This was a cross-sectional descriptive study using publicly available online data from CPSC’s NEISS emergency department data query system (U.S. Consumer Product Safety Commission 2020a). This data set contains a nationally representative validated probability sample from about 100 of the > 5000 U.S hospitals providing emergency services. The NEISS sample is stratified based on hospital size and age served (non-specialized versus children’s hospitals). Detailed NEISS data collection and sampling procedures, changes over time, and statistical handling aspects have been reported elsewhere (U.S. Consumer Product Safety Commission 2020a; U.S. Consumer Product Safety Commission 2020b; Schroeder and Ault 2001).
Data source and variables
Cases from the ten-year period 2010 to 2019 (unweighted n = 3,782,633) were downloaded from NEISS in comma separated value (CSV) format and imported into statistical packages for query, review, filtering, formatting, recoding, sub-setting, and analysis. Variables examined included treatment date, age, gender, body-part affected, primary diagnosis, disposition, location type and the involved product/activity (product codes). Each NEISS record also contains a 400-character narrative text field containing descriptive comments derived from the emergency department (ED) record about the patient, involved product(s)/activity, sequence of events, associated diagnoses, affected body part(s), and other information (U.S. Consumer Product Safety Commission 2019). The narrative text field was used to find pickleball cases since there is no specific product/activity code for pickleball, refine case selection, and to assign the mechanism of injury.
Case selection
Because NEISS does not capture fatalities well, (Acton et al. 2019) and deaths in this sport-specific study were few and rarely if ever injury related, fatalities were excluded from this study. Case selection procedures are summarized in Fig. 1.
Pickleball
Pickleball does not have a specific NEISS product/activity code. Most pickleball-related cases were assigned a non-specific product code of 3235 (“Other ball sports, activity/apparel/equipment”). Potential cases were selected with a computer assisted search of the narrative text field for the following case-insensitive strings: “PICKLEBALL” or “PICKLE BALL” (n = 585), the common misspellings “PICKELBALL” or “PICKEL BALL” (n = 5) and “PICKLE RACKET”’ or “PICKLE RACQUET” (n = 1).
Records were excluded if: a) fatal (unweighted n = 1), b) if more than one sport or activity was mentioned in the narrative related to the injury (unweighted n = 13) and c) a special case containing the word “PICKLE” (described in the limitations section). These criteria resulted in the initial selection of group P1 from the 2010–2019 NEISS data of n = 577 unweighted records (weighted N = 37,521, 95% CI = 25,005–56,301).
An important case definition issue with NEISS data, especially pertinent to an older cohort, is that NEISS may contain non-injury events if patients were engaged in the sport/activity at the time an acute non-injury medical condition appeared. For example, chest pain can be caused by either an injury or a cardiovascular condition acquired while playing pickleball. From a sports medicine perspective, emergency non-injury events related to a sport or activity are of interest. Given the predominance of older persons among pickleball participants, a significant number of diagnoses in both pickleball and older tennis age subsets included many non-injury medical conditions such as syncope (fainting), chest pain unrelated to trauma and cardiovascular events and related symptoms (such as atrial fibrillation, cardiac arrest, tachycardia, etc.).
These case-definition considerations were addressed by enumerating and briefly describing the cases with non-injury-related syncope (unweighted n = 9) and cardiovascular mentions (without indication of traumatic injury) in the record narrative (unweighted n = 45) but excluding them from the injury analyses. Note that mentions of dehydration and heat exhaustion were included as injuries since “heat-related illnesses” are considered injuries in traditional injury epidemiology classifications. Therefore, if the symptom of “syncope” was stated in the narrative as related to a body temperature elevation or dehydration, they were included and assigned to the heat-related illness mechanism. However, other cases with syncope or dizziness reported without mention of heat, dehydration, injury, or cardiovascular issues were excluded from the injury subsets. After these exclusions, the all-age pickleball injury cohort (group P2) consisted of 523 unweighted records (weighted N = 33,817, 95% CI = 22,942–49,847).
Lastly, pickleball injuries involving cases ≥60 years of age were selected for the primary analyses and comparisons. The age criteria was based on the knowledge that the average retirement age in the U. S, among living retirees in 2019 was 59.9 years (PK 2019) and the convenience of working with 10-year age groups for rate calculations. This left an unweighted n = 429 senior pickleball injury-related records (weighted N = 28,984, 95% CI = 19,463–43,163) in group P3.
Senior tennis comparison group
For comparison purposes, NEISS tennis injuries ≥60 years of age were selected over the same period by searching for the code “3284” (tennis activity/apparel/equipment) in the three fields indicating the products involved. This resulted in an unweighted n = 1413 + 1 false negative found by looking at all mentions of the word “TENNIS” in the narrative text field. However, the “3284” code incorporates not just the “activity” of “tennis” but also any tennis-related apparel or equipment involved in an injury. Therefore, we distinguished “playing tennis” as a recreational activity, from injuries involving tennis-related equipment while not engaged in playing tennis. Manual review of each senior tennis-related narrative for cases with a product code of “3284” led to many such exclusions. For example, a child struck in the eye at home with a “tennis ball”, lacerations involving “cutting tennis balls”, or falls related to “tennis balls” on an elderly assistive “walker” were all excluded.
Like the pickleball case selection, for the senior tennis-related cases we excluded fatalities (unweighted n = 13), one false positive and the multiple sport mentions (unweighted n = 35). This left an unweighted n = 1363 senior tennis related cases (weighted N = 74,932, 95% CI = 55,580-101,021) in group T1. Lastly, tennis cases with non-injury syncope (unweighted n = 82) and cardiovascular mentions without indication of traumatic injury in the record narrative (unweighted n = 172) were removed. This left n = 1109 records (weighted N = 58,836, 95% CI 44,736-77,381) in the senior tennis-related injury cohort, group T2.
