Abstract
Background
The impact of targeted next-generation sequencing (tNGS) in pediatric lower respiratory tract infections (LRTI) remains largely unknown. We aimed to investigate whether adding tNGS to standard conventional microbiological tests (CMTs), compared with CMTs alone, was associated with antimicrobial stewardship and clinical outcomes.
Methods
This retrospective cohort study was conducted at Tongji Hospital, Wuhan, between 2023-2024. Pediatric patients hospitalized with LRTI were categorized into tNGS (tNGS in addition to CMTs, typically as a second-line test) or control (CMTs alone) groups based on testing received during the first 4 days after admission. To address potential baseline confounding, we performed 1:2 propensity score matching to balance key clinical characteristics between groups. The primary outcomes were whether and time to provision of pathogen-directed treatment within 96 hours after the index time. Secondary outcomes included other antimicrobial stewardship measures and clinical outcomes. Between-group outcomes were assessed using generalized linear models to estimate risk differences (RDs, identity link) and odds ratios (ORs, logit link) for binary outcomes, while time-to-event analyses were performed using Fine-Gray subdistribution hazard models to account for competing risk of discharge, with results reported as hazard ratios (HRs) and corresponding 95% confidence intervals (CIs).
Results
After matching, 520 participants were included (182 tNGS group; 338 control group). During follow-up, 75 patients (41.21%) in the tNGS group and 48 (14.20%) in the control group received pathogen-directed treatment, with an absolute RD of 27.01% (95% CI: 18.95%, 35.07%) and an OR of 4.23 (95% CI: 2.78, 6.51). The mean time to treatment was 58.99 hours in the tNGS group vs. 82.95 hours in the control group (mean difference, −23.96 hours; 95% CI: −31.02, −16.90), with a corresponding HR of 3.55 (95% CI: 2.47, 5.09). Patients in the tNGS group also had lower rates of non-pathogen-directed antibiotic use (OR=0.52, 95% CI: 0.33, 0.79). Clinical outcomes were similar between the two groups.
Conclusions
Use of tNGS as a second-line test was associated with faster and more targeted antimicrobial treatment in pediatric LRTIs, but its impact on clinical outcomes requires further investigation.
Clinical trial number
Not applicable
Supplementary Information
The online version contains supplementary material available at 10.1186/s12931-025-03462-9.
Keyword: Targeted next-generation sequencing, Pediatric lower respiratory tractinfections, Antimicrobial stewardship, Retrospective cohort study
Introduction
Lower respiratory tract infections (LRTIs) were one of the leading cause of hospital admissions and mortality among children and adolescents younger than 14 years, accounting for 69.9 million episodes and 0.55 million deaths worldwide in 2021[1]. These infections present significant diagnostic challenges due to their diverse etiologies, with overlapping and indistinguishable clinical presentations. Accurate and timely identification of the causative pathogens is crucial for optimizing antimicrobial therapy, preventing antimicrobial resistance [2], and improving patient outcomes [3]. Although conventional microbiological tests (CMTs), including microscopy, culture, serology, and pathogen-specific polymerase chain reaction (PCR), remain the diagnostic cornerstone, they fail to identify a causative pathogen in more than half of patients with LRTI [4, 5]. This substantial diagnostic gap has created an urgent clinical need for broad-spectrum detection technologies capable of simultaneously identifying multiple pathogens with enhanced sensitivity.
Targeted next-generation sequencing (tNGS) has emerged as a promising solution to address these diagnostic limitations. By simultaneously screening for comprehensive panels of common pathogens (including bacteria, viruses, fungi, and atypical pathogens), tNGS offers enhanced diagnostic breadth and sensitivity compared to CMTs, making it an increasingly valuable complement to CMTs in LRTI diagnosis [6, 7]. The diagnostic performance advantages of tNGS over CMTs have been clearly demonstrated [7–11]. However, compelling evidence demonstrating its impact on patient management decisions and outcomes in real-world pediatric settings remains limited. While recent studies have investigated the potential impact of tNGS on antimicrobial use [10, 12, 13], they were constrained by before-after study designs and yielded ambiguous conclusions due to the lack of appropriate control groups.
Therefore, this study aimed to evaluate whether adding tNGS as a second-line test to standard CMTs, compared with CMTs alone, was associated with improved antimicrobial stewardship and clinical outcomes in pediatric patients hospitalized with LRTI.
Methods
Study design and participants
This retrospective cohort study was conducted at Tongji Pediatric Hospital (TJPH), affiliated with Tongji Hospital of Tongji Medical College, Huazhong University of Science and Technology, Wuhan. TJPH is a major pediatric medical center in central China, with annual volumes exceeding 1 million outpatients and 30,000 inpatients in 2024. We included patients who were diagnosed with LRTI and hospitalized in the Department of Pediatric Respiratory Medicine from January 1, 2023, to December 31, 2024. Participants aged > 14 years or with length of stay > 30 days were excluded (Fig. 1A). An episode of LRTI was defined as having two or more of the clinical features including clinical symptoms, physical signs, and chest imaging [14–16].
Fig. 1.
Flow diagrams and study design
Since over 95% of patients received CMT results within four days of admission, and tNGS was generally performed as a second-line test for patients with either negative CMT results or complex medical conditions where symptoms were not fully explained by initial positive CMT results [17], we established the first four days after admission as the evaluation window (Fig. 1B). The index time was set at 0:00 on day 4, with the exposure period defined as the time from hospital admission to index time. Participants were categorized into the tNGS group if they received tNGS testing in addition to CMT, or the control group if they received CMT alone during the exposure period. Participants who were discharged or initiated pathogen-directed treatment during the exposure period were further excluded (Fig. 1A).
The study complied with the Declaration of Helsinki principles and was approved by the Ethics Committee of Tongji Hospital (No.TJ-IRB202506028). Informed consent was waived due to the use of anonymous and de-identified secondary data.
