Diagnostic Performance and Cost-Effectiveness of BALF mNGS in Older Adults with Pulmonary Infections Authors: Kai-jie Liu, Yan Gao, Xirui Yang, Yong Xia, Chang Lu, Zhao-rong Li, Xiyu Chu, Hui Huang, Ping Xu, Mang Shi, Ke Yuan, Huilin Yang Published: 2026 ABSTRACT AbstractIntroductionPulmonary infections in elderly patients cause high morbidity and mortality. Conventional culture has low sensitivity and slow turnaround, delaying targeted therapy. Metagenomic next-generation sequencing (mNGS) is an emerging technology, but its diagnostic performance and cost-effectiveness are unclear. This study therefore aims to evaluate its diagnostic performance compared to conventional culture in older adults with pulmonary infections and to assess its cost-effectiveness.MethodsFrom March 2020 to March 2023, 522 patients (aged 55–69 years) diagnosed with pulmonary infections were enrolled at Peking University Shenzhen Hospital. Of these, 168 patients underwent simultaneous mNGS and conventional culture testing using bronchoalveolar lavage fluid (BALF) samples, while the remaining 354 patients received culture testing alone. Pathogen detection results were compared to assess the diagnostic performance of mNGS versus traditional culture methods. Additionally, cost-effectiveness analyses of the two diagnostic strategies—as well as the impact of mNGS testing timing post-admission—were conducted in the overall cohort and across stratified subgroups.ResultsAmong the 168 patients who underwent both tests, mNGS identified a greater diversity and abundance of microorganisms than culture (overall detection: 89.88% vs. 26.79%; pathogen detection: 67.86% vs. 18.45%, p < 0.001). mNGS testing yielded a net economic benefit of 1202.70 CNY per patient overall and 3831.15 CNY among pathogen-positive cases. Delaying mNGS testing tended to be associated with increased hospitalization length of stay (LOS) and costs, with the most pronounced difference observed around 6 days after admission (p < 0.001). Early mNGS testing (within 6 days of admission) provided a net benefit of 6346.00 CNY.ConclusionsBALF-based mNGS showed higher positivity rates and a broader pathogen detection spectrum compared to conventional culture methods in this study. Early implementation of mNGS shows strong potential to guide the treatment of pulmonary infections and reduce healthcare costs for elderly and aging patients.Supplementary InformationThe online version contains supplementary material available at 10.1007/s40121-026-01372-x.Keywords: Pulmonary infection, Older adults, mNGS, Cost-effectiveness analysis, Health economics Key Summary Points Why carry out this study? Pulmonary infections are a leading cause of mortality in the elderly, yet conventional culture methods often fail to identify pathogens due to low sensitivity and long turnaround times.This study aimed to evaluate the diagnostic performance of BALF mNGS compared to conventional culture and to assess the cost-effectiveness of its implementation and timing in middle-aged and older adults. What was learned from the study? mNGS demonstrated a significantly higher positivity rate (89.88% vs. 26.79%) than culture and provided a net economic benefit of 1202.70 CNY per patient, which increased to 6346.00 CNY when testing was performed within 6 days of admission.Early implementation of mNGS is a clinically superior and cost-effective strategy that reduces the economic burden for these aging patients by facilitating timely, targeted treatment and shortening hospital stays. Introduction Clinical Challenges of Pulmonary Infection in the Elderly Pulmonary infection is a common respiratory disease associated with high morbidity and mortality, particularly in middle-aged and older adults. This population is especially vulnerable due to age-related decline in lung function, immunosenescence, frailty, and the presence of multiple comorbidities [1, 2]. Geriatric pulmonary infection is often atypical, with 40–60% of patients present without fever or typical respiratory symptoms—a diagnostic challenge famously referred to by William Osler as “the natural enemy of the old man” [3]. In addition, older adults are more likely to experience recurrent infections and poorer outcomes due to diminished organ function and complex underlying conditions [4]. As aging populations grow in both middle- and high-income countries, there is an urgent need to reduce disease burden in this demographic. Advantages of Metagenomic Next-Generation Sequencing Despite advances in clinical care, diagnosing pulmonary infections remains limited because of the paucity of rapid and accurate pathogen detection methods [5]. Bronchoalveolar lavage fluid (BALF), obtained directly from infected pulmonary sites, is considered a reliable specimen due to its high diagnostic sensitivity and low risk of contamination [6]. However, conventional culture methods using BALF samples yield relatively low positive rates—approximately 30% in lower respiratory tract infection—which often fail to meet clinical demands [7]. In addition, their prolonged turnaround time can contribute to inappropriate antibiotic use and extended hospital stays [8]. Metagenomic next-generation