Article Text

Download PDFPDF

King-Devick concussion test performs poorly as a screening tool in elite rugby union players: a prospective cohort study of two screening tests versus a clinical reference standard
  1. Gordon Ward Fuller1,
  2. Matthew J Cross2,3,
  3. Keith A Stokes3,
  4. Simon P T Kemp2
  1. 1 Centre for Urgent and Emergency Care Research, University of Sheffield, Sheffield, UK
  2. 2 Rugby Football Union, London, UK
  3. 3 Department for Health, University of Bath, Bath, UK
  1. Correspondence to Dr Gordon Ward Fuller, Emergency Medicine Research in Sheffield Group, School of Health and Related Research, University of Sheffield,Sheffield, S1 4DA, UK ; g.fuller{at}sheffield.ac.uk

Abstract

Background The King-Devick (KD) test is an objective clinical test of eye movements that has been used to screen for concussion. We characterised the accuracy of the KD test and the World Rugby Head Injury Assessment (HIA-1) screening tools as methods of off-field evaluation for concussion after a suspicious head impact event.

Methods A prospective cohort study was performed in elite English rugby union competitions between September 2016 and May 2017. The study population comprised consecutive players identified with a head impact event with the potential to result in concussion. The KD test was administered off-field, alongside the World Rugby HIA-1 screening tool, and the results were compared with the preseason baseline. Accuracy was measured against a reference standard of confirmed concussion, based on the clinical judgement of the team doctor after serial assessments.

Results 145 head injury events requiring off-field medical room screening assessments were included in the primary analysis. The KD test demonstrated a sensitivity of 60% (95% CI 49.0 to 70) and a specificity of 39% (95% CI 26 to 54) in identifying players subsequently diagnosed with concussion. Area under the receiver operating characteristic curve for prolonged KD test times was 0.51 (95% CI 0.41 to 0.61). The World Rugby HIA-1 off-field screening tool sensitivity did not differ significantly from the KD test (sensitivity 75%, 95% CI 66 to 83, P=0.08), but specificity was significantly higher (91%, 95% CI 82 to 97, P<0.001). Although combining the KD test and the World Rugby HIA-1 multimodal screening assessment achieved a significantly higher sensitivity of 93% (95% CI 86% to 97%), there was a significantly lower specificity of 33% (95% CI 21% to 48%), compared with the HIA-1 test alone.

Conclusions The KD test demonstrated limited accuracy as a stand-alone remove-from-play sideline screening test for concussion. As expected with the addition of any parallel test, combination of the KD test with the HIA-1 off-field screening tool provided improved sensitivity in identifying concussion, but at the expense of markedly lower specificity. These results suggest that it is unlikely that the KD test will be incorporated into multimodal off-field screening assessments for concussion at the present time.

  • rugby
  • cohort study
  • testing
  • diagnosis
  • concussion

Statistics from Altmetric.com

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.

Introduction

Concussion is a common and high-profile injury in collision sports.1 Due to the variability and subtlety of symptoms and signs, and pressure on athletes to continue playing, identification of sports-related concussion is challenging and injuries may go unrecognised or be ignored.2 Elite sports, including rugby union, have introduced management systems to identify and manage head impact events with the potential for concussion during matches.3 These typically involve brief, off-field, initial screening for a possible concussion, rather than definitive diagnosis of a head injury. However, a recent systematic review supporting the 5th Consensus statement on concussion in sport was unable to make an evidence-based recommendation for any single screening test.1 4 A multimodal approach, based on the Sports Concussion Assessment Tool (SCAT),5 incorporating different subtests conducted in parallel was advocated.4 Elite rugby’s current operational solution is the Head Injury Assessment (HIA-1) off-field screening test, an abridged version of the SCAT-3.3

Visual and eye movement neuronal pathways may become impaired following brain trauma.6 The King-Devick (KD) test, an oculomotor test originally designed for reading evaluation, has been promoted as a concussion screening tool.7 Preliminary studies have demonstrated a worsening of performance from baseline in patients with concussion.6 However, a 2015 systematic review concluded that ‘The quality of evidence is not yet sufficient to warrant clinical recommendations for the use of oculomotor based vision measurement either as an indicator of mild traumatic brain injury or as a measure of recovery following mTBI’.8 Currently the HIA-1 off-field screening test does not include an assessment of oculomotor function.

