Derivation of a clinical-radiological scoring system to predict foreign body aspiration in children

Abstract

BACKGROUND

Foreign body aspiration (FBA) is a serious pediatric emergency in which delays in diagnosis and management may cause considerable morbidity. Although rigid bronchoscopy remains the procedure of choice, it is invasive and may be associated with complications.

AIM

To derive a clinical scoring system to predict FBA and aid in the selection of children for bronchoscopy.

METHODS

We performed a retrospective review of all children who underwent bronchoscopy for suspected FBA at our institute between 2013 and 2020. Clinical presentation, key historical features, examination findings, and radiological data were analyzed. Predictors of FBA were identified using regression analysis and weighted using adjusted odds ratios to develop a 17-point clinical-radiological score and an emergency decision-making algorithm.

RESULTS

A total of 274 bronchoscopies were performed, with a negative bronchoscopy rate of 34%. Regression modelling identified seven independent predictors that were incorporated into the 17-point score: Definite choking history (2 points), respiratory distress (1 point), absence of concurrent upper respiratory tract infection (4 points), no history of asthma/reactive airway disease (3 points), unilateral decreased air entry/wheeze (2 points), tachypnea (1 point), and unilateral hyperinflation on chest radiograph (4 points). A cut-off score of 7 yielded a sensitivity of 83% and a specificity of 63% for predicting FBA. Risk stratification showed an increasing probability of FBA with higher scores: ≤ 6 points (24%), 7-9 points (51%), and ≥ 10 points (85%).

CONCLUSION

While a low clinical threshold for bronchoscopy is often maintained in suspected FBA, the high negative bronchoscopy rate underscores the need for an objective tool to minimize unnecessary procedures without delaying intervention in high-risk children. The derived 17-point score demonstrated clinically useful discrimination ability and may support rational bronchoscopy selection.

Key Words: Foreign body aspiration; Bronchoscopy; Scoring system; Airway obstruction; Pediatric emergency; Risk stratification; Respiratory tract disease; Diagnostic prediction

Core Tip: Diagnosing pediatric foreign body aspiration remains challenging and is associated with high negative rigid bronchoscopy rates. We derived a novel 17-point clinical-radiological scoring system that achieved an area under the curve of 0.791 and a sensitivity of 83.4%. This score heavily emphasizes the absence of respiratory mimics (concurrent upper respiratory tract infections or asthma) alongside classic predictors. Accordingly, we propose a practical emergency algorithm in which children scoring ≥ 10 undergo immediate rigid bronchoscopy, while those scoring ≤ 6 may be initially evaluated with flexible bronchoscopy. This approach optimizes patient selection and reduces unnecessary invasive procedures and associated complications.

INTRODUCTION

Foreign body aspiration (FBA) remains a leading cause of accidental death and serious morbidity in preschool children, accounting for up to 5% of all accidental deaths in those under 4 years and a substantial proportion of emergency department visits for acute respiratory distress. The risk stems from the combination of a narrow, immature airway and the tendency of toddlers to place objects in their mouths while playing or eating, which often leads to sudden choking[1].

Diagnosing FBA on clinical grounds remains extremely challenging. A witnessed choking episode is reported in only approximately 70% of cases, and many children present with nonspecific or absent symptoms. In addition, up to 50% of patients have a normal chest radiograph, whereas physical findings such as diminished breath sounds are detected in fewer than half of children with confirmed aspiration[2]. Consequently, reported negative rigid bronchoscopy rates range from 10% to 60%, reflecting unnecessary exposure to general anesthesia and procedural risks[3]. Existing scoring systems have primarily focused on positive clinical and radiological predictors, with limited emphasis on differentiating FBA from common respiratory mimics, such as respiratory infections and reactive airway disease. This limitation forms the basis for the present study.

A structured scoring system may help overcome these limitations by quantifying the predictive value of historical, physical examination, and radiological findings. By integrating multiple variables into a single risk index, clinicians can more confidently identify children who truly require urgent bronchoscopy, limit unnecessary procedures, and expedite definitive management for those at highest risk. The objective of the present study was to derive and evaluate a clinical-radiological scoring system and an associated emergency decision algorithm to accurately predict the presence of an FBA and optimize the selection of pediatric patients for rigid bronchoscopy.

MATERIALS AND METHODS

Study design

We conducted a retrospective review of all children who underwent rigid bronchoscopy for suspected FBA at our tertiary care institute between January 2013 and December 2020. A total of 274 pediatric patients were included in the final analysis. The inclusion criteria comprised all children who underwent rigid bronchoscopy for suspected FBA during the study period. Patients with incomplete clinical or radiological data relevant to the analysis were excluded. Data collected included demographic characteristics, key historical features (e.g., choking history and history of asthma), clinical examination findings (e.g., tachypnea, respiratory distress, and unilateral decreased air entry), and radiological findings.

