Reliability of specific on-ice repeated-sprint ability test for ice-hockey players

Background: Repeated sprint ability tests are today widely used to evaluate the performance capability in team sports. Objective: The aim of this study was to determine the reliability of a specific ice hockey test, which indicates the agility and repeated-sprint ability of the players. Methods: Twenty four highly trained junior ice hockey players (age = 17.68 ± 1.52 years; BMI = 23.8 ± 1.92 kg · m) participated in the study. Each participant was assessed for specific on-ice repeated-sprint ability test 12 × 54 m with 30 s rest. Intraclass correlation coefficient (association between two repeated measurements) and coefficient of variation were calculated to assess the reliability of the test. Results: All intraclass correlation coefficients were .78 for sprint decrement and .98 for total time and the best time, the coefficient of variation was 1.52% for best sprint time, 1.31% for total time and 19.3% for sprint decrement variable. Conclusions: The results suggest the high reliability of the ice hockey agility test expressed by the best sprint time and repeated-sprint ability by the total time and less reliability of sprint decrement.


Introduction
Contemporary training process has witnessed a great deal of changes due to increasing scientific and expert knowledge about the nature of conditioning in team sports. The specificity of the physical training becomes a key factor in planning the training process, because of need to attain specific physiological adaptations as much as possible. That is why we should analyse the demands of the game by time motion analysis (Reilly, 2005) and adequately operate with a player's load intensity and activities during the training process.
Time motion analysis research has pointed out ice hockey as a high speed intermittent invasion game similar to the other games (Buchheit, Lefebvre, Laursen, & Ahmaidi, 2011;Montgomery, 1988). The game is characterised by irregular changes between high and low intensity activities over very short periods of time (up to 10 s) with high-impact body contact. Low intensity activities are related to the recovery processes (Balsom, Seger, Sjodin, & Ekblom, 1992;Quinney et al., 2008;Vescovi, Murray, & VanHeest, 2006). Each on-ice shift last between 30 and 80 s interspersed with 2 to 5 min of recovery between shifts (Arnett, 1996;Quinney et al., *Address for correspondence: Karel Hůlka, Department of Sports, Faculty of Physical Culture, Palacký University, tř. Míru 115, 771 11 Olomouc, Czech Republic. E-mail: karel.hulka@upol.cz a new performance test tools. Therefore tests of multiple sprints performance or repeated-sprint ability are becoming increasingly popular as a way of evaluating the performance capabilities of athletes involved in field and court sports (Glaister et al., 2009).
For ice hockey use (Leone, Leger, Lariviere, & Comtois, 2007) Power, Faught, Przysucha, McPherson, and Montelpare (2012) designed and validated the high intensity anaerobic intermittent tests with very short work and rest intervals for ice hockey players called Repeat Ice Skating Test (RIST), where the participants completed 6 × 49 m circles with 10s recovery between trials. As a first Reed et al. (1979) published this type of test called Reed Repeat Sprint Test (RRS) consists of 6 × 91 m sprints with 30 s recovery. This test protocol was criticized for being too exhausting (Nightingale, Miller, & Turner, 2012). Carey, Drake, Pliego, and Raymond (2007) used the test protocol with 5 × 1 lap sprints with a 30 s recovery period between laps.
We have constructed the specific on-ice repeated ability test (RAT) for ice hockey players. In this RAT we have tried to simulate the game demands on ice hockey players. Firstly, according to Léger, Seliger, and Bassard (1979) who showed lower mechanical efficiency (7.9%) of runners compare to skaters during the test is important to get valid data about repeatedsprint ability test for skaters. Secondly there is no evidence about performance test which is constructed to indicate repeated-sprint ability as well as use of the agility patterns (starts, repeated sprints with change of direction while sprinting, stops, and different types of skating) during the same sprint activity as during the ice hockey game. The main difference among our constructed protocol and the others test protocols (Carey, Drake, Pliego, & Raymond;Power et al., 2012;Reed et al., 1979) we see in higher specificity of the load thanks to recovery time, shorter sprints, more changes of direction, and different types of skating during each trial. In addition, before using the new performance test effectively, the researchers should determine reliability of the measurement Spencer, Fitzsimons, Dawson, Bishop, and Goodman (2006).
Many authors (Oliver, Williams, & Armstrong, 2006;Psotta, Bunc, Hendl, Tenney, & Heller, 2011) found the reliability of fatigue index as moderate. The reason can be pacing effect (Oliver, 2009) or small familiarisation (Oliver, 2007). According to Oliver (2007) we chose the sprint decrement as an indicator of fatigue. For reliability measurement design we followed by recommendations of Atkinson and Nevill (1998) and Hopkins, Schabort, and Hawley (2001), and used indicators of relative (intraclass correlation coefficient) and absolute reliability (coefficient of variation and Bland and Altman's 95% limits of agreement). The aim of this study was to construct and determine the reliability of the specific ice hockey test indicating repeated-sprint ability of the players.