The aim of this study was to validate the KD test for identifying players with concussion in elite adult male rugby union. The primary objective was to characterise the stand-alone accuracy of the KD test in identifying concussion. As set out in the World Rugby HIA protocol, concussion was determined by the team physician following clinical assessment at serial time-points.9 Secondary objectives, designed to inform the future composition of the HIA-1 off-field screening test, included comparing the sensitivity and specificity of the KD and HIA-1 off-field screening test and evaluating the combined performance of the KD and HIA-1 screening tests.

Methods

Study design, setting and study population

A prospective cohort study was performed in the top two English elite domestic rugby competitions (Premiership and Championship, 24 teams) in a single season between September 2016 and May 2017 to determine the accuracy of the KD screening test for concussion. To maximise internal validity, the study followed expert recommendations on the conduct and reporting of diagnostic accuracy and reliability studies.10–12

The source population comprised consecutive male adult players entering the World Rugby HIA process after identification of meaningful head impact events with the potential to cause concussion. The HIA process has been described previously.9 Briefly, players overtly demonstrating signs of concussion (eg, loss of consciousness, tonic posturing or ataxia) were immediately and permanently removed from the remainder of the match, without undergoing further off-field concussion screening. Where the consequences of a head impact event were not clear, players underwent an off-field screening assessment for possible concussion with the multimodality HIA-1 screening instrument, comprising Maddocks questions, tandem gait test, immediate and delayed recall, a symptom checklist, and brief evaluation of clinical signs. Any abnormality in the HIA-1 screening test mandates removal from play. The main study population included players undergoing off-field HIA-1 screening, as these are the players who could potentially benefit from KD testing within the HIA process. However, in other elite sports, all players undergo off-field screening following head impact events regardless of presenting signs. Players immediately and permanently removed from play were therefore also included in a subsequent combined analysis to increase the potential generalisability of the findings.

Index test

The KD test is an objective clinical test of rapid eye movements, primarily evaluating brain pathways involved in saccadic eye movements, attention and language.7 13 The test involves reading aloud a series of random single-digit numbers displayed in rows on three successive screens in a tablet application following familiarisation based on a practice screen. Athletes begin at the top left of each screen and read as quickly as possible from left to right across each row. The spacing between the rows of numbers becomes narrower on each successive screen, requiring increased concentration and more accurate eye movements to avoid errors. The time taken is automatically kept for each test and the KD summary score for the entire test is based on the cumulative time taken to read all three test screens. The number of uncorrected errors, defined as any addition, omission or reversal of the number pattern, is also recorded. A preseason baseline KD performance is established by the better of two consecutive trials. Post-head impact event results are then compared with the subject’s baseline. Any worsening of time and/or errors committed indicate an abnormal result.

Reference standard for the diagnosis of concussion

All players who entered the World Rugby HIA process underwent detailed medical assessments postmatch (HIA-2 assessment) and after two nights’ rest (HIA-3 assessment) to monitor clinical progress and to confirm (or refute) a diagnosis of concussion by the team doctor. The HIA-2 assessment consisted of a clinical evaluation including the SCAT-3 instrument. The HIA-3 assessment comprised a clinical evaluation, supported by an expanded SCAT-3 symptom checklist, a cognitive assessment (typically a computerised neurocognitive tool such as CogSport), and a balance assessment using the balance error scoring system and tandem gait balance tests. The reference standard, against which the accuracy of the KD test was compared, was a clinical diagnosis of concussion during the 48 hours post injury, based on abnormal HIA-2 and/or HIA-3 assessments, determined by the team doctor.9

