Trampolines Versus Playgrounds - A Comparative Assessment of Pediatric Fractures Sustained From Recreational Play

Abstract

Introduction:

Using the Pediatric Health Information System, this study compared the relative severity of fractures sustained from trampolines with those from other playground equipment.

Methods:

Pediatric patients were identified in the Pediatric Health Information System with trampoline-related injuries (TRIs) or playground-related injuries (PRIs) diagnosed as fractures. Adjustments were made for hospital, year of injury, sex, age, race, median household income, and rurality through propensity score weighting. Four injury-related outcome measures were examined as a proxy for injury severity.

Results:

A total of 133,232 patients met inclusion criteria. In unadjusted univariate analyses, TRIs were associated with greater odds of severe fracture and lower odds of receiving surgical treatment (OR = 0.954) compared with PRIs. After adjustment, TRIs sustained in late childhood and adolescence were more likely to receive surgical management (OR = 1.092 and OR = 1.192, respectively) while TRIs sustained in younger children were less likely (OR = 0.607) than PRIs.

Discussion:

Youths in late childhood and adolescence are at increased odds of undergoing surgical management after trampoline fractures. Beyond underscoring the risks of trampoline play, our results highlight the importance of considering age in recreational injury epidemiology and the public health safety initiatives aimed at specific age groups.

Trampoline-related injuries (TRIs) account for almost 100,000 emergency department visits annually and represent a large portion of pediatric accidental injuries.^1,2,3,4 The fracture incidence sustained on trampolines has also increased markedly, from 35 per 100,000 person-years in 2008 to 53 per 100,000 person-years in 2017.³ In response to the incidence and severity of injuries sustained from trampolines, both the American Academy of Pediatrics⁵ and the American Academy of Orthopaedic Surgeons⁶ have issued statements cautioning against their use. The effect of these public health campaigns on lasting reduction of pediatric injury burden remains unclear, especially in light of the growth of trampoline park facilities where the severity of injuries sustained appears to be substantially greater.^4,7,32

Several studies have described the incidence and risk factors of trampoline injury.^8,9,10,11 However, most are single-center studies where seasonal, regional, and hospital-specific factors may lead to findings that are not broadly generalizable. In addition, although TRIs have been compared with other recreational injuries in a small series,¹² they have yet to be evaluated in a national, multicenter data set against a reference group. Playground-related injuries (PRIs) may serve as a reasonable and common benchmark, accounting for over 200,000 ED visits annually.¹³ Among these injuries, monkey bars and climbing gyms cause most PRIs, frequently resulting in extremity fractures.^14,15

Among the existing studies using national databases that exist, many of the recent investigations have been conducted in other countries with different standards and policies surrounding trampoline use.^16,17,18,19 Of those conducted in the United States, most have used the National Electronic Injury Surveillance System,^1,7,20,21 which is useful in estimating injury occurrence but is limited in scope when examining several diagnoses, treatments, and outcomes. The Pediatric Hospital Information System (PHIS) is an administrative database providing a breadth of data regarding hospitalization, interventions, and complications and has not been previously used to examine injuries sustained from trampoline or playground use.

The primary aim of this study was to assess whether trampoline-related injuries are more at risk for needing surgical management compared with playground-related injuries. Secondary aims included identifying whether TRIs were more at risk for needing greater length of stay or more likely to have medical or surgical complications and greater administrative severity compared with playground injuries. It was hypothesized that TRI fractures would differ across all studied clinical realms and be more severe across all indicators of severity and that this effect would remain after confounding adjustment.

Methods

A retrospective review was conducted on the Pediatric Health Information System (PHIS). PHIS is an administrative database managed by the Children's Hospital Association that contains outpatient and inpatient encounters from over 49 not-for-profit, tertiary care pediatric hospitals in the United States. Its data are de-identified and undergo rigorous quality assurance testing on submission. The PHIS database was queried for pediatric patients (17 years old and younger) sustaining trampoline-related and playground-related orthopaedic injuries from January 1, 2010, to December 31, 2019. Patients were excluded if their fracture site was not the spine or appendicular skeleton (ie, skull, facial, and rib fractures) or the fracture site was not identified. Given the use of de-identified, publicly available data, this study was exempted according to our institutional review board.