sequencing (mNGS) provides rapid, accurate, comprehensive, and unbiased pathogen detection, making it particularly valuable for diagnosing rare, mixed, and opportunistic infections, especially in patients with weakened immunity [6]. Compared to traditional culture-based techniques, mNGS demonstrates superior diagnostic performance, higher sensitivity, and species-level resolution, especially for fastidious or uncommon pathogens [9]. For instance, one study reported that mNGS had a higher positivity rate than culture when detecting pathogens from BALF samples in patients with lower respiratory tract infections (69.60% vs. 25.00%, p < 0.05) [10]. Similarly, Gao and colleagues [11] showed that mNGS had significantly higher sensitivity (82.86% vs. 44.42%, p < 0.001), specificity (52.00% vs. 21.05%, p < 0.001), higher positive predictive value (96.71% vs. 79.55%, p < 0.01), and negative predictive value (15.12% vs. 5.19%, p < 0.01). Furthermore, mNGS has proven to be effective in identifying rare pathogens, such as Chlamydia psittaci, in cases of pneumonia of unknown etiology, thereby reducing delays in targeted treatment [12]. However, the ultra-high sensitivity of mNGS also increases the risk of false-positive results. Distinguishing true infectious pathogens from respiratory tract colonization or environmental contaminants remains a significant diagnostic challenge [13]. Such false positives can complicate clinical interpretation and potentially drive unnecessary antimicrobial use, underscoring the need to carefully evaluate its real-world clinical and economic utility. Knowledge Gap in Health Economic Evaluation Despite the notable diagnostic advantages of mNGS, health economic evaluations of this technology remain scarce. Studies from across the world have demonstrated that whole genome sequencing (WGS) can lead to substantial cost savings—ranging from hundreds to thousands of dollars—during infectious disease outbreaks [14]. For example, Kumar et al. [15] reported that WGS-based infection surveillance programs could reduce costs by approximately $200 per case. However, to date, there have been no comprehensive health economic analyses specifically focused on mNGS. To address this gap, the present study analyzed data from a cohort of 522 patients with pulmonary infections. We compared the diagnostic performance of mNGS versus conventional culture, and assessed the impact of mNGS on clinical decision-making and overall healthcare costs, with the goal of providing evidence to support its broader implementation in routine clinical practice. Methods Study Design This retrospective study was approved by the Ethics Committee of the Peking University Shenzhen Hospital (No. 2025052), and conducted in accordance with the Helsinki Declaration, including 522 patients (aged 55–69 years) with pulmonary infections treated at the hospital between March 2020 and March 2023. This target population was specifically selected because it captures a critical window where age-related thymic involution [16] and the progressive accumulation of clinical comorbidities [17] synergistically establish profound vulnerability to severe pneumonia. Additionally, we considered that patients aged over 70 years presented with a greater burden of underlying comorbidities, which could substantially influence the outcome analysis. The cohort encompassed all hospitalized cases, including both community-acquired (present upon admission) and hospital-acquired (developed during hospitalization) infections. Pulmonary infection was defined based on the presence of clinical manifestations of lower respiratory tract infection (such as fever accompanied by cough, sputum production, or chest pain), physical examination findings (including dry or moist rales, or signs of pulmonary consolidation), and radiographic evidence indicative of infection (such as thickened lung markings, patchy consolidations, or other infiltrative changes). As detailed in Fig. 1, this study population was established from an eligible cohort of 774 patients. These patients were categorized into two groups based on the testing method used on BALF samples. Specifically, all eligible patients in this cohort met the clinical indications for mNGS as outlined in the relevant expert consensus [18]. These guidelines recommend mNGS for high-risk populations, such as older adults with underlying comorbidities, particularly when traditional microbiological tests are inconclusive or when rapid pathogen identification is urgently required for severe infections. While mNGS was universally recommended to patients meeting these clinical criteria, the final decision to proceed was ultimately based on informed consent and patient preference. The mNGS group (n = 168) consisted of patients who underwent both mNGS and conventional culture testing on the same BALF samples. The non-mNGS group (n = 354) served as a comparator cohort for clinical and economic outcome analyses. To mitigate potential selection bias and ensure a balanced baseline, we employed a sex-stratified and age-stratified random sampling approach to select approximately 60% of the 600 eligible patients who received culture testing alone. Overview