Data variables

Clinical variables were standardized as follows: A definitive choking history required a witnessed aspiration event; concurrent upper respiratory tract infection was defined by the presence of acute nasal or pharyngeal symptoms; and reactive airway disease was identified by a documented history of recurrent wheeze or asthma. Chest radiographs were primarily interpreted by the treating pediatric surgical team as part of routine clinical assessment. Radiology consultation was sought in cases with diagnostic uncertainty or when further imaging clarification was required.

Statistical analysis

Statistical analysis was performed using SPSS software version 30 (IBM, Armonk, NY, United States). Univariate logistic regression analysis was initially used to identify clinical and radiological factors associated with confirmed FBA. Variables with strong clinical relevance were retained to enhance the clinical applicability and face validity of the scoring system. Adjusted odds ratios (aORs) derived from multivariate logistic regression were used to assign weights to each variable. Regression coefficients were scaled and rounded to the nearest integer to generate a practical point-based scoring system. A 17-point clinical-radiological weighted score was derived based on these aORs. Receiver operating characteristic curve analysis was performed to determine the optimal cut-off value and assess the predictive performance of the score.

RESULTS

Demographics

A total of 274 children who underwent rigid bronchoscopy for suspected FBA were analyzed. The cohort was predominantly male (69.3%). Foreign bodies were successfully recovered in 181 of the 274 procedures, yielding a positive bronchoscopy rate of 66%. Conversely, 93 procedures did not result in foreign body recovery, representing a negative bronchoscopy rate of 34%. Rigid bronchoscopy was associated with a low rate of major complications. One patient developed a pneumothorax requiring intercostal drainage, and one child required tracheostomy. Approximately 7% of patients required postprocedural intensive care unit admission with mechanical ventilation. No major airway injuries, significant hemorrhage, or mortality were observed.

Predictive variables

Various clinical, historical, and radiological parameters were evaluated to determine their association with FBA. Univariate logistic regression analysis identified several significant predictors of positive bronchoscopy (Table 1).

Table 1 Association of clinical history, examination findings, and radiologic features between children with foreign body recovery during bronchoscopy (positive) and those without foreign body aspiration (negative) by univariate logistic regression analysis, n (%).

Variable	FBA-negative, n = 93	FBA-positive, n = 181	Unadjusted OR (95%CI)	P value
Clinical history
Definitive history of choking	43 (46)	135 (75)	3.38 (2.00-5.72)	< 0.001
Cough	78 (84)	143 (79)	0.74 (0.38-1.41)	0.356
Noisy breathing/stridor	12 (13)	32 (18)	1.42 (0.70-2.87)	0.333
Respiratory distress	81 (87)	131 (72)	0.40 (0.20-0.79)	0.008
Wheeze	26 (28)	15 (8)	0.24 (0.12-0.47)	< 0.001
Absence of concurrent URI	28 (30)	131 (72)	5.98 (3.46-10.34)	< 0.001
No prior history of reactive airway disease	56 (60)	163 (90)	5.87 (3.11-11.05)	< 0.001
Examination findings
Unilateral decreased air entry/wheeze	44 (47)	126 (70)	2.54 (1.52-4.24)	< 0.001
Stridor	3 (3)	10 (6)	1.58 (0.46-5.45)	0.466
Tachypnea	74 (80)	105 (58)	0.36 (0.20-0.64)	< 0.001
Crepitations	33 (36)	34 (19)	0.42 (0.24-0.74)	0.003
SpO2 < 90	29 (31)	30 (17)	0.44 (0.25-0.79)	0.006
Radiological findings
Normal chest radiograph	28 (30)	36 (21)	0.61 (0.34-1.07)	0.085
Unilateral hyperinflation	23 (25)	94 (54)	3.52 (2.02-6.13)	< 0.001
Unilateral collapse/consolidation	29 (31)	39 (22)	0.64 (0.36-1.12)	0.116
Bilateral hyperinflation	8 (9)	4 (2)	0.27 (0.08-0.86)	0.027
Infiltrates	26 (28)	30 (17)	0.54 (0.30-0.98)	0.041

CI: Confidence interval; FBA: Foreign body aspiration; OR: Odds ratio; SpO₂: Peripheral oxygen saturation; URI: Upper respiratory tract infection.

Clinical history: A definitive history of choking (OR = 3.38; P < 0.001), absence of concurrent upper respiratory tract infection (OR = 5.98; P < 0.001), and absence of a prior history of asthma or reactive airway disease (OR = 5.87; P < 0.001) were strongly associated with FBA.

