Handball and movement screening – can non-contact injuries be predicted in adolescent elite handball players? A 1-year prospective cohort study

ABSTRACT

Introduction: The nine-test screening battery (9SB) consists of 11 tests used to assess injury risk in sports populations.

Objectives: To evaluate the predictive value of the composite score and underlying factors of the 9SB for sustaining non-contact injury in adolescent elite handball players.

Methods: Forty-five (23 females) adolescent elite handball players, median age 17 (range 16–18), pre-seasonally performed the 9SB, followed by weekly recordings of injuries for 52 weeks using a web-based questionnaire.

Results: The median value for seasonal substantial injury prevalence was 22% (25-75^th percentiles 6–41). An exploratory factor analysis extracted three factors, complex movements, mobility, and lower extremity control, that explained a cumulative variance of 56%, where each factor contributed with 13–26% of the total variance. Based on the identified cutoff values, none of the factors or the complete 9SB could predict the risk of a non-contact new injury as well as the risk of reporting a substantial injury. Area under the curve values were ranged 0.50 to 0.59, with the corresponding 95% CI including 0.50 for all factors.

Conclusion: Based on the limited predictive ability of the 9SB, it is not recommended that clinicians use the 9SB to predict injury in adolescent elite handball players.

KEYWORDS:

• Injury burden
• performance
• prediction
• youth athletes

Introduction

Handball is one of the team sports with the highest injury rates (Junge et al., 2006). In several studies, it has been reported an injury incidence during major international tournament above 100 injuries/1000 playing hours (Bere et al., 2015; Junge et al., 2006; Langevoort, Myklebust, Dvorak, and Junge, 2007). In youth handball, the injury rate has been reported to be 9.9 injuries/1000 competition hours and 0.9/1000 total hours including practice, using match and coach reports (Olsen, Myklebust, Engebretsen, and Bahr, 2006). Based on self-reported data, the overall injury rate in handball is reported to be 6.3/1000 match and training hours, and the injury rate seems to decrease by age (Moller, Attermann, Myklebust, and Wedderkopp, 2012). Further, male youths (18 years) have a 1.8 times higher risk of sustaining an injury, in any part of the body, compared to females (Moller, Attermann, Myklebust, and Wedderkopp, 2012). Recently, it was reported that 23% of adolescent handball players had substantial shoulder problems at some point over one season, of which almost half of the handball players reported complete inability to participate (Asker et al., 2018). The prevalence of shoulder injury was significantly higher in female, compared to male adolescents. Due to the high injury rates, especially in youth handball, screening tests are warranted to identify players with the highest injury risk.

Although sports-related injuries are common in youth sports (Stein and Micheli, 2010), few modifiable risk factors have been evaluated in children and adolescents (Caine, Maffulli, and Caine, 2008). In male handball players a reduced total range of glenohumeral motion, glenohumeral external rotation weakness and scapular dyskinesia have been found to increase the risk of shoulder injuries (Clarsen et al., 2014). In addition, decreased neuromuscular control of the core muscles and ability to reposition the trunk after the onset of an external force have been shown to predict knee injury in female collegiate athletes (Zazulak et al., 2007).

It has been argued whether screening tests can be a useful tool to predict injuries due to lack of sufficient accuracy (Bahr, 2016; Bonazza et al., 2017; Bunn, Rodrigues, and Bezerra da Silva, 2019; Moran, Schneiders, Mason, and Sullivan, 2017). One of the most frequently used movement screening tools to predict injury risk is the Functional Movement Screen (FMS™) (Bonazza et al., 2017; Moran, Schneiders, Mason, and Sullivan, 2017). In a meta-analysis by Bonazza et al. (2017), it was found that the odds of sustaining an injury in the future was 2.7 higher with a score ≤ 14 points on the FMS™. Their results were supported by Bunn, Rodrigues, and Bezerra da Silva (2019), concluding that participants classified as “high risk” by FMS™ were 1.5 more likely to be affected by injury than those classified as having low risk. In contrast to Bonazza et al. (2017) and Bunn, Rodrigues, and Bezerra da Silva (2019), Moran, Schneiders, Mason, and Sullivan (2017) concluded that the weak association between the composite score in FMS™ and subsequent injury questions its use as an injury prediction tool. The 9+ screening battery (9SB) is another functional screening test, recently developed from six of the seven movement tests of the FMS™ (Frohm et al., 2012). It has been shown to have good inter- and intra-rater reliability (Frohm et al., 2012). The 9SB differs from the FMS™, in term of five additional tests and stricter scoring criteria of each test. Recently two studies (Bakken et al., 2018; Leandersson et al., 2018) have been published on the predictive values of 9SB. In both studies, the authors concluded that the 9SB could not be used to predict lower extremity injuries in orienteering or soccer.

