Injury Risk Assessment and Surveillance in Professional Football

In professional football, the relationship between squad availability and competitive outcomes is well documented. Teams that maintain a healthy, available squad through the final stages of a competition consistently give themselves better conditions to perform. High injury burden correlates with underperformance across a season, and the effect is particularly evident during congested fixture periods. In this sense, injuries function as a collective performance problem as much as an individual clinical one, and addressing them effectively requires understanding them systematically.

Having spent the last decade at Manchester City, following earlier stages at Lazio, Palermo, and extensive work with national teams including Argentina at the 2022 World Cup and Ukraine at the 2020 European Championship, one observation recurs: the gap between practitioners who manage injuries reactively and those who manage them proactively tends to be a primary differentiator in performance outcomes at elite level. The frameworks described in this article apply well beyond professional football. Any practitioner working with athletes in a high-load, high-performance context will find these principles relevant.

The van Mechelen Framework: A Logical Structure for Prevention

The most important conceptual foundation for any injury management program is the four-step prevention sequence proposed by van Mechelen and colleagues (1992). It provides a structure that prevents the most common error in this field: jumping to interventions before adequately understanding the problem.

The first step is injury surveillance — the systematic, standardised, ongoing collection of data about the occurrence, type, severity, and circumstances of injuries. Without it, there is no baseline, no trend data, and no ability to evaluate whether any preventive intervention is producing measurable change. The second step is identification of etiology — understanding the factors that cause and contribute to injury, both intrinsic to the athlete and extrinsic to their environment. The third step is introduction of preventive strategies — the targeted interventions that etiology analysis reveals. The fourth step is evaluation of effectiveness — returning to the surveillance data to assess whether those strategies are working.

What makes this framework particularly useful is that steps one and four represents the same activity. Surveillance is not just a starting point — it is the feedback loop that closes the entire system. A program without ongoing, rigorous surveillance has no mechanism for knowing whether it is actually preventing anything. The cycle is only complete when data informs decisions and updated data evaluates those decisions.

Defining an Injury — and Why the Definition Matters

There is more conceptual confusion in injury recording than most practitioners realise. The foundational definition, established through the International Olympic Committee consensus in 2020 (Bahr et al., 2020), is broad: an injury is any physical complaint sustained by an athlete that results from sport-specific exposure, whether in training or competition. But recording systems vary considerably in how they operationalise this.

The most straightforward injuries to document are time-loss injuries — those that keep an athlete away from training or competition for at least one day. If a player cannot participate, their absence is documented. This is the layer most organisations record, but it represents only the most visible portion of a broader injury picture.

Below it lie medical attention injuries — problems that require consultation with a medical practitioner, imaging such as MRI or ultrasound, or treatment, but that do not prevent training or competition. These frequently go unrecorded, despite carrying real clinical significance. And below even that are athlete complaints — the minor sprains, the nagging discomfort, the compensatory pattern a player mentions in passing but never formally reports. These are almost never systematically captured, despite potentially representing the earliest signals of developing pathology.

The consequence of relying exclusively on time-loss recording is serious. A minor ankle sprain that goes unrecorded quietly alters the biomechanics of load distribution through the lower limb. A compensatory pattern develops. Months later, a hamstring injury occurs, and no one can connect it to the original complaint because the link was never captured. To interrupt this chain, practitioners need both a reporting culture in which athletes disclose minor issues without stigma and an athlete management system capable of storing, linking, and longitudinally analysing all levels of complaint — not only the ones that generate absences.

Five Core Variables Every Recording System Must Capture

Regardless of which surveillance system an organisation uses, five core variables must be recorded for every injury to ensure the data can actually inform clinical and performance decision-making.

1) Body area, tissue type, and pathology type, coded using the Orchard Sports Injury and Illness Classification System (OSICS). This provides a standardised nomenclature applicable across the entire body. Advances in MRI and diagnostic ultrasound have made the boundary between muscle and tendon injury increasingly nuanced, and the classification system accounts for this complexity. The pathology type within each tissue category — whether a hamstring injury is a partial tear, a complete rupture, or a myositis — determines both the clinical management approach and the realistic expected return-to-play timeline.