Mechanism of injury and variable regrouping
Certain variables were created or regrouped from the NEISS data as follows:
Primary injury mechanism assignment
Each tennis and pickleball-related case narrative text was used to assign an injury mechanism based on a modification of the coding scheme used by Gaw et al. (Gaw et al. 2014) to describe how the injury took place. In instances of narrative overlap, the first mechanism described was coded that most directly related to the injury. The case narrative text review resulted in assigning each case to a primary mechanism (or to an exclusion category) as follows:
-
1.
Slip/Trip/Fall/Dive. A combined category.
-
2.
Hit with racket or paddle (or “bat” if pickleball was mentioned).
-
3.
Hit with ball.
-
4.
Other specified mechanism. Commonly included a movement of some type (i.e., sudden stop, lunging, running, bending over, hyperextending, dislocation, sprain, twist, strain, bump, tear, pull, sudden pop or snap, inverted or rolled ankle), and less common mechanisms such as jammed body part, cutting a finger, abrasion, or insect sting.
-
5.
Play/playing tennis or pickleball. Encompassed injuries that incurred during the activity where the mechanism could not be determined or was unknown (i.e., musculoskeletal pain, contusion, or epistaxis).
-
6.
Heat-related illness. Assigned only if syncope/dizziness and hot/heat or dehydration was mentioned.
-
7.
Hit with, by, tripped or ran into other object or slipped on other equipment. Includes hitting fence, net, wall, chair, bench, tree, tripped or fell over ball, racquet, or paddle.
-
8.
Hit with, by or ran into another player.
Exclusions were assigned as follows:
-
The case involved multiple sports mentions or was not a tennis or pickleball “playing” activity.
-
Other syncope, dizziness, or dyspnea without mention of heat, injury, or cardiovascular issue.
-
Possible cardiovascular event. This includes heart rhythm issue, angina, chest pain, deep vein thrombosis (dvt), blood in urine, altered mental status (AMS), pleural effusion, gastro-intestinal (GI) bleed, ataxia, weakness, or abnormal blood pressure without mention of traumatic injury.
Special cases:
-
A subarachnoid hemorrhage can be due to a ruptured aneurysm, an arteriovenous malformation (AVM), or a traumatic head injury. One such tennis-related case was included as a Slip/Trip/Fall/Dive since the narrative said they fell while playing tennis, but it is acknowledged that it was not well differentiated whether the hemorrhage preceded or followed the fall.
-
It was assumed if the tennis-related case was coded as “3284” and the narrative said the person had been hit by a tennis racquet that they were playing tennis if there was no place of injury mentioned.
The primary mechanisms for all pickleball and senior tennis-related records were assigned independently by authors HW and JD. For pickleball cases, initial (unweighted) inter-rater reliability for the mechanism recoding was 94.9%. The cases in which the primary mechanism was assigned differently (discordant pairs) were all resolved by joint discussion. For tennis cases, inter-rater reliability for the mechanism recoding for the first independent review was 94.3% with all discordant pairs jointly resolved.
Body region
Body region injured was regrouped into: (1) upper extremity (including the NEISS categories of shoulder, elbow, upper arm, lower arm, wrist, hand, and finger); (2) lower extremity (including knee, upper leg, lower leg, ankle, foot, and toe); (3) trunk (including upper trunk, lower trunk, and pubic region); (4) head/neck (including head, face, eye, mouth, neck, and ear); and (5) other (including internal organs and injury to greater than 25% of the body) (Gaw et al. 2014).
An eye injury flag was assigned based on both the body part code (77 - EYEBALL) plus a search in the narrative text field for eye injury since sometimes the body part was assigned to the face or another body part code while the narrative text indicated an eye injury had occurred. Each narrative was reviewed, and the case included as an eye injury if it appeared the person has been struck in the eye resulting in injury (senior pickleball n = 4, senior tennis n = 20).
Disposition from the ED
Disposition from the ED was regrouped into 3 categories: (1) released; (2) hospitalized (including NEISS variables of treated and transferred, treated, and admitted, and held for < 24 h for observation); and (3) left against medical advice.
Location of injury
Location of injury was regrouped into school/public property, sports/recreation place, and other (including the NEISS categories of home, farm, apartment/condo, and street/highway).
Analysis
Descriptive analyses consisted of unweighted record counts (n) and weighted counts (N) where indicated. Weighted stratified survey specific analyses, confidence interval calculations and table and graphics preparation were performed using the R statistical programming language and the ‘survey’, ‘lubridate’, ‘vroom’, ‘segmented’, and ‘ggpubr’ add-on packages (R Core Team 2013).
The methods used to identify statistically significant trends in senior pickleball injuries were drawn from Thomas Yokota’s example reproducing the Centers for Disease Controls’ (CDC) guide on conducting statistical trend tests with multiple years of complex survey data (Yokota n.d.). The R ‘segmented’ package was used to estimate breakpoints in the trend analysis, and data manipulation was performed using packages in the ‘tidyverse’ ecosystem (Wickham et al. 2019). Confidence intervals for case counts were calculated on a log scale which produces intervals close to the Coefficient of Variation referenced in the NEISS research guide (U.S. Consumer Product Safety Commission Division of Hazard and Injury Data Systems 2000; Lumley 2020).
Simple logistic regression was used to compute the Odds Ratios (ORs) to estimate the strength of associations between binomial outcome variables for senior pickleball versus senior tennis-related injuries. They were calculated using survey adjusted general linear models with a logit link as per the methods of DiMaggio et al. (DiMaggio et al. 2019). Trends were visualized by plotting annual rates of injury counts per yearly census population estimates by age group. Population estimates by age group were obtained using the cencusapi R package (Recht 2020).