Participants
Twenty four highly trained junior ice hockey players from one team of the highest junior division in the Czech Republic (age = 17.7 ± 1.5 years; height = 174.5± 6.1 cm; weight = 72.5 ± 6.8 kg; BMI = = 23.8 ± 1.9 kg · m -2 ) volunteered to participate in the present study. The players trained on average five times a week plus matches. The aims and objectives of the present study have been clarified to all participants and all of them gave a voluntary written informed consent. Participation was voluntary and players could withdraw from the study at any time. Goalkeepers were excluded from the reliability analysis because of their entirely different type of performance during the ice hockey match.

Procedures
All participants completed two repetitive measurements of the specific repeated-sprint ability test for ice hockey with the full ice hockey equipment. Two measurements were made one week apart always after day off. The research was held after the regular competition season. All trials were performed at the same day-time to minimise the effect of diurnal variability. The subjects were instructed not to take any food two hours before testing. All measurements were held at an ice hockey arena with standardized size of the ice hockey rink (60 × 28 m).

Specific repeated-sprint ability test protocol
A week before the testing all participants were required to perform trial sprints to become familiar with the track and two sprints to determine the length of the recovery period between the sprints. To develop good specific RSA test, it is necessary to know the work rate profiles of players during the match (Psotta, Blahus, Cochrane, & Martin, 2005). The work rest ratio was stated from the results of time motion analysis (Noonan, 2010) to work/rest ratio = 0.43. The average sprint time of the trials was 12.74 ± 0.78 s and recovery period was determined by 30 s.
The average duration of one short sprint during the match was from 2.0 to 3.5 s (Green et al., 1976). That is why three changes of direction were set. Standard warm up procedures on both test days consisted of 15 minutes active warm up and thereafter five short sprints (length of the court) followed by five minute recovery time during which was done the test protocol description.
The participants were required to perform maximal 12 × 80 m sprint trials according to Three variables from the raw data were calculated: the best sprint time (s), sprint decrement (%) as an indicator of repeated-sprint ability and total time (s). Sprint decrement (S dec ) was calculated according to Girard, Mendez-Villanueva, and Bishop (2011): where S 1-12 are times of ten single sprints, S best is the best time from the single sprints. Total time was calculated as a sum of twelve single sprint times.
Statistical analyses SPSS software (Version 17; SPSS Inc., Chicago, IL, USA) was used to process the data. Descriptive statistic mean and standard deviation was used to describe participant's performance during the test. The homoscedasticity was assessed by using the Pearson product moment correlation coefficient (the absolute difference Figure 1. Graphic illustration of the specific on-ice repeatedsprint ability test between two repeated assessments vs. the mean of the two repeated assessments). The intraclass correlation coefficient ICC = (MSS-MSE) · MS -1 , where MSS is the mean square of the subjects, MSE is the mean square of the error, and MS is the mean square, was calculated as an indicator of intra-subject reliability. The coefficient of variation and Bland and Altman's 95% limits of agreement for two repeated measurements were used to express and assess the reliability (absolute reliability) in the original measurement unit (Atkinson & Nevill, 1998). The coefficient of variation (CV) was calculated as a percentage expression of typical error of measurement (TEM) according to Hopkins, Schabort, and Hawley (2001). To calculate TEM, the following formula was used (Hopkins et al., 2001): TEM = S diff /√2, where S diff is the standard deviation of difference score between two measurements.
To determine the reliability of changes of times during the all tests the standard error of measurement (SEM) was used. To calculate the SEM, the following formula was used (Thomas, Nelson, & Silverman, 2005): SEM = SD · √(1−ICC) where SD is the standard deviation of the sample and ICC is the calculated intraclass correlation coefficient (the unbiased error). The ICC was calculated from the differences (D i ) between average times of each trial (X i ) in first and second measurement by D i = X i+1 -X i , where i =1, 2, …, 10. The statistical significances of all parts of the analysis were determined at an alpha level of p < .05.

Results
In the first measurement the mean of the total time from ten sprints (TT) was 129.8 ± 9.9 s, the mean of the best sprint time (BT) was 12.45 ± 0.9 s and the mean of the sprint decrement 5.4 ± 1.7% (opposite) compare to TT = 131.5 ± 11.2 s, BT = 12.43 ± 0.9 s and S dec = 4.7 ± 1.3% from the second measurement. The individual times are presented in Table 1. Table 2 presents the results of the reliability analysis. The results of correlation analysis show high relative reliability of the repeated measurements. Especially the high value of the ICC indicates very high level of intrasubject reliability in all monitored variables. Similarly Pearson's correlation coefficient is high in all variables. It indicates a high association between the results of both repeated measurements.
Similar tendency show the indicators of absolute reliability. The change in the mean was minimal in all variables. The participants achieved better values in BT and S dec in the second measurement, and TT had negative tendency. Coefficient of variation was 1.52% for total sprint time, 0.19% for best time and 0.69% for sprint decrement. Similarly low differences were found between results expressed by Bland and Altman's 95% limits of agreement which refer to ± 4.77 s, ± 0.53 s and ± 1.91% respectively.
The results of standard error of measurement SEM as an indicator of the reliability of changes of times during the all test was 0.11 s. It represented only 0.88% of the average measured time (13.07 s).

