Clinical Insights into Pediatric Foreign Bodies: A Cluster Analysis of Symptoms and Assessment of Caregiver Reports

doi:10.21203/rs.3.rs-4831473/v1

Foreign body (FB) events are prevalent in emergency room consultations, with 107,000 cases annually in the United States (80% in the pediatric population). They result in an estimated 1,500 mortalities each year. A cross-sectional study between February 2008 and September 2015, employing hierarchical cluster analysis of signs and symptoms at presentation to the emergency room was performed. Sensitivity, specificity, and decision curve analysis for a caregivers’ report of events was calculated with confirmed FB diagnosis as a reference. 1,032 patients were included, 886 suspected ingestion and 166 suspected aspiration patients. For aspiration, a higher frequency of complications (14.3%) and mortalities (4.76%) were observed in the cluster exhibiting most respiratory symptoms. For ingestion, associations for complications (3.6%), longer stays (mean, 55.1 hours) and mortality (n = 2) were identified in a cluster with respiratory pattern. Increased respiratory symptoms at presentation are potential predictors of severe outcomes, intuitively in cases of aspiration, but notably in ingestion. Although definitive diagnostic procedures are still necessary, the severity of respiratory symptoms helps identify patients at risk of complications and helps prioritize resources when necessary. Caregivers’ certainty of FB aspiration yielded a 72.5% sensitivity and 45.4% specificity. Certainty of FB ingestion resulted in 86.8% sensitivity and 19.8% specificity.

Health sciences/Signs and symptoms/Digestive signs and symptoms

Health sciences/Health care/Paediatrics/Paediatric research

Health sciences/Medical research/Paediatric research

Foreign Bodies

Pediatric Emergency Medicine

Cluster Analysis

Caregivers

Symptom Assessment

The ingestion or aspiration of FBs frequently leads to pediatric emergency room visits and poses a significant risk of morbidity and mortality. Patient prognosis relies on several factors, including age, type of FB, lodging location, whether there is an associated obstruction, and the time elapsed since the event. In the United States, 107,000 annual cases are seen (80% involving pediatric patients) resulting in approximately 1,500 deaths [1]. Notably, in 2011, 87,197 cases were reported among children under the age of 5, with the average cost of extraction via rigid endoscopy being USD 1,445 per patient [2]. Complications arise in 1% of ingestion cases, with an increase to 35% when the ingested body is sharp [1]. For aspiration incidents, complications are seen for 15% of cases [3]. Mortality rates from the aspiration of FBs range from 0.6–3%, highlighting a significant concern in pediatric health care [4, 5].

Regarding causality, 98% of cases are accidental and often involve common household items [1, 6]. Although caregivers visually observe incidents in up to 88.2% of cases, there is a perceived underutilization of this information, as seen in the lack of utilization or reporting in nearly all studies [7]. Previous studies on emergency room visits due to ingestion or aspiration have rarely transcended purely descriptive characterizations, and they often fail to consolidate cases of both conditions. Due to the nonspecific symptoms associated with both conditions, a clustering approach is appropriate. Clustering allows for the identification of distinct patterns which can uncover hidden associations and phenotypes in conditions with broad and overlapping symptoms [8]. This study aims to describe clinical clusters of pediatric patients presenting with either condition in the emergency room and to assess the validity of information received from caregivers. The analysis is based on data collected from a single pediatric institution over a six-year period.

Methods

This study employed a descriptive cross-sectional design with retrospective data collection, gathering information from February 2008 to September 2015 at Fundación Hospital de la Misericordia, a tertiary-level pediatric institution. Patients under 18 years of age with suspected ingestion or aspiration of FBs were included. Patients with incomplete clinical records or those admitted in a severe condition, who required emergency endoscopic or surgical management and consequently could not undergo a comprehensive clinical assessment upon arrival, were excluded. To calculate the sample size, it was considered that 80% of foreign body events occur in patients under 18. In 2018, Bogotá had approximately 2,187,000 children, according to a national census. A sample size of 425 patients was calculated for a 99% confidence level. However, due to the availability of data, all eligible patients admitted during the study period were included.

Suspected ingestion or aspiration was defined based on caregivers' report of suspected or certainty of occurrence of the event, or clinical features suggestive of either ingestion or aspiration of an FB, based on initial interaction at arrival to the emergency department. Data was retrospectively collected from patients' clinical records, including sociodemographic characteristics, clinical features, in-hospital outcomes (complications, mortality, hospital stay, treatment approach), and caregiver report of the episode. Patients were followed for the entire duration of their admission until discharge or transfer to another institution. Information on clinical records was cross-checked with other primary sources of information such as imaging software. Due to its retrospective design and the sources of information used, this study received a "no risk" designation for participants under Article 11 of Colombian Resolution 8430. The study received approval from the ethics committees of Fundación Hospital de la Misericordia (approval no. CEI – 53 − 15, dated November 18, 2015) and Universidad Nacional de Colombia (approval no. 018-177-15, dated October 22, 2015). Waiver of informed consent was approved by both ethics committees. This study was conducted in accordance with the Declaration of Helsinki.

Statistical analysis: A univariate analysis was conducted to process the collected data, utilizing absolute and relative frequencies for nominal and ordinal variables. Variable distributions were assessed via Shapiro Wilk tests. Measures of central tendency (mean or median) and dispersion (standard deviation or interquartile range) were applied to quantitative variables. The data were categorized into two groups: suspected aspiration and suspected ingestion of FBs.

