156).

Data availability

The datasets used and analyzed during the current study are available from the corresponding author on reasonable request.

Declarations

Ethics approval and consent to participate

The retrospective study was approved by the Ethics Committee of KingMed (Approval No. 2025085). This study utilized fully anonymized clinical data from hospitalized patients who were tested at KingMed between 2022 and 2024. These data were collected as part of routine clinical care prior to the conception of this research. Due to the retrospective nature of the study and the use of anonymized data, the requirement for informed consent was formally waived by the Ethics Committee of KingMed. All procedures performed in this study were in accordance with the ethical standards of the institutional research committee and with the Helsinki declaration and its later amendments.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.