Sample collection and transport
Nasopharyngeal or oropharyngeal swabs and blood samples were collected within 12 hours after admission. Because children are often unable to produce adequate sputum for laboratory testing [18], nasopharyngeal aspirate (NPA) specimens were collected using a disposable suction catheter under negative pressure. Flexible bronchoscopy was performed to collect bronchoalveolar lavage fluid (BALF) in participants presenting with severe radiological findings (e.g., consolidation, atelectasis, emphysema) or when initial, less invasive specimens failed to yield a pathogen diagnosis. All samples were transported to the laboratory immediately after collection and subsequently stored at −80℃ until processing.
tNGS workflow construction
NPA or BALF samples were transported to one of two local commercial in vitro diagnostics laboratories (KingMed Diagnostics in 2023 and HUGO Biotech in 2024). Detailed methodologies are provided in the Supplementary Methods. Briefly, following nucleic acid extraction, specific primers are designed to target-specific sequences, as well as some selected antimicrobial resistance sequences for enrichment of multiplex targets, followed by library preparation for NGS sequencing and bioinformatics analysis. The multiplex PCR-tNGS (mp-tNGS) panel covered pathogens commonly encountered in clinical scenarios (Table S1 and Table S2).
Conventional microbiological tests
CMTs included microscopy, culture, immunoassays, and real-time PCR. NPA or BALF samples underwent microscopic examination after smear preparation and staining, as well as culture for detection of both bacteria and fungi. When mycobacterial infection was suspected, mycobacterial culture was also performed. Additionally, (1,3)-β-D-glucan and galactomannan tests were conducted to assist in the diagnosis of fungal infections. Serological antibody detection was performed for 4 atypical pathogens and 5 viruses (Table S3). Nasopharyngeal or oropharyngeal swabs were analyzed using real-time PCR to detect 2 atypical pathogens and 11 viruses (Table S4). Microscopy was performed on NPA samples according to guidelines. A purulent-looking portion was gram stained, and samples with < 10 squamous epithelial cells (SECs), > 25 leukocytes, or a ratio of leukocytes/SECs > 2.5 per field at a 100× magnification, were classified as valid [19]. Attending physicians determined which CMT should be performed based on the patient’s clinical manifestations. The proportions of each CMT performed in the tNGS and control groups are detailed in Table S5.
Covariates
Covariates were selected a priori as potential confounders and collected from the electronic data warehouse of TJPH. Demographics included age (months) and sex. Duration of symptoms prior to admission (days), comorbidities (chronic heart and lung diseases, preterm birth, and cesarean delivery), and vital signs (temperature [°C], respiratory rate [breaths per minute], and pulse rate [beats per minute]) were assessed at admission. A diagnosis of severe pneumonia was established if the patient met at least one of the following criteria: poor general condition; refusal to feed or signs of dehydration; impaired consciousness; significantly increased respiratory rate (infants ≥ 70 breaths/min, older children ≥ 50 breaths/min); cyanosis; dyspnea; oxygen saturation < 92% at sea level or < 90% in high-altitude areas; multi-lobar involvement or > 2/3 involvement of a single lobe; pleural effusion; or extrapulmonary complications [20]. Information on laboratory results (white blood cells [WBC, ×10⁹/L], lymphocytes [LYM, ×10⁹/L], platelets [PLT, ×10⁹/L], hemoglobin [Hb, g/L], albumin [Alb, g/L], alanine aminotransferase [ALT, IU/L], prothrombin time [PT, seconds], and serum creatinine [Scr, µmol/L]), chest radiological examinations (computed tomography [CT] and X-ray), microbiological results (bacteria, viruses, fungi, and atypical pathogens) by CMTs, antimicrobial treatment (antibiotics, antivirals, and antifungals), and oxygen administration was obtained during the exposure period. When multiple measurements of the same parameter were recorded for the same individual, only the earliest recorded value was considered. The calendar year of admission was also included as a predefined covariate. Given the low proportion of missing covariates (< 10%, Table S6), single imputation methods were employed for data processing [21]. Continuous variables were imputed using median values, while categorical variables were imputed by assigning the most frequent category.
Outcomes and follow-up
The primary outcomes were (1) binary outcome: provision of pathogen-directed treatment based on relevant microbiological test results and (2) time-to-event outcome: time to provision of pathogen-directed treatment. This approach has been used in a previous randomized controlled trial examining the impact of PCR-based panels on diagnostic stewardship in adult community-acquired pneumonia [22]. Follow-up started from the index time until receiving pathogen-directed treatment, discharge, or 96 hours after index time, whichever came first.
Two of our co-authors (D.X. and X.Y.Z.) reviewed the medical records and assessed whether and when a patient received pathogen-directed antimicrobial treatment based on microbiological test results, following the key principles outlined in the Supplementary Methods and Table S7. In cases of disagreement, a third senior physician (Y.J.H.) arbitrated. Pathogen-directed antimicrobial treatment was further categorized into seven patterns: antibiotic de-escalation, antibiotic escalation, antibiotic adjustment, initiated antiviral treatment with prior empirical antibiotic treatment, initiated antifungal treatment with prior empirical antibiotic treatment, initiated pathogen-directed antimicrobial treatment without prior empirical antibiotic treatment, and continuation of appropriate empirical antimicrobial treatment. According to the antibiotic spectrum index developed by Jeffrey and colleagues [23], we assigned frequently used antibiotics point values ranging from 2 to 13 (Table S8).
Secondary outcomes were assessed between index time and the time of discharge. These included antimicrobial stewardship measures: provision of any antibiotics, provision of intravenous antibiotics, provision of broad-spectrum antibiotics (defined as antibiotic spectrum index ≥ 7), provision of non-pathogen-directed antibiotics, provision of any antivirals, provision of non-pathogen-directed antivirals, provision of any antifungals, and provision of non-pathogen-directed antifungals. Clinical outcomes including PICU admission, oxygen administration, and time from index time to discharge were also reported.