Further, a bivariate analysis was performed to explore the relationships between independent and dependent variables. Patients were analyzed in two groups: confirmed ingestion and confirmed aspiration. Clinical characteristics between these groups were compared using Chi-square, Fisher's Exact Test, Kruskal-Wallis test or Mann-Whitney U Test, depending on the distribution of the variables. A p-value of less than 0.05 was considered statistically significant. The sensitivity and specificity of caregivers' certainty regarding FB aspiration or ingestion were calculated against the confirmation of an FB following a complete patient workup, with results reported using 95% confidence intervals. A Receiver Operating Characteristic (ROC) curve was graphed, but not presented due to figure restraints, and the Area Under the Curve (AUC) was calculated for both sensitivity and specificity. Positive predictive values and negative predictive values, along with their 95% confidence intervals, were also reported.

Decision Curve Analysis

To assess the clinical utility of caregiver certainty in predicting foreign body (FB) aspiration or ingestion among pediatric patients, we performed a decision curve analysis (DCA) for two distinct groups. The net benefit was calculated by comparing the true positives' proportion against the false positives, adjusting these findings by the odds of risk associated with each outcome. This analysis aimed to quantify the net benefit of precisely identifying patients at risk of FB aspiration or ingestion while considering the potential harms arising from false-positive predictions. This evaluation spanned a broad spectrum of threshold probabilities. Given that our model employed a dichotomous variable, a demonstrated net positive benefit at any threshold probability would be deemed advantageous compared to a treat all approach [9] [10].

Cluster analysis

Clusters regarding clinical symptoms in suspected ingestion and aspiration of FBs in pediatric patients were formed. Variables were created for each reported pain location in each group, identifying seven locations in the suspected aspiration group and eight in the suspected ingestion group. To reduce dimensionality and eliminate redundancy across the 15 recorded symptoms and pain locations, Multiple Correspondence Analysis (MCA) was performed. This analysis provided factor scores for the two dimensions that explained the highest percentage of variance in the data.

Clustering was performed using Ward's linkage method, employing factor scores from the two dimensions obtained through Multiple Correspondence Analysis. Various numbers of group clusters were tested to determine the best fit and distribution of patients. To validate the clustering results, the K-means clustering method was applied. Subsequently, the Kappa coefficient was calculated to assess the agreement between the Ward's linkage and K-means clustering methods. The resulting Kappa coefficient of 0.8643 indicated a very high agreement between the clustering methods for the suspected aspiration group, demonstrating consistency across methods.

For the suspected ingestion group, a similar process was employed. Initial clustering using Ward's linkage followed by K-means clustering and validation through the Kappa coefficient revealed only fair agreement between clusters. Consequently, variables where less than 5% of patients exhibited positive observations and were deemed clinically insignificant were excluded from further analysis. This adjustment resulted in considering 10 symptoms and 8 additional variables for pain location in the new clustering approach. The resulting clusters demonstrated a Kappa coefficient of 0.7168, indicating substantial agreement between clustering methods. Given the non-parametric distribution observed in the length of hospital stays, initial analyses used the median and interquartile range (IQR). However, these measures were insufficient to capture significant variations within the clusters for a clinically relevant variable. Thus, both median/IQR and mean/standard deviation were reported for each cluster to provide a more comprehensive view of data distribution and better capture the influence of outliers. To ensure the robustness of our findings, statistical significance was assessed using appropriate non-parametric tests, tailored to the data's distribution characteristics.

Following the clustering of patient data, additional bivariate analyses were conducted to further investigate the relationships between variables of interest. All statistical analyses, including these bivariate tests, were performed using StataCorp’s 2023 release of Stata Statistical Software: Release 18.0, based in College Station, TX.

The study initially included 1,262 admissions that met the inclusion criteria. However, due to the inaccessibility of medical records for 210 patients, these individuals were excluded from the analysis. Consequently, the study proceeded with an analysis of 1,052 patients.

The study categorized clinical characteristics of pediatric patients into two groups based on suspected FB incidents: ingestion (n = 886) and aspiration (n = 166), as detailed in Table 1. The median age was 3.9 years (IQR 2.3–6.4) in the ingestion group and 2.2 years (IQR 1.2–5.8) in the aspiration group. Males predominated in both groups, constituting 54.2% (n = 480) of the ingestion group and 68.7% (n = 114) of the aspiration group. Respiratory system-related comorbidities were notably prevalent, occurring in 20% (n = 177) of the ingestion cases and 75% (n = 24) of the aspiration cases. Caregivers of 84.9% (n = 752) of the ingestion cases were certain about the incident, compared to 62.1% (n = 103) in the aspiration cases. Radiography was utilized in 85.7% (n = 759) of ingestion cases and 15.2% (n = 23) of aspiration cases.