of patients enrolled and number of patients tested by each method. Flowchart of patient inclusion pathway and the stratification into mNGS (n = 168) and non-mNGS (n = 354) testing groups Data on patient demographics, clinical outcomes, and medical costs were obtained from the hospital’s electronic medical records system. We used the term “sex” to refer to sex assigned at birth. Ethnicity data were not formally analyzed in this study because the vast majority of patients visiting our center are Han Chinese. Experimentation Sample Processing and DNA Extraction BALF samples (1.5–3 ml) were collected according to standard clinical protocols. For host depletion, a commercial kit from BGI Biotechnology (Wuhan) Co., Ltd. was used. 450 μl of each sample was treated with saponin (final concentration: 0.025%), vortexed for 15 s, and incubated at room temperature for 5 min. Then, 75 μl of Host Depletion Mix was added, followed by vortexing (15 s) and incubation at 37 °C for 10 min. After centrifugation at 18,000×g for 5 min, 450 μl of the supernatant was removed, leaving ~ 70–80 μl. The pellet was washed with 800 μl PBS, vortexed, centrifuged again (18,000×g, 5 min), and the supernatant was discarded. Subsequently, 370 μl TE buffer and 7.20 μl lyticase were added for enzymatic cell wall disruption. Mechanical lysis was performed using 250 μl of 0.50 mm glass beads on a horizontal vortex mixer at 2800–3200 rpm for 30 min. Finally, 300 μl of the lysate was transferred to a new microcentrifuge tube, and DNA was extracted using the TIANamp Micro DNA Kit (Cat. No. DP316, TIANGEN Biotech) following the manufacturer’s instructions. Library Construction and Sequencing Extracted nucleic acids were processed through enzymatic fragmentation, end repair, adapter ligation, and PCR amplification for library preparation. The resulting libraries were assessed for fragment size (~ 300 bp) using an Agilent 2100 Bioanalyzer, and quantified using the Qubit dsDNA HS Assay Kit (Thermo Fisher Scientific Inc) [19]. The constructed libraries were then pooled in equimolar amounts based on the measured concentrations. The pooled libraries were circularized to form single-stranded circular DNA molecules, which subsequently underwent rolling circle amplification (RCA) to produce DNA nanoballs (DNB). These DNBs were loaded onto sequencing flow cells and sequenced on the MGISEQ-2000 platform [20]. Data Analysis Bioinformatic Analysis Following sequencing, low-quality sequences were filtered out to obtain high-quality data. Reads mapped to the human reference genome sequence were removed using the Burrows–Wheeler Alignment (BWA) algorithm (version 0.7.17) [21]. The remaining non-human, high-quality reads were further filtered to eliminate low-complexity reads and then mapped against the BGI PMDB pathogen database. Microbial abundance was quantified as reads per million (RPM) of the total reads in the library. A microorganism was considered positively detected if its RPM > 2. Identification of Pathogens To objectively distinguish potential causative pathogens from respiratory colonization or background contamination, a rule-based filtering approach was applied based on the expert consensus on lower respiratory tract infections [18]. Specifically, all detected microorganisms were cross-referenced against this consensus. Only microorganisms explicitly recognized as strict pathogens or opportunistic pathogens were retained and collectively identified as causative pathogens for analysis. In addition, the final determination of causative pathogens was supported by comprehensive clinical evaluation, incorporating clinicians’ assessments based on patient symptoms and imaging findings. These confirmed pathogens were ranked based on their individual detection frequencies across all samples. Cumulative detection percentages were then calculated by summing the detection proportions in descending order. Based on pre-defined cumulative thresholds, the pathogens were stratified into three tiers: common pathogens (species contributing to < 75.00% cumulative detection), uncommon pathogens (75.00–92.20%), and rare pathogens (remaining species) (detailed criteria in Supplementary Table S1 and Figure S1). For example, Klebsiella pneumoniae was the most frequently detected pathogen (rank 1), followed by Pseudomonas aeruginosa (rank 2), and so forth, until the cumulative cutoffs were reached. Cost-Effectiveness Analysis The cost-effectiveness analysis was conducted from a societal perspective, considering both direct and indirect costs were considered. Direct costs encompassed expenditures directly associated with the diagnostic and treatment process, including technology, consumables, and labor. Specifically, hospitalization costs comprised diagnostic costs (laboratory testing, microbial cultures, pathological examinations, imaging procedures such as X-rays and CT scans, and other evaluations including electrocardiograms), treatment expenses (medication, surgical procedures, therapeutic interventions, and medical equipment use), and other related charges (nursing care, bed occupancy, and meals). Indirect costs referred to productivity losses associated with the hospitalization