Using a Bayesian network to classify time to return to sport based on football injury epidemiological data
==========================================================================================================

* Kate K.Y. Yung
* Paul P.Y. Wu
* Karen aus der Fünten
* Anne Hecksteden
* Tim Meyer

## Abstract

**Objective** The return-to-sport (RTS) process is multifaceted and complex, as multiple variables may interact and influence the time to RTS. These variables include intrinsic factors of the player, such as anthropometrics and playing position, or extrinsic factors, such as competitive pressure. Providing an individualised estimation is often challenging, and yet clinical decision support tools are often rare in the industry. This study aims to demonstrate the functions of a Bayesian network by the use of a set of basic epidemiological data.

**Methods** To exemplify the use of Bayesian network in sports medicine, such as providing an individualised estimation time to RTS for individual players, we applied Bayesian network to a set of basic epidemiological data. Bayesian network was used as a decision support tool to model the epidemiological data and to integrate clinical data, non-clinical factors and expert knowledge. Specifically, we used the Bayesian network to capture the interaction between variables in order to 1) classify days to RTS and 2) injury severity (minimal, mild, moderate and severe).

**Results** Retrospective injury data of 3374 player seasons and 6143 time-loss injuries from seven seasons of the professional German football league (Bundesliga, 2014/2015 through 2020/2021) were collected from public databases and media resources. A total of twelve variables from three main categories (player’s characteristics and anthropometrics, match information and injury information) were included. The key response variables are 1) **days to RTS** (1-3, 4-7, 8-14, 15-28, 29-60, >60, and 2) **severity** (minimal, mild, moderate and severe). As there are more than two categories, producer’s and user’s accuracy was used to reflect the sensitivity and specificity of the model. The producer’s accuracy of the model for days to RTS ranges from 0.24 to 0.97, while for severity categories range from 0.73 to 1.00. The user’s accuracy of the model for days to RTS ranges from 0.52 to 0.83, while for severity categories, it ranges from 0.67 to 1.00.

**Conclusions** The Bayesian network can help to capture different types of data to model the probability of an outcome, such as days to return to sports. In our study, the result from the BN may support coaches and players in predicting days to RTS given an injury, 2) support team planning via assessment of scenarios based on player’s characteristics and injury risk and 3) provide evidence-based support of understanding relationships between factors and RTS. This study shows the key functions and applications of the Bayesian network in RTS, and we suggest further experimenting and developing the Bayesian network into a decision-supporting aid.

## 1. Background

### 1.1. Challenges in return to sport decision making

Return to sport (RTS) is when an injured athlete can return to full unrestricted team training without modifications in duration and/ or activities(1-3). Forecasting or estimating the return date of an injured athlete is crucial for team planning, performance optimisation, and game strategy development. By having an estimated return date, medical staff can create an individualised rehabilitation plan to gradually improve the athlete’s condition and mitigate re-injury risk. The predicted timeframe also helps coaching staff to adjust their game plans and strategies, maximising the team’s chances of success.

There is a growing acknowledgement that athletes operate within a complex system and are subject to the influences of a multitude of variables(4, 5). These factors include previous injury history, current injury, body mass index, playing positions, sociological factors, psychological status, and the nature of the sports event(4). Accurately predicting an athlete’s RTS can be a challenging task for medical staff due to the intricate interactions between various variables and their influence on injuries. The human body’s complexity and the system’s highly intricate nature further complicate the task. Understanding these interactions and their impact on injuries becomes crucial for providing accurate prognoses and projecting a range of the likely RTS time (e.g., there is a 75% chance for the player to return between 4-5 weeks). This could help medical staff and athletes to prepare for both ends of the scenario.

Large-scale epidemiological studies have made promising strides in providing valuable insights into the expected time to RTS for major injuries in football at the population level(6-9). These studies describe the risk factors for a disease or injury and the extent of the problem(10). Medical staff can use them as a starting point for estimating the time to RTS(11). This approach aligns with the “anchor and adjust” strategy(12), which can be used to optimise predictive accuracy, especially when making decisions in uncertain scenarios (e.g., injury). However, typical epidemiological studies are intended to offer correlations and data based on groups to reflect population-level insights and trends. While this data is valuable for highlighting general patterns and insights, a tool to translate population-level data into individual-level assessments and allow for personalised estimations would be helpful.

Due to the complex systems, forecasting the time to RTS based on individual characteristics can be challenging. There are several reasons why this process is not straightforward. First, personal characteristics and other factors, such as age and playing position, are recommended to be considered in the Strategic Assessment of Risk and Risk Tolerance framework(13). However, quantifying and synthesising these information in the clinical reasoning process can be difficult. For example, what is the difference between the rehabilitation time required for a 20-year-old striker versus a 30-year-old defender with a same type of hamstring strain injury? Second, synthesising useful information from broad population-level epidemiological data covering entire leagues and multiple seasons and then personalising it to a specific player-context is almost cognitively infeasible for a single individual or even a small group of individuals. There is a lack of tool to help translate the sheer volume and complexity of the population-level data and tailor them on an individual level. Third, medical staff, like any other humans, are susceptible to various decision-making challenges. They have limited information processing capacity (14) and face potential cognitive biases(15, 16). Human judgment can be vulnerable, particularly when it comes to statistical and probabilistic reasoning(17). Without an objective and reliable decision tool specifically designed to provide individualised RTS estimations, medical staff often rely heavily on their clinical experience and population-level epidemiological data.

Data scientists have the potential to develop decision support tools that can effectively synthesise broad epidemiological data and personalise it to specific player-contexts, overcoming the cognitive limitations faced by individuals. These tools can provide valuable additional information that would otherwise be challenging for a single human to process. In practical setting, data scientists can take charge of setting up and maintaining the decision support tool, while users such as medical staff and coaches utilise the tool in planning for an athlete’s RTS, making necessary adjustments for the team, and ensuring the team remains competitive. This collaboration between data scientists and end users can lead to more informed decision-making and strategic planning.

Developing a computer-based decision support tool that uses a complex systems approach may be helpful to overcome the previous challenges and offer a competitive advantage in forecasting RTS(18). Specifically, the key differentiator of a complex systems approach in RTS is the explicit modelling of factor interactions that could be queried and used by medical staff in multiple ways. This includes but is not limited to understanding the strength of influence of different variables and predicting the outcome based on custom RTS scenarios.

Computer-based decision support tools can be divided into predictive and descriptive modelling(19). Predictive modelling can be used for injury diagnosis, severity estimation, and rehabilitation planning. In particular, the Bayesian network (BN) is well suited for providing injury prognosis in sports, due to its capacity to model complex systems and to integrate clinical data, non-clinical factors and expert knowledge. However, it has not been used for such purpose.

### 1.2. About the Bayesian network

A BN is a graph-based modelling method where the relationships between variables (nodes) are represented with arrows (arcs). The presence of an arc denotes the influence of one node on another, and the absence of one assumes conditional independence(20). In Figure 1, we illustrated how the parent nodes and a child node can be related using graph-based modelling. While real data often contains a mixture of discrete and continuous variables, BN structure learning algorithms often assume the random variables are discrete. This type of BN is called the discrete BN, which involves discretising continuous variables in the dataset into categories. Although some information is lost when continuous data are categorised(21), there are merits of using the discrete Bayesian network that worth discussing.

![Figure 1](http://medrxiv.org/https://www.medrxiv.org/content/medrxiv/early/2024/11/08/2024.11.08.24316966/F1.medium.gif)

[Figure 1](http://medrxiv.org/content/early/2024/11/08/2024.11.08.24316966/F1)

Figure 1 
A graph-based modelling showing nodes (A to F) and arcs (arrows). A is the parent of B, while B is the child of A; B is the parent of C and D, while C and D are the children of B; and so on. A is the ancestor of B, C, D, E and F, while B, C, D, E and F are the descendants of A; and so on. The arrows indicate the direction of influence.

In practical settings, people often find it easier to work with discrete representations rather than continuous data. Discrete variables tend to be more interpretable, facilitating abstract reasoning(22). For example, word tokens often enable fast and exact processing(22). Consider the comparison between the words “tall” and “short” versus the numerical values “183cm” and “150cm”. Some may find the discrete terms are easier to grasp and apply in reasoning. Given the complexity of the human body and interacting processes involved in RTS, discretisation may help capture the resolution of data available and relevant to the decision support scenario at hand(23).

However, it’s important to note the disadvantages to discretising continuous variables(24). Discretisation can lead to information loss, as the finer details and nuances captured by continuous data may be overlooked. By converting continuous variables into discrete categories, we sacrifice the precision and granularity inherent in the original data. Additionally, the choice of how to discretise the variables can introduce subjectivity and bias into the analysis and therefore, it is crucial to ensure transparency in how variables are discretised to mitigate these potential issues.