We used International Classification of Diseases (ICD) coding to identify the fracture mechanism, which we defined as either TRI (ICD-9 E005.3 or ICD-10 Y93.44) or PRI (ICD-9 E884.0 or ICD-10 W09.X). Four different injury-related outcome variables were analyzed: length of stay in days, the presence of surgical management during that encounter (our primary outcome), the presence of medical or surgical complications, and administrative severity as defined by 3M's All Patient Refined Diagnosis Related Groups (APR DRGs). APR-DRGs are divided into four levels of illness: (1) minor; (2) moderate; (3) major; and (4) extreme, based on “the extent of physiologic decompensation or organ system loss of function”.²² Severity was then further dichotomized into two categories, minor (minor level) and severe (moderate, major, and extreme), given that less than half of hospitals had multiple patients with major and extreme severity scores from each mechanism (n = 24, 48%) and these small cells would make our planned mixed models unreliable.²³ Although this variable was not developed solely for orthopaedic injuries, it was included, given its clinical prognostic capabilities. The following patient characteristics were also collected: year of injury, age (infancy and early childhood: 0 to 4 years old; late childhood: 5 to 12yo; and adolescence: 13 to 17yo), sex (male and female), race (Asian, Black, White, and other), fracture site (spine, shoulder or arm, elbow or forearm, wrist or hand, pelvis, hip or thigh, knee or leg, and ankle or foot), zip-code median household income in 2010 (US dollars), and zip-code rurality (urban or rural). Median income was used as a proxy for individual and neighborhood socioeconomic status, given the available variables in the database, and summarized as quintiles to allow for nonlinearity.

Statistical Analyses

Limited missing data were observed in sex (n = 16, 0.01%), median income (n = 3638, 2.7%), and rurality (n = 3638, 2.7%), so these could be justifiably handled using listwise deletion.²⁴ Descriptive statistics were first calculated and reported as either medians with interquartile ranges or the number of participants and percentages of the sample. The unadjusted or baseline models were then estimated. The unadjusted analysis consisted of generalized linear mixed models for severity, surgical management, and complications, with mechanism as a fixed effect and hospital as a random effect. Next, to account for possible confounding, adjusted models were fit using propensity score weighting to estimate the average treatment effect. Weights were obtained from a mixed logistic regression of mechanism on patient age, sex, race, zip-code rurality, and zip-code median income, with treating discharge year and treating hospital as random effects. Given that the fracture site may mediate the relationship between the mechanism and outcome, it was not included in the propensity scoring model. The previously described generalized linear mixed models were then weighted to adjust for confounding by those demographic variables. Because play and skeletal development vary throughout growth, effect modification by age using likelihood-ratio testing was examined. All analyses were conducted in R version 4.1.1. Significance was set a priori at ɑ = 0.05, and balance was defined at a standardized mean difference (SMD) less than 0.10.

Results

Overall, 133,232 pediatric patients sustaining axial or appendicular orthopaedic fractures associated with trampolines or playground activities were included. A total of 33,030 TRIs (25%) and 100,202 PRIs (75%) were included. Patient characteristics of the entire cohort by mechanism are presented in Table 1.

Table 1.

