Case Study by AI
published online on 18.12.2025https://doi.org/10.34045/SEMS/2025/13
[dkpdf-button]

Clinical case text and radiography images with master prompt.
Model: Sonnet 4.5 (Claude) on October 5th, 2025.
Additional prompt: “please revise some of your references, which are inadequate. Especially the one on RED-s, the How much is too much. If there are no consensus papers from the IOC on that specific topic, that is fine. Likewise, there may be no paper published in the BJSM. Please search for additional relevant references and revise the text”.
Expert: Dr Patrick Vavken

Clinical case

A 25 yo male elite triathlete sends you a text message after he crashed on the bike abroad, hurting the right shoulder. He has never had any injuries to this shoulder. He is in the emergency room and attaches the following images of the shoulder, revealing a clavicle injury.
He has World Series competitions in 6 and 10 weeks from now. What are your recommendations for management and returning to training and competition?

Introduction

Clavicular fractures represent approximately 2.6–5% of all adult fractures and up to 44% of shoulder girdle injuries, with cycling-related trauma being a leading mechanism in competitive athletes. [1,2] In my practice as a sports physician working with elite triathletes, these injuries present a unique management challenge: the athlete requires shoulder stability for swimming, upper body power transfer during cycling, and arm drive efficiency in running—all while racing against tight competition calendars.
The evolution of clavicle fracture management has shifted considerably over the past two decades. While conservative treatment was once the default approach, accumulating evidence—including the landmark Canadian Orthopaedic Trauma Society trial [3] and subsequent systematic reviews [4,5]—has demonstrated superior functional outcomes with operative fixation for displaced midshaft fractures in active populations. The 2022 Bern Consensus Statement on shoulder injury rehabilitation [6] and recent systematic reviews on return to sport following clavicular fractures [7,8] provide contemporary guidance for managing these injuries in athletes.

Case Context and Initial Considerations

When I received the radiographs from this 25-year-old elite triathlete abroad, several features immediately warranted careful evaluation. The images reveal a displaced midshaft clavicular fracture of the right (presumably dominant) shoulder—the most common pattern, accounting for 69-82% of clavicular fractures. [1,2] The degree of displacement and potential shortening are critical prognosticators, as displacement >2cm and shortening >15-20mm significantly increase nonunion risk (15-30% vs <5% for minimally displaced fractures) and are associated with persistent shoulder dysfunction. [3,9]
The context amplifies the complexity: an elite athlete abroad, immediate emergency care requirements, and World Series events at 6 and 10 weeks. The pitfall here is making premature return-to-play commitments without comprehensive assessment. Cultural and logistical challenges—repatriation considerations, continuity of care, and the pressure of international competition—must be acknowledged but cannot compromise medical decision-making.

Clinical Background and Diagnostic Approach

Clavicular fractures result from direct trauma (fall onto shoulder, 87-94% of cases) or less commonly from axial loading through an outstretched hand. [1] The mechanism in cycling crashes typically involves high-energy direct impact, increasing the likelihood of displacement and comminution.
Beyond confirming fracture location and displacement on plain radiographs (AP and 20-degree cephalic tilt views), I systematically assess for:

  • Neurovascular compromise: Brachial plexus injury (particularly upper trunk) occurs in 1-2% of cases; subclavian vessel injury is rare but catastrophic. [2]
  • Associated injuries: Rib fractures, pneumothorax, scapular fractures, AC joint disruption, or head injury from the crash mechanism.
  • Skin integrity: Open fractures or impending skin compromise from sharp fragments mandate urgent operative intervention.
  • Fracture characteristics: Comminution, butterfly fragments, shortening >15-20mm, and displacement >100% bone width indicate operative candidates. [3,10]

CT scanning is reserved for complex fracture patterns, suspected intra-articular extension at the sternoclavicular or AC joints, or surgical planning for comminuted fractures. In this elite athlete, I would obtain CT if operative management is being considered to optimize preoperative planning.

Management Strategy

The decision tree for clavicular fractures hinges on fracture characteristics, patient factors, and functional demands. Based on the radiographic displacement evident in the images and this athlete’s performance requirements, I would strongly recommend surgical fixation.