Data collection and procedures

Medical staff completed a web-based training session led by King-Devick Technologies prior to participating in the study. Following this training, ongoing technical support was provided by King-Devick Technologies, with study-specific support given by the research team. Players from included teams received baseline KD testing preseason by recording the best time (fastest) of two consecutive trials in a representative off-field setting during a training session. Following a meaningful head impact event, the KD test was repeated. The KD test was performed by the team doctor in a dedicated medical room, after completion of the usual World Rugby HIA-1 screening test or following immediate and permanent removal with clear signs of concussion.9 KD test time and errors were recorded using a proprietary tablet application. The KD test was used non-operationally after the conclusion of the standard HIA-1 assessment process. Results were not displayed immediately, but due to the KD application design were accessible to clinicians. Team doctors were specifically instructed not to look at the results, or allow findings to influence return-to-play decisions. KD data were recorded contemporaneously using tablets and the web-based proprietary KD software platform. HIA process data were routinely collected at the point of assessment using the tablet-based, web-hosted, CSx data platform,14 with data subsequently linked to the World Rugby and English Rugby Football Union HIA databases. KD and HIA data were linked deterministically using unique player identifiers.

Analyses

Sample characteristics and the distribution of baseline KD scores were initially examined using descriptive statistics. Repeatability of baseline KD testing was evaluated using repeatability coefficients.15 The accuracy of an abnormal KD test result (prolonged time from baseline and/or errors) for detecting concussion was then assessed in the primary analyses. Each case was coded according to the index test and reference standard result, with a 2×2 contingency table constructed to determine true positives, false positives, true negatives and false negatives. Prevalence of concussion, sensitivity and specificity, positive and negative predictive values, positive and negative likelihood ratios, and diagnostic ORs with their 95% CIs were calculated. The accuracy of prolonged KD time in isolation was examined by calculating the area under the receiver operating curve (AUROC). Youden’s J statistic was used to attempt to identify an optimal threshold cut-point for prolonged KD times.16 17 Accuracy of errors alone was also examined independently through calculation of sensitivity and specificity. This primary analysis was initially performed for players requiring off-field screening following head impact events where the consequences were not clear, but was also repeated in a combined sample also including players immediately and permanently removed from play after demonstrating clear signs of concussion.

A number of secondary analyses were performed to compare agreement and accuracy between the KD test and current World Rugby HIA-1 screening tests (raw agreement/Fleiss’s kappa and McNemar’s test, respectively); demonstrate the combined performance of the KD and HIA-1 screening tests when performed in parallel (sensitivity and specificity); and evaluate the reproducibility of preseason and postseason KD testing (Bland-Altman limits of agreement analysis).18 Additional sensitivity analyses investigated the potential influence of clustered (clustered sandwich estimator for SEs) and missing data (scenarios with different assumptions for cases with missing data).19

Sample size, statistics and ethics

A sample size calculation of 207 players undergoing off-field concussion screening assessments was calculated for the primary analysis, using Bruderer’s method based on a conventional α of 0.05 and the following assumptions from previous HIA data: a prevalence of concussion of 30% in players with meaningful head impact events requiring HIA-1 concussion screening assessment3 20; a sensitivity of 90%; a specificity of 75% for prolonged KD test times to identify concussion; and a desired precision of ±7.5% for the 95% CI of the sensitivity estimate. This sample size would provide a 95% CI precision of ±7.0% for specificity and ±0.05 for an AUROC of 0.83.

Available case analyses were performed with sample size determined by the number of players with complete data for each analysis. Statistical analyses were carried out in Stata V.13.1 with a conventional significance level (α) of 0.05 used. All players provided informed consent for participation prior to the start of the season. All data were anonymised. The KD application, technical support and KD test data management were provided free of charge by King-Devick Technologies. Statistical analyses were performed independently of the RFU and KD at the University of Sheffield according to a prespecified protocol.

Results

Derivation and characteristics of study participants

A total of 274 consecutive head impact events with the potential to cause concussion were detected in 261 players (ie, 13 players had two head impact events) during 264 matches in the 2016/2017 season. Of these 73 incidents (in 67 players) were associated with overt signs or symptoms of concussion requiring immediate and permanent removal from play. The remaining 201 incidents (in 196 players), where it was unclear if a meaningful head impact event had occurred, underwent off-field medical room screening assessments. Figure 1 presents a flow chart describing the derivation of study participants.