BNs provide a platform for inferring state probabilities given observations, referred to as evidence, of one or more nodes in the network. In the discrete BN, the relationship between parent nodes and a child node can be quantified using conditional probability tables (CPTs). The CPT reflects the probability of child node states (or outcomes) given every possible combination of parent node states(25). As new evidence comes in, changes may be brought to the node’s marginal probability(26), which is known as the posterior probability. The posterior probability is the updated probability of a hypothesis or event after considering new evidence. They combine prior beliefs with the likelihood of the data to provide an updated estimate of the event’s probability.

BNs can perform both predictive and diagnostic inference. For example, medical staff can use the former to predict the outcome of an injury for a given clinical diagnosis, anthropometric and match factors (predictive inference); but can also enter the injury severity as an observation to examine what injury factors could explain that observation (diagnostic inference). As an important feature to end users who may not be familiar with statistics, the BN provide visuals to facilitate understanding and supporting decision making across teams of users, such as, among athletes, coaching staff, medical and non-medical personnel.

Another key feature of a BN is that medical staff can integrate and visualise data from multiple sources into a single BN, such as clinical data, empirical evidence and expert knowledge(27), or a combination of both them(28). Expert knowledge can be invaluable when empirical data is scarce or unavailable. It also plays a significant role in developing models, selecting data or variables, estimating parameters, interpreting results and determining the uncertainty characteristics. Since the BN can be customised based on the experts’ knowledge(29, 30), it may be appealing for small sample size research in elite sports research(31-33), where some data may be missing or not be feasible to collect (e.g., limitations in applied settings)(27). BNs have been used to support clinical decisions(34-37), analyse complex systems in ecology (23, 38, 39) and logistics(40).

### 1.3. Study objectives

The objective of this study was to demonstrate the use of a Bayesian network and its potential use in becoming a decision support tool for medical staff for RTS. Specifically, a discrete BN was modelled based on a set of basic epidemiological data to demonstrate how medical staff can use BN to understand (i) the most influential variable to the outcomes (section 3.3), (ii) the strength of influence of different variables (section 3.3), and to (iii) demonstrate its use with a case scenarios (section 4.2). The paper is structured as follows: Section 2 describes the methodology, including study design and steps to construct a BN; Section 3 describes the results; Section 4 discusses the results and application of the BN in a clinical setting; Section 5 provides an overall conclusion of the study.

## 2. Method

### 2.1. Study design

The study is a retrospective analysis of prospectively collected injury data from the German professional men’s highest football league (Bundesliga) between 2014/2015 - 2020/2021. Neither research ethics board approval nor a trial registration was required as all data were collected from publicly available sources(9, 41). We reported the result with reference to the Transparent Reporting of a multivariable prediction model for Individual Prognosis Or Diagnosis (TRIPOD) statement(42).

### 2.2. Participants

Players who played in the Bundesliga in the above-mentioned seven seasons were included in the study. Injured players who did not return to the Bundesliga in the same season and those without complete data were excluded from the analysis. All participants were identified using a publicly available database, including Kicker SportmagazinTM, and clubs’ official websites. Data collection was performed via methods established in previous investigations(9, 43). The data has been collected according to the Fuller et al. consensus statement for football injury research and the same definitions have been the same throughout the 7 seasons(44). 3374 player Bundesliga seasons were registered over the seven seasons from 2014/2015-2020/2021. A total of 6653 time-loss injuries were recorded. After removing injuries without complete data, 6143 time-loss injuries remained.

### 2.3. Construction of the Bayesian network

The modelling process and validation were performed in GeNIe 2.0 (Bayes Fusion, LLC)(45). There are four main steps to creating and validating a BN: 

1.  Identify the variables (nodes) of the BN and the key response variables. In this study, the key response variables are the (i) severity of the injury *(severity)* and (ii) days to RTS *(days_rts)*. They are used to evaluate the performance of the BN. The nodes and sources of information are outlined in Table 1.

2.  Learn the graphical structure of the BN and display the relationships between all the nodes.

3.  Learn the probability distribution governing the relationships between the nodes.

4.  Validate the BN to evaluate how well the BN can classify injury based on the key response variables: (i) *severity* and (ii) *days_rts*.

View this table:
[Table 1](http://medrxiv.org/content/early/2024/11/08/2024.11.08.24316966/T1)

Table 1 Summary of data type and source of information.

#### 2.3.1. Step 1: Identify the main variables

This model is intended to capture the conditional probabilistic relationships among personal characteristics, match and injury information. The variables in this study are commonly included in epidemiologic studies (46) and we have summarised them in Table 1.

All time-loss injuries that occurred during football competitions and training sessions were included, and the day of the injury is counted as day zero. A time-loss injury is when a player cannot fully participate in training or competition due to injury(44). RTS is defined as when the player has received medical clearance to allow full participation in training and is available for match selection(2). Details of the injury (type of injury, body region, contact/non-contact injury, training or match) were labelled based on the Fuller et al. consensus of data collection in football(44). The time of the season was based on an injury epidemiologic study in European football(47). Personal characteristics and anthropometrics, match information and the above information are used as the explanatory variables for constructing the model.

We are interested in determining the days to RTS, therefore, the two key response variables in this model are injury severity (***severity****)* and days to RTS *(**days_rts**)*. Injury severity (*severity*) was categorised according to the days of absence in match or training as outlined by the Fuller et al. consensus statement on football injury studies(44): minimal (1–3 days), mild (4–7 days), moderate (8–28 days), or severe (>28 days). We further create a new variable based on days to RTS *(days_rts)* to evaluate if the model could classify the days to RTS into more precise categories. There are two additional categories in *days_rts* as compared to *severity*: 1-3, 4-7, 8-14, **15-28**, 29-60, **>60** days. In context, these variables form the nodes in the BN. In particular, *days*_rts directly maps to *severity*, except that severity has fewer states, combining the last two in *days_rts* into one (i.e., any injuries taking more than 28 for RTS is classified as severe). A summary of the variables and source of the data are presented in Table 1.

#### 2.3.2. Step 2: Define the graphical structure

We incorporated expert knowledge to constrain the search to help ensure graph structures produced are consistent with clinical science (See Supplementary Information Figure 1). We first specified the temporal order of the variables, ensuring there were no arcs from variables that occurred later (e.g., injury) to nodes happening earlier (e.g., main playing position). Following this, KY helped establish the relationships (arcs) within the graphical structure. Notably, our key response variable, days_rts, was directly linked to severity in the graphical structure. This mapping was based on the definition of severity outlined in the Fuller et al. consensus statement on football injury studies, which defines severity by the duration of unavailability for full training(44). This mapping was done manually based on clinical knowledge and could not be learnt from algorithms. Finally, with the constraints in place, we use the Bayesian Search algorithm in GeNIe 2.0 to find the best-fitted network for the optimal network configuration that aligns with the collected data and expert knowledge(20, 48).

#### 2.3.3. Step 3: Define the probability distribution

There are multiple ways to discretise continuous data, including manual, unsupervised and supervised. Each method offers distinct advantages, such as improved model performance, easier interpretation, and computational efficiency(49). Manual discretisation involves manually defining thresholds or categories to discretise the continuous data. This approach provides flexibility and allows domain experts to incorporate their knowledge, ensuring that the categories align with relevant domain-specific consensus statements and the specific needs of the analysis. Unsupervised discretisation involves using clustering algorithms to identify natural groupings or patterns in the data. These clusters can then be treated as discrete categories. This is particularly useful when prior knowledge or predefined categories are unavailable(50). Supervised discretisation utilises labelled data to guide the discretisation process. Machine learning algorithms or decision trees can be employed to learn optimal thresholds or categories that maximise predictive performance. This method can enhance both model performance and interpretability(50).

We manually discretised the continuous data based on domain-specific decision categories to maximise usability across various practitioner types and clubs. Specifically, the continuous data are discretised based on categories that are easy to apply practically or with reference to relevant consensus statements. We discretised *age, height,* and *weight* into three categories that were representative of the typical values observed in the sample, to create uniform count categories that were both meaningful and easy to interpret. *severity* was discretised based on the Fuller et al. consensus of epidemiological data collection into four categories (minimal, mild, moderate and severe)(44). *days_rts* was discretised into six categories: 1-3, 4-7, 8-14, 15-28, 29-60, >60 days, with reference to the Fuller et al. consensus of epidemiological data collection(44). Descriptions of the nodes and categories are summarised in Table 2. We used the Expectation Maximisation (EM) algorithm in GeNIe 2.0 to determine the probability distribution (parameter learning) of the dataset(51, 52).

View this table:
[Table 2](http://medrxiv.org/content/early/2024/11/08/2024.11.08.24316966/T2)

Table 2 Description of data set variables.