Table 1

Clinical characteristics of pediatric patients with suspected foreign body ingestion or aspiration. (n = 1052)
		Suspected ingestion. (n = 886)	Suspected aspiration. (n = 166)
Age group
	Infant (1–12 months)	43 (4.9)	31 (18.7)
	Toddler (First year of life)	137 (15.5)	39 (23.5)
	Preschooler (2–5)	456 (51.5)	58 (34.9)
	School-age child (6–11)	206 (23.3)	27 (16.3)
	Adolescent (12–18)	44 (5.0)	11 (6.6)
Age (years), median (IQR)		3.9 (2.3–6.4)	2.2 (1.2–5.8)
Sex
	Female	406 (45.8)	52 (31.3)
	Male	480 (54.2)	114 (68.7)
Weight (kg), median (IQR)		15 (12–20)	12 (9.7–19)
Caregiver account
	Certainty	752 (84.9)	103 (62.1)
	Suspected	86 (9.7)	44 (26.5)
	Not suspected	40 (4.5)	18 (10.8)
	Unavailable	8 (0.9)	1 (0.6)
Confirmed FB after workup		628 (70.9)	69 (41.6)
Failiure in endoscopic management		16 (2.6)	1 (1.5)
Complications, confirmed FB		11 (1.8)	9 (13.0)
Mortality		2 (0.2)	3 (1.8)
Comorbidities		177 (20.0)	32 (19.3)
Type of comorbidity*		N = 223	N = 43
	Respiratory	71 (40.1)	24 (75.0)
	Gastrointestinal	64 (36.2)	8 (25.0)
	Neurologic	33 (18.6)	7 (21.9)
	Others	55 (31.1)	4 (12.5)
State of conciousness
	Alert	886 (100)	154 (92.8)
	Stupor	0	1 (0.6)
	Coma	0	1 (0.6)
	Sedated/Mechanical ventilation	0	9 (5.4)
	Brain death	0	1 (0.6)
Radiography performed		759 (85.7)	151 (91.0)
CT-scan performed		11 (1.2)	16 (9.6)
Radiography findings		N = 759	N = 151
	Confirmatory for FB	474 (62.5)	23 (15.2)
	Suspected FB	22 (2.9)	40 (26.5)
	Rules out FB	263 (34.7)	88 (58.3)
CT-scan findings		N = 11	N = 16
	Confirmatory for FB	5 (45.5)	10 (62.5)
	Suspected FB	2 (18.2)	5 (31.3)
	Rules out FB	4 (36.4)	1 (6.3)
FB, Foreign Body. * Multiple patients had more than one comorbidity.

In terms of FB composition, metal was the most frequent in the ingestion group, involved in 71% of cases (n = 451), with coins being the most ingested item (55.1%, n = 346). For the aspiration group, organic materials were predominant, affecting 59.4% of patients (n = 41), with grains as the most aspirated item (20.3%, n = 14). Complications were confirmed in 1.8% (n = 11) of ingestion cases, including two fatalities. In the aspiration group, there were nine complications (13%) and three deaths (1.8%). Minimally invasive extraction methods were highly effective, with only 16 cases (2.6%) in the ingestion group and one case (1.5%) in the aspiration group requiring alternative interventions, as seen in Table 1.

A total of 628 cases of FB ingestion were identified. Among these, a higher prevalence of ingestion over aspiration was observed in preschool-aged children, with 59.1% in the ingestion group compared to 29.0% in the aspiration group (p-value < 0.001). Significant differences were noted across other age groups as well, although preschool ages predominated in aspirated FB cases too. The toddler age group was the second most common and showed a stronger association with aspirated rather than ingested FBs (27.5% vs. 16.2%, p-value 0.019). Regarding comorbidities related to the systems affected, FB aspiration was notably associated with respiratory comorbidities, which represented 87.5% of comorbidities in the aspiration group, compared to 34.8% in the ingestion group (p-value 0.005). In the confirmed FB ingestion group, gastrointestinal comorbidities accounted for 39.3% (n = 44) of all comorbidities, and in the confirmed aspiration group, they constituted 25.0% (n = 2); however, these differences were not statistically significant. In cases where X-rays were performed, the findings were confirmatory for the presence of a FB in 84.2% of confirmed ingestion cases (n = 470), which was significantly higher compared to the aspiration group, where confirmatory results were seen in only 37.7% of cases (n = 23) (p-value < 0.001), as shown in Table 2.

Table 2

Clinical outcomes in pediatric patients with confirmed foreign body ingestion or aspiration. (n = 697)
		Confirmed ingestion. (n = 628)	Confirmed aspiration. (n = 69)	p-value
Age group
	Infant (1–12 months)	31 (4.9)	10 (14.5)	0.001
	Toddler (First year of life)	102 (16.2)	19 (27.5)	0.019
	Preschooler (2–5)	371 (59.1)	20 (29.0)	< 0.001
	School-age child (6–11)	106 (16.9)	14 (20.3)	0.476
	Adolescent (12–18)	18 (2.9)	6 (8.7)	0.012
Age in years, median (IQR)		3.6 (2.2–5.3)	2.9 (1.3–6.7)	0.234ᵇ
Sex				0.013ᵃ
	Female	290 (46.2)	21 (30.4)
	Male	338 (53.8)	48 (69.6)
FB composition				< 0.001ᵃ
	Organic	142 (22.6)	41 (59.4)
	Metal	451 (71.8)	15 (21.7)
	Polymer	17 (2.7)	11 (15.9)
	Other	18 (2.9)	2 (2.9)
Comorbidities		112 (17.8)	8 (11.6)	0.192ᵃ
Type of comorbidity*		N = 152	N = 20
	Respiratory	39 (34.8)	7 (87.5)	0.005ᶜ
	Gastrointestinal	44 (39.3)	2 (25.0)	0.709ᶜ
	Neurologic	20 (17.9)	1 (12.5)	> 0.999ᶜ
	Others	38 (33.9)	1 (12.5)	0.272ᶜ
Complications		11 (1.8)	9 (13.0)	< 0.001ᶜ
Radiography findings		N = 558	N = 61	< 0.001ᵃ
	Confirmatory for FB	470 (84.2)	23 (37.7)
	Suspected FB	9 (1.6)	16 (26.2)
	Rules out FB	79 (14.2)	22 (36.1)
FB, Foreign Body. * Multiple patients had more than one comorbidity. ᵃ Chi-square test of independence. ᵇMann-Whitney U test. ᶜ Fisher's exact test.