Figure 1

Derivation of study participants in primary analysis. FN, false negative; FP, false positive; HIA-1, Head Injury Assessment; KD test, King-Devick test; TN, true negative; TP, true positive.

The mean age of players in the complete sample was 27.6 years (SD 2.6), with a mean height of 187 cm (SD 6.9) and mean weight of 105 kg (SD 11.8). Of the players, 61.1% were forwards, with 38.9% backs. A wide range of mechanisms of injury were observed, with head contact during tackles predominating (n=225, 60.7%, either being tackled or tackling). The distribution of baseline KD results was slightly positively skewed, with a median time of 44.3 s (IQR 38.5–50.9, range 28.3–73.9 s, n=207). The KD test demonstrated a small improvement on average in preseason testing, with a mean decrease of 1.75 s across the two baseline trials (paired t-test, P<0.001). The repeatability coefficient was 13.9 s, indicating that the absolute difference between the two baselines differs up to this value on 95% of occasions. The second baseline trial was slower in 24.1% of players. Player characteristics are shown in table 1.

Table 1

Characteristics of participants in primary analysis

Baseline KD test data were missing in 20.1% of the 261 included players. Across the overall sample of 274 head impact events, variable-wise missing data rates were as follows: HIA-1 test: 10.5%; KD index test: 27.4%; reference standard: 7.3%. Case-wise missing data rates for each analysis are shown in table 2.

Table 2

Statistical metrics describing the accuracy of KD and HIA-1 tests

Primary analysis

Of the 201 incidents requiring off-field medical room screening assessments, there were missing data on index test or reference standard results, in 56 (28%), leaving 145 head impact events for inclusion in an available case analysis. Ninety-four of the included events had a confirmed final clinical diagnosis of concussion, giving a target disorder prevalence of 65.0% (95% CI 56.0% to 72.6%). The distribution of post-head impact event KD test times did not differ significantly between concussed and non-concussed players (median increase in KD test time from baseline +1.15 s, IQR −3.9 to +5.0 vs +0.7 s, IQR −2.8 to +6.4, respectively; P=0.62). The proportion of KD test errors was also not significantly different between concussed and non-concussed players  (11.7% vs 13.7%; P=0.72).

Of players with concussion, 56 had an abnormal KD tests (true positives) resulting in a sensitivity of 59.6% (95% CI 49.0% to 72.6%). Fifty-one players were reference standard-negative with no confirmed concussion, of which 20 cases were classified as true negatives with normal KD results. The specificity to correctly identify players without concussion in this study group was therefore 39.2% (95% CI 25.8% to 53.9%). The positive and negative predictive values of the KD test were 64.4% (95% CI 53.4 to 74.4) and 34.5% (22.5% to 48.1%), respectively. Figure 2 and table 2 summarise the performance of the KD test and present the point estimates of metrics of test accuracy with their precision. There were no obvious distinguishing features of false-negative cases.

Figure 2

Forest plots summarising sensitivity and specificity results for different screening tests and study populations. HIA-1, Head Injury Assessment-1; KD test, King-Devick test.

Prolonged KD test times were unable to discriminate between concussion and no concussion in players undergoing off-field screening following meaningful head impact events, with the receiver operating characteristic curve close to the identity line and an area under the curve of 0.51 (95% CI 0.41 to 0.61). No optimal cut-point for prolonged KD test time was evident, with a Youden’s Index of 0.11 (95% CI 0.0 to 0.28) at the best empirical cut-point of 2.15 s. The KD test conventionally measures both time and number of errors. However, ignoring errors, sensitivity and specificity of prolonged KD time alone for concussion were 54.3% (95% CI 43.7% to 64.6%) and 45.1% (95% CI 31.1% to 59.7%), respectively. Sensitivity of errors in isolation was low (11.7%, 95% CI 6.0% to 20.0%), but specificity was higher at 86.3% (95% CI 73.7% to 94.3%).