Statistical analysis
We employed a propensity score (PS) matching approach to minimize potential confounding. We calculated the PS for receiving tNGS vs. control using a logistic regression model with the potential confounders listed above as independent variables. Patients in the tNGS group were then matched 1:2 with the control group using the nearest neighbor approach with a caliper width equal to 0.2 standard deviations of the PS scale.
Baseline characteristics were summarized using median (interquartile range, IQR) for continuous variables and counts (percentage) for categorical variables. We used standardized mean differences (SMD) to assess the balance of covariates, with differences greater than 0.1 indicating meaningful imbalances [24].
For binary outcomes, we used generalized linear models with binomial distribution, applying an identity link to estimate absolute risk differences or a logit link to estimate odds ratios (ORs) with their 95% confidence intervals (CIs) [25]. Multinomial logistic regression models were employed to estimate the relative risk ratios (RRRs) and 95% CIs for associations between tNGS and patterns of pathogen-directed treatment, with non-provision of pathogen-directed treatment as the reference group. For the time-to-event primary outcome, we used restricted mean survival time models to estimate differences in mean time to provision of pathogen-directed treatment [26], as well as Fine-Gray competing risk models to estimate subdistribution hazard ratios (HRs) and 95% CIs for tNGS vs. control, with follow-up duration as the underlying time scale and discharge as the competing event. The proportional hazards assumption was evaluated using cumulative sums of residuals [27] and there were no violations (p-value > 0.05). We also used the Aalen-Johansen estimator to plot cumulative incidence of outcomes and Gray’s tests to compare differences between groups. For continuous secondary outcomes, the bootstrap method with 1,000 resampling iterations was used to calculate median differences and 95% CIs.
The robustness of our results was examined through multiple sensitivity analyses. First, we performed complete case analyses by excluding participants with at least one missing covariate. Second, we excluded participants who received tNGS testing during the follow-up period. Third, we defined the index time as 00:00 on day 3 to mitigate potential bias associated with arbitrarily specified exposure windows. Fourth, we incorporated specific CMTs performed during the exposure period into the PS matching process to further control for residual confounding introduced by differences in the intensity of initial microbiological investigation between groups. Fifth, we excluded participants with a disease duration > 7 days prior to admission to minimize bias related to prolonged illness and potential prior antibiotic exposure that could significantly alter the respiratory microbioe profile. Prior to conducting these analyses, re-estimation of PS and re-matching were performed to ensure the robustness of intergroup balance.
All statistical analyses were performed using R software version 4.5.1 (https://www.R-project.org/). All statistical tests were two-sided, and a p-value < 0.05 was considered statistically significant.
Results
Study population and baseline characteristics
After exclusions, a total of 1,509 eligible pediatric patients were identified, with 185 in the tNGS group and 1,324 in the control group. The median (IQR) age was 39.16 (13.10, 69.43) months, and 585 (38.77%) were girls. Following PS matching, the analytical cohort comprised 520 participants (tNGS group: n = 182; control group: n = 338), and all covariates achieved balance (SMD < 0.1) (Table 1). Of the samples subjected to tNGS testing, 89 were BALF and 93 were NPA. In the tNGS group, 146 of 182 (80.22%) participants had pathogens detected by tNGS plus CMTs, compared with 160 of 338 (47.34%) in the control group with CMTs alone (Table S9).
Table 1.
Baseline characteristics of participants in full and propensity-score matched cohorts
| Full cohort | Propensity score–matched cohort | |||||
|---|---|---|---|---|---|---|
| tNGS | Control | SMD a | tNGS | Control | SMD a | |
| Number of participants | 185 | 1324 | 182 | 338 | ||
| Age median (IQR), months | 38.32 (7.37, 80.50) | 39.21 (13.96, 68.66) | 0.027 | 38.74 (7.37, 80.21) | 37.53 (8.32, 74.27) | 0.027 |
| Girls (n, %) | 80 (43.24) | 505 (38.14) | 0.104 | 80 (43.96) | 143 (42.31) | 0.033 |
| Duration of symptoms prior to admission, median (IQR), days | 8.00 (6.00, 16.80) | 7.00 (4.00, 11.25) | 0.273 | 8.00 (6.00, 15.00) | 7.00 (5.00, 15.00) | 0.069 |
| Comorbidities (n, %) | ||||||
| Chronic heart and lung diseases | 46 (24.86) | 277 (20.92) | 0.094 | 43 (23.63) | 87 (25.74) | 0.049 |
| Preterm birth | 21 (11.35) | 133 (10.05) | 0.042 | 20 (10.99) | 40 (11.83) | 0.027 |