#### 2.3.4. Step 4: Validation of the BN

A crucial element of learning is to validate the model. Validation was performed on the two target nodes, i.e., s*everity* and *days_rts*, which are the main outcomes of interest. Ten-fold cross-validation was performed where the dataset was split into ten parts of equal probability(53). The model was trained on nine parts and tested on the remaining tenth part of the unseen data (holdout test sets). The process was repeated ten times, with a different part of the data being used for testing. The model evaluation technique implemented in GeNIe keeps the model structure fixed and re-learns the model parameters during each of the folds. We compiled the results of the test splits to report the producer’s and user’s accuracy(54). The producer’s accuracy is calculated by dividing the total number of correctly classified injury severity by the total number of actual occurrences of injury of that severity. Producer’s accuracy is analogous to sensitivity in binary classification, as it represents the true positive rate for each category. ![Formula][1]</img>  The user’s accuracy is calculated by dividing the total number of correctly classified injury severity by the total classified true occurrences of injury of that severity. Similar to specificity, it represents the correctness of the model’s predictions within each category. User’s accuracy provides insights into the precision of the model’s predictions for each category. ![Formula][2]</img>  By considering both the producer’s and user’s accuracy, we can evaluate the model’s performance in terms of both sensitivity (ability to classify correctly) and specificity (accuracy of predictions within each category). This is particularly relevant when dealing with models that have multiple categories, such as our current model’s injury severity classification.

## 3. Results

### 3.1. Demographics

The demographics and the main playing positions of injured players are shown in Supplementary Table 1. The breakdown of injury by nature of injury (contact or non-contact), event (match or training), body region, and types of injuries are available in Supplementary Table 2.

### 3.2. Bayesian network

The network and the probability distribution of each variable are presented in Figure 2. The model’s producer’s accuracy in classifying *days_rts* and *severity* is presented in Table 3. In terms of categorising *days_rts*, the producer’s accuracy ranges from 0.24 to 0.97, with the best performance for shorter days (i.e., below 3 days) and the worst performance for the mid-range category (i.e., 8-14 days). In classifying the injury’s severity, the producer’s accuracy ranges from 0.73 to 1.00, with the best performance for severe and the worst for minimal. In terms of categorising *days_rts*, the user’s accuracy ranges from 0.52 to 0.83, with the best performance for days 3-7 and the worst performance for the mid-range category (i.e., 8-14 and 15-28 days). In classifying the injury’s severity, the producer’s accuracy ranges from 0.67 to 1.00, with >0.90 in all categories except minimal.

![Figure 2](http://medrxiv.org/https://www.medrxiv.org/content/medrxiv/early/2024/11/08/2024.11.08.24316966/F2.medium.gif)

[Figure 2](http://medrxiv.org/content/early/2024/11/08/2024.11.08.24316966/F2)

Figure 2 
Structure and network probability of the Bayesian network. The percentages indicate the distribution of values under each variable, while the blue arrows indicate how factors influence one another.

View this table:
[Table 3](http://medrxiv.org/content/early/2024/11/08/2024.11.08.24316966/T3)

Table 3 The model’s producer’s and user’s accuracy of classifying days to RTS and severity of the injuries.

In our model, the type of injury and the injured body region are directly connected to *days_rts* (Figure 2). The percentages indicate the distribution of values under each variable, while the blue arrows indicate how factors influence one another (Figure 2). Based on the sensitivity analysis, the type of injury and the injured body region are most influential to the *days_rts*, followed by age, contact or non-contact injury and the nature of the event (training or match) (see Figure 2). Time of the season, weight, and height only had a minor influence on the result. The time of the season is also associated with the nature of the event (training or match) and the *type of injury*, which is supported by empirical evidence(55).

### 3.3. Sensitivity analysis, feature selection and strength of influence

A sensitivity analysis was performed using built-in functionality in GeNIe to determine the influence of the individual nodes. Sensitivity analysis helps determine the influence of observing the states of specific nodes (i.e., prior and conditional probabilities) on the output variables (i.e., posterior probabilities)(56), which in this case, are the days to RTS and severity of the injury (Figure 3). This can help to support the selection of key variables to be included in a model (feature selection). Results of the sensitivity analysis are visually summarised in the form of tornado charts in Supplementary Information 2.

![Figure 3](http://medrxiv.org/https://www.medrxiv.org/content/medrxiv/early/2024/11/08/2024.11.08.24316966/F3.medium.gif)

[Figure 3](http://medrxiv.org/content/early/2024/11/08/2024.11.08.24316966/F3)

Figure 3 
Feature selection of the Bayesian network

Red-coloured nodes: contain variables important for calculating posterior probability distributions in *days_rts*. Key: A darker red colour indicates a higher degree of influence. Grey-coloured nodes: the node has no influence on the posterior probability distributions of *days_rts*.

Highly sensitive variables affect the inference results more significantly. Identifying the highly sensitive variables directs medical staff to specific areas to focus on to affect the BN’s outcome. As an example, in Figure 3, we have set *days_rts* as the key response variable. Nodes coloured in red contain variables important for calculating posterior probability distributions in *days_rts*. Grey-coloured nodes have no influence on the posterior probability distributions of *days_rts*, as determined by data and knowledge. Based on the sensitivity analysis, *body_region*, *injury_type* and *contact_noncontact* are the most influential to the outcome in *days_rts* (Figure 3). A numeric form of the strength of influence of different variables can be found in Supplementary Information Table 3.

## 4. Discussion

This study demonstrates the applicability of a BN in an epidemiological dataset to project rehabilitation timelines and provides evidence-based support to RTS decisions. We integrated personal characteristics and anthropometrics, match information and injury information to construct a BN that may inform the days to RTS and injury severity.

### 4.1. The BN in the context of RTS

In this model, we use a hybrid approach in constructing the BN; that is, we have combined clinical knowledge and data-driven learning (Bayesian Search algorithm) when constructing the graphical structure. This is because constructing the BN using pure algorithmic approaches (i.e., unsupervised learning) can sometimes produce graph structures with unreasonable arcs, such as the model attempting to explain age with severity. However, if we use a pure domain approach, we may miss some of the patterns and linkages between variables that were not observed by clinicians’ eyes. Therefore, we opted to use a hybrid approach when constructing the BN.

The BN model may help medical staff working in the Bundesliga to view athletes and injuries with a complex systems approach. For example, the model captured the complex relationships between the time of the season, the injury type and injury occurrence (Figure 2). Specifically, in our BN model, the time of the season correlates with the weather, which is then associated with the ground condition (55) and injury occurrence(57, 58). As a key feature of a complex systems approach, the BN can explicitly model the factor interactions/relationships and query in multiple ways, such as comparing the outcomes of custom RTS scenarios (section 4.2), understanding the most influential factors (section 3.3) and the strength of influence (section 3.3).

To our knowledge, no study in the Bundesliga has studied the correlations of the above factors, so we use an example from the English Premier League (EPL) to illustrate the possible correlations. In the EPL, the ground condition tends to be drier in the preseason. Warm, dry and hard surfaces may be associated with higher injury occurrence(57, 58), possibly due to a higher level of shoe-surface traction influence (59) and faster running speed(60). On the contrary, wet and muddy ground is associated with lower injury occurrence(58), possibly due to changes in playing style (e.g., more tackles) and reduced shoe-surface traction. While the time of the season affects the injury occurrence, there has been no direct effect on days to RTS. Analysing these complex relationships would be difficult without the use of computers and advanced statistical modelling.

The model is more accurate for classifying injuries into 4 categories under *severity* (sensitivity = 0.75 – 1.00), compared to 6 categories *days_rts* (sensitivity = 0.24 – 0.97). This is not surprising because the increase in the number of categories challenged the model to provide a higher accuracy. The model is most accurate in classifying injuries with shorter *days_rts* (days below 3, sensitivity = 0.97) and least accurate in classifying mid-range injuries (days 8-14, sensitivity = 0.24). A possible explanation is that injuries with minimal days of absence have particular injury patterns, for example, they may be upper body injuries and hematoma/contusion/bruise injuries. The model is least accurate for classifying mid-range injuries (8-14 days), possibly because the model lacks information that may differentiate the prognoses. This information may be the extent of tissue damage (e.g., the sub-classification of muscle injury) and the specific location of the injury (e.g., involvement of central tendon injury).

### 4.2. Illustrative application of the BN

Here, we use a hypothetical case study to demonstrate the practicability of using the BN in classifying the days to RTS and the injury severity of a player. We input the player’s characteristics and anthropometrics, match information and injury information into the BN constructed earlier. The probability distribution is shown in Figure 4

![Figure 4](http://medrxiv.org/https://www.medrxiv.org/content/medrxiv/early/2024/11/08/2024.11.08.24316966/F4.medium.gif)

[Figure 4](http://medrxiv.org/content/early/2024/11/08/2024.11.08.24316966/F4)

Figure 4 
Bayesian network of the case study.

*A player (age: 25, weight: 77kg, height: 185cm, attacker) playing in the German professional football league sprained his ankle ligament in a non-contact injury and pulled out of the preseason training session. The coach would like to know when the player is available for selection*.

The BN indicates the joint multivariate probability distribution: the likelihood of the injury to be minimal is 63%, mild 24%, moderate 12% and severe 1% (Figure 4). The likelihood of RTS below 3 days is 52%, 3-7 days is 33%, 7-14 days is 10%, 14-28 days is 4%, 28-60 days and more than 60 days are less than 1%, respectively. This information, in combination with clinical assessment, may be used to support coaches and players in predicting days to RTS and support team planning by assessing the number of players available for training and competition.