The accuracy of caregivers' initial certainty regarding FB aspiration or ingestion at admission was evaluated by comparing it to FB confirmation after patient workup, focusing on measures of sensitivity and specificity. The sensitivity of caregivers’ certainty of FB aspiration was 72.5% (95% CI 65.7–79.3), while the specificity was 45.4% (95% CI 37.8–52.9). Similarly, the sensitivity for caregivers’ certainty of FB ingestion was reported as 86.8% (95% CI 84.6–89.0), with a specificity of 19.8% (95% CI 17.2–22.4). The positive and negative predictive values for this assessment are described in Table 3.

Table 3

Sensitivity and Specificity of caregiver´s certainity of foreign body aspiration or ingestion
		Caregiver certainty of FB aspiration	Caregiver certainty of FB ingestion
Diagnostic test characterstics
	Senstivity % (95% CI)	72.5 (65.7–79.3)	86.8 (84.6–89.0)
	Specificity % (95% CI)	45.4 (37.8–52.9)	19.8 (17.2–22.4)
	AUC (95% CI)	58.9 (51.6–66.2)	53.3 (50.5–56.0)
	PPV % (95% CI)	48.5 (40.9–56.2)	72.5 (69.5–75.4)
	NPV % (95% CI)	69.8 (62.9–76.8)	38.1 (34.9–41.3)

Three clusters were identified among patients with suspected FB aspiration, spanning 166 cases. Cluster 1 included 63 patients, showing a high frequency of symptoms such as cough (n = 46, 73%), dyspnea (n = 42, 66.7%), and asphyxia (n = 19, 30.2%). This cluster also reported a higher frequency of complications (n = 9, 14.3%) compared to the other two clusters, Cluster 2 (n = 4, 4.8%) and Cluster 3 (n = 1, 5.3%); however, the differences in complication rates were not statistically significant (p-value = 0.105). Notably, all mortality cases within the study were reported in Cluster 1 (n = 3, 4.76%). Cluster 2, the largest group, consisted of 84 patients. The predominant symptoms reported in this cluster were coughing (n = 43, 51.2%), choking (n = 30, 35.7%), and vomiting (n = 19, 22.6%). Cluster 3, the smallest, comprised 19 patients, primarily characterized by pain (n = 19, 100%), cough (n = 15, 79%), and dysphagia (n = 6, 31.6%), as shown in Table 4.

Table 4

Pediatric patients with suspected foreign body aspiration (n = 166)
	Cluster 1. (n = 63)	Cluster 2. (n = 84)	Cluster 3. (n = 19)	p-value
Age (years), median (IQR)	1.9 (1.0-3.3)	2.1 (1.1–5.2)	7.8 (3.0-10.1)	0.0001ᵃ
Sex				0.277ᵇ
Female	24 (38.1)	24 (28.6)	4 (21.1)
Male	39 (61.9)	60 (71.4)	15 (79.0)
Weight (kg), median (IQR)	11 (9-14.5)	12 (9.75-19)	29 (12–34)	0.0003ᵃ
Length of hospital stay in days, median (IQR)	4 (1–8)	1 (0–3)	1 (0–3)	0.0001ᵃ
Comorbidities	13 (20.6)	16 (19.1)	3 (15.8)	0.893ᵇ
Confirmed FB after workup	35 (55.6)	25 (29.8)	9 (47.4)	0.006
Asphyxia	19 (30.2)	9 (10.7)	3 (15.8)	0.011ᵇ
Vomiting	13 (20.6)	19 (22.6)	2 (10.5)	0.498ᵇ
Choking	8 (12.7)	30 (35.7)	2 (10.5)	0.002ᵇ
Coughing	46 (73.0)	43 (51.2)	15 (79.0)	0.008ᵇ
Dysphagia	0	5 (6.0)	6 (31.6)	< 0.001ᵇ
Wheezing	17 (27.0)	3 (3.6)	2 (10.5)	< 0.001ᵇ
Stridor	16 (25.4)	3 (3.6)	3 (15.8)	0.001ᵇ
Lung field hypoventilation	17 (27.0)	4 (4.8)	0	< 0.001ᵇ
Rhonchi	21 (33.3)	0	1 (5.3)	< 0.001ᵇ
Complete airway obstruction	7 (11.1)	2 (2.4)	0	0.037ᵇ
Dyspnea	42 (66.7)	0	5 (26.3)	< 0.001ᵇ
Pain	2 (3.2)	0	19 (100)	< 0.001ᵇ
Complications	9 (14.3)	4 (4.8)	1 (5.3)	0.105ᵇ
Mortality	3 (4.76)	0	0	0.082ᵇ
ᵃKruskar-Wallis test ᵇChi-square test of independence

Three distinct clusters were identified among pediatric patients with suspected foreign body (FB) ingestion. Cluster 1, the largest, included 359 patients and predominantly exhibited symptoms such as vomiting (n = 121, 33.7%), dysphagia (n = 127, 35.4%), and sialorrhea (n = 129, 35.9%). Cluster 2, the smallest, included 220 patients, with symptoms most frequently reported being vomiting (n = 138, 62.7%), dyspnea (n = 106, 48.2%), and sialorrhea (n = 78, 35.5%). Notably, this cluster had a significantly higher number of complications (n = 8, 3.6%) compared to Cluster 1 (n = 2, 0.6%) and Cluster 3 (n = 3, 1.0%) (p-value = 0.008). It also reported longer hospital stay (55.1 hours) (p-value 0.0307) and accounted for all mortality cases (n = 2) (p-value 0.048). Cluster 3 contained 307 patients, with all reporting pain (n = 307, 100%), and other frequent symptoms including dysphagia (n = 150, 48.9%) and sialorrhea (n = 87, 28.3%). This group reported three complications but no fatalities. Detailed comparisons and additional data are presented in Table 5.