Disscussion
Reliability is a theoretical concept that is used to describe the quality of a measurement instrument. The aim of this study was to assess the reliability of Note. BEST1 = best time in the first measurement, BEST2 = best time in the second measurement, TT1 = total time in the first measurement, TT2 = total time in the second measurement, S dec1 = sprint decrement in the first measurement, S dec2 = sprint decrement in the second measurement.
the specific repeated-sprint ability test protocol for icehockey players. The small differences were found between the means of two measurements. The players achieved worse result during the second measurement in TT, but better results in BT and S dec . These changes can be explained due to the previous experience with the test (known as the learning effect). Nevertheless we consider the changes in the means of two measurements as very small and acceptable for all indicators.
To rate BT, the relative reliability (expressed by ICC) shows very close relation between two measurements (r = .98). Results are widely comparable with other repeated-sprint ability studies. Oliver (2009) presented CV ranging 0.7 to 2.7% for BT. Glaister et al. (2009) presented very good reliability r = .91 for 40 m Shuttle run test and r = .94 for RMAT by Haj-Sassi et al. (2011). The within subject variation expressed by coefficient of variation show small differences between two measurements too. The CV = 1.52% shows very small error of measurement similarly as CV = 0.82% of Wragg et al. (2000) for The Maximal Anaerobic Running Test (MART). Oliver, Armstrong, and Williams (2009) recommended CV for BT lower than 3%. The researchers and practitioners should utilize results of Bland and Altman's 95% limits of agreement during the interpretation of the results. Therefore it is necessary to take into account determined value of ± 0.53 s. Its magnitude should be considered as negligible difference between two measurements.
The sprint trial was made as an agility pattern of ice hockey performance. We think that BT could represent an indicator of ice skating (locomotion) skills of the players. To use this, verification is needed. Our test protocol, where we put together the measurement of ice skating (locomotion) skills and repeated-sprint ability, could save the time during practice.
Assessment of fatigue was expressed by S dec as a fatigue index and TT as a global RSA performance -ability of the player to sustain his power output at high-intensity, there was a very close relation of the ICC between two measurements too. The absolute reliability values of TT seems to be high and satisfactory as the results of Haj-Sassi et al. (2011), who had gained 0.93% for RMAT test and Spencer et al. (2006) 0.70% for field hockey test respectively. Moreover the value of CV equal to 5% is referred (Buchheit et al., 2011) as a criterion of good reliability and sensitivity, Hopkins, Hawley, and Burke (1999) refer 3%. Total time variable of our study accomplished this criterion like above mentioned studies. The ICC shows similar reliability as Boddington, Lambert, Gibson, and Noakes (2001) .98 for 5 m shuttle run test respectively. Similarly the results of relative reliability indicate high relationship between both measurements. We can use total time of our test protocol for repeated-sprint ability measurement.
The value of CV for S dec is quite higher (19.3%) than other variables because of small percentage of sprint decrement, but similar to Gabbett (2010) and Spencer et al. (2006) and lower than 38.7% of Glaister et al. (2009).
To reach relevant and credible results, the maximal effort of players during the measurements is needed. For the control of the effort the maximum of difference among every sprint during each measurement and the Note. TT = total time, BT = best time, S dec = sprint decrement. Mean diff = mean of difference between test-retest measurements, s diff = standard deviation of difference between test-retest measurements, ICC = intraclass correlation coefficient, 95% CI = 95% confidence interval, R = Pearson product moment correlation coefficient, TEM = typical error of measurement, CV = coefficient of variation. mean of two trial sprints a week before first measurement had to be less than 10%. The observed value of the standard error of measurement of 0.11 s indicates the satisfactory quality of this instrument. This value appears to be insignificant in view of the purpose of the test protocol. The SEM shows relative stability of fatigue during skating and tendency to decrease the effort of participants during the test.
As a limitation of this study we considered only two performed repeated measures and the possible changeability of the distance covered due to skating backwards.

Conclusions
In this study we proved the high reliability of our test for ice hockey players expressed by the best sprint time variable. When analysing test as specific repeatedsprint ability indicator, we found out that the total time variable was highly reliable. On the other hand, the sprint decrement variable is a less reliable indicator for repeated-sprint ability similar to other RSA test studies. Our RSA test can be used as a specific ice hockey repeated-sprint ability test on the ice.