| Cesarean delivery | 103 (55.68) | 751 (56.72) | 0.021 | 100 (54.95) | 188 (55.62) | 0.014 |
| Observations at admission, median (IQR) | ||||||
| Temperature, 。C | 36.70 (36.60, 37.00) | 36.80 (36.60, 37.10) | 0.071 | 36.70 (36.60, 37.00) | 36.80 (36.60, 37.00) | 0.025 |
| Respiratory rate, breaths per min | 25.00 (23.00, 28.00) | 25.00 (23.00, 30.00) | 0.006 | 25.00 (23.00, 28.00) | 25.00 (24.00, 30.00) | 0.043 |
| Pulse rate, bpm | 108.00 (100.00, 123.00) | 110.00 (100.00, 125.00) | 0.062 | 108.00 (100.00, 123.75) | 110.00 (100.00, 126.00) | 0.026 |
| Diagnosis of severe pneumonia at admission (n, %) | 8 (4.32) | 47 (3.55) | 0.04 | 8 (4.40) | 15 (4.44) | 0.002 |
| Laboratory results, median (IQR) | ||||||
| WBC, ×109/L | 8.62 (6.61, 12.40) | 8.31 (7.01, 11.35) | 0.096 | 8.59 (6.56, 12.32) | 8.96 (7.07, 11.88) | 0.003 |
| LYM, ×109/L | 3.64 (2.12, 6.07) | 2.90 (2.16, 4.85) | 0.245 | 3.59 (2.12, 6.06) | 3.70 (2.34, 6.19) | 0.002 |
| PLT, ×109/L | 370.00 (281.00, 453.00) | 324.00 (282.75, 406.25) | 0.189 | 367.00 (281.00, 452.50) | 336.00 (293.00, 450.75) | 0.019 |
| Hb, g/L | 118.00 (110.00, 125.00) | 120.00 (113.00, 125.00) | 0.170 | 118.00 (110.00, 125.00) | 119.00 (112.00, 124.00) | 0.038 |
| Alb, g/L | 43.00 (40.90, 45.50) | 43.80 (42.00, 45.80) | 0.155 | 43.00 (40.95, 45.50) | 43.40 (41.32, 45.10) | 0.052 |
| ALT, IU/L | 15.00 (11.00, 22.00) | 13.00 (10.00, 19.00) | 0.051 | 15.00 (11.00, 22.00) | 14.00 (11.00, 22.00) | 0.028 |
| PT, s | 11.30 (10.70, 12.00) | 11.40 (10.80, 12.00) | 0.040 | 11.30 (10.70, 12.00) | 11.40 (10.70, 12.00) | 0.05 |
| Scr, µmol/L | 27.00 (22.00, 34.00) | 28.00 (22.00, 34.00) | 0.011 | 27.00 (22.00, 34.00) | 27.00 (22.00, 34.00) | 0.004 |
| Chest radiological examination (n, %) | 73 (39.46) | 696 (52.57) | 0.265 | 72 (39.56) | 144 (42.60) | 0.062 |
| Microbiological results by CMTs (n, %) | ||||||
| Bacteria | 5 (2.70) | 45 (3.40) | 0.04 | 5 (2.75) | 7 (2.07) | 0.044 |
| Virus | 59 (31.89) | 691 (52.19) | 0.42 | 59 (32.42) | 112 (33.14) | 0.015 |
| Fungi | 0 (0.00) | 4 (0.30) | 0.078 | 0 (0.00) | 0 (0.00) | < 0.001 |
| Atypical pathogens | 35 (18.92) | 331 (25.00) | 0.147 | 35 (19.23) | 63 (18.64) | 0.015 |
| Antimicrobial treatment (n, %) | ||||||
| Antibiotics | 165 (89.19) | 1084 (81.87) | 0.209 | 162 (89.01) | 290 (85.80) | 0.097 |
| Antivirals | 33 (17.84) | 391 (29.53) | 0.278 | 33 (18.13) | 63 (18.64) | 0.013 |
| Antifungals | 7 (3.78) | 21 (1.59) | 0.136 | 6 (3.30) | 10 (2.96) | 0.019 |
| Oxygen administration (n, %) | 107 (57.84) | 370 (27.95) | 0.634 | 104 (57.14) | 184 (54.44) | 0.054 |
| Admission year (n, %) | 0.304 | 0.011 | ||||
| 2023 | 115 (62.16) | 625 (47.21) | 119 (62.09) | 208 (61.54) | ||
| 2024 | 70 (37.84) | 699 (52.79) | 69 (37.91) | 130 (38.46) | ||
Abbreviations: tNGS targeted Next-Generation Sequencing, SMD standard mean difference, IQR interquartile range, WBC white blood cell, LYM lymphocyte, PLT platelet, Hb Hemoglobin, Alb albumin, ALT alanine aminotransferase, PT prothrombin time, Scr serum creatinine, CMT conventional microbiological tests
a SMD less than 0.1 indicates good balance between 2 groups
Primary outcomes
During follow-up, 75 of 182 (41.21%) patients in the tNGS group and 48 of 338 (14.20%) patients in the control group received pathogen-directed treatment (Fig. 2), with a difference in absolute risk of 27.01% (95% CI: 18.95%, 35.07%) and an OR of 4.23 (95% CI: 2.78, 6.51) (Table 2). Pathogens detected by means of tNGS and/or CMTs in patients with pathogen-directed treatment are shown in Table S10. Pathogen-directed treatment resulted in changes in antibiotic prescriptions, with de-escalation in 22 (12.09%) patients in the tNGS group and 19 (5.62%) patients in the control group (RRR = 3.14, 95% CI: 1.63, 6.03), and escalation in 12 (6.59%) patients in the tNGS group and 4 (1.18%) patients in the control group (RRR = 8.13, 95% CI: 2.57, 25.76) (Table 2, Table S11). However, sole antibiotic adjustment was relatively rare and comparable between the two groups (RRR = 2.71, 95% CI: 0.17, 43.72). In addition, patients in the tNGS group were more likely to initiate antifungal treatment with prior empirical antibiotic treatment (RRR = 13.55, 95% CI: 2.92, 62.85) and pathogen-directed antimicrobial treatment without prior empirical antibiotic treatment (RRR = 5.42, 95% CI: 1.60, 18.37), but not for antiviral treatment with prior empirical antibiotic treatment (RRR = 2.71, 95% CI: 0.93, 7.91), compared with their counterparts in the control group. Continuation of appropriate empirical treatment occurred in 15 of 182 patients (8.24%) in the tNGS group and 11 of 338 patients (3.25%) in the control group, resulting in an RRR of 3.70 (95% CI: 1.65, 8.30).
Fig. 2.
Competing risk cumulative incidence in match cohort. Aalen-Johansen cumulative probability of provision of pathogen-directed treatment among tNGS vs. control
Table 2.