Table 5

Pediatric patients with suspected foreign body ingestion (n = 886)
	Cluster 1. (n = 359)	Cluster 2. (n = 220)	Cluster 3. (n = 307)	p-value
Age (years), median (IQR)	3.7 (2.4-6.0)	2.5 (1.4–4.2)	5.5 (3.7–7.7)	0.0001
Sex				0.047
Female	152 (42.3)	96 (43.6)	158 (51.5)
Male	207 (57.7)	124 (56.4)	149 (49.5)
Weight (kg), median (IQR)	15 (12-18.3)	12 (10–16)	18 (14–25)	0.0001
Length of hospital stay in hours, median (IQR)	12 (12–24)	12 (12–24)	12 (12–24)	0.0307
Length of hospital stay in hours, mean (SD)	39.7 (99.5)	55.1 (144.9)	34.4 (95.5)
Comorbidities	70 (19.5)	53 (24.1)	54 (17.6)	0.176
Confirmed FB after workup	283 (78.8)	162 (73.6)	183 (59.6)	< 0.001
Choking	36 (10.0)	64 (29.1)	42 (13.7)	< 0.001
Vomiting	121 (33.7)	138 (62.7)	111 (36.2)	< 0.001
Hemoptysis	0	35 (15.9)	22 (7.2)	< 0.001
Dyspnea	0	106 (48.2)	28 (9.1)	< 0.001
Coughing	0	73 (33.2)	33 (10.8)	< 0.001
Pain	23 (6.4)	18 (8.2)	307 (100)	< 0.001
Dysphagia	127 (35.4)	58 (26.4)	150 (48.9)	< 0.001
Anorexia	10 (2.8)	38 (17.3)	24 (7.8)	< 0.001
Sialorrhea	129 (35.9)	78 (35.5)	87 (28.3)	0.083
Complications	2 (0.6)	8 (3.6)	3 (1.0)	0.008
Mortality	0	2 (0.9)	0	0.048
ᵃKruskar-Wallis test ᵇChi-square test of independence

The decision curve analysis showed that treating based on caregiver certainty provides a net benefit in the management of foreign body aspiration and ingestion in pediatric patients compared to treating all cases, as seen in Fig. 1A and Fig. 1B.

Figure 1A and Fig. 1B. Decision Curve Analysis of Caregiver Certainty as a Model for Treatment Decisions in Suspected Cases of Aspiration and Ingestion.

This study employed a descriptive cross-sectional design with retrospective data collection, gathering information from February 2008 to September 2015 at Fundación Hospital de la Misericordia, a tertiary-level pediatric institution. Patients under 18 years of age with suspected ingestion or aspiration of FBs were included. Patients with incomplete clinical records or those admitted in a severe condition, who required emergency endoscopic or surgical management and consequently could not undergo a comprehensive clinical assessment upon arrival, were excluded. To calculate the sample size, it was considered that 80% of foreign body events occur in patients under 18. In 2018, Bogotá had approximately 2,187,000 children, according to a national census. A sample size of 425 patients was calculated for a 99% confidence level. However, due to the availability of data, all eligible patients admitted during the study period were included.

Suspected ingestion or aspiration was defined based on caregivers' report of suspected or certainty of occurrence of the event, or clinical features suggestive of either ingestion or aspiration of an FB, based on initial interaction at arrival to the emergency department. Data was retrospectively collected from patients' clinical records, including sociodemographic characteristics, clinical features, in-hospital outcomes (complications, mortality, hospital stay, treatment approach), and caregiver report of the episode. Patients were followed for the entire duration of their admission until discharge or transfer to another institution. Information on clinical records was cross-checked with other primary sources of information such as imaging software. Due to its retrospective design and the sources of information used, this study received a "no risk" designation for participants under Article 11 of Colombian Resolution 8430. The study received approval from the ethics committees of Fundación Hospital de la Misericordia (approval no. CEI – 53 − 15, dated November 18, 2015) and Universidad Nacional de Colombia (approval no. 018-177-15, dated October 22, 2015). Waiver of informed consent was approved by both ethics committees. This study was conducted in accordance with the Declaration of Helsinki.

Statistical analysis: A univariate analysis was conducted to process the collected data, utilizing absolute and relative frequencies for nominal and ordinal variables. Variable distributions were assessed via Shapiro Wilk tests. Measures of central tendency (mean or median) and dispersion (standard deviation or interquartile range) were applied to quantitative variables. The data were categorized into two groups: suspected aspiration and suspected ingestion of FBs.

Further, a bivariate analysis was performed to explore the relationships between independent and dependent variables. Patients were analyzed in two groups: confirmed ingestion and confirmed aspiration. Clinical characteristics between these groups were compared using Chi-square, Fisher's Exact Test, Kruskal-Wallis test or Mann-Whitney U Test, depending on the distribution of the variables. A p-value of less than 0.05 was considered statistically significant. The sensitivity and specificity of caregivers' certainty regarding FB aspiration or ingestion were calculated against the confirmation of an FB following a complete patient workup, with results reported using 95% confidence intervals. A Receiver Operating Characteristic (ROC) curve was graphed, but not presented due to figure restraints, and the Area Under the Curve (AUC) was calculated for both sensitivity and specificity. Positive predictive values and negative predictive values, along with their 95% confidence intervals, were also reported.