Injury screening tests are mostly warranted in elite handball and especially in adolescents due to the high injury risk. Further, it is debated in the literature (Butler et al., 2013) whether some tests might be more valid than others in terms of predictability. It may therefore be important to investigate if both the complete or parts of the 9SB could be used as an injury prediction tool. This is also recommended by Flodstrom, Heijne, Batt, and Frohm (2016) and Frohm et al. (2012), who suggested that future studies explore and identify the most relevant tests of the 9SB in a specific population. To date, no study has explored the 9SB’s ability to predict injury in elite adolescent handball players. The aim of this study was therefore to evaluate the predictive value of the composite score and underlying factors of the 9SB for sustaining a non-contact injury in young elite handball players.

Methods

This prospective cohort study is part of a larger KASIP research program (Karolinska Athlete Screening Injury Prevention). In the KASIP research program, Swedish adolescent elite athletes were prospectively followed over 2 years in order to gain and deepen the knowledge about injury occurrence in youth elite sports. In the present study, data from the first year are reported.

Inclusion procedure

The National Federation in Handball in Sweden (NFHS) was invited as one of the several federations to participate in the KASIP study. After the agreement with the NFHS, the principal of the National Handball High School in Gothenburg the only National High school in handball was contacted and given oral and written information about the purpose and procedure of the study. One of the authors visited the school and arranged a meeting where all students received a full explanation of the research that participation was voluntary and that the students at any time could discontinue their engagement in the study. The students e-mail addresses were obtained and an e-mail asking if they were willing to participate was sent out. Out of 48 possible students, 45 accepted the invitation while 3 declined or did not respond at all.

Participants

The sample consisted of 45 (female/male 23/22) adolescent elite handball players (median age 17, range 16–18), competing at high national level of their age group. The average weekly training volume for the handball players was 9.2 h (SD 4.6) and 1.9 h (SD 1.4) exposure to match play.

The 9+ screening battery

The athletes were screened in October 2013 using the 9SB. The 9SB is designed to assess movement quality and to challenge an athlete’s neuromuscular control, mobility, and stability when performing fundamental movements (Frohm et al., 2012). The test consists of 11 tests (Deep squat, In-line lunge, Push up, Active hip flexion, Diagonal lift, Functional shoulder mobility, One legged squat, Seated rotation, Straight leg raise, Deep one-legged squat, and Vertical drop jump test). Pictures, detailed instructions and scoring criteria of each exercise have previously been described in a study on recreational athletes (Flodstrom, Heijne, Batt, and Frohm, 2016). The 9SB have shown good inter- and intra-reliability (0.80) (Frohm et al., 2012).

Screening procedure

The athletes were allowed a short warm-up of 5–10 min at a self-selected pace, of running or cycling, before the screening was performed. Prior to testing, the athlete was given standardized verbal instructions and shown pictures on starting position/ending position and expected performance. The athlete then performed the test three times and the best result was recorded, except for the vertical drop jump test, where the lowest score was recorded, according to Frohm et al. (2012). Additional verbal correction was given individually if necessary. Each test is rated on a scale from 3 to 0 where 3 represents a test performed “correct with no compensatory movements”, 2 “correct but with presence of compensatory movements,” 1 “not correct despite compensatory movements” and 0 if pain is present (Frohm et al., 2012). The maximum total score of the 9SB is 33, the higher score on the 9SB, the better function. For the tests that were performed unilaterally the left side was tested first. For unilateral tests, the lowest score from left and right tests is used when calculating the total score. The vertical drop jump test was performed with the athletes using their preferred shoes. All other tests were performed barefoot. The person conducting the screening was one physiotherapist with formal education from the center of elite sports in Sweden and who had approximately 1 year of experience in screening with the 9SB.