2) Relationship to activity — whether the injury occurred in training or competition, and within training, what type of session was involved: strength and conditioning, technical-tactical, recovery, or otherwise. If surveillance data consistently shows that 80% of injuries occur during training, and within those sessions the majority arise in high-intensity technical work, that is a direct signal to review session design and load distribution.

3) Mode of onset and mechanism — whether the injury was acute or overuse in nature, and for acute injuries, the specific sport action involved. Football-specific consensus guidelines identify the key actions to record: acceleration, deceleration, change of direction, contact with an opponent, kicking, and landing. Each action points toward different areas of the physical preparation program that may require attention. A hamstring injury during maximum-velocity sprinting has different prevention implications than one occurring during a deceleration or a cutting movement.

4) Multiple injury recording — documenting the complete timeline of injuries across an athlete's career and analysing how each relates to those that preceded and followed it. This longitudinal dimension is where clinically relevant patterns tend to emerge, and where computational models can identify configurations that manual inspection would not reliably detect.

5) Time loss and availability impact — quantifying severity by recording the number of days, training sessions, and matches missed, and noting whether that absence was continuous or intermittent. This last distinction is particularly important for overuse injuries such as tendinopathies, which characteristically produce recurring flares rather than a single defined absence.

The Overuse Injury Problem

Overuse injuries deserve special attention because they are systematically underrepresented in standard surveillance systems, despite their disproportionate impact on both availability and performance quality. A player managing patellar tendinopathy may never miss a complete match yet may train at consistently reduced capacity for months, competing in pain and with compromised movement mechanics. A time-loss recording system will show zero absences for this athlete and conclude there is no injury problem — missing the clinical and performance reality entirely.

The Oslo Sports Trauma Research Center Overuse Injury Questionnaire addresses this gap directly (Clarsen, Myklebust and Bahr, 2013). Four questions, submitted to athletes on a weekly basis, capture the presence, severity, and performance impact of ongoing complaints independently of whether those complaints generate any training absence. Over the course of a season, this produces a longitudinal dataset of overuse injury burden that is otherwise completely invisible to conventional recording. The practical barrier to implementing it is low: the questionnaire takes under two minutes per player per week. The informational return is substantial. Every practitioner managing athletes in high-load sports should treat weekly overuse monitoring as a standard component of their surveillance system.

From Linear Risk Factors to Complex Systems

Between approximately 2010 and 2015, sports injury risk factor research was largely linear and reductionist. The dynamic model proposed by Meeuwisse and colleagues (2007) described a pathway in which internal risk factors predispose an athlete, external factors make them susceptible, and a specific inciting event triggers injury. This generated a generation of studies that isolated individual variables: hip internal rotation range of motion in groin injury cases, single-leg landing mechanics in ACL injury populations, ankle mobility in Achilles tendinopathy patients.

The fundamental limitation of this approach is that it treats risk factors as independent, when in reality the musculoskeletal system operates as an integrated complex system in which multiple variables interact simultaneously. Bittencourt and colleagues (2016) proposed the first explicitly non-linear model of sports injury etiology, recognising that individual risk factors do not simply add up — they form networks, and it is the configuration of the entire network at a given moment, combined with an inciting event, that determines whether an injury occurs.

A player with restricted hip rotation may or may not sustain a groin injury depending on how that restriction interacts with their current training load, their movement history, their muscle balance profile, and many other contextual variables. Two athletes with identical ROM measurements can have completely different injury trajectories. This is why risk factor studies with small sample sizes and single-variable designs have produced such inconsistent findings across the literature: they are studying isolated variables within a system whose behaviour is determined by interactions, not individual components.

The practical implication for practitioners is equally significant: structure and function are mutually influential and both are shaped by injury history. Every injury an athlete has sustained leaves a permanent structural trace, because tissue repair does not produce perfect anatomical restitution. Altered motor recruitment patterns, areas of increased tissue stiffness, and compensatory loading strategies can persist across years and accumulate across careers. Assessing an athlete's current risk profile without understanding their injury history means assessing an incomplete system.