length of stay (LOS). Given that the time horizon was restricted to the single hospital admission (< 1 year), neither costs nor health outcomes were discounted. Patients were stratified by sex, age, and type of pathogens. Hospitalization LOS and total hospitalization costs were compared between those who underwent mNGS testing and those who underwent culture only. For pathogen-based stratification, only the pathogen with the highest abundance was considered in cases of mixed infections. To assess the influence of testing timing, patients were further divided into five subgroups based on when mNGS was performed—on approximately days three, four, five, six, and seven post-admission—to compare differences in hospitalization LOS and costs. Health economic analyses were then conducted to evaluate the cost-effectiveness of mNGS testing regarding both its implementation and timing. The analysis incorporated hospitalization expenses and productivity losses. Productivity losses were calculated using the year-specific per capita disposable income (CNY/day) corresponding to each patient’s actual period of hospitalization, derived from the Shenzhen Statistical Yearbook (2021, 2023) and the 2023 Shenzhen National Economic and Social Development Statistical Bulletin [22–24]. For hospital stays spanning across calendar years, a time-weighted average income was applied based on the number of days spent in each year. Specific cost figures are presented in Supplementary Table S2. Given the skewed distribution of hospitalization costs and LOS, with some extreme values, the expected benefit was calculated using the median cost, weighted by the patient proportion in each group. Assuming the infection pattern of the 522 enrolled patients reflects the hospital’s overall patient profile, a cost-effectiveness analysis was performed accordingly (detailed in Eqs. 1–2) [25, 26]. A net benefit > 0 indicates that mNGS is more cost-effective than culture from an economic perspective, and that earlier mNGS testing yields greater economic advantages. Note: Ctotal: Total costs, the expected total economic burden per for a specific diagnostic strategy; Cdirect(i): Direct costs, the median direct hospitalization cost for patients with infection type i; T(i): Time, the median hospitalization LOS for infection type i; AvgIncome: Average income per capita; P(i): Prevalence, the proportion of infection type i in the specific cohort (used as weighting factor). NMB: Net monetary benefit. Statistical Methods Statistical and health economic analyses were conducted using R version 4.4.2. The Kolmogorov–Smirnov method was used to assess data normality. Non-normally distributed continuous variables are presented as median (first quartile, third quartile) [M (Q1, Q3)], while categorical variables were expressed as frequencies and percentages [n (%)]. Comparisons between groups were performed using χ2 tests for categorical data and Wilcoxon rank-sum tests for non-parametric data. Median differences and their 95% confidence intervals (CIs) were estimated using the Hodges–Lehmann method. To ensure that the clinical and economic advantages of mNGS were independent of baseline variations, multivariable generalized linear models (GLMs) with a Gamma distribution and log link function were employed to adjust for age, sex, symptoms, and comorbidities. A p value < 0.05 was considered statistically significant. Stratified analyses were performed across age, sex, and pathogen subgroups to assess the robustness and consistency of the results. Given the retrospective design of this study, all subgroup analyses were pre-defined as exploratory. Accordingly, nominal p values are reported without multiplicity adjustments. Results Patient Baseline Characteristics This study enrolled 522 patients aged 55–69 with pulmonary infections. The two groups—the mNGS group and the non-mNGS group—had comparable sex distributions and shared a median age of 62 years (Table 1). Over half of the patients had no underlying comorbidities. Multivariable GLMs analysis, adjusting for age, sex, symptoms, and comorbidities, confirmed mNGS as a significant independent predictor of lower costs (p < 0.001) and a downward trend in hospitalization LOS (Supplementary Table S3–S4). This indicates that the observed economic and clinical advantages are robust and independent of baseline variations. Table 1Demographic and clinical baseline characteristics of older adults with pulmonary infectionPatient characteristicsmNGS group (n = 168)non-mNGS group (n = 354)Sex, n (%) Female88 (52.38)172 (48.59) Male80 (47.62)182 (51.41)Age (M, IQR)62 (8.00)62 (8.00) 55–59, n (%)61 (36.31)132 (37.29) 60–64, n (%)56 (33.33)98 (27.68) 65–69, n (%)51 (30.36)124 (35.03)ICU patients, n (%)5 (2.87)0Hospitalization LOS, (M, IQR)10 (6.00)11 (8.00) 1–10, n (%)91 (54.17)160 (45.20) 11–30, n (%)70 (41.67)168 (47.46) ≥ 30, n (%)7 (4.17)26 (7.34)Comorbidity, n (%) Hypertension44(26.19)109(30.79) Hyperglycemia12(7.14)47(13.28) Hyperlipidemia13(7.74)18(50.85) Coronary artery disease3(1.79)35(9.89) COPD9(5.36)13(3.67) Tumor1(0.60)11(3.11)Clinical symptoms, n (%) Fever31 (18.45)33 (9.30) Cough112 (66.67)235 (66.38) Chest pain23 (13.69)53 (14.93)M median, IQR interquartile range, COPD chronic obstructive pulmonary disease Demographic and clinical baseline characteristics of older adults with pulmonary infection M median, IQR interquartile range, COPD chronic obstructive pulmonary disease Microbial Detection Results from mNGS and Culture Among the 168 patients who received both simultaneous mNGS and culture testing (the mNGS group), mNGS demonstrated a significantly higher diagnostic yield compared to conventional culture. This substantial superiority was consistently observed both in the overall detection of any microorganisms and in the specific identification of causative pathogens (Table 2). Table 2Comparison of positivity rate between mNGS and culture in the mNGS group [n (%)]MethodAll microorganismsPathogensPositiveNegativeχ2p valuePositiveNegativeχ2p valuemNGS151 (89.88)17 (10.12)139.91< 0.001114 (67.86)54 (32.14)81.58< 0.001Culture44 (26.19)124 (73.81)< 0.00131 (18.45)137 (81.55)< 0.001 Comparison of positivity rate between mNGS and culture in the mNGS group [n (%)] Patient-level comparisons revealed that mNGS identified significantly more microbial species than culture (Fig. 2A, p < 0.001). In total, 259 species were detected by mNGS, including 184 bacteria, 64 fungi, and six viruses, with Streptococcus pneumoniae being the most frequently detected microorganism. Detection and pathogen spectrum by mNGS and culture, and concordance within the mNGS group. A Comparison of pathogen species number detected per patient by each method. B Detailed detection frequency of each pathogen species by mNGS and culture. C Distribution of detected pathogens by frequency (common, uncommon, rare) for each method. Blue: Common pathogens, Orange: uncommon pathogens, Red: rare pathogens. Pathogens with no detection (Klebsiella aerogenes, Corynebacterium striatum, Cryptococcus neoformans, Burkholderia cepacia, Klebsiella oxytoca, Proteus penneri, Pseudomonas oryzihabitans, Streptococcus pyogenes, Aspergillus terreus) are excluded from the figure. D Diagnostic concordance and discordance between mNGS and culture results. Concordant: mNGS and non-mNGS identified identical pathogens, Overlapping: mNGS detected all non-mNGS-identified pathogens plus additional species, Discordant: no pathogen overlap A comparison of pathogen detection between culture and mNGS revealed that 66.20% (47/71) of these pathogens were exclusively identified by mNGS. The most frequently detected species included Streptococcus pneumoniae, Acinetobacter baumannii, and Acinetobacter johnsonii. In contrast, only three pathogens—Pseudomonas stutzeri, Aggregatibacter segnis, and Serratia marcescens were uniquely detected by culture, likely reflecting their preferential growth in culture conditions (Fig. 2B). Based on the weighted statistics of pathogen species, the distribution of detection results for the two methods is shown in Fig. 2C. Most pathogens identified by culture (82.35%) and mNGS (72.93%) were classified as common pathogens. Twenty-six patients (15.48%) were positive in both tests, and mNGS results overlapped with culture results in 16 (61.54%) of them, with 15 (93.75%) cases exhibited polymicrobial infections (Fig. 2D). Cost-Effectiveness Analysis Comparison of mNGS and non-mNGS in Cost-Effectiveness To evaluate the cost-effectiveness of mNGS testing, we compared hospitalization LOS and associated costs between the mNGS and non-mNGS groups after stratifying patients by type of pathogens (Supplementary Table S1 and Figure S1). In patients aged 55–59 years, mNGS significantly reduced hospitalization LOS (median difference: − 2.00 days; 95% CI − 3.00 to 0.00; p = 0.020) and costs (median difference: − 2185 CNY; 95% CI − 4590 to − 376; p = 0.018). Similarly, for patients with common pathogens infection, the mNGS group exhibited significantly lower costs (median difference: − 3067 CNY; 95% CI − 7283 to − 572; p = 0.014) (Fig. 3, detailed in Supplementary Tables S5–S6). The most substantial benefits were observed in the uncommon or rare pathogens subgroup, with mNGS yielding a median difference of − 4.00 days in LOS (95% CI − 9.00 to 0.00; p = 0.036) and − 10,715 CNY in costs (95% CI − 54,722 to − 2084; p = 0.004). Comparison of hospitalization costs and LOS: mNGS vs. non-mNGS in the overall stratified groups. The three groups differ in terms of age, gender, and infection type. A Comparison of Hospitalization LOS Across Different Groups (excluding the 5th and 95th percentile outliers). B Comparison of Hospitalization Costs Across Different Groups (excluding the 5th and 95th percentile outliers). Detailed statistical values are provided in the main text. *p < 0.05, **p < 0.01. The reported p values are derived from the original dataset and are nominal and unadjusted for multiple comparisons Based on previously published economic parameters and the clinical outcome data (including hospitalization LOS and costs) generated in this study, which are detailed in “Cost-Effectiveness Analysis”, we conducted a cost–benefit analysis comparing mNGS and non-mNGS groups. The total cost per patient was 13,901.81 CNY in the mNGS group and 15,104.51 CNY in the non-mNGS group, resulting in a net