The association of tNGS with antimicrobial stewardship
| tNGS, n (%) (n = 182) |
Control, n (%) (n = 338) |
tNGS vs. Control | ||
|---|---|---|---|---|
| Difference, % (95% CI) a | Ratio (95% CI) b | |||
| Provision of pathogen-directed treatment | 75 (41.21) | 48 (14.20) | 27.01 (18.95, 35.07) | OR: 4.23 (2.78, 6.51) |
| Antibiotic de-escalation | 22 (12.09) | 19 (5.62) | NA | RRR: 3.14 (1.63, 6.03) |
| Antibiotic escalation | 12 (6.59) | 4 (1.18) | NA | RRR: 8.13 (2.57, 25.76) |
| Antibiotic adjustment | 1 (0.55) | 1 (0.30) | NA | RRR: 2.71 (0.17, 43.72) |
| Initiated antiviral treatment with prior empirical antibiotic treatment | 7 (3.85) | 7 (2.07) | NA | RRR: 2.71 (0.93, 7.91) |
| Initiated antifungal treatment with prior empirical antibiotic treatment | 10 (5.49) | 2 (0.59) | NA | RRR: 13.55 (2.92, 62.85) |
| Initiated pathogen-directed antimicrobial treatment, without prior empirical antibiotics | 8 (4.40) | 4 (1.18) | NA | RRR: 5.42 (1.60, 18.37) |
| Continuation of appropriate empirical treatment | 15 (8.24) | 11 (3.25) | NA | RRR: 3.70 (1.65, 8.30) |
| Time to provision of pathogen-directed treatment, mean (95% CI), hours | 58.99 (52.80, 65.18) | 82.95 (79.56, 86.34) | −23.96 (−31.02,−16.90) | HR: 3.55 (2.47, 5.09) |
Abbreviations: tNGS targeted Next-Generation Sequencing, CI confidence interval, OR odds ratio, RRR relative risk ratio, HR hazard ratio, NA not applicable
a Differences for the binary outcomes (on use) were estimated using generalized linear model with a binomial probability distribution and an identity link function and differences for the event-time outcomes were estimated using restricted mean survival time methods
b The ORs for binary outcomes, RRRs for multiple categorical outcomes, and HRs for event-time outcomes were estimated using binary logistic regression models, multinomial logistic regression models, and Fine-Gray competing risk models, respectively
The mean time to provision of pathogen-directed treatment within 96 hours was 58.99 (95% CI: 52.80, 65.18) hours in the tNGS group and 82.95 (95% CI: 79.56, 86.34) hours in the control group (mean difference,−23.96 hours; 95% CI: −31.02, −16.90). The corresponding HR for the tNGS group compared with the control group was 3.55 (95% CI: 2.47, 5.09).
Secondary outcomes
Provision of any antibiotics, provision of intravenous antibiotics, provision of broad-spectrum antibiotics, provision of any antivirals, and provision of non-pathogen-directed antivirals were comparable between the two groups (Table 3). However, patients in the tNGS group had a lower risk of provision of non-pathogen-directed antibiotics (absolute difference=−12.47%, 95% CI: −20.11%, −4.83%; OR = 0.52, 95% CI: 0.33, 0.79). Notably, although the tNGS group compared with the control group had a higher proportion of antifungal provision (absolute difference = 4.95%, 95% CI: 0.35%, 9.54%; OR = 2.41, 95% CI: 1.13, 5.21), the proportion of non-pathogen-directed antifungal provision was comparable between the two groups. For secondary clinical outcomes of PICU admission, oxygen administration, and time from index time to discharge, no significant differences were found.
Table 3.
The association of tNGS with secondary outcomes
| tNGS, n (%) (n = 182) |
Control, n (%) (n = 338) |
tNGS vs. Control | ||
|---|---|---|---|---|
| Difference, % (95% CI) a | OR (95% CI) b | |||
| Antimicrobial stewardship | - | |||
| Provision of any antibiotics | 84 (46.15) | 145 (42.90) | 3.25 (−5.71, 12.22) | 1.14 (0.79, 1.64) |
| Provision of intravenous antibiotics | 21 (11.54) | 38 (11.24) | 0.30 (−5.44, 6.03) | 1.03 (0.58, 1.80) |
| Provision of broad-spectrum antibiotics | 16 (8.79) | 33 (9.76) | −0.97 (−6.16, 4.22) | 0.89 (0.47, 1.64) |
| Provision of non-pathogen-directed antibiotics | 36 (19.78) | 109 (32.25) | −12.47 (−20.11, −4.83) | 0.52 (0.33, 0.79) |
| Provision of any antivirals | 15 (8.24) | 22 (6.51) | 1.73 (−3.05, 6.52) | 1.29 (0.64, 2.53) |
| Provision of non-pathogen-directed antivirals | 13 (7.14) | 20 (5.92) | 1.23 (−3.28, 5.73) | 1.22 (0.58, 2.50) |
| Provision of any antifungals | 16 (8.79) | 13 (3.85) | 4.95 (0.35, 9.54) | 2.41 (1.13, 5.21) |
| Provision of non-pathogen-directed antifungals | 5 (2.75) | 8 (2.37) | 0.38 (−2.49, 3.26) | 1.17 (0.35, 3.55) |
| Clinical outcomes | ||||
| PICU admission | 2 (1.10) | 0 (0.00) | NA | NA |
| Oxygen administration | 38 (20.88) | 89 (26.33) | −5.45(−13, 2.09) | 0.74 (0.48, 1.13) |
| Days with oxygen administration, median (IQR), days | 0 (0, 0) | 0 (0, 1) | 0 (0, 0) | NA |
| Time from index time to discharge, median (IQR), hours | 36.39 (12.53, 60.92) | 36.60 (12.70, 60.77) | 0.21 (−0.67, 23.24) | NA |
Abbreviations: tNGS targeted Next-Generation Sequencing, OR odds ratio, CI confidence interval, PICU pediatric intensive care unit, IQR interquartile range, NA not applicable
a Differences for the binary outcomes (on use) were estimated using generalized linear model with a binomial probability distribution and an identity link function, and difference for continuous outcomes were estimated using bootstrap method with 1000 times resampling
b The ORs for binary outcomes were estimated using binary logistic regression models
Sensitivity analyses outcomes
In sensitivity analyses excluding participants with at least one missing covariate, the results were consistent in that the magnitude and direction of point estimates with those in the primary analyses (Table S12). When excluding participants who received tNGS testing during the follow-up period, defining the index time as 00:00 on day 3, matching for specific CMTs performed during the exposure period, or excluding participants with a disease duration > 7 days prior to admission, the associations between tNGS and provision of pathogen-directed treatment became more pronounced, with ORs increased to 5.32 (95% CI: 3.42, 8.38), 5.76 (95% CI: 4.14, 8.08), 5.28 (95% CI: 3.40, 8.34), and 6.07 (95% CI: 3.21, 11.85), and HRs increased to 4.29 (95% CI: 2.91, 6.34), 4.20 (95% CI: 3.23, 5.46), 4.37 (95% CI: 2.97, 6.42), and 4.82 (95% CI: 2.76, 8.41), respectively.