The Oslo sports trauma research center overuse injury questionnaire

For injury registration, the Norwegian to Swedish-translated version of the web-based Oslo Sports Trauma Research Center Overuse (OSTRC) Injury Questionnaire was used (Clarsen, Myklebust, and Bahr, 2013; Ekman et al., 2015). All new injuries were recorded and defined as “all new physical complaints resulting in reduced training volume, experience of pain, difficulties participating in normal training or competition, or reduced performance in sports”. New injuries that were “caused by collision with other player or object” were excluded. A substantial injury was defined as “all physical complaints leading to moderate or severe reductions in training volume, performance, or complete inability to participate in sports”.

The injury questionnaire was e-mailed to the participants every week for 52 consecutive weeks using the Questback online survey software (Questback V. 9.9, Questback AS, Oslo, Norway). Data from each questionnaire were stored as an individual record in a single database. If no response was registered, the participant received a reminder e-mail four days later. Upon entering the study participants were asked to fill out a questionnaire regarding demographics data such as age, sex, height, dominant throwing hand, etc. The response rate over the 52 weeks was 59.6%.

Data analysis

The substantial injury prevalence measure was determined by dividing the number of athletes reporting substantial injury with the number of questionnaire responses for each week. The average substantial injury prevalence across all weeks was then calculated to constitute the seasonal substantial injury prevalence. The proportion of athletes reporting new injury across the 52 weeks was also determined.

Factor analysis using the principal component analysis was used to examine the psychometric properties of the 9SB and thereby determine the factor structure of the 9SB (Williams, Brown, and Onsman, 2010). Varimax rotation was utilized to enhance the interpretation of the factors and to ensure the resulting factors are independent of each other. To assess the suitability of the respondent data, Kaiser-Meyer-Olkin (KMO) measure of sampling adequacy and Bartlett’s test of sphericity were conducted. Based on the anti-image correlation matrix, the test “diagonal lift” was excluded from the principal component analysis since the item was not considered to share common variance with the other items (KMO < 0.4). Factors were extracted based on eigenvalues (>1) and the scree plot analysis (factors above the break in the curve), resulting in that three factors were retained. ROC curves were generated for the three factors and for the 9SB and corresponding area under the curve (AUC) values were determined. The cutoff values determined to be optimal for injury screening were calculated using the formula: (1-sensitivity)² + (1-specificity)², where the cutoff score with the lowest value was chosen (Perkins and Schisterman, 2006).

Due to the non-normally distributed (positively skewed) nature of seasonal substantial injury prevalence data, as assessed by Shapiro-Wilk’s test (p > .05), a Mann–Whitney U test was conducted to explore differences at each cutoff value. Odds ratio for the proportion of athletes reporting new injury, with a 95% CI, was calculated for each cutoff value. A p-value < 0.05 was considered statistically significant. All analyses were performed using the SPSS software for Windows, version 24.0 (SPSS, Evanston, IL).

Results

Injury data and composite score on the 9SB

Of all athletes (n = 45), 64% (n = 29) reported at least one new injury over the season, with 29% (n = 13) of the handball players reporting more than two new injuries (range 0–6 new injuries). The median value for the seasonal substantial injury prevalence was 22% (25–75^th percentiles 6–41%), where female handball players reported a higher degree of substantial injury (median 39%, 25–75^th percentiles 15–59%) compared to male handball players (median 17%, 25–75^th percentiles 2–24%). For the total sample, the composite score of the 9SB was 23.5 (SD 3.4), where females had a mean value of 23 (SD 3.0) and males a score of 24 (SD 3.8).

Factor analysis

The KMO measure of sample adequacy was met (0.63) and Bartlett’s test of sphericity was significant (p = 0.005), indicating the sample was adequate for factor analysis. Consequently, exploratory factor analysis extracted three factors that explained a cumulative variance of 56%, where each factor contributed with 13–26% of the total variance (Table 1). The first factor had high loadings on trunk stability tests as well as lower extremity and trunk strength tests and was therefore considered to measure complex movements. The second factor had high loadings on flexibility tests such as trunk rotation and shoulder range of motion and was therefore considered to measure mobility. The third factor loaded three items, measuring lower extremity control based on squat and jump tests.

