In this systematic review and meta-analysis of 68 studies comprising more than 10,000 hospitalised patients with microbiologically confirmed scrub typhus, ARDS occurred in 13% of patients across all age groups, with a higher pooled proportion among adults (15%) than paediatric patients (8%). Mortality among patients with scrub typhus-associated ARDS was 21% overall, 28% among adults, and 16% among paediatric patients. In the direct within-study comparison, patients who developed ARDS had an approximately 13-fold higher observed risk of death than those without ARDS (RR 12.75, 95% CI 8.48–19.17). Reported factors associated with ARDS varied between studies but included delayed presentation or treatment, eschar, hepatomegaly, thrombocytopenia, hypotension, respiratory symptoms, metabolic acidosis, and hypoalbuminaemia. Limited available data also suggested that respiratory failure or ARDS commonly developed several days into illness, often during or after the second week. These associations were inconsistently reported and should be considered hypothesis-generating.

The mortality comparison should be interpreted with caution because it was derived from unadjusted aggregate data. ARDS in scrub typhus is likely to occur in patients with more severe systemic illness, including shock, acute kidney injury, hepatic dysfunction, thrombocytopenia, altered sensorium, delayed presentation, or need for ICU-level care. The higher mortality among patients with ARDS may therefore partly reflect confounding by baseline severity, comorbidity burden, referral patterns, ICU admission status, organ dysfunction, and access to respiratory support, rather than the independent effect of ARDS alone. In this context, ARDS should be viewed as a clinically important marker of severe scrub typhus and high mortality risk. The higher ARDS proportion and ARDS-specific mortality observed among adults may also reflect differences in comorbidity burden, physiological reserve, referral patterns, thresholds for hospitalisation, and access to intensive care. These subgroup findings should be interpreted cautiously because they were derived from study-level rather than individual-patient data, and adult and paediatric studies differed in ARDS definitions, severity thresholds, and outcome reporting.

Orientia tsutsugamushi primarily infects endothelial cells, leading to vasculitis, increased vascular permeability, and tissue injury [13, 27]. This underlies the capillary leak, pulmonary inflammation, and multi-organ involvement seen in severe disease [1, 28, 29]. Systemic inflammatory mediators, including interleukin-8 (IL-8), IL-10, macrophage inflammatory protein-1 alpha (MIP-1α), and tumour necrosis factor-alpha (TNF-α), have been associated with disease severity and mortality in scrub typhus [30]. IL-8 is also a recognised marker of hyperinflammatory ARDS subphenotypes in other settings [31]. A dysregulated host inflammatory response may therefore partly explain differences in ARDS severity and outcomes across scrub typhus cohorts. However, this remains hypothesis-generating given the absence of patient-level cytokine or sub-phenotype data in the included studies [32].

The pooled ARDS-specific mortality of 21% in this review of hospitalised scrub typhus patients is lower than the approximately 35–45% reported previously in cohorts of all-cause ARDS [33, 34]. However, this comparison should be interpreted cautiously, as scrub typhus-associated ARDS may differ from all-cause ARDS in patient age, comorbidity burden, reversibility of infection, timing of antimicrobial therapy, severity of lung injury, and availability of respiratory support. Early empiric or targeted doxycycline or azithromycin may reduce the burden of the organism and systemic inflammation when scrub typhus is recognised promptly [35]. Nevertheless, once ARDS develops, particularly in association with shock or multi-organ dysfunction, mortality remains substantial. Therefore, timely diagnosis, early appropriate anti-rickettsial therapy, and escalation of respiratory or intensive care support remain clinically important, although their independent effect on ARDS-specific mortality is not established by a large body of direct evidence and requires prospective evaluation. The newer global ARDS definition, which extends recognition to non-intubated patients receiving high-flow nasal oxygen, may facilitate earlier identification of at-risk patients before progression to severe respiratory failure [36]. This is particularly relevant because early pulmonary involvement in scrub typhus may present non-specifically as pneumonitis, pleural effusion, or hypoxaemia.