Of the 73 incidents where players were immediately and permanently removed from play with clear signs of concussion, 54 (including 19 incidents with loss of consciousness, 4 with tonic posturing, 4 with ataxia and 17 with confusion) underwent immediate off-field KD testing. Of these, 21 players (38.9%, 95% CI 26.6% to 52.8%) passed the KD test with a quicker than baseline time and no errors. Across the combined available case sample of consecutive meaningful head impact events with the potential to cause concussion (including both incidents with clear signs of concussion and those where the consequences of the head impact event were unclear, n=199), the sensitivity and specificity of KD test in diagnosing concussion were 60.1 (95% CI 51.8 to 68.1) and 39.2 (95% CI 25.8 to 53.9), respectively.

Secondary analyses

Of the 201 incidents requiring off-field concussion screening, 21 had missing index test or reference standard data for assessment of HIA-1 screening test accuracy, giving an available case sample of 180 head impact events. Sensitivity of the HIA-1 screening test was higher than the KD test at 74.8% (95% CI 65.6% to 82.5%), although this did not reach statistical significance (McNemar’s test, P=0.08). Conversely, HIA-1 specificity was significantly better than the KD test at 91.3% (95% CI 82.0% to 96.7%, McNemar’s test, P<0.001).

The HIA-1 and KD tests, conducted in parallel, showed no agreement beyond chance (raw agreement 46.2%, Fleiss’s kappa −0.08, P=0.34). Combining HIA-1 and KD test performance, into a parallel joint off-field assessment, generated a sensitivity of 92.6% (95% CI 85.9% to 96.7%) with a specificity of 33.3% (95% CI 20.8% to 47.9%, n=159). This combined sensitivity was significantly better compared with either the KD test (McNemar’s test, P=<0.001) or HIA-1 test (McNemar’s test, P<0.001) alone. Combined specificity was significantly lower than the HIA-1 test alone (McNemar’s test, P<0.001), but did not differ significantly from the KD test used in isolation (McNemar’s test, P=0.25). Separately combining KD errors or prolonged time individually with HIA-1 screening results revealed sensitivities of 80.6% (95% CI 71.8 to 87.5) and 88.9% (95% CI 81.4 to 94.1), and specificities of 76.5% (95% CI 62.5 to 87.2) and 37.3% (95% CI 24.1 to 51.9), respectively (n=159).

Bland-Altman limits of agreement analysis revealed a mean improvement of 1.69 s (95% CI −3.2 to −0.1 s) and 95% limits of agreement of −11.4 to +8.0 s between baseline and postseason tests in non-concussed players (single team, n=40). Sixty-five per cent (95% CI 48.4% to 78.6%) of these healthy players ‘failed’ their postseason KD test with a slower time.

Scenario analyses investigating the potential influence of missing data indicated that KD performance remained lower than the HIA-1 test even when assuming a missing data pattern most favourable to KD test performance (ie, all missing KD test results being correct, prevalence of concussion 60%). The ‘best case’ sensitivity and specificity for the KD test estimates were 70.3% (95% 61.6% to 78.1%) and 57.5% (95% CI 45.4 to 69.0). Further sensitivity analyses exploring clustered data did not alter point estimates and negligibly affected 95% CI coverage.

Discussion

The KD test demonstrated a sensitivity of 59.6% and specificity of 39.2% for the presence of clinically diagnosed concussion in elite rugby players. Given the reported prevalence, team doctors would be between 35% and 48% sure that a player did not have concussion following a negative KD test at the 95% confidence level. This performance compared less favourably than the current World Rugby HIA-1 off-field screening tool (sensitivity 74.8%, P=0.08; specificity 91.3%, P<0.001). Combining the KD test and the HIA-1 tool in parallel provided a multimodal assessment with a higher sensitivity of 92.6%, but significantly lower specificity of 33.3% than the HIA-1 test alone (P<0.001).

Strengths and limitations

This study is the largest prospective investigation of the KD screening test for sports-related concussion published to date, and has a number of strengths. Consecutive players were recruited following suspicious head impact events, avoiding the bias inherent in a diagnostic case–control study designs commonly used in previous KD studies. The index tests and reference standard were independently applied with no potential for incorporation, partial or differential verification biases. Furthermore, the reference standard was determined after serial standardised examinations by experienced team physicians, minimising the risk of reference standard misclassification.