In Figure 4, the thickness of the arcs indicates the strength of influence. Medical staff can use the strength of influence analysis to understand the local relationships among the variables in a network and how they contribute to the posterior marginal probability (i.e., outcome) for each node in the network (Figure 4). The strength of influence calculates the average effect of changing the state of a parent node on the probability distribution of states in the child node(61). The arcs are normalised, thus the thickest arc indicates the one with the highest influence. From Figure 4, the nature of the injury (contact or non-contact) is related to the injury type and the body region. Injury type and body region are directly associated with the *days_rts*. Medical staff may use *days_rts* in conjunction with the clinical knowledge to project the rehabilitation timeline.

### 4.3. Practical application and future developments

The software we used for constructing the BN, GeNIe 2.0, could be used as a decision support tool by sports practitioners, including but not limited to coaches, team managers, sports science and medical staff. Once the BN has been constructed, for example, with the use of GeNIe 2.0, the end users do not need any machine-learning knowledge or advanced computer skills to use the BN. Instead, the only input from the sports practitioners is to collect and input the players’ characteristics into the BN. The new data from every week or season can be input into the system’s database by data scientists (either in-house or third party) to ensure the model is up-to-date. As technology continues to advance, the process of constructing and utilising BN is expected to become easier and more efficient. This increased ease of use may encourage medical staff to invest their efforts in learning and adopting BNs to support their decision-making processes.

When planning for RTS for an athlete, the BN can provide a personalised risk analysis from broad epidemiological data, as exemplified in section 4.2 “Illustrative application of the BN”. While the results of the BN for estimating time to RTS align with existing clinical systems and therefore some practitioners may question the need for a computational decision support tool, it is important to see the potential benefits of integrating BN into sports medical practice in future. Conventional clinical guidelines often apply universally to all individuals regardless of age, sex, sport, or level of play. Consequently, they are not intended to offer an individualised approach to RTS estimation. Given that athletes’ diverse needs and circumstances, coupled with the growing emphasis on personalised and precision medicine, a more tailored approach to estimating RTS becomes essential. In this regard, the incorporation of BN can serve as a valuable addition to sports medical practice to improve decision making.

In our previous example in section 4.2, we explained the use of BN with a return-to-sport (RTS) scenario and estimated the time to RTS based on a player’s basic characteristics. However, it is important to note that this study is a pilot for future studies, and the potential applications of BN extend beyond this specific use case. They can be leveraged for various purposes, such as forecasting expected performance or injury risk in a game by considering factors like personal characteristics, opponent playing style, and recent training and game performance. While in this study, the BN yield similar results to existing clinical systems, its true benefits lie in its ease of use, efficiency, and the capacity to incorporate a wider range of data into the decision-making process. Moreover, BN’s ability to identify the strength of influence allows medical staff to address the variables that significantly impact outcomes, a task that cannot be adequately addressed by conventional clinical guidelines alone. By leveraging BN, medical staff may enhance their decision-making processes and provide more tailored and effective care to athletes.

Computational models are more likely to be implemented in applied sports settings if their accuracy, interpretability and functionality fit with the operation framework of a sports organisation(62). In this case, BN provides an intuitive visualisation of the complex relationship of injuries, which medical staff may understand even with little or no experience in computer analytics. This may increase the model’s transparency and may improve the medical staff’s trust in the model(63). In terms of practical applicability, medical staff can use the model to create scenarios that facilitate the evaluation of different management options or the development of rehabilitation protocols for players with various positions or injuries occurring at different times during the season. These scenarios enable the assessment of the combined effects of risk factors(27). This may enable them to proactively manage injury risk, and in case of injury, make more accurate predictions regarding the days to RTS for individual athletes based on epidemiological evidence. The BN can be modified to suit the specific context as determined by the user and updated with new information when available. In summary, BNs seem promising for modelling the relationships of variables in a complex system and may be further explored to support clinical decision making.

The epidemiological data we used in this study is static; thus, we use a BN to represent the system as a time-aggregated model. However, most systems, including athletes, have been well recognised to change over time(4, 64). To capture the change over time and the feedback loops, a BN can be further extended into a dynamic Bayesian network (DBN)(65). DBN is an extension of BNs and replicates the BN model at discrete points in time (time slices) and captures temporal relationships between the variables. DBNs can represent complex questions, such as how changes in rehabilitation training load affect the time to RTS in athletes with different demographic and anthropometric characteristics. Modifiable variables, such as rehabilitation training exposure, can be collected continuously over time as time-series data(66). Although DBN are more complex than BNs, DBNs have also been applied in medical diagnosis or prognosis (67-69) and are a promising avenue for further investigation in an injury modelling context.

### 4.4. Limitations

The proposed BN model has limitations that should be considered when interpreting its results. Firstly, it is important to recognise that the model was constructed using data solely from the Bundesliga, and therefore, the generalisability of the results to other leagues or levels of play is limited. Second, the quality of the input variables directly impacts the model’s predictive accuracy. In the case of our BN, it was constructed using basic epidemiological data sourced from public databases. While this approach maximises the availability of injury events for the model, it lacks pertinent information related to prognosis, such as diagnosis (e.g., structural damage or functional disorder) and the precise location of the injury (e.g., involvement of central tendon in hamstring injury)(70). An accurate diagnosis of an injury plays a vital role in determining prognosis(11, 71). For instance, the recovery time to RTS can vary significantly between a minor partial tear in the hamstring muscle (requiring 17 days to RTS) and a moderate partial tear (requiring a longer time of 36 days to RTS). Therefore, the lack of detailed diagnostic information can impact the model’s graphical structure and prediction accuracy.

Further studies may consider including more information into the model, such as quality and duration of rehabilitation training, RTS performance, reinjury incidence, the importance of upcoming competition, the remaining contract length, the club’s geographical regions and the players’ transfer value. Potentially, depending on the purpose of the model, we can integrate epidemiological data from other larger datasets into the model to enlarge the database to reduce the chance of overfitting and improve result transportability.

Comparing the machine-learned BN with the existing scientific understanding of injury and recovery would also be a valuable future avenue of research. This comparison can provide insights into the performance and predictive capabilities of a complex model that incorporates a wide range of variables in contrast to a simpler model, such as an epidemiological one. By doing so, researchers can evaluate the added value of BN as a decision-support tool.

## 5. Conclusion

This discrete BN provides a decision support tool to help medical staff, coaches and players manage injury. The BN has a high producer accuracy, ranging from 0.73 to 1.00 in predicting severity and provides a graphical representation of the investigated interdependencies. Medical staff can use BN to understand the strength of influence of different variables on the outcome and analyse the outcome based on custom RTS scenarios. This information may help medical staff evaluate different injury scenarios and better respond to individual player’s rehabilitation and team planning. BNs seem promising for modelling the relationships of variables from multiple sources and can be further explored to support clinical decision making.

## Data Availability

Dataset cannot be shared publicly because it is currently in use for another research project. However, the data would be available on reasonable request.

## Ethics approval and consent to participate

As the study data stems from publicly available media data, no ethical approval was needed. A consent to participate was not necessary as the study does not contain any person’s individual data.

## Consent for Publication

Not applicable as the study does not use any persons’ individual data.

## Conflicts of interest/ Competing interests

KY, PW, KadF, AH and TM declare that they have no competing interests.

## Funding

No funding was received for this research.

## Authors’ contribution

KY, PW and TM conceptualised and designed the study. KadF led the data collection, checked all data for medical plausibility, and refined parts of the database. KY analysed, constructed, and interpreted the Bayesian network with technical support from PW. AH, PW and TM supervised the study. All authors read and approved the final manuscript.

## Acknowledgement

Our Bayesian network was built using GeNIe Modeler (BayesFusion 2019), available free of charge for academic research and teaching use from [https://www.bayesfusion.com/](https://www.bayesfusion.com/). The authors would also like to thank Tobias Tröß and Abed Hadji for the data collection and injury database management.

*   Received November 8, 2024.
*   Revision received November 8, 2024.
*   Accepted November 8, 2024.


*   © 2024, Posted by Cold Spring Harbor Laboratory

This pre-print is available under a Creative Commons License (Attribution 4.0 International), CC BY 4.0, as described at [http://creativecommons.org/licenses/by/4.0/](http://creativecommons.org/licenses/by/4.0/)

## 6. Reference

1.  1.Waldén M, Mountjoy M, McCall A, Serner A, Massey A, Tol JL, et al. Football-specific extension of the IOC consensus statement: methods for recording and reporting of epidemiological data on injury and illness in sport 2020. British Journal of Sports Medicine. 2023;57(21):1341–50.
    