Discussion
In this retrospective-matched cohort study of 520 pediatric patients with LRTIs, we found that integrating tNGS into standard pathogen diagnostic workflows was associated with enhanced antimicrobial stewardship through increased the probability of and reduced the time to provision of pathogen-directed treatment, as well as reduced non-pathogen-directed antibiotic use compared with the CMTs alone. However, the addition of tNGS was not associated with short-term clinical outcomes.
Multiple observational studies have shown that tNGS offers improved diagnostic performance compared with CMTs for detecting pulmonary pathogens [7–11]. Consistent with these findings, our study demonstrated a significantly higher pathogen detection rate in the tNGS group compared to the control group using CMTs only. However, improved detection alone does not ensure clinical benefit, as it may not translate into changes in patient management. Recent studies have begun to assess the clinical relevance of tNGS. For example, Li et al. reported treatment modifications in 33 of 336 pediatric patients after receiving tNGS results [10]. Similarly, Qin et al. found that tNGS prompted empirical antibiotic escalation in 35.6% and de-escalation in 8.9% of patients [12]. These studies provide early signals of clinical impact but are limited by before-after designs and absence of control groups. In the present study, we included patients who underwent both tNGS and control groups and used PS matching to minimize indication bias [28]. Although differences in exclusion due to early discharge or initiation of targeted therapy were a concern, exclusion rates were similar across groups (72.4% vs. 72.2%), suggesting minimal selection bias. We found that addition of tNGS to CMTs was associated with a 3.23-fold increase in the likelihood of receiving pathogen-directed therapy and a nearly 24-hour reduction in time to treatment initiation. These findings not only reinforce the diagnostic advantages of tNGS but also demonstrate a measurable impact on clinical decision-making.
Because tNGS is primarily used as a targeted, second-line test for a small subset of complex cases [17], rather than as a first-line diagnostic tool, the proportion of the total hospitalized cohort who received pathogen-directed therapy based on its results is modest. Consequently, its overall impact on antimicrobial stewardship within the broader pediatric LRTI population remains limited. However, recent advancements have led to shorter turnaround times and reduced costs, thereby improving its feasibility for wider clinical adoption [7]. This could enhance its potential contribution to antimicrobial stewardship on a larger scale, a possibility that warrants further investigation.
This study also identified distinct patterns of antimicrobial modification associated with tNGS use. Patients in the tNGS group had significantly higher rates of both antibiotic de-escalation and escalation, suggesting that tNGS may support more precise antimicrobial decision-making. Notably, the study period overlapped with a resurgence of Mycoplasma pneumoniae infections in China [29], with about one third of participants testing positive. The ability of tNGS to detect Mycoplasma pneumoniae resistance genes [30] may have facilitated targeted therapy, particularly for macrolide-resistant strains, contributing to changes in treatment patterns. Although overall antibiotic use was similar between groups, patients in the tNGS group had lower rates of non–pathogen-directed antibiotic use, indicating a potential role for tNGS in minimizing inappropriate antimicrobial exposure—a key goal in antimicrobial stewardship. In addition, the tNGS group demonstrated a higher rate of pathogen-directed antifungal therapy initiation following empirical antibiotic treatment. This finding suggested improved identification of fungal pathogens that may be missed by conventional testing and is consistent with previous reports supporting the utility of tNGS in diagnosing fungal infections [31, 32].
To our knowledge, only 2 studies have investigated the association between tNGS use and clinical outcomes in patients with LRTI [12, 33]. Qin et al. reported that most patients showed clinical improvement following tNGS testing, including defervescence, resolution of respiratory symptoms, and reductions in inflammatory markers, regardless of whether antibiotic therapy was adjusted [12]. Similarly, an ICU–based study found that 123 of 142 patients (86.6%) showed clinical benefit from antibiotic treatment, defined as improvement or stabilization of infection-related symptoms, with higher benefit observed among patients in whom tNGS identified causative or possibly causative pathogens (113 of 123 [91.9%]). However, neither study included a comparator group receiving CMTs alone; thus, it remains uncertain whether the observed benefits were attributable to tNGS use. In the present study, we found no significant differences in short-term clinical outcomes between the tNGS and control groups. One possible explanation is the relatively brief 96-hour follow-up period, which may have limited the ability to detect downstream effects. Longer-term studies are needed to more fully assess the clinical impact of tNGS and its potential role in improving patient outcomes.
This study has several notable strengths. First, the evaluation of tNGS as a second-line diagnostic tool reflects real-world clinical practice, enhancing the relevance of the findings. Second, the use of comprehensive clinical data and PS matching allowed for the creation of comparable groups and reduced the potential for indication bias. Third, the study focused on assessing the clinical utility of tNGS beyond diagnostic performance, offering insights into its impact on patient management. However, several limitations should be acknowledged. As a retrospective observational analysis, the possibility of residual confounding from unmeasured variables cannot be excluded. Specifically, data on vaccination status against pneumococcus and influenza were not systematically available, which could potentially influence pathogen distribution and outcomes. However, the overall low coverage of these vaccines in the study region, along with the inclusion of calendar year in our PS matching, likely mitigates some of these concerns. The single-center design, while ensuring consistency in clinical protocols and laboratory practices, may limit the generalizability of findings to other pediatric populations or health care settings with differing tNGS platforms or workflows. In addition, the limited sample size precluded subgroup analyses, such as comparisons between different specimen types (e.g., BALF vs. sputum) or among patients with varying clinical characteristics.