Table 1. Rotated for each test

Tests	Complex movement tests	Mobility tests	Lower extremity control tests
Deep one-legged squat	0.62	0.17	0.20
In-line lunge	0.49	0.49	0.26
Straight leg raises	0.76	0.22	0.16
Push up	0.84	0.13	0.24
Seated rotation	0.04	0.64	0.19
Functional shoulder mobility	0.02	0.81	0.22
Active hip flexion	0.13	0.48	0.07
Deep squat	0.01	0.06	0.82
One-legged squat	0.02	0.23	0.66
Drop jump	0.29	0.23	0.55

Bolding indicates tests with high loadings.

Injury risk

The 9SB and each factor’s ability to predict injury in handball players are presented as ROC curves (Figure 1). Based on the identified cutoff values, none of the factors or the complete 9SB could significantly (p = 0.43–0.98) predict the risk of a non-contact new injury or the risk of reporting substantial injury (Table 2). AUC values were ranged from 0.50 to 0.59, with the corresponding 95% CI including 0.50 for all factors.

Table 2. Injury risk and area under the curve (AUC) values, for the 9SB and for the detected factors alone

Total test and factors	Cutoff value	Odds ratio new injury (95% CI)	p-value	Median value (25–75^th percentiles) for seasonal substantial injury prevalence		p-value	AUC (95% CI)
				Lower cutoff value (x ≤)	Higher cutoff value (> x)
9SB	≤22	1.02 (0.29–3.59)	0.98	0.49 (0.15–0.60)	0.55 (0.31–0.79)	0.50	0.56 (0.38–0.73)
Complex movement tests	≤9	1.64 (0.48–5.62)	0.43	0.51 (0.33–0.78)	0.35 (0.17–0.79)	0.76	0.50 (0.32–0.68)
Mobility tests	≤6	1.20 (0.35–4.09)	0.78	0.52 (0.19–0.75)	0.49 (0.24–0.78)	0.79	0.59 (0.42–0.76)
Lower extremity control tests	≤6	0.74 (0.20–2.72)	0.66	0.49 (0.17–0.64)	0.58 (0.33–0.82)	0.49	0.53 (0.36–0.71)
9SB = 9+ screening battery; AUC = area under the curve

Figure 1. ROC curves for the 9SB and the three factors

Discussion

The main findings in the present study were that neither the composite score in the 9SB nor the scoring related to the three different factors: 1) complex movement; 2) mobility; and 3) lower extremity control could be used to predict non-contact injuries in young elite handball players. The AUC values indicate that the three different factors or the complete 9SB are “no better than chance” at discriminating injury in adolescent handball players.

Our results are in agreement with the results from studies of the 9SB (Bakken et al., 2018; Leandersson et al., 2018) and the FMS™ (Mokha, Sprague, and Gatens, 2016; Moran, Schneiders, Mason, and Sullivan, 2017; Smith and Hanlon, 2017), concluding no predictive validity of injury risk. For instance, the predictive value of the 9SB in orienteers and in soccer players has been found to be low or non-existing (Bakken et al., 2018; Leandersson et al., 2018). To date, no meta-analysis or systematic review on the 9SB can be found in the scientific literature. In one of the first studies on the FMS™ by Kiesel, Plisky, and Voight (2007) it was suggested that a composite FMS™ score of ≤14 could predict serious injury. This study has been widely cited and used in a clinical setting. Recently, their result was confirmed in a systematic and meta-analysis on the FMS™, concluding an almost threefold increased injury risk if an athlete score ≤ 14 (Bonazza et al., 2017). In contrast, meta-analyses by Dorrel, Long, Shaffer, and Myer (2015) and Moran, Schneiders, Mason, and Sullivan (2017) found no association between the composite score of the FMS™ and the risk of injury. The contradictory results are probably due to different inclusion criteria, for example Moran, Schneiders, Mason, and Sullivan (2017) excluded studies with cross-sectional and retrospective study designs.

The use of different injury definition and statistical methods is challenging when comparing the results of studies on predictive value and screening methods (Bahr, 2016) and may partly explain the different results on the predictive validity of the FMS™ (Dorrel, Long, Shaffer, and Myer, 2015). For instance, in Kiesel, Plisky, and Voight (2007) a time-loss definition was used, whereas in studies on the 9SB by Bakken et al. (2018) and Leandersson et al. (2018), injury was defined as “any physical complaint”. A time-loss injury definition probably identifies the most serious injuries but may also result in the likelihood that the majority of injuries sustained by athletes are missed (Bahr, 2009; Clarsen, Myklebust, and Bahr, 2013). In the present study, an “any physical complaint” injury definition was used, probably resulting in that a wide range of injuries were recorded, which should be considered as a strength rather than a weakness.

