Independence of Predictors for Aspiration Pneumonia in Long‐Term Care Facilities

We read with great interest the nationwide facility-level study by Wu et al. examining factors associated with aspiration pneumonia (AP) in Japanese geriatric health service facilities [1]. The authors reported that a history of AP (OR = 45.14) and the occurrence of aspiration events (OR = 9.28) were strongly associated with current AP, and that the absence of occupational therapists (OTs) also increased risk. While the study addresses an important clinical issue, we wish to highlight two closely related methodological concerns—circularity/surveillance bias and reverse causation—that together challenge the validity of causal interpretation of the facility-level findings. First, the strong association between “aspiration events” and AP suffers from circularity and surveillance bias. In the study, both aspiration events and AP were measured over the same preceding 3-month period using facility-reported data. However, aspiration is not only a risk factor but also a core diagnostic element of AP—witnessed or strongly suspected aspiration is part of the definition adopted by the authors (Japanese Respiratory Society guidelines). When both variables are collected from the same time window and the same population, the observed OR of 9.28 may largely reflect overlapping case definitions rather than a true temporal or causal relationship. Facilities with more rigorous surveillance of mealtime choking are also more likely to label a pneumonia episode as aspiration-related, creating an artificial inflation known as surveillance bias or differential misclassification [2, 3]. This inflation can substantially bias the odds ratio upward, making a non-causal association appear strong. Although the authors acknowledged this possibility in the discussion, no sensitivity analysis (e.g., excluding facilities where aspiration and AP counts were identical, or using a clear temporal lag) was performed to quantify its impact. Without such steps, the magnitude of this predictor should be interpreted with caution. Second, the variable “history of AP” (occurring > 3 months before the survey) is subject to reverse causation and persistent confounding by unmeasured resident characteristics. In a cross-sectional design, a prior AP episode may reflect not facility-level care quality but, more plausibly, a high baseline risk of residents (e.g., chronic dysphagia, neurodegenerative disease, poor oral health) that persists over time. These factors were adjusted for only by the mean care-need level, a facility-level aggregate measure that cannot capture individual variation in swallowing function, cognitive status, or oral hygiene. Moreover, “history of AP” and “current AP” are often different episodes of the same recurrent condition [4, 5]; including the former as a predictor essentially models recurrence, which tells us little about modifiable facility practices. The enormous odds ratio (45.14) with a wide confidence interval (16.94–120.29) further suggests model instability and low precision, possibly due to sparse data or near-complete separation [6]. These two issues are not independent—they both stem from using predictors that are either non-independent (aspiration events overlapping with AP definition) or time-confounded (history of AP predicting current AP) in a cross-sectional model. Consequently, the final multivariable model (see tab. 1, model 2, reproduced from [1]) shows that after adding “history of AP” and “aspiration events,” facility type ceased to be significant, suggesting that the two problematic predictors absorbed most of the explained variance (Table 1). The association with OTs (OR = 4.88) might therefore represent residual confounding rather than a genuine protective effect. A more valid approach would have been to (a) use AP incidence density (events per 1000 resident-days) rather than binary occurrence, (b) exclude aspiration events that were part of the AP definition, (c) perform sensitivity analyses excluding facilities where aspiration and AP counts were identical, and (d) apply longitudinal designs with clear temporal separation between predictors and outcomes [7, 8]. We agree that OTs may play a supportive role in dysphagia care, but the current evidence from this low-response (12.8%) cross-sectional survey is insufficient to infer causality. Our aim is not to dismiss the valuable work by Wu et al. but to highlight the need for cautious interpretation and more rigorous future study designs. Studies with harmonized definitions, prospective follow-up, and explicit adjustment for surveillance intensity are urgently needed. Y.S. and X.L. conceived the letter concept, performed the literature analysis, and drafted the manuscript. Q.Y., H.D., and Y.P. critically revised the manuscript for important intellectual content and provided supervision. All authors approved the final version and agree to be accountable for all aspects of the work. The authors have nothing to report. The authors have nothing to report. The authors have nothing to report. The authors have nothing to report. The authors declare no conflicts of interest. Data sharing is not applicable to this article as no new data were created or analyzed in this correspondence.

Causation · Choking · Guideline · Odds · Odds ratio · Pneumonia · Dysphagia Assessment and Management · Nosocomial Infections in ICU · Nutrition and Health in Aging

• Facility‐Level Factors Associated With Aspiration Pneumonia in Japanese Geriatric Health Service Settings

  Open Access•Xinze Wu, Tatsuma Okazaki et al.•Geriatrics and gerontology…•2026

  Add to list