monetary benefit of 1202.70 CNY for mNGS group based on our economic model (Eqs. 1 and 2). Stratified analysis further confirmed the economic advantages of mNGS across every sex and age subgroups. All sex and age groups showed positive net benefits (> 0 CNY, Fig. 4). The smallest benefit was observed in the 65–69 age group, with a net gain of 514.37 CNY (detailed in Supplementary Table S7–S8). Furthermore, among pathogen-positive patients who were categorized into three groups based on pathogen type, the net benefit associated with mNGS testing was 3831.15 CNY (Supplementary Table S9). These findings indicated that mNGS testing can substantially reduce the economic burden of pulmonary infections in older adults. Net benefit of mNGS vs. culture across different groups. Net benefit (CNY) of mNGS versus culture, stratified by sex and age group Further analysis of the pathogens-infected population revealed that mNGS group had consistently lower median hospitalization LOS and costs than the non-mNGS group across various subgroups. Specifically, across female and 60–64-year-old subgroups, the non-mNGS group exhibited significantly higher median hospitalization LOS compared to the mNGS group. Regarding costs, significant differences were found in both sexes, in patients aged 60–64 years, and in patients with infections from common and uncommon pathogens. More details are shown in Supplementary Figure S2, Supplementary Table S9–S10. Additionally, we analyzed net benefit by infection types, namely, common, uncommon, and rare pathogens. The highest benefit was observed in patients infected with uncommon pathogens (1654.52 CNY), followed by those infected with rare pathogens (1349.14 CNY). Infections caused by common pathogens yielded relatively modest benefits (< 1000 CNY). While initial broad-spectrum empirical therapy was frequently maintained across the cohort due to disease severity, the pronounced economic advantages observed in cases involving fastidious pathogens highlight the critical value of mNGS in specific complex scenarios. As detailed in Supplementary Table S11, in instances where conventional cultures were negative or provided incomplete pathogen profiles, mNGS uniquely identified key causative agents (e.g., Pneumocystis jirovecii and Actinomyces israelii). This facilitated a timely transition to definitive targeted therapies, such as SMZ-TMP or Penicillin G. Such real-world management adjustments in specific patient subsets directly contributed to the overall reduction in hospitalization LOS and medical costs quantified in our economic model. Cost-Effectiveness of mNGS Testing Timing To quantify the economic and clinical value of early mNGS implementation, we analyzed 168 mNGS-tested patients using sequential daily cutoffs from 3 to 7 days post-admission, creating five comparison pairs. The distribution of patients was relatively balanced between those tested before versus after 5 days (76 vs. 92 patients) and 6 days (98 vs. 70 patients), whereas other time-stratified groups showed two-to-three fold differences in sample sizes (detailed in Supplementary Table S12). Patients who underwent mNGS testing at later timepoints experienced significantly longer median hospital stays and incurred higher hospitalization costs (by sex, age, and infection type; p < 0.01 or p < 0.001 for most comparisons) (Fig. 5). This trend remained consistent in the subgroup analysis of pathogen-infected patients, where delayed mNGS testing was similarly associated with prolonged hospitalization and higher costs (p < 0.01 or p < 0.001). Supplementary Figure S3 and Supplementary Table S13–S14 in supplementary material file showed more details. Comparison of hospitalization costs and LOS: early vs. late mNGS in overall population. A Comparison of hospitalization LOS in patients with different mNGS testing times. B Comparison of total medical costs in patients with different mNGS testing times (excluding the 5th and 95th percentile outliers). *p < 0.05, **p < 0.01, ***p < 0.001. The reported p values are derived from the original dataset and are nominal and unadjusted for multiple comparisons The magnitude of these differences varied across testing windows. Based on Wilcoxon rank-sum tests, a 6-day post-admission threshold maximized the disparities in both hospitalization LOS and costs between early and late testing groups. Specifically, early testing (≤ 6 days) significantly reduced hospitalization LOS (median difference: − 5.00 days; 95% CI − 6.00 to − 4.00; p < 0.001) and direct costs (median difference: − 2527 CNY; 95% CI − 3998 to − 1156; p < 0.001) compared to delayed testing. Furthermore, when incorporating productivity loss into our deterministic economic model, early testing (≤ 6 days) demonstrated substantially more favorable outcome, showing 6346.00 CNY (median total costs: 11,420.59 vs. 17,766.59 CNY). This suggests that the early implementation of mNGS offers greater economic advantages and clinical value. Discussion This study systematically evaluated the diagnostic performance and health economic impact of mNGS in older adults with pulmonary infection. The findings demonstrated that mNGS yielded significantly higher microbial