Conversely, there are a number of limitations that could challenge internal validity. First, there were missing data on baseline, off-field tests and reference standard results. These were predominantly secondary to non-systematic reasons such as missing baseline KD times in injured, absent or transferred players. Furthermore, there were no distinguishing characteristics of excluded head impact events, and diagnostic accuracy metrics for the HIA-1 off-field screen are consistent with previous studies. Sensitivity analyses indicated that the KD test may have improved diagnostic accuracy metrics if there were systematic reasons for missing data. However, the findings of the comparison between the KD test and HIA-1 screening tool would not be materially altered, even in a best case scenario assuming a missing data pattern most favourable to KD test performance. Taken together this suggests that the findings are robust to selection bias.

Second, there is the possibility of diagnostic review bias. Although KD test results were not initially displayed until after the completion of the HIA-1 process, it was possible for team doctors to access these data later, or form a subjective opinion based on qualitative KD test performance, potentially influencing their diagnostic assessment. Unfortunately due to operational and competitive imperatives, completely separate index and reference standard assessment was not possible. The KD test was conducted after the HIA-1 tool, but prior to communicating return-to-play decisions. There was minimal agreement between HIA-1 and KD test results, indicating that it is unlikely that interpretation of the KD test was influenced by the preceding findings; however, it is possible that pending return to play decisions were anticipated by players, influencing their subsequent KD test performance.

Third, as acknowledged in the Berlin consensus document, the diagnosis of concussion may be challenging. Misclassification of the reference standard by inaccurate clinical assessment could therefore lead to errors in the reported accuracy metrics. Furthermore reference standard misclassification could have arisen from players deliberately concealing symptoms to avoid missing games through graduated return-to-play protocols. Finally, the study is relatively underpowered with imprecise results.

Comparison with previous studies 

Three systematic reviews have previously examined the performance of the KD test in sports-related concussion, including 10 individual studies.4 7 8 More recently Molloy and colleagues21 performed a diagnostic case–control study in semiprofessional rugby union. Baseline KD results from the current study (44.3 s) were consistent with the 43.8 s (95% CI 40.1 to 47.5) reported in the recent meta-analysis by Galetta.7 The observed improvement in times between baseline KD trials was also very similar to those previously reported. Published KD accuracy results were imprecise and heterogeneous, with sensitivity estimates ranging from 53% to 100%.4 These studies were at high or unclear risk of bias secondary to case–control study designs, test review bias, inaccurate reference standards, or inappropriate interval between index test and reference standard, making comparison of results difficult.

Galetta and colleagues performed an individual patient meta-analysis using original data from a subset of nine diagnostic case–control studies.7 This pooled analysis reported a value for the sensitivity of the KD time in detecting concussion on the sidelines at 86% (96/112 athletes with concussion with any worsening of baseline KD time; 95% CI 78 to 92). Pooled specificity was 90% (181/202 non-concussed control athletes with no worsening of baseline KD times; 95% CI 85 to 93).7 Differences in study methodology are likely to explain the discordance with the current findings; for example, diagnostic case–control studies are known to exaggerate diagnostic accuracy metrics.11

Interpretation of results

The source population from the top tiers of professional English rugby should ensure that these results are generalisable throughout elite rugby union competitions that use the HIA process. External validity to the elite level of other sports with different frameworks for evaluating head impact events is less certain. The extent to which direct sight, or video review, of observable signs of concussion is used to immediately diagnose and definitively remove players with concussion without the need for off-field screening assessment will influence the predictive values of the KD test and could introduce spectrum effects. However, given the reported performance of the KD test, these factors are unlikely to substantially alter the conclusions, and KD accuracy remained low in players removed with clear observable signs of concussion, for example, loss of consciousness or tonic posturing.