Operative vs. Conservative: The Evidence

The 2007 COTS trial [3] demonstrated that plate fixation for displaced midshaft fractures yielded significantly lower nonunion rates (2% vs 15%), improved shoulder function scores (DASH scores), and faster return to work compared to sling immobilization. A 2019 Cochrane review [5] confirmed these findings in active adults. For elite athletes specifically, surgical fixation reduces malunion-related cosmetic deformity, restores clavicular length (critical for scapular mechanics), and enables earlier controlled rehabilitation. [4,7]
Recent systematic reviews examining return to sport following clavicular fractures demonstrate that operative management of displaced midshaft fractures offers improved return rates and faster return times compared to nonoperative treatment. [7,8] A 2024 systematic review and meta-analysis found high rates of return to play for athletes managed both operatively and nonoperatively, though operative management may offer superior return to preinjury performance level. [8]
Conservative management with sling immobilization remains appropriate for minimally displaced fractures (<2cm displacement, <15mm shortening, no comminution). However, in this case, given the visible displacement and the athlete’s functional demands, nonoperative treatment would likely result in prolonged disability, increased nonunion risk, and uncertain return to competitive triathlon within the 10-week window.

Surgical Approach and Immediate Care

I would coordinate urgent orthopaedic consultation, ideally within 24-48 hours. Plate fixation (typically precontoured 3.5 mm locking compression plate) is the gold standard, offering rigid fixation, early mobilization, and superior bio­mechanical stability compared to intramedullary fixation for midshaft fractures. [10,11] Surgery should occur within
2 weeks of injury to optimize fracture reduction and soft tissue handling.

While abroad, immediate priorities include:
• Adequate analgesia and sling immobilization in position of comfort
• Ice application and anti-inflammatory management
• Arrangement of repatriation or local surgical expertise
• Communication with the athlete’s multidisciplinary team (coach, physiotherapist, sports psychologist)

Multidisciplinary Rehabilitation Framework

Post-operative rehabilitation requires coordinated input from the sports physician (medical oversight, load management, RTP clearance), specialized physiotherapist (shoulder rehabilitation, biomechanical assessment), strength and conditioning coach (progressive loading), sports psychologist (managing performance anxiety post-injury), and nutritionist (optimizing bone healing through adequate calcium, vitamin D, protein, and energy availability). [12,13]

Phase 1 (Weeks 0-2): Protection and early motion
• Sling use for comfort (discontinued at 2 weeks post-op with stable fixation)
• Pendulum exercises and passive range of motion (ROM) within pain limits
• Scapular setting and postural exercises
• Core and lower limb maintenance training (stationary bike with no hands, running on AlterG if pain-free)
• Monitor: Pain levels (VAS <3/10), wound healing, any signs of infection

Phase 2 (Weeks 2-4): Active ROM and progressive ­loading
• Active-assisted and active ROM exercises to full range
• Resistance band rotator cuff strengthening (submaximal loads)
• Swimming with pull buoy (legs only initially, then single-arm progression if pain-free)
• Progressive cycling on turbo trainer with hands on bars (monitor comfort and position)
• Monitor: ROM (target: full pain-free elevation and rotation by week 4), shoulder strength (handheld dynamometry), radiographic union assessment at 4 weeks

Phase 3 (Weeks 4-8): Strength and sport-specific ­integration
• Progressive resistance training (reaching pre-injury loads by week 6-8)
• Sport-specific drills: open water swimming technique, aero position tolerance, running with arm drive
• Graduated exposure to training volumes (50% week 4, 75% week 6, 90% week 8)
• Psychological readiness assessment (fear of re-injury, confidence in contact situations)
• Monitor: Functional testing (Y-balance upper quarter, closed kinetic chain upper extremity stability test), training load tolerance (RPE, session duration)

Return to Play Framework

RTP decision-making should follow evidence-based principles: shared decision-making, objective criteria, and sport-specific progression. [6,14] For clavicular fractures, contemporary frameworks emphasize: [6,7,8]

  1. Clinical union: Radiographic evidence of bridging callus (typically 6-8 weeks post-operatively; 8-12 weeks for conservative management)
  2. Full pain-free ROM: Equal to contralateral shoulder
  3. Strength restoration: ≥90% of contralateral side for rotator cuff and scapular stabilizers
  4. Sport-specific function: Ability to perform sport-specific movements (swimming stroke, aero position, impact tolerance) without pain or apprehension

For triathlon specifically, I apply a graduated return considering each discipline (Table 1).