    [Abstract/FREE Full Text](http://medrxiv.org/lookup/ijlink/YTozOntzOjQ6InBhdGgiO3M6MTQ6Ii9sb29rdXAvaWpsaW5rIjtzOjU6InF1ZXJ5IjthOjQ6e3M6ODoibGlua1R5cGUiO3M6NDoiQUJTVCI7czoxMToiam91cm5hbENvZGUiO3M6ODoiYmpzcG9ydHMiO3M6NToicmVzaWQiO3M6MTA6IjU3LzIxLzEzNDEiO3M6NDoiYXRvbSI7czo1MDoiL21lZHJ4aXYvZWFybHkvMjAyNC8xMS8wOC8yMDI0LjExLjA4LjI0MzE2OTY2LmF0b20iO31zOjg6ImZyYWdtZW50IjtzOjA6IiI7fQ==) 

2.  2.Ardern CL, Glasgow P, Schneiders A, Witvrouw E, Clarsen B, Cools A, et al. 2016 Consensus statement on return to sport from the First World Congress in Sports Physical Therapy, Bern. British Journal of Sports Medicine. 2016;50(14):853-64.
    
    [Abstract/FREE Full Text](http://medrxiv.org/lookup/ijlink/YTozOntzOjQ6InBhdGgiO3M6MTQ6Ii9sb29rdXAvaWpsaW5rIjtzOjU6InF1ZXJ5IjthOjQ6e3M6ODoibGlua1R5cGUiO3M6NDoiQUJTVCI7czoxMToiam91cm5hbENvZGUiO3M6ODoiYmpzcG9ydHMiO3M6NToicmVzaWQiO3M6OToiNTAvMTQvODUzIjtzOjQ6ImF0b20iO3M6NTA6Ii9tZWRyeGl2L2Vhcmx5LzIwMjQvMTEvMDgvMjAyNC4xMS4wOC4yNDMxNjk2Ni5hdG9tIjt9czo4OiJmcmFnbWVudCI7czowOiIiO30=) 

3.  3.Pol R, Balagué N, Ric A, Torrents C, Kiely J, Hristovski R. Training or Synergizing? Complex Systems Principles Change the Understanding of Sport Processes. Sports Medicine - Open. 2020;6(1):28.
    
    [PubMed](http://medrxiv.org/lookup/external-ref?access_num=32661759&link_type=MED&atom=%2Fmedrxiv%2Fearly%2F2024%2F11%2F08%2F2024.11.08.24316966.atom) 

4.  4.Yung KK, Ardern CL, Serpiello FR, Robertson S. Characteristics of Complex Systems in Sports Injury Rehabilitation: Examples and Implications for Practice. Sports Medicine - Open. 2022;8(1):24.
    
    [PubMed](http://medrxiv.org/lookup/external-ref?access_num=35192079&link_type=MED&atom=%2Fmedrxiv%2Fearly%2F2024%2F11%2F08%2F2024.11.08.24316966.atom) 

5.  5.Balague N, Torrents C, Hristovski R, Davids K, Araújo D. Overview of complex systems in sport. Journal of Systems Science and Complexity. 2013;26(1):4–13.
    
    [CrossRef](http://medrxiv.org/lookup/external-ref?access_num=10.1007/s11424-013-2285-0&link_type=DOI) 

6.  6.Hallen A, Ekstrand J. Return to play following muscle injuries in professional footballers. Journal of Sports Sciences. 2014;32(13):1229–36.
    
    [CrossRef](http://medrxiv.org/lookup/external-ref?access_num=10.1080/02640414.2014.905695&link_type=DOI) 
    
    [PubMed](http://medrxiv.org/lookup/external-ref?access_num=24784885&link_type=MED&atom=%2Fmedrxiv%2Fearly%2F2024%2F11%2F08%2F2024.11.08.24316966.atom) 
    
    [Web of Science](http://medrxiv.org/lookup/external-ref?access_num=000340107900005&link_type=ISI) 

7.  7.López-Valenciano A, Ruiz-Pérez I, Garcia-Gómez A, Vera-Garcia FJ, De Ste Croix M, Myer GD, et al. Epidemiology of injuries in professional football: a systematic review and meta-analysis. British Journal of Sports Medicine. 2020;54(12):711.
    
    [Abstract/FREE Full Text](http://medrxiv.org/lookup/ijlink/YTozOntzOjQ6InBhdGgiO3M6MTQ6Ii9sb29rdXAvaWpsaW5rIjtzOjU6InF1ZXJ5IjthOjQ6e3M6ODoibGlua1R5cGUiO3M6NDoiQUJTVCI7czoxMToiam91cm5hbENvZGUiO3M6ODoiYmpzcG9ydHMiO3M6NToicmVzaWQiO3M6OToiNTQvMTIvNzExIjtzOjQ6ImF0b20iO3M6NTA6Ii9tZWRyeGl2L2Vhcmx5LzIwMjQvMTEvMDgvMjAyNC4xMS4wOC4yNDMxNjk2Ni5hdG9tIjt9czo4OiJmcmFnbWVudCI7czowOiIiO30=) 

8.  8.Lu D, McCall A, Jones M, Kovalchik S, Steinweg J, Gelis L, et al. Injury epidemiology in Australian male professional soccer. Journal of Science and Medicine in Sport. 2020;23(6):574–9.
    
    [PubMed](http://medrxiv.org/lookup/external-ref?access_num=32008909&link_type=MED&atom=%2Fmedrxiv%2Fearly%2F2024%2F11%2F08%2F2024.11.08.24316966.atom) 

9.  9.aus der Fünten K, Tröß T, Hadji A, Beaudouin F, Steendahl IB, Meyer T. Epidemiology of Football Injuries of the German Bundesliga: A Media-Based, Prospective Analysis over 7 Consecutive Seasons. Sports Medicine - Open. 2023;9(1):20.
    
    [PubMed](http://medrxiv.org/lookup/external-ref?access_num=36867257&link_type=MED&atom=%2Fmedrxiv%2Fearly%2F2024%2F11%2F08%2F2024.11.08.24316966.atom) 

10. 10.Hägglund M, Waldén M, Bahr R, Ekstrand J. Methods for epidemiological study of injuries to professional football players: developing the UEFA model. British Journal of Sports Medicine. 2005;39(6):340–6.
    
    [Abstract/FREE Full Text](http://medrxiv.org/lookup/ijlink/YTozOntzOjQ6InBhdGgiO3M6MTQ6Ii9sb29rdXAvaWpsaW5rIjtzOjU6InF1ZXJ5IjthOjQ6e3M6ODoibGlua1R5cGUiO3M6NDoiQUJTVCI7czoxMToiam91cm5hbENvZGUiO3M6ODoiYmpzcG9ydHMiO3M6NToicmVzaWQiO3M6ODoiMzkvNi8zNDAiO3M6NDoiYXRvbSI7czo1MDoiL21lZHJ4aXYvZWFybHkvMjAyNC8xMS8wOC8yMDI0LjExLjA4LjI0MzE2OTY2LmF0b20iO31zOjg6ImZyYWdtZW50IjtzOjA6IiI7fQ==) 

11. 11.Ekstrand JK, Werner; Spreco, Armin; van Zoest, Wart; Roberts, Craig; Meyer, Tim; Bengtsson, Håkan. Time before return to play for the most common injuries in professional football: a 16-year follow-up of the UEFA Elite Club Injury Study. British Journal of Sports Medicine. 2019:bjsports-2019-100666.
    
    

12. 12.Hecksteden A, Keller N, Zhang G, Meyer T, Hauser T. Why Humble Farmers May in Fact Grow Bigger Potatoes: A Call for Street-Smart Decision-Making in Sport. Sports Med Open. 2023;9(1):94.
    
    [PubMed](http://medrxiv.org/lookup/external-ref?access_num=37837528&link_type=MED&atom=%2Fmedrxiv%2Fearly%2F2024%2F11%2F08%2F2024.11.08.24316966.atom) 

13. 13.Shrier I. Strategic Assessment of Risk and Risk Tolerance (StARRT) framework for return-to-play decision-making. British Journal of Sports Medicine. 2015;49(20):1311–5.
    
    [Abstract/FREE Full Text](http://medrxiv.org/lookup/ijlink/YTozOntzOjQ6InBhdGgiO3M6MTQ6Ii9sb29rdXAvaWpsaW5rIjtzOjU6InF1ZXJ5IjthOjQ6e3M6ODoibGlua1R5cGUiO3M6NDoiQUJTVCI7czoxMToiam91cm5hbENvZGUiO3M6ODoiYmpzcG9ydHMiO3M6NToicmVzaWQiO3M6MTA6IjQ5LzIwLzEzMTEiO3M6NDoiYXRvbSI7czo1MDoiL21lZHJ4aXYvZWFybHkvMjAyNC8xMS8wOC8yMDI0LjExLjA4LjI0MzE2OTY2LmF0b20iO31zOjg6ImZyYWdtZW50IjtzOjA6IiI7fQ==) 

14. 14.Miller GA. The magical number seven, plus or minus two: Some limits on our capacity for processing information. Psychological review. 1956;63(2):81–97.
    