detection rates and broader spectrum coverage than conventional culture approach. Additionally, mNGS testing was associated with a notable reduction in both hospitalization LOS and related costs. These results are consistent with previous studies [8] which evaluated the performance of mNGS in detecting pathogens using a large clinical cohort, highlighting the potential of mNGS to optimize clinical strategies to treat patients with pulmonary infections through early and comprehensive pathogen identification. By facilitating timely and targeted therapy, mNGS may help reduce the reliance on prolonged empirical treatment and the associated healthcare resource burden [27, 28]. Cost-effectiveness analysis revealed substantial net benefits from implementing mNGS, with particularly pronounced advantages in complex clinical cases. Compared to the gains of less than 1000 CNY in the treatment of common pathogens, infections involving uncommon or rare pathogens yielded greater gains, reflecting the ability of mNGS to expedite accurate diagnosis and reduce prolonged diagnostic delays. This underscores the value of mNGS in managing atypical microbial infections, where it enhances patient outcomes by minimizing unnecessary treatments and optimizing resource utilization. Moreover, these findings highlight mNGS’s potential relevance to precision medicine initiatives and public health responses to emerging pathogens. Therefore, while often positioned as a second-line or “rescue” diagnostic tool, our data suggest a pragmatic, high-value approach would be to prioritize mNGS earlier in the diagnostic algorithm for this vulnerable aging population, who are at high risk of deterioration. mNGS offers critical diagnostic value in severe cases such as complicated pneumonia, where rapid pathogen identification is essential. Age-stratified analysis revealed notable differences in economic returns: patients aged 55–59 years had the highest benefit (2458.81 CNY), whereas those aged 65–69 years showed more modest gains (514.37 CNY). In clinical practice, this disparity may reflect the higher prevalence of underlying comorbidities in the older subgroup [29], which could potentially lead to more comprehensive diagnostic evaluations—such as CT imaging—and prolonged hospital stays for close monitoring due to their increased vulnerability. Moreover, clinicians managing patients in this age group may supplement mNGS results with empirical judgment, frequently opting for broader-spectrum or higher-tier antibiotics to prevent potential deterioration. These combined factors likely contribute to longer hospitalization LOS and increased medical expenses in this subgroup, thereby diluting the relative economic benefit of mNGS. Our analysis further demonstrates that early implementation of mNGS—within 6 days of admission—confers significant health economic benefits, with a net saving of 6346.00 CNY compared to later testing. The benefits of early testing are likely driven by two main facts: (i) shortening the critical window between diagnosis and initiation of targeted therapy, thereby avoiding prolonged use of broad-spectrum antibiotics [30]; and (ii) reducing downstream complications and the associated costs linked to progressive organ damage [31]. This study validated mNGS as a highly sensitive diagnostic tool for BALF samples, aligning with previously published reports [32]. Quantitative comparisons affirmed the superior detection capability of mNGS for most pathogens, particularly for fastidious organisms such as Streptococcus pneumoniae [33] and Haemophilus influenzae [34], highlighting its unique advantage in identifying difficult-to-culture pathogens. More critically, mNGS provided a substantial time advantage, with a median detection time of 26.11 h compared to 74.5 h for culture, thus accelerating clinical decision-making in the management of severe infections [35]. However, these comprehensive advantages must be interpreted cautiously due to the inherent limitation of our DNA-focused protocol that excludes RNA viruses. Given the clinical importance of respiratory RNA viruses [36, 37], our study could not assess the diagnostic yield for this specific class of pathogens, likely leading to a conservative estimate of the true economic benefits. Several limitations should be acknowledged when interpreting these findings. First, the single-center design and modest sample size may limit the generalizability of our findings, as the infection patterns observed may only reflect our local epidemiological characteristics. Second, the retrospective design introduces inherent selection bias, as mNGS allocation was influenced by clinical judgment and socioeconomic factors, which also precluded the systematic evaluation of cohort-wide changes in clinical management. While this limited our control over confounding variables like disease severity, we employed sex- and age-stratified random sampling and multivariable GLMs to ensure a more balanced and robust baseline comparison. Third, the retrospective design precluded prospective, individual-level adjudication of colonization versus true infection, which is a well-known clinical challenge of mNGS due to its