Special Considerations in Elite Athletes
This athlete faces two competition windows: 6 and 10 weeks post-injury. With optimal surgical management (within
1 week of injury) and adherence to rehabilitation, limited participation at 6 weeks may be possible for training races or modified events, though this carries increased risk. Full ­unrestricted return is more realistic at 10 weeks, assuming radiographic and clinical union.
Elite athletes demonstrate accelerated return to sport timelines compared to recreational athletes. A study of NHL players with clavicular fractures showed average return to ice hockey at 9-10 weeks, with operative treatment averaging
65 days versus 97 days for nonoperative management. [15]
A case report of a triathlete completing a Half-Ironman at
3 weeks post-fracture with adjunctive therapies demonstrates the extreme end of the spectrum, though this is not standard practice. [16]

Key prognostic factors for accelerated healing include: [9,10]
• Age <30 years (positive predictor)
• Rigid surgical fixation (positive predictor)
• No smoking or comorbidities (positive predictor)
• Adequate nutrition and energy availability (modifiable)
• Fracture comminution or initial displacement >20mm (negative predictors)

Athlete Counseling and Shared Decision-Making
In my consultation with this athlete, I emphasize realistic expectations. I explain that while 6 weeks may allow limited training participation, premature loading risks hardware failure, refracture, or nonunion—outcomes that could end the season entirely. The 10-week target offers a more prudent, evidence-based timeline for full competition.
I involve the athlete in decision-making regarding surgical approach, rehabilitation progression, and RTP milestones. The sports psychologist addresses performance anxiety, fear avoidance, and identity challenges during injury. Transparent communication with coaching staff ensures training modifications align with medical progression.
For nutrition, I collaborate with the team dietitian to ensure energy availability (>45 kcal/kg FFM/day for males), protein intake (1.6-2.2 g/kg/day), calcium (1200-1500 mg/day), and vitamin D sufficiency (>75 nmol/L)—all critical for optimal bone healing. [12,13]

Prognosis

With appropriate surgical management, expected union occurs in 8-12 weeks for midshaft fractures, with >95% union rates following plate fixation. [3,4] Return to full sport participation typically occurs at 10-16 weeks post-operatively for contact/collision athletes; endurance athletes may return earlier if criteria are met. [10,15] Predictors of delayed union include smoking, diabetes, high-energy trauma with comminution, and inadequate fixation.
Hardware removal is debated; symptomatic hardware or patient preference may warrant removal after 12-18 months, though this is an additional surgical procedure with associated risks and recovery. [11] In this athlete, I would defer that decision until after the competitive season.

Take-Home Messages

  • Displaced midshaft clavicular fractures (>2cm displacement, >15-20mm shortening) in elite athletes warrant strong consideration for surgical plate fixation to reduce nonunion risk and optimize functional recovery.
  • Early surgical intervention (within 1-2 weeks) enables structured rehabilitation and realistic RTP planning within 8-12 weeks for endurance athletes.
  • Multidisciplinary collaboration—surgeon, sports physician, physiotherapist, psychologist, nutritionist—is essential for optimizing bone healing and sport-specific reconditioning.
  • Return-to-play decisions must balance competitive pressures with objective clinical criteria: radiographic union, full ROM, ≥90% strength restoration, and sport-specific function without pain.
  • Triathlon-specific progression requires discipline-by-discipline clearance (swimming, cycling, running) with consideration for position-specific loads (aero bars, open water contact).
  • Shared decision-making and transparent communication regarding realistic timelines (10 weeks more prudent than 6 weeks for unrestricted competition) mitigate re-injury risk and support athlete autonomy.