Decision Curve Analysis

To assess the clinical utility of caregiver certainty in predicting foreign body (FB) aspiration or ingestion among pediatric patients, we performed a decision curve analysis (DCA) for two distinct groups. The net benefit was calculated by comparing the true positives' proportion against the false positives, adjusting these findings by the odds of risk associated with each outcome. This analysis aimed to quantify the net benefit of precisely identifying patients at risk of FB aspiration or ingestion while considering the potential harms arising from false-positive predictions. This evaluation spanned a broad spectrum of threshold probabilities. Given that our model employed a dichotomous variable, a demonstrated net positive benefit at any threshold probability would be deemed advantageous compared to a treat all approach [9] [10].

Cluster analysis

Clusters regarding clinical symptoms in suspected ingestion and aspiration of FBs in pediatric patients were formed. Variables were created for each reported pain location in each group, identifying seven locations in the suspected aspiration group and eight in the suspected ingestion group. To reduce dimensionality and eliminate redundancy across the 15 recorded symptoms and pain locations, Multiple Correspondence Analysis (MCA) was performed. This analysis provided factor scores for the two dimensions that explained the highest percentage of variance in the data.

Clustering was performed using Ward's linkage method, employing factor scores from the two dimensions obtained through Multiple Correspondence Analysis. Various numbers of group clusters were tested to determine the best fit and distribution of patients. To validate the clustering results, the K-means clustering method was applied. Subsequently, the Kappa coefficient was calculated to assess the agreement between the Ward's linkage and K-means clustering methods. The resulting Kappa coefficient of 0.8643 indicated a very high agreement between the clustering methods for the suspected aspiration group, demonstrating consistency across methods.

For the suspected ingestion group, a similar process was employed. Initial clustering using Ward's linkage followed by K-means clustering and validation through the Kappa coefficient revealed only fair agreement between clusters. Consequently, variables where less than 5% of patients exhibited positive observations and were deemed clinically insignificant were excluded from further analysis. This adjustment resulted in considering 10 symptoms and 8 additional variables for pain location in the new clustering approach. The resulting clusters demonstrated a Kappa coefficient of 0.7168, indicating substantial agreement between clustering methods. Given the non-parametric distribution observed in the length of hospital stays, initial analyses used the median and interquartile range (IQR). However, these measures were insufficient to capture significant variations within the clusters for a clinically relevant variable. Thus, both median/IQR and mean/standard deviation were reported for each cluster to provide a more comprehensive view of data distribution and better capture the influence of outliers. To ensure the robustness of our findings, statistical significance was assessed using appropriate non-parametric tests, tailored to the data's distribution characteristics.

Following the clustering of patient data, additional bivariate analyses were conducted to further investigate the relationships between variables of interest. All statistical analyses, including these bivariate tests, were performed using StataCorp’s 2023 release of Stata Statistical Software: Release 18.0, based in College Station, TX.

In this study, we utilized cluster analysis to assess the clinical syndromes of patients with suspected ingestion or aspiration, aiming to identify potential patterns in their presentation. Our analysis demonstrated a significant association between the prevalence of respiratory symptoms upon arrival at the emergency room and increased mortality, complications, and prolonged hospital stays in patients suspected of ingestion. A similar pattern was noted among patients with suspected aspiration, who presented with pronounced respiratory symptoms. This innovative research approach represents the first application of cluster analysis for evaluating clinical symptoms in pediatric FB events and the first integration of sensitivity/specificity analysis with decision curve analysis to assess caregiver narratives. Additionally, this study is among the largest to concurrently analyze aspiration and ingestion cases.

Patient data indicate that FB events primarily occur in children aged 2–5 years, with metals being the most ingested material (71.8%) and organics the most frequently aspirated (59.4%). This is consistent with previous research showing that organic materials are found in 66.7% of aspiration cases and metals in 60–84% of ingestion cases, predominantly affecting children in this age group [2, 4, 11]. Specifically, coins accounted for 55.1% of ingestion incidents, while grains were involved in 20.3% of aspiration cases, followed by other organic objects. Although imaging studies are crucial in patient workup, their effectiveness varies with the composition of the FB [12]. Radiographs confirmed FBs in 62.5% of suspected ingestion cases, compared to 37.7% for aspiration cases. Tseng et al.'s review of imaging studies for FBs reported similar findings, with negative radiographs occurring in up to two-thirds of aspiration cases [13]. Despite CT scans offering a higher detection rate of 70–100% for FBs, their routine use in children is discouraged due to potential risks associated with radiation exposure [14]. The attraction to coins and other visually appealing objects during the preschool years is often linked to a developmental phase characterized by oral curiosity, a tendency to explore their environment, and the absence of molars necessary for mastication [15].