The lower heterogeneity observed for ARDS-specific mortality in paediatric studies compared with adult studies may reflect a more uniform underlying population and fewer competing mortality risks. Paediatric cohorts generally include fewer chronic comorbidities than adult cohorts, and deaths may be more closely related to acute illness severity and organ dysfunction rather than to background cardiopulmonary, renal, hepatic, or metabolic disease. In addition, paediatric studies may have had more similar thresholds for hospitalisation or PICU referral within the available literature. However, this interpretation remains cautious because the number of paediatric ARDS deaths was small, ARDS definitions were not standardised, and paediatric studies also differed in diagnostic methods, supportive-care capacity, and outcome reporting. The lower pooled ARDS proportion among paediatric patients should also be interpreted in the context of variable and often non-standardised paediatric ARDS ascertainment. Only a small number of paediatric studies explicitly applied the Paediatric Acute Lung Injury Consensus Conference (PALICC) criteria. Underuse of paediatric-specific criteria may have influenced the pooled estimate in either direction. Some children with hypoxaemia, evolving lung injury, or non-intubated ARDS may not have been captured if oxygenation indices, positive end-expiratory pressure requirements, or serial imaging were not systematically reported. Conversely, studies relying on broad clinical or radiological descriptions may have included children with severe pneumonia, fluid overload, or mixed-aetiology respiratory failure under the ARDS label. Thus, the paediatric pooled estimate should be viewed as an approximate hospital-based burden estimate rather than a definitive measure of paediatric ARDS incidence in scrub typhus.

The substantial heterogeneity across pooled estimates warrants careful interpretation and reflects the diversity of the included evidence base. Studies differed in design, geographic setting, age distribution, illness severity at presentation, hospital level, and timing of referral. ARDS ascertainment varied considerably; only 32 of 68 studies explicitly stated the diagnostic criteria used, with some applying the Berlin definition and others relying on clinical or radiological assessment without specifying oxygenation thresholds, imaging criteria, or exclusion of cardiogenic pulmonary oedema [15]. This inconsistency means that some patients labelled as ARDS may have had severe pneumonia, pulmonary oedema, or mixed-aetiology respiratory failure. At the same time, patients with early or non-intubated ARDS may not have been captured in studies without systematic oxygenation assessment or serial imaging. To address this concern, we conducted a restricted analysis of studies with clearly defined ARDS criteria. The pooled ARDS proportion in this subgroup was similar to the overall pooled estimate, suggesting that the main burden estimate was not driven solely by studies with poorly specified ARDS definitions. Nevertheless, substantial residual heterogeneity persisted in this restricted analysis, reinforcing that the pooled estimate should be interpreted as a hospital-based summary estimate rather than a precise population-level incidence estimate.

The descriptive analysis of studies with extreme ARDS proportions further supports the role of case mix and severity enrichment in explaining heterogeneity. Studies reporting very high ARDS proportions (≥ 40%) were largely restricted to ICU or HDU cohorts or included patients with delayed presentation, shock, multi-organ dysfunction, mechanical ventilation, or other markers of severe disease [12]. In contrast, studies reporting very low ARDS proportions (≤ 3%) generally included broader hospital-based or non-ICU-selected populations, less severely ill patients, or paediatric cohorts. These differences in referral setting, illness severity, timing of presentation, and ARDS ascertainment are likely to have contributed substantially to the very high heterogeneity observed for the pooled ARDS proportion. Separate pooling by disease severity, for example, ICU versus non-ICU cohorts, or patients meeting versus not meeting Sepsis-3 criteria, would have been the most informative approach to address this source of heterogeneity, but was not feasible because severity classification was inconsistently defined and reported across studies, with most not providing sufficient data to categorise patients by a standardised severity threshold [37]. Similarly, subgroup analyses by ARDS definition type, diagnostic method for scrub typhus, or geographic region were limited by the small number of studies in each potential subgroup and the absence of individual patient data. Delayed hospital presentation, delayed initiation of anti-rickettsial therapy, prior antibiotic exposure, and limited access to respiratory support are additional clinical factors likely to have varied across cohorts and contributed to outcome heterogeneity, but were inconsistently reported and therefore not amenable to formal evaluation. The geographic concentration of the evidence base, with most studies from India, limits generalisability to other scrub typhus-endemic regions, including Southeast Asia, the western Pacific, and parts of East Africa, where differences in circulating O. tsutsugamushi strains, healthcare infrastructure, and patient populations may produce different ARDS rates and outcomes. Future studies should use standardised adult and paediatric ARDS definitions and consistently report oxygenation indices, imaging criteria, ventilatory support, illness severity, ICU/HDU status, timing of respiratory deterioration, timing of anti-rickettsial therapy, and outcomes. The pooled estimates should therefore be interpreted as summary estimates from heterogeneous hospital-based studies rather than as precise population-level incidence or mortality figures [1, 29].