Building the Injury Risk Profile: Assessment and Monitoring

The practical process of profiling athletes for injury risk operates across two streams that complement rather than duplicate each other.

The first is the periodical health evaluation — a comprehensive assessment conducted two to three times per season that examines structure, function, and history in detail. These evaluations involve objective measurements of joint mobility, tissue stiffness, and postural alignment, functional movement assessments, and sport-specific performance tests. They are time-intensive but generate the high-resolution baseline data that continuous monitoring alone cannot provide.

The second is continuous athlete monitoring — the daily and weekly collection of data on training load, physical output, wellness, recovery quality, and subjective wellbeing. Monitoring data is less granular than evaluation data but captures fluctuations across time that a periodic assessment would miss entirely. An athlete's RPE trend over three consecutive weeks, overnight HRV trajectory, or GPS-derived high-speed running accumulation across a congested fixture period — these variables are limited in isolation but become powerful when modelled against injury history and baseline evaluation data.

Both streams should integrate three types of data: objective data from reliable measurement tools (force platforms, GPS, heart rate monitoring, strength testing); subjective athlete-reported data (RPE, player-reported outcome measures, wellness questionnaires); and qualitative practitioner data (hands-on tissue assessment, movement observation, clinical reasoning). No single source is sufficient on its own. The injury risk profile emerges from their integration.

A useful operational distinction is between Level 1 testing — standardised assessments conducted across all players, calibrated to the predominant injury patterns of the sport and the specific team — and Level 2 testing — individualised additions for players whose history or emerging signals suggest specific vulnerabilities not addressed by the team-wide protocol. A player who has sustained three ankle sprains over four seasons requires additional ankle-specific assessment and monitoring above and beyond whatever the group protocol requires.

Data Integration: Where Most Organisations Fall Short

All of the assessment and monitoring activity described above generates real value only if the data is brought together in a single integrated system and analysed as a whole. This is the point at which most organisations, including many at elite level, fall short.

In the typical setup, strength data lives in force plate software, injury history in the medical records system, body composition in a DEXA report, GPS output in a training management platform, and wellness data in a separate questionnaire application. When these streams are siloed — and they usually are — the information cannot be analysed in relation to itself. The practitioner is forced to draw on fragments rather than the complete picture, making decisions on partial evidence and missing the multi-factor interaction patterns that most reliably predict injury risk.

Connecting all data streams into a single integrated system and applying appropriate analytical models — whether statistical or machine learning-based — is a necessary condition for identifying the multi-factor configurations that precede injury. Models trained on longitudinal integrated datasets can recognise patterns that no human analyst would identify through manual inspection, and can flag emerging risk in real time, before injury occurs rather than after.

The practical choice is between a reactive approach — waiting for injuries to occur and responding — and a proactive one that commits to the full cycle: surveillance, etiology analysis, risk profiling, continuous monitoring, integrated data, and targeted intervention. The second approach requires sustained investment and interdisciplinary collaboration. The evidence from elite football and other high-performance sports indicates that it produces meaningful reductions in injury burden, supports athlete availability across a full season, and reduces the gap between a squad's theoretical performance potential and what it actually delivers on the pitch.

---

Frequently Asked Questions

Q: Why do time-loss recording systems alone give an inaccurate picture of a team's injury burden?

Time-loss recording only captures the visible surface of a much larger problem. Below it lie medical attention injuries — those requiring clinical consultation or imaging but not causing absence — and below those, athlete complaints that may never be formally reported at all. A player managing a chronic tendinopathy or recurring muscle soreness who never misses training will show zero time-loss data, yet their performance capacity may be substantially compromised for months. A complete picture requires recording all three levels and using specific tools — such as the Oslo Overuse Injury Questionnaire — to capture what absence-based systems miss.

Q: What distinguishes the complex systems approach to injury etiology from earlier linear models?