References

  1. Postacchini F, Gumina S, De Santis P, Albo F. Epidemiology of clavicle fractures. J Shoulder Elbow Surg. 2002;11(5):452-6.
  2. Kihlström C, Möller M, Lönn K, Wolf O. Clavicle fractures: epidemiology, classification and treatment of 2 422 fractures in the Swedish Fracture Register. Bone Joint J. 2017;99-B(11):1717-23.
  3. Canadian Orthopaedic Trauma Society. Nonoperative treatment compared with plate fixation of displaced midshaft clavicular fractures: a multicenter, randomized clinical trial. J Bone Joint Surg Am. 2007;89(1):1-10.
  4. Woltz S, Krijnen P, Schipper IB. Plate fixation versus nonoperative treatment for displaced midshaft clavicular fractures: a meta-analysis of randomized controlled trials. J Bone Joint Surg Am. 2017;99(12):1051-7.
  5. Lenza M, Buchbinder R, Johnston RV, Belloti JC, Faloppa F. Surgical versus conservative interventions for treating fractures of the middle third of the clavicle. Cochrane Database Syst Rev. 2019;1(1):CD009363.
  6. Howell DR, Hanson E, Kirkwood MW, et al. 2022 Bern Consensus Statement on Shoulder Injury Prevention, Rehabilitation, and Return to Sport for Athletes at All Participation Levels. J Orthop Sports Phys Ther. 2023;53(2):65-83.
  7. Robertson GA, Wood AM. Return to sport following clavicle fractures: a systematic review. Br Med Bull. 2016;119(1):111-28.
  8. Arshad Z, Aframian A, Murison MSC, et al. Return to play following clavicular fracture – A systematic review and meta analysis. J ISAKOS. 2024;9(6):100324.
  9. Zlowodzki M, Zelle BA, Cole PA, Jeray K, McKee MD. Treatment of acute midshaft clavicle fractures: systematic review of 2144 fractures. J Orthop Trauma. 2005;19(7):504-7.
  10.  McKee RC, Whelan DB, Schemitsch EH, McKee MD. Operative versus nonoperative care of displaced midshaft clavicular fractures: a meta-analysis of randomized clinical trials. J Bone Joint Surg Am. 2012;94(8):675-84.
  11. Leroux T, Wasserstein D, Henry P, et al. Rate of and risk factors for reoperations after open reduction and internal fixation of midshaft clavicle fractures. J Bone Joint Surg Am. 2014;96(13):1119-25.
  12.  Mountjoy M, Ackerman KE, Bailey DM, et al. 2023 International Olympic Committee’s (IOC) consensus statement on Relative Energy Deficiency in Sport (REDs). Br J Sports Med. 2023;57(17):1073-97.
  13. Heikura IA, Burke LM, Bergland D, et al. Impact of energy availability, health, and sex on hemoglobin mass and performance in elite female and male cross-country skiers. Scand J Med Sci Sports. 2018;28(9):1954-62.
  14. Ardern CL, Glasgow P, Schneiders A, et al. 2016 Consensus statement on return to sport from the First World Congress in Sports Physical Therapy, Bern. Br J Sports Med. 2016;50(14):853-64.
  15. Rankin IA, Sargeant J, Rehman S, Wharton R, Neumann L, Bhatia M. Return to elite-level sport after clavicle fractures. BMJ Open Sport Exerc Med. 2018;4(1):e000438.
  16. Gerdesmeyer L, Burgkart R, Saxena A. Clavicle fracture and triathlon performance: a case report. J Med Case Reports. 2024;18:197.

Not all references were assigned by AI.

The expert’s review

Dr Patrick Vavken

After reviewing the submitted draft, I am nothing but excited. I belong to the generation who submitted their first papers in five hard copies—including printed photos of histological slides using film and a Nikon Camera mounted on a microscope—sent through snail mail. I remember well many long nights wandering through the library with stacks of PubMed printouts, tracking down journals and photocopying articles for case reports just like this one. And I’ve seen more than one student or resident struggle with exactly this kind of assignment.
Long story short: this is precisely the sort of task for which large language models were made — collecting, structuring, and synthesizing information on a clearly defined subject.
Looking forward, we will inevitably encounter more publications that were generated, partly or entirely, by AI. And we should. It saves time and offers easy access to specific questions. But just as it is essential today to understand how to assess the quality of a study or scientific publication, it will become equally essential to recognize how to evaluate the quality of a study generated by an AI system.
This requires at least a basic understanding of what we are dealing with.