Conclusions
In this retrospective cohort study, use of tNGS as a second-line diagnostic test was associated with more timely and targeted antimicrobial therapy among hospitalized children with LRTI. These findings support current consensus that recommend tNGS as a complementary tool to CMTs. However, randomized clinical trials are needed to validate these results and to assess the long-term clinical outcomes associated with tNGS implementation.
Supplementary Information
Acknowledgements
We would like to thank all staff of the Department of Pediatrics and Department of Clinical Laboratory, Tongji Hospital, Tongji Medical College, Huazhong University of Science and Technology, KingMed Diagnostics, and HUGO Biotech for data collection.
Role of the Funder/Sponsor
The funders had no role in the design and conduct of the study; collection, management, analysis, and interpretation of the data; preparation, review, or approval of the manuscript; and decision to submit the manuscript for publication.
Authors’ contributions
Yongjian Huang and Xiaoping Luo are guarantors of this work and, as such, had full access to all the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis. Zongming Yang, Dong Xu, and Xiuyun Zhou contributed equally to the work as co-first authors. Concept and design: Zongmin Yang, Yongjian Huang, and Xiaoping Luo. Acquisition, analysis, or interpretation of data: All authors. Drafting of the manuscript: Zongming Yang and Yongjian Huang. Critical revision of the manuscript for important intellectual content: All authors. Statistical analysis: Zongming Yang, Dong Xu, and Xiuyun Zhou. Obtained funding: Xiaoping Luo. Administrative, technical, or material support: Wankai Xue, Wenjing Zhang, Xiaopei Cao, Rui Shi, and Lin He. Supervision: Yongjian Huang and Xiaoping Luo.
Funding
This study was supported by the National Key R&D Program of China (2023YFC2706300).
Data availability
The data that support the findings of this study are not openly available due to reasons of sensitivity and are available from the corresponding author upon reasonable request.
Declarations
Ethics approval and consent to participate
The study was approved by the Ethics Committee of Tongji Hospital (No.TJ-IRB202506028). Informed consent was waived due to the use of anonymous and de-identified secondary data.
Consent for publication
Not applicable. All patient data used in this manuscript have been thoroughly anonymized to ensure that no individual can be identified.
Competing interests
The authors declare no competing interests.
Footnotes
Publisher’s Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Zongming Yang, Dong Xu and Xiuyun Zhou contributed equally to this work.
Contributor Information
Xiaoping Luo, Email: xpluo@tjh.tjmu.edu.cn.
Yongjian Huang, Email: yjhuang003@163.com.
References
- 1.Global regional. National incidence and mortality burden of non-COVID-19 lower respiratory infections and aetiologies, 1990–2021: a systematic analysis from the global burden of disease study 2021. Lancet Infect Dis Sep. 2024;24(9):974–1002. 10.1016/s1473-3099(24)00176-2. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 2.Bassetti S, Tschudin-Sutter S, Egli A, Osthoff M. Optimizing antibiotic therapies to reduce the risk of bacterial resistance. Eur J Intern Med May. 2022;99:7–12. 10.1016/j.ejim.2022.01.029. [DOI] [PubMed] [Google Scholar]
- 3.Caliendo AM, Gilbert DN, Ginocchio CC, et al. Better tests, better care: improved diagnostics for infectious diseases. Clin Infect Dis Dec. 2013;57(Suppl 3):S139–70. 10.1093/cid/cit578. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 4.Dong Y, Chen Q, Tian B, Li J, Li J, Hu Z. Advancing microbe detection for lower respiratory tract infection diagnosis and management with metagenomic Next-Generation sequencing. Infect Drug Resist. 2023;16:677–94. 10.2147/idr.S387134. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 5.Besser J, Carleton HA, Gerner-Smidt P, Lindsey RL, Trees E. Next-generation sequencing technologies and their application to the study and control of bacterial infections. Clin Microbiol Infect Apr. 2018;24(4):335–41. 10.1016/j.cmi.2017.10.013. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 6.Chen Q, Yi J, Liu Y, et al. Clinical diagnostic value of targeted next–generation sequencing for infectious diseases (Review). Mol Med Rep Sep. 2024;30(3). 10.3892/mmr.2024.13277. [DOI] [PubMed]
- 7.Yin Y, Zhu P, Guo Y, et al. Enhancing lower respiratory tract infection diagnosis: implementation and clinical assessment of multiplex PCR-based and hybrid capture-based targeted next-generation sequencing. EBioMedicine Sep. 2024;107:105307. 10.1016/j.ebiom.2024.105307. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 8.Da H, Meng T, Xu Y. Application of targeted next-generation sequencing for detecting respiratory pathogens in the sputum of patients with pulmonary infections. Infect Genet Evol Mar. 2025;128:105722. 10.1016/j.meegid.2025.105722. [DOI] [PubMed] [Google Scholar]
- 9.He S, Xue W, Wu X, et al. The application value of targeted next-generation sequencing using Bronchoalveolar lavage fluid samples in community-acquired pneumonia in children. J Infect Chemother Mar. 2025;31(3):102610. 10.1016/j.jiac.2025.102610. [DOI] [PubMed] [Google Scholar]
- 10.Li C, Zhang X, Liu P, et al. Targeted next-generation sequencing characterization of respiratory pathogens in children with acute respiratory infection. BMC Infect Dis May. 2025;8(1):675. 10.1186/s12879-025-11051-w. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 11.Zhang P, Liu B, Zhang S, et al. Clinical application of targeted next-generation sequencing in severe pneumonia: a retrospective review. Crit Care Jul. 2024;8(1):225. 10.1186/s13054-024-05009-8. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 12.Qin L, Liang M, Song J, et al. Utilizing targeted Next-Generation sequencing for Rapid, Accurate, and Cost-Effective pathogen detection in lower respiratory tract infections. Infect Drug Resist. 2025;18:329–40. 10.2147/idr.S494558. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 13.Wang L, Zhong H, Yang L, Yang X. Clinical utility and outcomes of targeted Next-Generation sequencing in pediatric respiratory tract infections: A Nine-Month retrospective study at a North Chinese hospital. MedRxiv. 2024. 2024.09.18.24313214.