Demographic Characteristics by Fracture Mechanism

	Overall	Other Playground Equipment	Trampoline	P
	(N = 134,487)	(N = 101,038)	(N = 33,449)
Year of discharge, median [IQR]	2015 [2012, 2017]	2014 [2012, 2017]	2016 [2013, 2018]	<0.001
Age, n (%)				<0.001
Infancy and early childhood (0-4 years)	32288 (24.2)	21771 (21.7)	10517 (31.8)
Late childhood (5-12 years)	96853 (72.7)	76872 (76.7)	19981 (60.5)
Adolescence (13-17 years)	4091 (3.1)	1559 (1.6)	2532 (7.7)
Sex, n (%)				<0.001
Male	71358 (53.6)	54607 (54.5)	16751 (50.7)
Female	61874 (46.4)	45595 (45.5)	16279 (49.3)
Race, n (%)				<0.001
Asian	3809 (2.9)	3382 (3.4)	427 (1.3)
Black	13242 (9.9)	11157 (11.1)	2085 (6.3)
Multiracial/other	23813 (17.9)	18778 (18.7)	5035 (15.2)
White	93598 (70.3)	67854 (67.7)	25744 (77.9)
2010 median household income in US dollars, median [IQR]	45,049 [35,074, 59,144]	45,233 [35,188, 59,555]	44,274 [34,921, 57,841]	<0.001
Residential location, n (%)				<0.001
Rural	10943 (8.2)	7268 (7.3)	3675 (11.1)
Urban	122289 (91.8)	92934 (92.7)	29355 (88.9)
Fracture site				<0.001
Spine	943 (0.7)	717 (0.7)	226 (0.7)
Arm and shoulder	6900 (5.2)	5351 (5.3)	1549 (4.7)
Forearm and elbow	66825 (50.2)	53032 (52.9)	13793 (41.8)
Hand and wrist	38468 (28.9)	33032 (33.0)	5436 (16.5)
Pelvis	117 (0.1)	76 (0.1)	41 (0.1)
Hip and thigh	1970 (1.5)	1328 (1.3)	642 (1.9)
Leg and knee	11251 (8.4)	3611 (3.6)	7640 (23.1)
Foot and ankle	6758 (5.1)	3055 (3.0)	3703 (11.2)

Patients who sustained fractures from trampolines were more likely to be older, White, and female (P < 0.001) overall. However, subgroup analysis showed that there were actually more young children (<5yo) and children from rural zip codes with lower median incomes injured by a trampoline than other playground equipment (P < 0.001). The most common fracture site for both mechanisms was the forearm and elbow (n = 53,032 [52.9%] in TRIs and 13,793 [41.8%] in PRIs). However, 36% of trampoline fractures and 8% of playground fractures occurred to the lower extremity and 16% of trampoline fractures and 33% of playground fractures occurred to the hand or wrist. Post hoc testing revealed significant heterogeneity in surgical rates by mechanism and fracture site in an adjusted model (P < 0.001). Upper extremity trampoline fractures were more likely to undergo surgical management (upper extremity OR = 1.264 [1.203, 1.329]), but lower extremity trampoline fractures were less likely (lower extremity OR = 0.413 [0.367, 0.465]).

The unadjusted models demonstrated a higher risk of severe fracture (odds ratio = 1.597 95% confidence interval [1.263, 1.931]) but a lower risk of surgical management (OR = 0.954 [0.926, 0.984]). No notable differences were observed in length of stay (Exp(B) = 1.009 [0.997, 1.022]) and complication rates (OR = 1.174 [0.998, 1.38]). With propensity score weighting, balance was achieved on all covariates (max SMD<0.10). After adjustment, the difference in surgical management remained notable (OR = 0.923 [0.883, 0.964]) and the difference in administrative severity reversed, now matching surgical management (OR = 0.900 [0.850, 0.950]).

Turning to effect modification by age, the effect of mechanism varied by age of severity (P < 0.01) and surgical management (P < 0.01). Fractures sustained from trampolines were more likely to be treated surgically in adolescents (OR = 1.192 [1.05, 1.354]) and older children (OR = 1.092 [1.024, 1.166]) but less likely to be treated surgically in younger children (OR = 0.607 [0.561, 0.657]). Similarly, fractures from trampolines were less likely to be severe in infancy and early childhood (OR = 0.763 [0.688, 0.837]) but did not differ in severity in late childhood (OR = 0.967 [0.888, 1.045]) and adolescence (OR = 1.142 [0.969, 1.314]). Model results are further provided in Table 2.

