Beyond the break: development and validation of a new classification system for acute traumatic rib fractures

Abstract

Background

Despite the high incidence and significant injury burden of rib fractures, a universally agreed-upon taxonomy for acute traumatic rib fractures (TARF) has not been established. We propose a novel TARF with three components: (1) the rib that is fractured, (2) the anatomic sector of the fracture, and (3) the fracture severity. We hypothesized that each of these components can be interpreted on chest CT by radiologists and trauma surgeons with acceptable interobserver agreement (IOA), providing a consistent method for describing traumatic rib fractures.

Methods

Between August and November 2020, 100 patients with reported rib fractures from our Level 1 trauma database were randomly selected for study inclusion. Two lead surgeons created a key map on specific fracture variable data on chest CT. Four surgeons, one chief surgical resident, and one radiologist completed the interpretation of 50 CT images (Set 1). Each physician was provided feedback and asked to verify and interpret a second set of 50 CT chest scans (Set 2) to determine the final IOA.

Results

The baseline IOA between the lead surgeons on Set 2 was 86.3%. The radiologist scored 83.9% agreement on Set 1 and was excluded from further review. Four surgeons and one chief resident had a mean IOA of 41.9% for Set 1. Subsequently, they had a mean IOA of 72% on Set 2.

Conclusions

Our experience suggests that this taxonomy can be used among surgical and radiology experts to better describe the true rib fracture burden, allowing for highly detailed information in a concise clinically relevant format. Furthermore, this TARF may allow for optimal comparison of patients with rib fractures and potentially provide improved insight into fracture patterns and patient outcomes with or without surgical rib fixation. Further investigations utilizing this TARF are needed.

Level of evidence

IV.

WHAT IS ALREADY KNOWN ON THIS TOPIC

• Although rib fracture classification systems exist, a universally agreed-upon taxonomy for acute traumatic rib fractures (TARF) has yet to be established.

WHAT THIS STUDY ADDS

• A three-component TARF that offers a consistent method for describing traumatic rib fractures with highly detailed information in a concise and clinically relevant format.

HOW THIS STUDY MIGHT AFFECT RESEARCH, PRACTICE OR POLICY

• This study offers a novel, clinically validated TARF that holds promise for universal adoption among trauma surgeons to standardize acute rib fracture classification.

Introduction

Approximately 22,000–45,000 people per year sustain rib fractures in the USA.¹ Despite the high incidence and significant injury burden of rib fractures, there is no universally agreed-upon taxonomy for acute traumatic rib fractures. Publications from the Chest Wall Injury Society (CWIS) were an important step toward framing a taxonomy.^{2 3} In 2020, Edwards et al proposed an initial taxonomy based on the Delphi method, but it has not yet gained widespread acceptance. In 2023, Nguyen et al released a refined consensus proposal, “Improved Chest Wall Trauma Taxonomy”, which was another step forward with support from the American Society of Emergency Radiology along with improved descriptive terminology.³ Despite this, the CWIS working group could not reach a consensus for the descriptive anatomic sector of rib fractures based on chest CT imaging.

Patients with bicortically displaced rib fractures are more likely to develop pneumonia, acute respiratory distress syndrome, and need for tracheostomy.⁴ An offset of 50% or greater axially has traditionally been the standard measure of displacement⁵ and as a basis for CWIS recommendations for rib fixation.⁶ However, evidence suggests that rib fractures are likely to displace further over time.^{7 8} Therefore, we think a new taxonomy should be developed that aims at clinical relevance whereas remaining easily communicable. Thus, we developed a novel taxonomy for acute rib fractures (TARF) with three components: (1) the rib that is fractured (including the laterality (right or left)), (2) the sector of the fracture (costochondral (CC), anterior (A), lateral (L), posterior (P), or transverse process (TP) associated), and (3) the severity of the fracture (single cortical (1), bicortical non-displaced (2), or displaced/distracted/comminuted (3)).

This TARF provides a simplistic, but finely discrete, descriptive model that can be used as a map for rib fractures. This map is useful for clinical communication with the team about the severity of rib fractures and is a tool for preoperative planning for rib stabilization. Additionally, this approach can be used in research to better compare patient groups whose rib fractures, to date, have been characterized simplistically as “rib fractures”, ignoring which rib was broken, where the fracture was located on the rib, or its severity.

Methods

Development of taxonomy

This novel TARF was developed by two lead trauma surgeons and agreed on by eight trauma surgeons and one radiologist from the HonorHealth hospital trauma center network (Phoenix Metropolitan Area, Arizona, USA). The TARF identifies the rib number, the fracture sector on the rib, and the severity of the fracture. It was designed to be as simplistic as possible whereas also conveying the greatest amount of useful information.