    [CrossRef](http://medrxiv.org/lookup/external-ref?access_num=10.1037/h0043158&link_type=DOI) 
    
    [PubMed](http://medrxiv.org/lookup/external-ref?access_num=13310704&link_type=MED&atom=%2Fmedrxiv%2Fearly%2F2024%2F11%2F08%2F2024.11.08.24316966.atom) 
    
    [Web of Science](http://medrxiv.org/lookup/external-ref?access_num=A1956WG39300001&link_type=ISI) 

15. 15.Blumenthal-Barby JS, Krieger H. Cognitive Biases and Heuristics in Medical Decision Making:A Critical Review Using a Systematic Search Strategy. Medical Decision Making. 2015;35(4):539–57.
    
    [CrossRef](http://medrxiv.org/lookup/external-ref?access_num=10.1177/0272989X14547740&link_type=DOI) 
    
    [PubMed](http://medrxiv.org/lookup/external-ref?access_num=25145577&link_type=MED&atom=%2Fmedrxiv%2Fearly%2F2024%2F11%2F08%2F2024.11.08.24316966.atom) 

16. 16.Yung KK, Ardern CL, Serpiello FR, Robertson S. A Framework for Clinicians to Improve the Decision-Making Process in Return to Sport. Sports Medicine - Open. 2022;8(1):52.
    
    [PubMed](http://medrxiv.org/lookup/external-ref?access_num=35416633&link_type=MED&atom=%2Fmedrxiv%2Fearly%2F2024%2F11%2F08%2F2024.11.08.24316966.atom) 

17. 17.Tversky A, Kahneman D. Judgment under Uncertainty: Heuristics and Biases. Science. 1974;185(4157):1124-31.
    
    [PubMed](http://medrxiv.org/lookup/external-ref?access_num=17835457&link_type=MED&atom=%2Fmedrxiv%2Fearly%2F2024%2F11%2F08%2F2024.11.08.24316966.atom) 

18. 18.Kawamoto K, Houlihan CA, Balas EA, Lobach DF. Improving clinical practice using clinical decision support systems: a systematic review of trials to identify features critical to success. BMJ. 2005;330(7494):765.
    
    [Abstract/FREE Full Text](http://medrxiv.org/lookup/ijlink/YTozOntzOjQ6InBhdGgiO3M6MTQ6Ii9sb29rdXAvaWpsaW5rIjtzOjU6InF1ZXJ5IjthOjQ6e3M6ODoibGlua1R5cGUiO3M6NDoiQUJTVCI7czoxMToiam91cm5hbENvZGUiO3M6MzoiYm1qIjtzOjU6InJlc2lkIjtzOjEyOiIzMzAvNzQ5NC83NjUiO3M6NDoiYXRvbSI7czo1MDoiL21lZHJ4aXYvZWFybHkvMjAyNC8xMS8wOC8yMDI0LjExLjA4LjI0MzE2OTY2LmF0b20iO31zOjg6ImZyYWdtZW50IjtzOjA6IiI7fQ==) 

19. 19.Han J. Data mining concepts and techniques, third edition. 3rd ed ed. Kamber M, Pei J, editors. Waltham, Mass.: Morgan Kaufmann Publishers; 2012.
    
    

20. 20.Heckerman D, Geiger D, Chickering DM. Learning Bayesian Networks: The Combination of Knowledge and Statistical Data. Machine Learning. 1995;20(3):197–243.
    
    

21. 21.Binney ZO, Mansournia MA. Methods matter: (mostly) avoid categorising continuous data – a practical guide. British Journal of Sports Medicine. 2024;58(5):241–3.
    
    [FREE Full Text](http://medrxiv.org/lookup/ijlink/YTozOntzOjQ6InBhdGgiO3M6MTQ6Ii9sb29rdXAvaWpsaW5rIjtzOjU6InF1ZXJ5IjthOjQ6e3M6ODoibGlua1R5cGUiO3M6NDoiRlVMTCI7czoxMToiam91cm5hbENvZGUiO3M6ODoiYmpzcG9ydHMiO3M6NToicmVzaWQiO3M6ODoiNTgvNS8yNDEiO3M6NDoiYXRvbSI7czo1MDoiL21lZHJ4aXYvZWFybHkvMjAyNC8xMS8wOC8yMDI0LjExLjA4LjI0MzE2OTY2LmF0b20iO31zOjg6ImZyYWdtZW50IjtzOjA6IiI7fQ==) 

22. 22.Cartuyvels R, Spinks G, Moens M-F. Discrete and continuous representations and processing in deep learning: Looking forward. AI Open. 2021;2:143–59.
    
    

23. 23.Wu PP-Y, Mengersen K, McMahon K, Kendrick GA, Chartrand K, York PH, et al. Timing anthropogenic stressors to mitigate their impact on marine ecosystem resilience. Nature Communications. 2017;8(1):1263.
    
    [PubMed](http://medrxiv.org/lookup/external-ref?access_num=29093493&link_type=MED&atom=%2Fmedrxiv%2Fearly%2F2024%2F11%2F08%2F2024.11.08.24316966.atom) 

24. 24.Kim SK, Frisby CL. Gaining from discretization of continuous data: The correspondence analysis biplot approach. Behav Res Methods. 2019;51(2):589–601.
    
    [PubMed](http://medrxiv.org/lookup/external-ref?access_num=30406507&link_type=MED&atom=%2Fmedrxiv%2Fearly%2F2024%2F11%2F08%2F2024.11.08.24316966.atom) 

25. 25.Fenton N, Neil M. Risk assessment and decision analysis with Bayesian networks. Boca Raton: Chapman and Hall/CRC.; 2018.
    
    

26. 26.Eugene C. Bayesian Networks without Tears. AI Magazine. 1991;12(4).
    
    

27. 27.Fuster-Parra P, Tauler P, Bennasar-Veny M, Ligęza A, López-González AA, Aguiló A. Bayesian network modeling: A case study of an epidemiologic system analysis of cardiovascular risk. Computer Methods and Programs in Biomedicine. 2016;126:128–42.
    
    [PubMed](http://medrxiv.org/lookup/external-ref?access_num=26777431&link_type=MED&atom=%2Fmedrxiv%2Fearly%2F2024%2F11%2F08%2F2024.11.08.24316966.atom) 

28. 28.Howard RA. The Foundations of Decision Analysis Revisited. In: von Winterfeldt D, Miles Jr RF, Edwards W, editors. Advances in Decision Analysis: From Foundations to Applications. Cambridge: Cambridge University Press; 2007. p. 32-56.
    
    

29. 29.Heckerman DE, Nathwani BN. Toward normative expert systems: Part II. Probability-based representations for efficient knowledge acquisition and inference. Methods of information in medicine. 1992;31(2):106–16.
    
    [PubMed](http://medrxiv.org/lookup/external-ref?access_num=1635462&link_type=MED&atom=%2Fmedrxiv%2Fearly%2F2024%2F11%2F08%2F2024.11.08.24316966.atom) 
    
    [Web of Science](http://medrxiv.org/lookup/external-ref?access_num=A1992JB36600004&link_type=ISI) 

30. 30.Constantinou A, Fenton N. Things to know about Bayesian networks. Significance. 2018;15:19–23.
    
    

31. 31.Hecksteden A, Kellner R, Donath L. Dealing with small samples in football research. Science and Medicine in Football. 2021:1–9.
    
    

32. 32.Mengersen KL, Drovandi CC, Robert CP, Pyne DB, Gore CJ. Bayesian Estimation of Small Effects in Exercise and Sports Science. PLOS ONE. 2016;11(4):e0147311.
    
    [PubMed](http://medrxiv.org/lookup/external-ref?access_num=27073897&link_type=MED&atom=%2Fmedrxiv%2Fearly%2F2024%2F11%2F08%2F2024.11.08.24316966.atom) 

33. 33.Hecksteden A, Forster S, Egger F, Buder F, Kellner R, Meyer T. Dwarfs on the Shoulders of Giants: Bayesian Analysis With Informative Priors in Elite Sports Research and Decision Making. Frontiers in Sports and Active Living. 2022;4.
    
    

34. 34.Olmedilla A, Rubio VJ, Fuster-Parra P, Pujals C, García-Mas A. A Bayesian Approach to Sport Injuries Likelihood: Does Player’s Self-Efficacy and Environmental Factors Plays the Main Role? Frontiers in Psychology. 2018;9(1174).
    
    

35. 35.Dimitrova L, Petkova KG, editors. Bayesian Network - Based Causal Analysis of Injury Risk in Elite Rhythmic Gymnastics2014.
    
    

36. 36.Lau CL, Mayfield HJ, Sinclair JE, Brown SJ, Waller M, Enjeti AK, et al. Risk-benefit analysis of the AstraZeneca COVID-19 vaccine in Australia using a Bayesian network modelling framework. medRxiv. 2021:2021.09.30.21264337.
    
    

37. 37. El Khediri S, Thaljaoui A, Alfayez F. A Novel Decision-Making Process for COVID-19 Fighting Based on Association Rules and Bayesian Methods. The Computer Journal. 2021.
    