Table 2.

Descriptive Statistics and Effect Estimates of Mechanism on Fracture Outcomes

		Overall	Other Playground Equipment	Trampoline	Overall Effects		Patient Age-Specific Effects
	(N = 134487)	(N = 101038)	(N = 33449)	Unadjusted	Adjusted ATT	Infancy and Early Childhood	Late Childhood	Adolescence
Management				0.954 [0.926, 0.984]	0.923 [0.883, 0.964]	0.607 [0.561, 0.657]	1.092 [1.024, 1.166]	1.192 [1.05, 1.354]
Nonsurgical	100577 (74.8)	75089 (74.3)	25488 (76.2)
Surgical	33910 (25.2)	25949 (25.7)	7961 (23.8)
Length of stay	1.00 [1.00, 1.00]	1.00 [1.00, 1.00]	1.00 [1.00, 1.00]	1.009 [0.997, 1.022]	1.007 [0.992, 1.022]	0.989 [0.963, 1.016]	1.012 [0.993, 1.032]	1.039 [0.989, 1.091]
Severity				1.166 [1.119, 1.213]	0.900 [0.850, 0.950]	0.763 [0.688, 0.837]	0.967 [0.888, 1.045]	1.142 [0.969, 1.314]
Minora	119536 (88.9)	90075 (89.1)	29461 (88.1)
Moderateb	14519 (10.8)	10669 (10.6)	3850 (11.5)
Severeb	404 (0.3)	278 (0.3)	126 (0.4)
Extremeb	28 (0.0)	16 (0.0)	12 (0.0)
Complication status				1.174 [0.998, 1.380]	0.977 [0.766, 1.246]	1.117 [0.782, 1.596]	1.197 [0.766, 1.871]	0.804 [0.433, 1.493]
Uncomplicated	133716 (99.4)	100481 (99.4)	33235 (99.4)
Complicated	771 (0.6)	557 (0.6)	214 (0.6)

^aIncluded in the low-severity category.

^bIncluded in the high-severity category.

Descriptive statistics presented as n (%) or median [IQR]. ATT = average treatment effect on the treated. Effects presented with 95% confidence intervals.

Discussion

Our study assesses whether TRIs are more at risk of surgical management, longer length of stays, complications, and greater administrative severity compared with PRIs. In summary, our results support the understanding that trampolining represents a high-risk activity for young individuals and highlight the importance of recognizing the context in which an injury occurs.²⁵ Notable differences were observed in demographic characteristics and clinical outcomes between pediatric patients presenting with fractures from trampolines and those with fractures from other playground equipment. Youths in late childhood and adolescence were more likely to require surgical management for fractures sustained by trampolines, while younger children injured by trampolines were less likely to require surgical treatment. Although fractures from trampolines were more severe before adjustment, this difference became insignificant after adjustment.

Our study identified an increased rate of surgical management for trampoline fractures for youths in late childhood and adolescence, a finding echoed in other studies as well.¹⁰ This increased risk may be due to either the higher forces youths may generate on trampolines compared with other playground equipment or even the risky play styles encouraged on trampolines. The emergence of trampoline parks nationally over the past decade and the associated increased severity of injuries sustained in these environments^4,32 likely plays a major role in the observed differences in our cohort. Differences in fracture site distribution may also provide some insight into how fractures sustained after trampoline injuries are more likely to undergo surgical management. Elbow and forearm injuries overall were the most prevalent regardless of the mechanism, a finding echoed by several other studies.^{3,13,15,26,27} Similarly, there was an elevated rate of lower limb fractures secondary to trampolines, a pattern also reflected in the literature.²⁸ Kim and colleagues²⁸ offered that these injuries may have occurred secondary to a varus shearing or axial compression force. These forces likely mediate the effect of mechanism on fracture pattern, implied by the different rates of surgical management in trampoline fractures.