In lower levels of competition where trained medical staff are not available, off-field concussion screening tests are contraindicated, and a ‘recognise and remove’ strategy is recommended with immediate withdrawal from play when there is any degree of suspicion of concussion.22 Previously administered as test cards, the KD test is now currently available only as a proprietary tablet application. Preceding studies have suggested differential baseline performance between these formats, and although unlikely it is possible that diagnostic accuracy could also vary across these configurations.23

The KD test requires vision, eye movements (saccades, convergence and accommodation), attention and language function. Neuronal pathways for these systems are widely distributed throughout cortical and subcortical cerebral areas, cerebellum and the brainstem; vulnerability to functional or structural damage in concussion could imply content validity for the KD test.24 However, concussion may manifest as a diverse range of somatic, cognitive, behavioural or emotional symptoms, and/or physical signs such as loss of consciousness and ataxia. It would therefore be surprising if a single test would be able to reliably and consistently detect such a complex pathology that is recognised to affect different clinical domains.

Incorporating a test of oculomotor function test within a multimodal screening test for concussion assessment (such as the HIA-1), with the ability to evaluate a greater number of clinical domains, could offer a more rational approach. With simultaneous, parallel testing a net gain in sensitivity usually occurs at the expense of a net loss in specificity.25 The overall accuracy of the aggregated screening test is strongly influenced by the test accuracy of the individual subcomponents. The limited ability of the KD test as a stand-alone test to accurately identify players with concussion reduces the value it can add to a multimodal assessment. Although a favourable sensitivity of 93% was achieved when combining the KD test and the HIA-1 tool, specificity was reduced to 33%. A key concept in off-field assessment is rapid screening for a suspected concussion, rather than the definitive diagnosis of a head injury, and perfect accuracy is therefore implausible.4 While it is unlikely that false-negative and false-positive cases are equally important, the limited ability of the KD test as a stand-alone test to accurately identify players with concussion reported in this study makes it unlikely that it will be incorporated into multimodal off-field screening assessments at the present time.

Conclusions

This study suggests that the KD test has limited accuracy as a stand-alone remove-from-play sideline screening test for concussion. The low specificity observed when combined with the HIA-1 test suggests it is unlikely that the KD test will be incorporated into multimodal off-field screening assessments at the present time.

What are the findings?

  • The King-Devick (KD) test has been promoted as a remove-from-play sideline screening test for sports-related concussion.

  • This is the largest prospective investigation of the KD screening test for sports-related concussion in professional sport published to date.

  • The diagnostic accuracy design with novel inclusion of consecutive head impact events with the potential to cause concussion maximises internal validity.

  • The KD test demonstrated limited accuracy as a stand-alone off-field screening test for concussion.

How might it impact on clinical practice in the future?

  • The KD test should be used with caution as a stand-alone remove-from-play sideline screening test in professional sport, pending further research.

  • As expected with the addition of a parallel test, combining the KD test with the Head Injury Assessment-1 off-field screening tool provided improved sensitivity in identifying concussion, but at the expense of markedly lower specificity.

  • There is no consensus currently on the acceptable standards for concussion screening tests in professional sports; however, the implications of varying sensitivities and specificities on false-negative and false-positive case rates require careful consideration.

Acknowledgments

The authors would like to acknowledge the invaluable support provided by team doctors and physiotherapists from the participating teams who provided KD and HIA data used within this study.

References

Footnotes

  • GWF and MJC contributed equally.

  • Contributors GWF conceived and designed the study in collaboration with MJC, SPTK and KAS. MJC coordinated data collection and was responsible for data management. GWF processed, analysed and interpreted the data, and wrote and prepared the manuscript for publication. Analyses were checked for accuracy by MJC. All authors critically revised the manuscript for important intellectual content and gave final approval of the version to be published.

  • Funding The study was funded by the Rugby Football Union and supported by King-Devick Technologies.

  • Competing interests None declared.

  • Patient consent Detail has been removed from this case description/these case descriptions to ensure anonymity. The editors and reviewers have seen the detailed information available and are satisfied that the information backs up the case the authors are making.

  • Ethics approval The study protocol received ethical approval from the University of Bath.

  • Provenance and peer review Not commissioned; externally peer reviewed.

  • Data sharing statement No data sharing agreements in place at present.