This review has several limitations. Substantial between-study heterogeneity was present across most pooled estimates, reflecting differences in study design, referral setting, ICU or HDU recruitment, severity enrichment, age distribution, ARDS definitions, scrub typhus diagnostic methods, timing of presentation, and supportive-care availability; therefore, the pooled estimates should be interpreted as summary estimates from heterogeneous hospital-based studies rather than precise population-level incidence or mortality figures. ARDS ascertainment was inconsistent: although the pooled ARDS proportion was similar when restricted to studies with explicitly reported ARDS criteria, only 32 of 68 studies clearly stated the definition used, and studies without formal criteria may have misclassified severe pneumonia, pulmonary oedema, or mixed-aetiology respiratory failure as ARDS, while early or non-intubated ARDS may have been missed in settings without systematic oxygenation assessment or serial imaging. This limitation is particularly relevant to paediatric studies, in which few explicitly used paediatric-specific ARDS criteria, and the underuse of PALICC criteria may have contributed to under- or over-ascertainment or inconsistent classification of paediatric ARDS. Scrub typhus diagnostic methods also varied; most studies used IgM ELISA, IFA, PCR, or combinations of these methods, but some used Weil-Felix testing as part of the diagnostic approach, which has limited diagnostic performance, particularly poor sensitivity and variable specificity depending on setting and cut-off, and may therefore have introduced diagnostic misclassification. Because the paediatric subgroup included only a limited number of studies and outcome events, we did not perform a separate sensitivity analysis by diagnostic method, as such an analysis would have produced unstable estimates and would not reliably indicate the direction or magnitude of bias. ICU and referral hospital cohorts probably over-represented severe disease and may have overestimated ARDS frequency and mortality, whereas non-ICU cohorts may have under-detected ARDS when oxygenation indices, serial imaging, or structured respiratory monitoring were not routinely performed. The comparison of mortality between patients with and without ARDS was based on unadjusted aggregate study-level data, and most included studies did not consistently identify or adjust for critical confounders such as baseline illness severity, shock, multi-organ dysfunction, comorbidities, ICU admission status, timing of presentation, timing of anti-rickettsial therapy, prior antibiotic exposure, or access to mechanical ventilation and other respiratory support; therefore, the pooled risk ratio should not be interpreted as demonstrating an independent causal effect of ARDS on mortality, but rather as indicating that ARDS identifies a subgroup with substantially higher observed mortality. ARDS-specific mortality was not extractable from all studies, and where available, some estimates were based on very small numbers of ARDS cases, yielding imprecise extreme values of 0% or 100%. Reliance on aggregate published data rather than individual patient data prevented patient-level subgroup analyses and limited exploration of treatment timing, ventilatory management, illness severity, comorbidity burden, inflammatory sub-phenotypes, cytokine profiles, and strain-level variation in Orientia tsutsugamushi. Publication and reporting bias cannot be excluded, as studies with higher ARDS proportions or more severe outcomes may have been preferentially published; exclusion of grey literature may also have contributed to reporting bias. Although preprints were included only when retrieved through the database search and not superseded by a peer-reviewed publication at the time of the search, their inclusion may have introduced additional uncertainty because they had not undergone peer review. Finally, although the revised risk-of-bias assessment used design-specific JBI tools and most studies were classified as moderate to high quality, the evidence base remained limited by observational study designs, variable reporting of ARDS definitions, incomplete reporting of illness severity and ICU admission status, limited adjustment for confounding, and inconsistent reporting of follow-up completeness in some cohort studies. Despite these limitations, this review provides a comprehensive synthesis of available evidence on ARDS in scrub typhus across adult, paediatric, and mixed-age populations.