- 14.Shah SN, Bachur RG, Simel DL, Neuman MI. Does this child have pneumonia? The rational clinical examination systematic review. JAMA. 2017;318(5):462–71. 10.1001/jama.2017.9039. [DOI] [PubMed] [Google Scholar]
- 15.Rambaud-Althaus C, Althaus F, Genton B, D’Acremont V. Clinical features for diagnosis of pneumonia in children younger than 5 years: a systematic review and meta-analysis. Lancet Infect Dis. 2015;15(4):439–50. 10.1016/S1473-3099(15)70017-4. [DOI] [PubMed] [Google Scholar]
- 16.[Guidelines for the management of community-acquired pneumonia in children. (2024 revision)]. Zhonghua Er Ke Za Zhi. 2024;62(10):920–930. 10.3760/cma.j.cn112140-20240728-00523 [DOI] [PubMed]
- 17.[Expert consensus on the. Application and practice of targeted next-generation sequencing in infectious diseases]. Zhonghua Yi Xue Za Zhi Dec. 2024;24(48):4375–83. 10.3760/cma.j.cn112137-20240927-02208. [DOI] [PubMed] [Google Scholar]
- 18.Grant LR, Hammitt LL, Murdoch DR, O’Brien KL, Scott JA. Procedures for collection of induced sputum specimens from children. Clin Infect Dis Apr. 2012;54(Suppl 2):S140–5. 10.1093/cid/cir1069. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 19.Bartlett JG, Breiman RF, Mandell LA, File TM. Jr. Community-acquired pneumonia in adults: guidelines for management. The infectious diseases society of America. Clin Infect Dis Apr. 1998;26(4):811–38. 10.1086/513953. [DOI] [PubMed] [Google Scholar]
- 20.[Guidelines for the management of community-acquired pneumonia in children. (2024 revision)]. Zhonghua Er Ke Za Zhi. Oct 2 2024;62(10):920–930. 10.3760/cma.j.cn112140-20240728-00523 [DOI] [PubMed]
- 21.Dong Y, Peng CY. Principled missing data methods for researchers. Springerplus Dec. 2013;2(1):222. 10.1186/2193-1801-2-222. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 22.Markussen DL, Serigstad S, Ritz C, et al. Diagnostic stewardship in Community-Acquired pneumonia with syndromic molecular testing: A randomized clinical trial. JAMA Netw Open Mar. 2024;4(3):e240830. 10.1001/jamanetworkopen.2024.0830. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 23.Gerber JS, Hersh AL, Kronman MP, Newland JG, Ross RK, Metjian TA. Development and application of an antibiotic spectrum index for benchmarking antibiotic selection patterns across hospitals. Infect Control Hosp Epidemiol Aug. 2017;38(8):993–7. 10.1017/ice.2017.94. [DOI] [PubMed] [Google Scholar]
- 24.Austin PC. Balance diagnostics for comparing the distribution of baseline covariates between treatment groups in propensity-score matched samples. Stat Med Nov. 2009;10(25):3083–107. 10.1002/sim.3697. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 25.Yorlets RR, Lee Y, Gantenberg JR. Calculating risk and prevalence ratios and differences in R: developing intuition with a hands-on tutorial and code. Ann Epidemiol. 2024;99:48–53. 10.1016/j.annepidem.2024.10.004. 2024/11/01. [DOI] [PubMed] [Google Scholar]
- 26.Kloecker DE, Davies MJ, Khunti K, Zaccardi F. Uses and limitations of the restricted mean survival time: illustrative examples from cardiovascular outcomes and mortality trials in type 2 diabetes. Ann Intern Med Apr. 2020;21(8):541–52. 10.7326/m19-3286. [DOI] [PubMed] [Google Scholar]
- 27.Li J, Scheike TH, Zhang MJ. Checking fine and Gray subdistribution hazards model with cumulative sums of residuals. Lifetime Data Anal Apr. 2015;21(2):197–217. 10.1007/s10985-014-9313-9. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 28.Joseph KS, Mehrabadi A, Lisonkova S. Confounding by Indication and Related Concepts. Current Epidemiology Reports. 2014/03/01 2014;1(1):1–8. 10.1007/s40471-013-0004-y
- 29.Yan C, Xue GH, Zhao HQ, Feng YL, Cui JH, Yuan J. Current status of Mycoplasma pneumoniae infection in China. World J Pediatr Jan. 2024;20(1):1–4. 10.1007/s12519-023-00783-x. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 30.Dekyi XY, Wang X, et al. Predominance of A2063G mutant strains in the Mycoplasma pneumoniae epidemic in children: A clinical and epidemiological study in 2023 in Wuhan, China. Int J Infect Dis. 2024;2024/08/01:145:107074. 10.1016/j.ijid.2024.107074. [DOI] [PubMed] [Google Scholar]
- 31.Larkin PMK, Lawson KL, Contreras DA et al. Amplicon-Based Next-Generation Sequencing for Detection of Fungi in Formalin-Fixed, Paraffin-Embedded Tissues: Correlation with Histopathology and Clinical Applications. The Journal of Molecular Diagnostics. 2020/10/01/ 2020;22(10):1287–1293. 10.1016/j.jmoldx.2020.06.017 [DOI] [PubMed]
- 32.Naik S, Kashyap D, Deep J, et al. Utilizing Next-Generation sequencing: advancements in the diagnosis of fungal infections. Diagnostics. 2024;14(15):1664. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 33.Chen C, Wang R, Wei B, et al. Performance of broad-spectrum targeted next-generation sequencing in lower respiratory tract infections in ICU patients: a prospective observational study. Crit Care Jun. 2025;4(1):226. 10.1186/s13054-025-05470-z. [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
Data Availability Statement
The data that support the findings of this study are not openly available due to reasons of sensitivity and are available from the corresponding author upon reasonable request.