Building on the diagnostic approach to FBs, clinical interrogation of both the patient and caregiver plays a crucial role during patient workup. While previous studies have often focused solely on the frequency of caregiver reports without assessing their validity, our evaluation highlights their significant value. Caregiver certainty about the occurrence of an event demonstrates strong performance in detecting confirmed cases, with sensitivities of 72.5% for aspiration and 86.8% for ingestion. Furthermore, DCA indicates that acting on caregiver certainty provides a net benefit gain in managing both suspected aspiration and ingestion, with this advantage persisting across a broad spectrum of decision thresholds. Notably, although the net benefit for foreign body ingestion is more subtle, it remains beneficial. Considering the low-risk profile of diagnostic interventions like radiography and endoscopic procedures, and the severe potential consequences of undetected foreign bodies, our findings justify a relatively higher tolerance for false positives. This approach contrasts with previous predictive models for FB aspiration that relied predominantly on witnessed choking incidents, which have shown limited performance, with some studies reporting a sensitivity of 45.9% and specificity of 52.4% [16, 17]. However, relying solely on caregiver information is insufficient for conclusively ruling out the presence of FBs in cases of both aspiration and ingestion, as evidenced by low specificity values. Ultimately, a comprehensive assessment of these events requires the integration of various sources of information. Despite perceived historical underutilization, our findings advocate for a more balanced approach that gives due consideration to caregiver insights alongside clinician intuition.

Regarding clinical presentation upon arrival to the emergency room, clinical syndromes for aspiration cases typically include a combination of cough, dyspnea, rhonchi, wheezing, stridor, and decreased breathing sounds, as observed in other studies [17][18][19]. In our study, the three predominant symptoms were cough, dyspnea, and vomiting. Notably, patients in Cluster 1 exhibited a higher prevalence of symptoms not encompassed by these three main symptoms, with increased frequencies for stridor, wheezing, lung field hypoventilation, rhonchi, and complete airway obstruction. This cluster was significantly associated with a longer hospital stay compared to other groups. Although no significant associations were observed, possibly due to a small sample size and low frequencies for both events, there was a trend towards higher mortality and complications in this group, likely related to a greater degree of obstruction leading to severe disease.

For ingestion cases, clinical syndromes are characterized by symptoms such as vomiting, sialorrhea, dysphagia, cough, abdominal pain, FB sensation, and hematemesis [19]. In our study, the three most observed symptoms were pain, vomiting, and dysphagia. Interestingly, Cluster 2 patients followed this trend but also presented with marked respiratory symptoms. Dyspnea was the second most common symptom, occurring in 48.2% of these patients, with other respiratory and obstructive symptoms like cough and choking also being more frequent in this cluster. The significant associations with mortality, complications, and length of stay, along with the marked respiratory symptoms at presentation observed in Cluster 2, suggest that a greater degree of respiratory symptoms at presentation is predictive of severe disease in ingestion cases. While this finding aligns with widely accepted disease mechanisms, it has not previously been discussed in the context of FB ingestion at time of presentation.

Our study results face limitations initially by a disproportionately high number of ingestion cases in pre clustering analysis. While this distribution is consistent with clinical practice, it resulted in unbalanced comparison groups. To address this issue, we employed cluster analysis within segregated groups for ingestion or aspiration, facilitating comparisons of variables of interest among patients with the same type of suspected event. Additionally, although the clustering was consistent across various methods, the generalizability of our findings might be limited due to the single-center nature of our study. Clinical presentations may vary across different settings, yet no specific reasons for such variations have been identified by the authors at this time. Measurement bias may have been introduced by the retrospective collection of data and the reliance on clinical records, which may contain inaccuracies. This was addressed by cross-referencing medical records with other sources of information, such as imaging software, to ensure the accuracy of the clinical records. Lastly, our study was also limited by the lack of further descriptions of complications and mortality, even though these outcomes were reported. However, our objective was to explore a perspective different from what has been reported so far.

In conclusion, our study underscores the importance of respiratory symptoms at arrival to the emergency room as potential predictors of severe outcomes, intuitively in cases of aspiration, but notably also in ingestion. Although definitive diagnostic procedures are still necessary, assessing the severity of respiratory symptoms can help identify patients at a higher risk of complications and help prioritize resources when necessary. The findings also highlight the importance of actively including caregiver information in patient workup. While caregiver reports cannot conclusively rule out the presence of an FB, they demonstrate good performance in detecting FB cases, with a net benefit observed when treatment decisions are guided by caregiver certainty rather than a treat all approach. Confirming widely discussed clinical practices, the low incidence of mortality, complications, and failures in endoscopic management underscores the effectiveness of endoscopic approaches as the cornerstone of both patient workup and treatment of pediatric FBs. Future studies focusing on respiratory symptoms at presentation for ingestion cases might further elucidate the relationship discussed in this study.

Conflict of Interest:

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Funding:

This research did not receive any specific grant from funding agencies in the public, commercial, or not-for-profit sectors.

Author Contribution

SA-R: Data analysis, interpretation of data, writing of manuscript. EDAG: Data acquisition, interpretation of data, writing of manuscript. SCS: Data acquisition, interpretation of data, writing of manuscript. IEG: Study design, data analysis, data interpretation. MMB: Study design, data analysis, data interpretation. FFA: Conception of study, interpretation of data. All authors approved the final version of this manuscript.

Data Availability

The datasets generated and/or analyzed during the current study are not publicly available due to regulations in Colombia that restrict sharing patient information but are available from the corresponding author on reasonable request.