Physical examination: Significant findings included unilateral decreased air entry or wheeze (OR = 2.54; P < 0.001), while generalized findings such as respiratory distress (OR = 0.40; P = 0.008) and tachypnea (OR = 0.36; P = 0.001) were paradoxically less frequent among patients with confirmed FBA.

Radiology: Unilateral hyperinflation on chest radiography was the strongest radiological predictor of FBA (OR = 3.52; P < 0.001).

Scoring system

Seven independent predictors were incorporated into a multivariate logistic regression model to derive a weighted 17-point clinical-radiological score. The weights assigned to each variable, based on their aORs, are summarized in Table 2. Interestingly, generalized respiratory findings such as respiratory distress and tachypnea were negatively associated with FBA, suggesting that these features may be more indicative of alternative diagnoses such as infectious or reactive airway conditions. The diagnostic performance of the 17-point score was evaluated using receiver operating characteristic curve analysis, which yielded an area under the curve (AUC) of 0.791 (95% confidence interval: 0.738-0.838; P < 0.0001). An optimal cut-off score of > 7 was identified, yielding a sensitivity of 83.43% and a specificity of 63.44% (Figure 1). Risk stratification further demonstrated the clinical utility of the score. Patients with a score ≤ 6 had a low probability of FBA (24.07%), while scores of 7-9 indicated an intermediate probability (50.88%). A high score of ≥ 10 was strongly predictive of FBA, with a probability of 85.28%.

Figure 1 Receiver operating characteristic curve demonstrating the diagnostic performance of the weighted 17-point clinical-radiological risk score for predicting foreign body aspiration. ROC: Receiver operating characteristic; AUC: Area under curve.

Table 2 Multivariate logistic regression model to identify significant predictors and derivation of the weighted risk score for foreign body aspiration.

Clinico-radiological variables	Regression coefficient	Adjusted OR (95%CI)	P value	Weight score
Definitive history of choking	0.58	1.79 (0.94-3.41)	0.077	2
Respiratory distress	-0.08	0.92 (0.37-2.31)	0.864	1
Absence of concurrent URI	1.44	4.21 (2.11-8.40)	< 0.001	4
No prior history of reactive airway disease	1.01	2.77 (1.29-5.92)	0.009	3
Unilateral decreased air entry/wheeze	0.86	2.37 (1.25-4.51)	0.008	2
Tachypnea	-0.91	0.40 (0.19-0.87)	0.021	1
Unilateral hyperinflation	1.32	3.73 (1.91-7.31)	< 0.001	4
Total score				17

CI: Confidence interval; OR: Odds ratio; URI: Upper respiratory tract infection.

DISCUSSION

Diagnosing pediatric FBA remains challenging, as classic symptoms are often absent, and up to 25% of chest radiographs appear normal[4]. This diagnostic uncertainty leads to a high incidence of negative rigid bronchoscopies, historically reported to range from 20% to 47% in the literature, comparable to the 34% negative bronchoscopy rate observed in our cohort[5]. Unnecessary bronchoscopies expose children to significant intrinsic procedural risks, including bronchospasm, airway edema, bleeding, and anesthesia-related adverse events, which occur in 8%-17% of cases[6]. To mitigate these avoidable harms and associated resource burdens, objective and quantitative decision-making tools are needed.

The most recurrent variables across the reviewed literature are a witnessed choking episode; wheezy chest or new/recurrent wheeze; unilateral diminished breath sounds; respiratory distress (including stridor or cyanosis); fever; age > 3 years; and specific radiographic findings such as hyperinflation, atelectasis, mediastinal shift, or a visible foreign body[7]. In our cohort, multivariate analysis identified seven independent predictors, confirming the strong diagnostic value of classical findings such as a definitive history of choking, unilateral decreased air entry, and unilateral hyperinflation, with the latter demonstrating a strong association with FBA (aOR = 3.73; 4 points).

The negative association of tachypnea and respiratory distress with FBA likely reflects the predominance of these findings in infectious or reactive airway conditions, which represent common clinical mimics of aspiration. In contrast, FBA more often produces localized airway obstruction with asymmetric clinical findings. Inclusion of these variables as negative predictors enhances the model’s ability to distinguish FBA from alternative diagnoses. Notably, the strongest clinical predictors in our model were negative historical findings, namely the absence of a concurrent upper or lower respiratory tract infection (aOR = 4.21; 4 points) and the absence of a prior history of asthma or reactive airway disease (aOR = 2.77; 3 points). Consequently, our scoring system places substantial emphasis on the absence of these confounding histories to effectively identify patients whose symptoms are unlikely to be attributable to underlying medical conditions, thereby reducing false-positive referrals for bronchoscopy.