ultra-high sensitivity and risk of false positives. Consequently, our pathogen identification relied on a rule-based filtering approach (RPM > 2 and expert consensus guidelines) to exclude typical colonizing flora and background noise. Finally, due to the study design, we lacked systematic long-term follow-up data to formally evaluate post-discharge outcomes such as readmissions or complications. Although a preliminary review of electronic medical records revealed very few documented readmissions in this cohort, our formal analysis was necessarily confined to a short-term hospitalization period (LOS and direct medical costs). We did not model long-term outcomes, such as quality-adjusted life years (QALYs), long-term complications, or intangible costs (e.g., pain and suffering), which likely resulted in a conservative estimate of the total value of mNGS. BALF-based mNGS offers a superior diagnostic approach for pulmonary infections, providing higher positivity rates and detecting a broader range of pathogens than conventional culture methods. Early implementation of mNGS not only enhances diagnostic accuracy and informs targeted antimicrobial therapies but also reduces healthcare costs for patients in this age bracket. Beyond the immediate clinical and economic benefits observed in this study, adopting a broader societal perspective, accounting for productivity loss and long-term care needs, might reveal even greater economic advantages. Conclusions Collectively, BALF-based mNGS demonstrates greater diagnostic yield and cost-effectiveness than culture, and its early implementation enhances targeted therapy while reducing costs, particularly in this aging demographic. These findings suggest that integrating this technology into clinical practice can improve patient outcomes while promoting more cost-effective management of respiratory infections. Supplementary Information Below is the link to the electronic supplementary material. Supplementary file1 (PDF 820 KB) Acknowledgements We gratefully acknowledge the contributions of the doctors and nurses at Peking University Shenzhen Hospital in the collection of clinical samples, as well as the hospital staff and students for their contribution to recording the clinical dataset. Author Contributions Ke Yuan, Mang Shi, and Ping Xu conceived the study, designed the experiments, and supervised the project. Kai-jie Liu, Yan Gao, Xiyu Chu, and Mang Shi performed the data acquisition, statistical analyses, interpreted the analytical results, and generated figures/tables. Huilin Yang collected the human specimens and associated clinical data obtained ethical review board approval. Zhao-rong Li, Ke Yuan, Chang Lu, and Huilin Yang provided critical project assistance, technical advice, and contributed to data interpretation. Kai-jie Liu, Yan Gao, and Xirui Yang drafted the initial manuscript and prepared the appendices. Yong Xia, Chang Lu, Hui Huang, Zhao-rong Li, Ping Xu, and Mang Shi revised the manuscript and provided critical comments. All authors reviewed the results, discussed the findings, commented on all versions of the manuscript, and approved the final version for submission. Funding This work was supported by the Shenzhen Science and Technology Innovation Commission Foundation (KQTD20200820145822023, JCYJ20230807095204008, and JCYJ20240813120110015), Key Program for Clinical Research at Peking University Shenzhen Hospital (No. LCYJ2021008). The journal’s Rapid Service Fee was funded by JCYJ20230807095204008. Data Availability The datasets generated during and/or analyzed during the current study are available from the corresponding author on reasonable request. The analytical codes and de-identified clinical data are available in the GitHub repository: https://github.com/NN-00/mNGS_economic. Declarations Conflict of Interest Kai-jie Liu, Yan Gao, Xirui Yang, Yong Xia, Chang Lu, Zhao-rong Li, Xiyu Chu, Hui Huang, Ping Xu, Mang Shi, Ke Yuan, and Huilin Yang each have nothing to disclose. Ethical Approval This study was approved by the Ethics Committee of the Shenzhen Hospital, Peking University (No. 2025052). This study was conducted in strict accordance with the 1964 Helsinki Declaration and its subsequent amendments. Informed consent was obtained from each patient or the immediate family. All sample processing and sequencing procedures were performed in accordance with the relevant guidelines and regulations of the biosafety facility of Shenzhen Hospital, Peking University. Footnotes Publisher's Note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. Kai-jie Liu, Yan Gao, and Xirui Yang contributed equally to this work. Contributor Information Ping Xu, Email: pingxu@hotmail.com. Mang Shi, Email: shim23@mail.sysu.edu.cn. Ke Yuan, Email: Yuanke@genomics.cn. Huilin Yang, Email: canesta@163.com. References Associated Data Supplementary Materials Supplementary file1 (PDF 820 KB) Data Availability Statement The datasets generated during and/or analyzed during the current study are available from the corresponding author on reasonable request. The analytical codes and de-identified clinical data are available in the GitHub repository: https://github.com/NN-00/mNGS_economic.