    

38. 38.Johnson S, Mengersen K, de Waal A, Marnewick K, Cilliers D, Houser AM, et al. Modelling cheetah relocation success in southern Africa using an Iterative Bayesian Network Development Cycle. Ecological Modelling. 2010;221(4):641–51.
    
    

39. 39.Wu PP-Y, McMahon K, Rasheed MA, Kendrick GA, York PH, Chartrand K, et al. Managing seagrass resilience under cumulative dredging affecting light: Predicting risk using dynamic Bayesian networks. Journal of Applied Ecology. 2018;55(3):1339–50.
    
    

40. 40.Wu PPY, Pitchforth J, Mengersen K. A Hybrid Queue-based Bayesian Network framework for passenger facilitation modelling. Transportation Research Part C-Emerging Technologies. 2014;46:247–60.
    
    

41. 41.Beaudouin F, Demmerle D, Fuhr C, Tröß T, Meyer T. Head Impact Situations in Professional Football (Soccer). Sports Med Int Open. 2021;5(02):E37–E44.
    
    

42. 42.Heus P, Damen JAAG, Pajouheshnia R, Scholten RJPM, Reitsma JB, Collins GS, et al. Uniformity in measuring adherence to reporting guidelines: the example of TRIPOD for assessing completeness of reporting of prediction model studies. BMJ Open. 2019;9(4):e025611.
    
    [Abstract/FREE Full Text](http://medrxiv.org/lookup/ijlink/YTozOntzOjQ6InBhdGgiO3M6MTQ6Ii9sb29rdXAvaWpsaW5rIjtzOjU6InF1ZXJ5IjthOjQ6e3M6ODoibGlua1R5cGUiO3M6NDoiQUJTVCI7czoxMToiam91cm5hbENvZGUiO3M6NzoiYm1qb3BlbiI7czo1OiJyZXNpZCI7czoxMToiOS80L2UwMjU2MTEiO3M6NDoiYXRvbSI7czo1MDoiL21lZHJ4aXYvZWFybHkvMjAyNC8xMS8wOC8yMDI0LjExLjA4LjI0MzE2OTY2LmF0b20iO31zOjg6ImZyYWdtZW50IjtzOjA6IiI7fQ==) 

43. 43.aus der Fünten K, Faude O, Lensch J, Meyer T. Injury Characteristics in the German Professional Male Soccer Leagues After a Shortened Winter Break. Journal of Athletic Training. 2014;49(6):786–93.
    
    [CrossRef](http://medrxiv.org/lookup/external-ref?access_num=10.4085/1062-6050-49.3.51&link_type=DOI) 
    
    [PubMed](http://medrxiv.org/lookup/external-ref?access_num=25365132&link_type=MED&atom=%2Fmedrxiv%2Fearly%2F2024%2F11%2F08%2F2024.11.08.24316966.atom) 

44. 44.Fuller CW, Ekstrand J, Junge A, Andersen TE, Bahr R, Dvorak J, et al. Consensus statement on injury definitions and data collection procedures in studies of football (soccer) injuries. British Journal of Sports Medicine. 2006;40(3):193–201.
    
    [Abstract/FREE Full Text](http://medrxiv.org/lookup/ijlink/YTozOntzOjQ6InBhdGgiO3M6MTQ6Ii9sb29rdXAvaWpsaW5rIjtzOjU6InF1ZXJ5IjthOjQ6e3M6ODoibGlua1R5cGUiO3M6NDoiQUJTVCI7czoxMToiam91cm5hbENvZGUiO3M6ODoiYmpzcG9ydHMiO3M6NToicmVzaWQiO3M6ODoiNDAvMy8xOTMiO3M6NDoiYXRvbSI7czo1MDoiL21lZHJ4aXYvZWFybHkvMjAyNC8xMS8wOC8yMDI0LjExLjA4LjI0MzE2OTY2LmF0b20iO31zOjg6ImZyYWdtZW50IjtzOjA6IiI7fQ==) 

45. 45.Druzdzel MJ, editor SMILE: Structural Modeling, Inference, and Learning Engine and GeNIe: a development environment for graphical decision-theoretic models. Aaai/Iaai; 1999.
    
    

46. 46.Ekstrand J, Hägglund M, Waldén M. Injury incidence and injury patterns in professional football: the UEFA injury study. British Journal of Sports Medicine. 2011;45(7):553–8.
    
    [Abstract/FREE Full Text](http://medrxiv.org/lookup/ijlink/YTozOntzOjQ6InBhdGgiO3M6MTQ6Ii9sb29rdXAvaWpsaW5rIjtzOjU6InF1ZXJ5IjthOjQ6e3M6ODoibGlua1R5cGUiO3M6NDoiQUJTVCI7czoxMToiam91cm5hbENvZGUiO3M6ODoiYmpzcG9ydHMiO3M6NToicmVzaWQiO3M6ODoiNDUvNy81NTMiO3M6NDoiYXRvbSI7czo1MDoiL21lZHJ4aXYvZWFybHkvMjAyNC8xMS8wOC8yMDI0LjExLjA4LjI0MzE2OTY2LmF0b20iO31zOjg6ImZyYWdtZW50IjtzOjA6IiI7fQ==) 

47. 47.Hägglund M, Waldén M, Ekstrand J. Risk Factors for Lower Extremity Muscle Injury in Professional Soccer:The UEFA Injury Study. The American Journal of Sports Medicine. 2013;41(2):327–35.
    
    [CrossRef](http://medrxiv.org/lookup/external-ref?access_num=10.1177/0363546512470634&link_type=DOI) 
    
    [PubMed](http://medrxiv.org/lookup/external-ref?access_num=23263293&link_type=MED&atom=%2Fmedrxiv%2Fearly%2F2024%2F11%2F08%2F2024.11.08.24316966.atom) 
    
    [Web of Science](http://medrxiv.org/lookup/external-ref?access_num=000314307500016&link_type=ISI) 

48. 48.Cooper GF, Herskovits E. A Bayesian method for the induction of probabilistic networks from data. Machine Learning. 1992;9(4):309–47.
    
    

49. 49.Beuzen T, Marshall L, Splinter KD. A comparison of methods for discretizing continuous variables in Bayesian Networks. Environmental Modelling & Software. 2018;108:61–6.
    
    

50. 50.Liu H, Hussain F, Tan CL, Dash M. Discretization: An Enabling Technique. Data Mining and Knowledge Discovery. 2002;6(4):393–423.
    
    [CrossRef](http://medrxiv.org/lookup/external-ref?access_num=10.1023/A:1016304305535&link_type=DOI) 

51. 51.Dempster AP, Laird NM, Rubin DB. Maximum Likelihood from Incomplete Data via the EM Algorithm. Journal of the Royal Statistical Society Series B (Methodological). 1977;39(1):1–38.
    
    [CrossRef](http://medrxiv.org/lookup/external-ref?access_num=10.1111/j.2517-6161.1977.tb01600.x&link_type=DOI) 
    
    [Web of Science](http://medrxiv.org/lookup/external-ref?access_num=A1977DM46400001&link_type=ISI) 

52. 52.Lauritzen SL. The EM algorithm for graphical association models with missing data. Computational Statistics & Data Analysis. 1995;19(2):191–201.
    
    

53. 53.Hastie T, Tibshirani R, Friedman J. Model Assessment and Selection. The Elements of Statistical Learning: Data Mining, Inference, and Prediction. New York, NY: Springer New York; 2009. p. 219-59.
    
    

54. 54.Congalton RG. A review of assessing the accuracy of classifications of remotely sensed data. Remote Sensing of Environment. 1991;37(1):35–46.
    
    [CrossRef](http://medrxiv.org/lookup/external-ref?access_num=10.1016/0034-4257(91)90048-B&link_type=DOI) 

55. 55.Woods C, Hawkins R, Hulse M, Hodson A. The Football Association Medical Research Programme: an audit of injuries in professional football—analysis of preseason injuries. British Journal of Sports Medicine. 2002;36(6):436–41.
    
    [Abstract/FREE Full Text](http://medrxiv.org/lookup/ijlink/YTozOntzOjQ6InBhdGgiO3M6MTQ6Ii9sb29rdXAvaWpsaW5rIjtzOjU6InF1ZXJ5IjthOjQ6e3M6ODoibGlua1R5cGUiO3M6NDoiQUJTVCI7czoxMToiam91cm5hbENvZGUiO3M6ODoiYmpzcG9ydHMiO3M6NToicmVzaWQiO3M6ODoiMzYvNi80MzYiO3M6NDoiYXRvbSI7czo1MDoiL21lZHJ4aXYvZWFybHkvMjAyNC8xMS8wOC8yMDI0LjExLjA4LjI0MzE2OTY2LmF0b20iO31zOjg6ImZyYWdtZW50IjtzOjA6IiI7fQ==) 

56. 56.Kjærulff U, Van Der Gaag LC, editors. Making sensitivity analysis computationally efficient. Conference on Uncertainty in Artificial Intelligence 2013; Stanford, CA.
    