Patients with fractures from trampolines were more likely to be older, White, and female and live in rural zip codes with a slightly lower median household income. It is difficult to identify the source of these demographic differences. This could be due to differences in factors such as trampoline availability, activity selection, play style, supervision, bone health, and interactions between them. It may reflect differences in playground safety between rural and urban areas, which increased risk of certain injuries.²⁹ Of note, young children (<5yo) were more likely to have been injured by a trampoline than other playground equipment in our sample. Because PHIS only represents what happens after children present to the hospital, we cannot untangle exposure to trampolines from risk on trampolines. Regardless of the reason, these findings suggest that trampoline injury prevention interventions might benefit from responding to the needs of rural communities.

Finally, no differences were found in length of stay or complication rates between the mechanistic groups. The source of this equivalence is not clear. If we had focused on a more high-risk population or injury pattern, other differences may have been identifiable; however, there remain limitations in doing this using PHIS because radiographs are not available for review for more fine-tuned diagnostic considerations. Similar explanations exist for our finding of a comparable complication rate. Because PHIS only records complications that occur during the original encounter, we cannot comment on long-term complications such as malunion, nonunion, or chronic pain.

The findings of this study may facilitate improved counseling by orthopaedic surgeons and musculoskeletal providers regarding risks associated with the studied play equipment. Injuries vary depending on the mechanism, with injuries of the wrist and elbow being common for both playground and trampoline mechanisms. Patients are also more likely to have lower extremity fractures related to trampoline use. Older children and adolescents had an increased odds of surgical management for trampoline injuries while younger children with playground injuries had a greater odds of surgical management compared with trampoline injuries. Public health measures should continue to prevent injuries in children related to recreational activities.

Analysis of this large database comes with inherent limitations. First, the fractures included were only those listed as the primary diagnosis. Second, this data set is not sampled and weighted in a way that can provide national estimates; thus, the results may be affected in part by the availability of studied play equipment in the regions treated by the contributing PHIS hospitals. This also limits the generalizability to national or international populations, and it is not possible to calculate the burden of injury from this study, other than to state that playground injuries accounted for approximately three times more fractures than those that came from trampolines at these hospitals. Third, the data are not granular enough to provide additional information on the mechanisms or whether injuries were recreational or related to participation in organized sports (ie, gymnastics).³⁰ The playground injuries were similarly not defined in relation to specific play equipment. Fourth, because sampling is only from tertiary care children's hospitals, there may be an oversampling of more severe cases that may have been transferred for more specialized or surgical care. Fifth, owing to lack of sampling and weighting to provide national estimates, it is not possible to determine from this study whether the rates of injury are increasing with time³¹ for either trampoline-related or playground-associated injuries. Miscoding and thus misclassification of injury type (ie, fracture) and etiology (trampoline or playground) based on ICD coding are also acknowledged as a limitation in this administrative study; however, there is no means of investigators cross-checking PHIS coding by medical record or radiographic review in this de-identified multicenter database. Finally, only fractures were considered, but trampolines are also associated with notable orthopaedic soft-tissue injuries²¹ as well as nonorthopaedic injuries, including dental injury, vascular injury, and concussions.^{21,33,34,35,36} These dimensions were beyond the scope of our study but should be taken into account when considering public health messaging around trampolines.

In conclusion, youths in late childhood and adolescence sustaining fractures from trampoline use are more likely to undergo surgical management compared with injuries resulting from other playground equipment. Younger children remain at greater risk of injuries from other playground equipment. Our study underscores the need for public health interventions to reduce TRIs and PRIs but also understand demographic factors affecting recreational choices that may affect pediatric orthopaedic epidemiologic findings.