Several clinical tools have been developed to improve the diagnostic accuracy of FBA, evolving from early decision trees proposed by Martinot et al[8] and computerized algorithms[9] to more structured, point-based predictive models. Among the most rigorously validated are the scoring systems proposed by Janahi et al[10] and Özyüksel et al[11]. Janahi et al[10] derived a 7-point scoring system incorporating both clinical history, including witnessed choking, and physical and radiological findings, achieving an AUC of 0.76, with a sensitivity of approximately 78% and a specificity of 85%[10]. In contrast, Özyüksel et al[11] excluded historical factors, arguing that they held poor predictive value, and focused exclusively on physical examination and radiological findings. Their model achieved a higher AUC of 0.816, with a sensitivity of 77.9% and a specificity of 74.8%.

Our derived 17-point scoring system bridges these methodologies by assigning considerable weight to the absence of confounding historical factors alongside definitive clinical and radiological findings. With an AUC of 0.791, the overall diagnostic performance of our model is comparable to that of previously established systems. More importantly, at our optimal cut-off score of > 7, our model demonstrated a sensitivity of 83.4%. In suspected pediatric FBA, where a missed diagnosis can lead to catastrophic airway compromise or chronic pulmonary morbidity, maximizing sensitivity is paramount. While the specificity of our model (63.4%) was slightly lower than that reported by Janahi et al[10] and Özyüksel et al[11], minimizing missed diagnoses was prioritized owing to the consequences of delayed recognition.

The primary clinical utility of our 17-point scoring system is its direct translation into a practical emergency decision-making algorithm. Recognizing that the absolute priority in suspected FBA is to never miss a retained foreign body because of a falsely reassuring low score, our pathway offers a safe middle-ground solution. Based on this risk stratification, patients scoring ≥ 10, corresponding to a high probability of FBA (85.28%), should be considered for immediate rigid bronchoscopy. Conversely, children scoring ≤ 6, corresponding to a low probability of FBA (24.07%), may be considered for further evaluation, including flexible bronchoscopy, depending on institutional practice. Finally, equivocal cases scoring between 7 and 9 (50.88% probability) should be managed on a case-by-case basis, relying on individualized clinical judgment and available institutional resources.

The present study does not directly evaluate the impact of the scoring system on reducing negative bronchoscopy rates, and prospective validation is required before routine clinical implementation. Our study has several limitations. First, the study cohort was limited to children who underwent bronchoscopy for suspected FBA, which introduces potential selection bias and may not fully represent the broader population of children with suspected aspiration. Patients managed conservatively without bronchoscopy were not included; therefore, the performance of the model in this lower-risk group remains uncertain. Consequently, the scoring system should be interpreted as a decision-support tool in patients in whom bronchoscopy is being considered, rather than as a screening tool for all presentations of suspected FBA.

Second, the retrospective design carries inherent risks of documentation bias and missing variables. In addition, reliance on parental recall for a “definitive choking history” introduces subjectivity. Furthermore, as this was a single-center study conducted at a tertiary care institute, our findings may reflect local referral patterns, potentially limiting broader generalizability. Importantly, our cohort exclusively comprised children who ultimately underwent bronchoscopy. We did not include data of patients initially suspected of FBA who were managed conservatively owing to low clinical suspicion and subsequently improved. Therefore, we cannot definitively determine whether this un-instrumented group would have consistently fallen in the low-risk (≤ 6) category.

Finally, while our 17-point score demonstrated high sensitivity (83.4%) and a strong AUC (0.791), the present study represents a derivation cohort without formal internal validation techniques such as bootstrapping or split-sample validation. This may result in overestimation of the model’s performance. Prospective validation of the scoring system is currently underway at our institution. Given the clinical context, a prospective observational study design is considered more appropriate than a randomized trial, allowing assessment of real-world applicability without compromising patient safety. Further multicenter studies are required to validate the model and confirm its broader clinical impact.

CONCLUSION

In conclusion, the high rate of negative rigid bronchoscopies in suspected pediatric FBA highlights the need for developing objective diagnostic tools. Our derived 17-point clinical-radiological scoring system offers a robust and highly sensitive (83.4%) method for risk stratification. By assigning substantial weight to the absence of respiratory mimics alongside classic findings such as unilateral hyperinflation, the score effectively discriminates true FBA from other diagnoses. The proposed algorithm provides a safe, middle-ground clinical pathway by expediting high-risk patients (score ≥ 10) to immediate rigid bronchoscopy while reserving initial flexible bronchoscopy for low-risk cases (score ≤ 6). However, prospective multicenter validation is necessary to confirm the reliability and clinical utility of the scoring system before its widespread adoption and implementation.