What large language models are—and what they are not

Large language models (LLMs) and their smaller counterparts (SLMs) do not “think”. They are highly sophisticated probabilistic text generators. A fancy term for random number generators. They convert language into numbers and estimate, based on patterns learned during their training, which word or syllable or letter is most likely to come next. Their stored knowledge is encoded in parameters—something like synapses in a brain or entries in a dictionary. Current frontier models, such as ChatGPT 5.0, operate near 2 trillion parameters, each contributing to the generation of every word. That is a 2 with 12 zeros.
Despite this enormous capacity, an AI system does not inherently treat the New England Journal of Medicine as more reliable than last week’s “20 Minuten” magazine. It also struggles with recognizing the boundaries of information. This is most obvious in its handling of citations: many are incorrect or internally inconsistent because the AI simply treats author names, journal titles, and dates found during training as if they were true references. Image quality is also an issue.
But most importantly: language models do not know — they guess (based on likelihood). And sometimes they guess wrong. These errors are called hallucinations.
The more we understand the system and the safeguards against hallucinations, the more reliable its output becomes.

How an ideal setup for generating a scientific case report AI model should look

It makes sense for all interested to leverage the potential of AI applications. A robust workflow however should not rely on a single model producing a complete report in one step. Instead, it would involve an ecosystem of agents operating across at least four levels:
1. A curated knowledge source: A deliberately limited and high-quality database containing only peer-reviewed primary evidence. This reduces noise, even if it increases the risk of omission bias.
2. A large language model generating the draft: The LLM writes the initial text based solely on the curated library, ensuring that the scientific foundation is sound.
3. A small language model acting as a verifier: SLMs — smaller, faster, and easily fine-tuned — would carry out four essential quality-control tasks.
– Source citation enforcement: Every citation must link to a concrete text passage. If not, the citation is flagged.
– Answer-only-if-supported prompting: The model responds only when evidence is available. Unsupported claims are suppressed.
– Two-pass reasoning: The model identifies sections where uncertainty is high and flags them for review.
– DOI cross-checking: Every reference is checked for consistency between DOI, title, journal, and year.
– Collectively, these mechanisms fall under retrieval-augmented generation (RAG). They ensure that a model does not generate text solely on its internal training but grounds its output in verifiable evidence. (As a side note: ChatGPT’s latest training data ends in June 2024. Everything beyond that is retrieved online — but always blended with its internal heuristics).
4. A human in the loop: Finally, human review remains essential for clinical judgment, plausibility, and ethical oversight.

Free online tools such as crewAI for example allow for easy setup of such chains of agents. Reading AI-generated tools we should at least assess the prompt used, the model or models involved, and whether the minimal quality safeguards described above were fulfilled (Yes/No). Just as we should critically review the materials and methods in any paper.
In summary, this is a fascinating submission. It shows the strengths of AI-generated scientific writing, but it also demonstrates the typical weaknesses — especially in imaging and citation accuracy. Looking ahead, we will encounter AI-generated text more and more often and we will need to develop the skills to evaluate quality and/or prompt reliable answers from AI.

Author profile

Dr Patrick Vavken
patrick.vavken@adus-klinik.ch

 

Affiliation: alphaclinic Zürich & ADUS Klinik
Specializations: Orthopedics, Sports Medicine, Interventional Pain Medicine
Year of my medical school graduation: 2005
Sports organizations/federations I work with: None
Fields/topics I am fond of: Shoulder and Elbow Surgery, Performance Management, Coaching
Outside of Sports & Exercise Medicine, I also love: Podcasting & raising trouble with my kids
How I use AI: A lot. Really. It was also the main topic of my MBA at ETH.



Comments are closed.