Cook, Burton, and Hoogenboom (2006) hypothesized that compensatory movement patterns could lead to poor biomechanics with the potential to cause micro- and macro-traumatic injuries and by using movement screening such compensatory movement patterns could be detected. In the present study, it was hypothesized that contact injuries are not a consequence of compensatory movement pattern, and therefore these injuries were excluded in the analysis. In seasonal play in handball, it has been reported that 40–60% of injuries are due to contact with another player or object (Giroto, Hespanhol Junior, Gomes, and Lopes, 2017; Seil, Rupp, Tempelhof, and Kohn, 1998; Wedderkopp et al., 1999). We explored the psychometric properties of the 9SB using two different injury definitions, new injury and substantial injury. However, regardless of injury definition, the predictive ability of the 9SB was limited.

The 9SB includes tests that demand flexibility, stability and also strength (Frohm et al., 2012). However, all exercises are performed in a self-chosen speed, which is usually slow, compared to the speed and rapid changes of directions, requiring accelerations and deceleration involved in most sports, including handball. Secondly, few tests involve shoulder strength, flexibility, and stability which, in handball, may have decreased the validity. On the contrary, the most frequently reported injuries in handball are lower extremity injuries (Olsen, Myklebust, Engebretsen, and Bahr, 2006). We believe that the tests of the 9SB are not discriminative enough for most elite athletes, involved in sports that require high speed and contact with other players. The drop jump test, included in the present study, is the only test not involving self-chosen speed. Recently, this test was used in a study by Krosshaug et al. (2016) to evaluate its predictive value. They concluded that the vertical drop jump test indicated a poor-to-failed combined sensitivity and specificity of the test, even when including previously injured players.

Methodological considerations

As proposed by Fuller et al. (2006) we used a prospective study design following the athletes over 52 weeks. Together with the weekly injury registration, this design minimizes the risk of recall bias (Bahr, 2009). By using a web-based self-reported injury registration questionnaire we also managed to survey the athletes over season at low cost and without the need of on-site medical teams to register injuries. Although we monitored the athletes over a complete year, the major limitation of this study is the response rate, since in average only 60% of all questionnaires were collected, affecting the internal validity of the study. There are no criteria for what constitutes a “high” response rate and monitoring this young population weekly using online questionnaires over a year is challenging. Based on Clarsen, Myklebust, and Bahr (2013), we believe the response rate may have underestimated the true substantial injury prevalence for the handball players. Importantly, we have no reason to think this had a different effect on substantial injury prevalence for players with score above/below the identified cutoff values for the 9SB.

A strength of the present study is the use of a broader definition of injury, suggested by Clarsen, Myklebust, and Bahr (2013) detecting more than 10 times as many injuries as using a “time-loss” definition. In addition, the cutoff values were calculated and not arbitrarily chosen. Still, there are some methodological considerations in the present study. No detailed diagnostic information on each injury was recorded and therefore no sub-group analysis of predicting different injury types could be performed. In addition, the cohort consisted of a limited sample of adolescent handball players at the only national sports high school in Sweden. The 9SB was only performed once and we do not know if the movement pattern of the handball players changed across the season. Since young handball players are rapidly improving their handball skills, performing the 9SB at multiple times across a season may provide a more accurate picture of movement patterns in young handball players. It is also suggested that the included tests involve high-speed component and tests for the upper extremity, which is lacking in the 9SB.

Conclusion

In summary, based on the identified cutoff scores on the individual factors (complex movements, mobility or lower extremity control) or the score on the complete 9SB could not predict the risk of a non-contact new injury as well as the risk of reporting substantial injury in adolescent elite handball players. It is therefore suggested that high-speed tests and more complex movements should be evaluated regarding their predictive values for injury risk in adolescent handball players.

Declaration of Interest

The authors declare no conflict of interest.

Additional information

Funding

This work was supported by the Centrum for Idrottsforskning [FO2016-0009].