Linear models identified individual risk factors and studied them in isolation — a specific range of motion measurement, a specific strength asymmetry — treating each as independently causative. The complex systems model proposed by Bittencourt and colleagues (2016) recognises that risk factors form interconnected networks, and that injury emerges from the configuration of the whole network at a given moment, not from any single variable. This explains why single-variable risk factor studies have produced inconsistent predictive validity: they are studying one node of a network rather than the network itself.

Q: How should organisations prioritise which assessments to include in a Level 1 vs. Level 2 protocol?

Level 1 assessments should be calibrated to the injury profile of the sport and the team — if hamstring injuries represent 25% of time-loss burden, hamstring-specific functional assessments should be in the universal battery. Level 2 additions are determined by individual history. A player with recurrent ankle sprains needs specific ankle assessment and monitoring; a player with a history of groin problems needs hip-region profiling beyond what the group protocol provides. The Level 1 / Level 2 distinction prevents both under-assessment (missing individual vulnerabilities) and over-assessment (spending evaluation time on low-probability risks for a given player).

Q: What is the practical value of monitoring HRV and subjective wellness alongside GPS load data?

Training load metrics tell you what the body was exposed to; wellness and recovery metrics tell you how the body responded. The same GPS output — say, 900m of high-speed running — carries very different physiological implications for an athlete who is well-recovered, sleeping well, and showing stable HRV versus one who is in accumulated fatigue with declining wellness scores and a suppressed HRV trend. Integrating both streams allows practitioners to distinguish load accumulation from load tolerance deterioration, which is the relevant distinction for injury risk management.

Q: At what point should machine learning models be applied to injury data, and what do they add over conventional analysis?

Machine learning becomes useful when the dataset has sufficient longitudinal depth — at minimum, one or two full seasons of multi-variable integrated data per athlete — and when the relationships between variables are too complex for conventional statistical methods to capture reliably. Their specific advantage is the ability to identify multi-factor interaction patterns: combinations of load, movement quality, wellness, and history variables that precede injury events in non-obvious ways. They do not replace clinical judgement; they surface signals that practitioners can then evaluate in context. The prerequisite is data integration — a machine learning model applied to fragmented, siloed data will produce fragmented, unreliable outputs.

References

• Bahr, R., Clarsen, B., Derman, W. et al. (2020) 'International Olympic Committee consensus statement: methods for recording and reporting of epidemiological data on injury and illness in sport 2020 (STROBE-SIIS)', British Journal of Sports Medicine, 54(7), pp. 372–389.

• Bittencourt, N.F.N., Meeuwisse, W.H., Mendonça, L.D., Nettel-Aguirre, A., Ocarino, J.M. and Fonseca, S.T. (2016) '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, 50(21), pp. 1309–1314.

• Clarsen, B., Myklebust, G. and Bahr, R. (2013) 'Development and validation of a new method for the registration of overuse injuries in sports injury epidemiology: the Oslo Sports Trauma Research Centre (OSTRC) overuse injury questionnaire', British Journal of Sports Medicine, 47(8), pp. 495–502.

• Meeuwisse, W.H., Tyreman, H., Hagel, B. and Emery, C. (2007) 'A dynamic model of etiology in sport injury: the recursive nature of risk and causation', Clinical Journal of Sport Medicine, 17(3), pp. 215–219.

• van Mechelen, W., Hlobil, H. and Kemper, H.C. (1992) 'Incidence, severity, aetiology and prevention of sports injuries: a review of concepts', Sports Medicine, 14(2), pp. 82–99.

Federico serves as Director of Performance at Manchester City FC, where he has been part of the performance and medical staff since 2016, supporting injury prevention, rehabilitation, and performance optimisation for elite players.

He brings more than fifteen years of experience in professional sport, having worked with top-level clubs such as S.S. Lazio and Palermo FC, as well as with national teams in Olympic preparation contexts.

He was part of the technical staff of the Argentina national football team during the 2018 FIFA World Cup and has collaborated with national federations across sports such as judo, wrestling, and karate, broadening his expertise across different competitive demands.