    

57. 57.Orchard J. Is there a relationship between ground and climatic conditions and injuries in football? Sports medicine (Auckland, NZ). 2002;32(7):419–32.
    
    

58. 58.Azubuike SO, Okojie OH. An epidemiological study of football (soccer) injuries in Benin City, Nigeria. British Journal of Sports Medicine. 2009;43(5):382–6.
    
    [Abstract/FREE Full Text](http://medrxiv.org/lookup/ijlink/YTozOntzOjQ6InBhdGgiO3M6MTQ6Ii9sb29rdXAvaWpsaW5rIjtzOjU6InF1ZXJ5IjthOjQ6e3M6ODoibGlua1R5cGUiO3M6NDoiQUJTVCI7czoxMToiam91cm5hbENvZGUiO3M6ODoiYmpzcG9ydHMiO3M6NToicmVzaWQiO3M6ODoiNDMvNS8zODIiO3M6NDoiYXRvbSI7czo1MDoiL21lZHJ4aXYvZWFybHkvMjAyNC8xMS8wOC8yMDI0LjExLjA4LjI0MzE2OTY2LmF0b20iO31zOjg6ImZyYWdtZW50IjtzOjA6IiI7fQ==) 

59. 59.Thomson A, Whiteley R, Bleakley C. Higher shoe-surface interaction is associated with doubling of lower extremity injury risk in football codes: a systematic review and meta-analysis. British Journal of Sports Medicine. 2015;49(19):1245–52.
    
    [Abstract/FREE Full Text](http://medrxiv.org/lookup/ijlink/YTozOntzOjQ6InBhdGgiO3M6MTQ6Ii9sb29rdXAvaWpsaW5rIjtzOjU6InF1ZXJ5IjthOjQ6e3M6ODoibGlua1R5cGUiO3M6NDoiQUJTVCI7czoxMToiam91cm5hbENvZGUiO3M6ODoiYmpzcG9ydHMiO3M6NToicmVzaWQiO3M6MTA6IjQ5LzE5LzEyNDUiO3M6NDoiYXRvbSI7czo1MDoiL21lZHJ4aXYvZWFybHkvMjAyNC8xMS8wOC8yMDI0LjExLjA4LjI0MzE2OTY2LmF0b20iO31zOjg6ImZyYWdtZW50IjtzOjA6IiI7fQ==) 

60. 60.Stanitski CL, McMaster JH, Ferguson RJ. Synthetic turf and grass: a comparative study. The Journal of sports medicine. 1974;2(1):22–6.
    
    [PubMed](http://medrxiv.org/lookup/external-ref?access_num=4468319&link_type=MED&atom=%2Fmedrxiv%2Fearly%2F2024%2F11%2F08%2F2024.11.08.24316966.atom) 

61. 61.Koiter JR. Visualizing inference in Bayesian networks: Delft University of Technology; 2006.
    
    

62. 62.Fernández J, editor Decomposing the Immeasurable Sport: A deep learning expected possession value framework for soccer. MIT Slogan Sports Analytics Conference; 2019; Boston, MA.
    
    

63. 63.Bullock GS, Hughes T, Arundale AH, Ward P, Collins GS, Kluzek S. Black Box Prediction Methods in Sports Medicine Deserve a Red Card for Reckless Practice: A Change of Tactics is Needed to Advance Athlete Care. Sports Medicine. 2022.
    
    

64. 64.Bittencourt NFN, Meeuwisse WH, Mendonça LD, Nettel-Aguirre A, Ocarino JM, Fonseca ST. Complex systems approach for sports injuries: moving from risk factor identification to injury pattern recognition—narrative review and new concept. British Journal of Sports Medicine. 2016;50(21):1309–14.
    
    [Abstract/FREE Full Text](http://medrxiv.org/lookup/ijlink/YTozOntzOjQ6InBhdGgiO3M6MTQ6Ii9sb29rdXAvaWpsaW5rIjtzOjU6InF1ZXJ5IjthOjQ6e3M6ODoibGlua1R5cGUiO3M6NDoiQUJTVCI7czoxMToiam91cm5hbENvZGUiO3M6ODoiYmpzcG9ydHMiO3M6NToicmVzaWQiO3M6MTA6IjUwLzIxLzEzMDkiO3M6NDoiYXRvbSI7czo1MDoiL21lZHJ4aXYvZWFybHkvMjAyNC8xMS8wOC8yMDI0LjExLjA4LjI0MzE2OTY2LmF0b20iO31zOjg6ImZyYWdtZW50IjtzOjA6IiI7fQ==) 

65. 65.Murphy KP, Russell SJ, editors. Dynamic bayesian networks: representation, inference and learning2002.
    
    

66. 66.Nielsen RØ, Malisoux L, Møller M, Theisen D, Parner ET. Shedding Light on the Etiology of Sports Injuries: A Look Behind the Scenes of Time-to-Event Analyses. Journal of Orthopaedic & Sports Physical Therapy. 2016;46(4):300–11.
    
    [PubMed](http://medrxiv.org/lookup/external-ref?access_num=26954269&link_type=MED&atom=%2Fmedrxiv%2Fearly%2F2024%2F11%2F08%2F2024.11.08.24316966.atom) 

67. 67.Sandri M, Berchialla P, Baldi I, Gregori D, De Blasi RA. Dynamic Bayesian Networks to predict sequences of organ failures in patients admitted to ICU. Journal of Biomedical Informatics. 2014;48:106–13.
    
    [PubMed](http://medrxiv.org/lookup/external-ref?access_num=24361388&link_type=MED&atom=%2Fmedrxiv%2Fearly%2F2024%2F11%2F08%2F2024.11.08.24316966.atom) 

68. 68.van Gerven MAJ, Taal BG, Lucas PJF. Dynamic Bayesian networks as prognostic models for clinical patient management. Journal of Biomedical Informatics. 2008;41(4):515–29.
    
    [CrossRef](http://medrxiv.org/lookup/external-ref?access_num=10.1016/j.jbi.2008.01.006&link_type=DOI) 
    
    [PubMed](http://medrxiv.org/lookup/external-ref?access_num=18337188&link_type=MED&atom=%2Fmedrxiv%2Fearly%2F2024%2F11%2F08%2F2024.11.08.24316966.atom) 

69. 69.Galan SF, Aguado F, Dıez F, Mira J. NasoNet, modeling the spread of nasopharyngeal cancer with networks of probabilistic events in discrete time. Artificial Intelligence in Medicine. 2002;25(3):247–64.
    
    [PubMed](http://medrxiv.org/lookup/external-ref?access_num=12069762&link_type=MED&atom=%2Fmedrxiv%2Fearly%2F2024%2F11%2F08%2F2024.11.08.24316966.atom) 

70. 70.Mueller-Wohlfahrt H-W, Haensel L, Mithoefer K, Ekstrand J, English B, McNally S, et al. Terminology and classification of muscle injuries in sport: The Munich consensus statement. British Journal of Sports Medicine. 2013;47(6):342–50.
    
    [Abstract/FREE Full Text](http://medrxiv.org/lookup/ijlink/YTozOntzOjQ6InBhdGgiO3M6MTQ6Ii9sb29rdXAvaWpsaW5rIjtzOjU6InF1ZXJ5IjthOjQ6e3M6ODoibGlua1R5cGUiO3M6NDoiQUJTVCI7czoxMToiam91cm5hbENvZGUiO3M6ODoiYmpzcG9ydHMiO3M6NToicmVzaWQiO3M6ODoiNDcvNi8zNDIiO3M6NDoiYXRvbSI7czo1MDoiL21lZHJ4aXYvZWFybHkvMjAyNC8xMS8wOC8yMDI0LjExLjA4LjI0MzE2OTY2LmF0b20iO31zOjg6ImZyYWdtZW50IjtzOjA6IiI7fQ==) 

71. 71.Ekstrand J, Askling C, Magnusson H, Mithoefer K. Return to play after thigh muscle injury in elite football players: implementation and validation of the Munich muscle injury classification. British Journal of Sports Medicine. 2013;47(12):769.
    
    [Abstract/FREE Full Text](http://medrxiv.org/lookup/ijlink/YTozOntzOjQ6InBhdGgiO3M6MTQ6Ii9sb29rdXAvaWpsaW5rIjtzOjU6InF1ZXJ5IjthOjQ6e3M6ODoibGlua1R5cGUiO3M6NDoiQUJTVCI7czoxMToiam91cm5hbENvZGUiO3M6ODoiYmpzcG9ydHMiO3M6NToicmVzaWQiO3M6OToiNDcvMTIvNzY5IjtzOjQ6ImF0b20iO3M6NTA6Ii9tZWRyeGl2L2Vhcmx5LzIwMjQvMTEvMDgvMjAyNC4xMS4wOC4yNDMxNjk2Ni5hdG9tIjt9czo4OiJmcmFnbWVudCI7czowOiIiO30=)

 [1]: /embed/graphic-6.gif
 [2]: /embed/graphic-7.gif