Chung, S., Forte, V. & Campisi, P. A review of pediatric foreign body ingestion and management. Clin. Pediatr. Emerg. Med. 11, 225–230 (2010). doi:10.1016/j.cpem.2010.06.002.
Russell, R., Lucas, A., Johnson, J., Yannam, G., Griffin, R., Beierle, E., et al. Extraction of esophageal foreign bodies in children: rigid versus flexible endoscopy. Pediatr. Surg. Int. 30, 417–422 (2014). doi:10.1007/s00383-014-3481-2.
Foltran, F., Ballali, S., Passali, F.M., Kern, E., Morra, B., Passali, G.C., et al. Foreign bodies in the airways: a meta-analysis of published papers. Int. J. Pediatr. Otorhinolaryngol. 76, S12–S19 (2012). doi:10.1016/j.ijporl.2012.02.004.
Samkani, A., Lynge Larsen, K.V., Faber, C.E. & Godballe, C. Bronchoscopy should always be performed in children on suspicion of foreign body aspiration. Dan. Med. J. 60, A4715 (2013). PMID: 24083529.
Boufersaoui, A., Smati, L., Benhalla, K.N., Boukari, R., Smail, S., Anik, K., et al. Foreign body aspiration in children: experience from 2624 patients. Int. J. Pediatr. Otorhinolaryngol. 77, 1683–1688 (2013). doi:10.1016/j.ijporl.2013.07.026.
Jayachandra, S. & Eslick, G.D. A systematic review of paediatric foreign body ingestion: presentation, complications, and management. Int. J. Pediatr. Otorhinolaryngol. 77, 311–317 (2013). doi:10.1016/j.ijporl.2012.11.025.
Chinski, A., Foltran, F., Gregori, D., Ballali, S., Passali, D. & Bellussi, L. Foreign bodies in children: A comparison between Argentina and Europe. Int. J. Pediatr. Otorhinolaryngol. 76, S76–S79 (2012). doi:10.1016/j.ijporl.2012.02.018.
Loftus, T.J., Shickel, B., Balch, J.A., Tighe, P.J., Abbott, K.L., Fazzone, B., et al. Phenotype clustering in health care: A narrative review for clinicians. Front. Artif. Intell. 5, 842306 (2022). doi:10.3389/frai.2022.842306.
Rousson, V. & Zumbrunn, T. Decision curve analysis revisited: overall net benefit, relationships to ROC curve analysis, and application to case-control studies. BMC Med. Inform. Decis. Mak. 11, 45 (2011). doi:10.1186/1472-6947-11-45.
Piovani, D., Sokou, R., Tsantes, A.G., Vitello, A.S. & Bonovas, S. Optimizing clinical decision making with decision curve analysis: insights for clinical investigators. Healthcare (Basel) 11, 2244 (2023). doi:10.3390/healthcare11162244.
Navia-López, L.A., Cadena-León, J.F., Ignorosa-Arellano, K.R., Toro-Monjaraz, E.M., Zárate-Mondragón, F., Loredo-Mayer, A., et al. Foreign body ingestion and associated factors in pediatric patients at a tertiary care center. Rev. Gastroenterol. Mex. (Engl. Ed.) 87, 20–28 (2022). doi:10.1016/j.rgmxen.2021.09.004.
Pinto, A., Lanza, C., Pinto, F., Grassi, R., Romano, L., Brunese, L., et al. Role of plain radiography in the assessment of ingested foreign bodies in pediatric patients. Semin. Ultrasound CT MR 36, 21–27 (2015). doi:10.1053/j.sult.2014.10.008.
Tseng, H.-J., Hanna, T.N., Shuaib, W., Aized, M., Khosa, F. & Linnau, K.F. Imaging foreign bodies: ingested, aspirated, and inserted. Ann. Emerg. Med. 66, 570–582.e5 (2015). doi:10.1016/j.annemergmed.2015.07.499.
Sugawa, C., Ono, H., Taleb, M. & Lucas, C.E. Endoscopic management of foreign bodies in the upper gastrointestinal tract: a review. World J. Gastrointest. Endosc. 6, 475–481 (2014). doi:10.4253/wjge.v6.i10.475.
Torsello, M., Sicuranza, L., Meucci, D., Salvati, A., Tropiano, M.L., Santarsiero, S., et al. Foreign body aspiration in children: our pediatric tertiary care experience. Pediatr. Surg. Int. 40, 93 (2024). doi:10.1007/s00383-024-05679-z.
Pozailov, S., Goldbart, A., Aviram, M., Maimon, M.S., Dizitzer Hillel, Y., Gatt, D., et al. Foreign body aspiration score (FOBAS)—a prospectively validated algorithm for the management and prediction of foreign body aspiration in children. Eur. J. Pediatr. 183, 815–825 (2024). doi:10.1007/s00431-023-05347-9.
Lee, J.J.W., Philteos, J., Levin, M., Namavarian, A., Propst, E.J. & Wolter, N.E. Clinical prediction models for suspected pediatric foreign body aspiration: a systematic review and meta-analysis. JAMA Otolaryngol. Head Neck Surg.147, 787–796 (2021). doi:10.1001/jamaoto.2021.1548.
Na'ara, S., Vainer, I., Amit, M. & Gordin, A. Foreign body aspiration in infants and older children: a comparative study. Ear Nose Throat J. 99, 47–51 (2020). doi:10.1177/0145561319839900.
Khorana, J., Tantivit, Y., Phiuphong, C., Pattapong, S. & Siripan, S. Foreign body ingestion in pediatrics: distribution, management and complications. Medicina (Kaunas) 55, 686 (2019). doi:10.3390/medicina55100686.

No competing interests reported.

Clinical Insights into Pediatric Foreign Bodies: A Cluster Analysis of Symptoms and Assessment of Caregiver Reports

Status:

Version 1

Abstract

Figures

Introduction

Results

Methods

Discussion

Declarations