Four trauma surgeons, one chief surgical resident, and one radiologist participated in validating this taxonomy. 100 consecutive patients with reported acute rib fractures were randomly selected from the trauma database at an American College of Surgeons Level 2 trauma center. The inclusion criterion was at least one, reportedly acute, rib fracture and an age above 14 years. Eight patients in our database were identified with old rib fractures and excluded. The average age was 60.5 years (range 23–100 years), with 60% male patients, 82% white patients, 8% Hispanic patients and 7% Asian patients. The first set of patients had a total of 380 rib fractures, ranging from 0 to 24 fractures, with 13 patients having bilateral rib fractures. The second set of patients had a total of 342 rib fractures, ranging from 0 to 25 fractures, with 13 patients having bilateral rib fractures. The chest CT (Siemens 64-slice CT scanner) images of each of these patients were reviewed by the two lead surgeons (JFS and HL) who developed the TARF. The CT images were randomly selected from the trauma registry for inclusion. Although some patients in the dataset were later found to be incorrectly classified as having acute fractures, we chose to retain them in the study. This decision was made to better reflect real-world clinical scenarios and to gather valuable data on diagnostic agreement.

The first 50 patient CT studies (Set 1) were initially read independently by two lead surgeons and further discussed at a consensus meeting where a master key map for Set 1 was created. Subsequently, the lead surgeons independently interpreted the second set of 50 CT studies (Set 2) and held a consensus meeting where the key map for Set 2 was created. In addition, the formal radiology reports lacked detailed documentation of fracture sector locations and severity classifications and therefore were not used for confirming the original analysis.

The lead investigator (JFS) held 15 min training sessions individually with four additional trauma surgeons, one chief resident, and one radiologist. Participants underwent the 15 min training session before independently completing the interpretation of Set 1. For each physician, agreements and disagreements were logged against the master key map, and interobserver agreement (IOA) for Set 1 was calculated. After the interpretation of Set 1, participants received individualized feedback on each scan from the initial set to enhance their comprehension of the classification system before proceeding to the second set (Set 2) of CT images. Participants then proceeded to interpret Set 2 to determine the final IOA score. Having already achieved the baseline agreement threshold on Set 1, the radiologist was exempted from Set 2 analysis, given the anticipated consistency in radiologist performance when reevaluating rib fractures on CT. Participant variation was expected due to the subjective nature of image interpretation, particularly when determining fracture types and locations where clear distinctions may not always exist.

Description of taxonomy

Rib and laterality

The first component is the rib number and its laterality, whether right or left. The rib number is counted from the apex downward. Patients with “cervical ribs” are counted as the first rib. Additionally, some patients can have 13 ribs on one side. Although the TARF in this study does not include a designation for a 13th rib, the taxonomy is readily adaptable in cases of anatomic variation.

Fracture sector

The second component is the fracture sector (location). The sector of a rib fracture is the most difficult component to define owing to the complexity of the three-dimensional anatomy and the variety of angles and turns that different ribs can take based on differing body types and rib sectors (location of fracture on the rib). This component has proven particularly challenging for radiologists and chest wall surgeons, making it difficult to agree on how to describe the sectors. Therefore, we defined the fracture location to be optimally described in five sectors: CC, A, L, P, and TP. All five sectors can be used for ribs 1–7 (CC, A, L, P, and TP). Ribs 8–10 use only four sectors (CC, L, P, and TP), and ribs 11–12 use only two sectors (P and TP). The CC (from the sternum to 2 cm past the start of the bony rib) and TP (from the head of the rib to 2 cm past the TP) sectors are easy to describe, as they have discrete anatomic landmarks. The A, L, and P sectors are described based on the muscle group borders or by the local anatomy on the basis of rib number, as noted below (table 1 and figure 1).

Table 1. Summary of rib anatomy terms used in phoenix TARF.

	Anatomic sector
Rib #	CC	A	L	P	TP
1–2	Sternum—start of bony rib	Start of bony rib—angle of clavicle	Angle of clavicle—Turn toward spine	Angle turning toward spine—2 cm from TP	Head of rib—2 cm past TP
3–7	Sternum—2 cm past the start of the bony rib	2 cm past start of bony rib—lateral border of pectoralis	Lateral border of pectoralis—lateral border of latissimus	Lateral border of latissimus—2 cm from TP	Head of rib—2 cm past the TP
8–10	Confluence of costochondral cartilage—2 cm past start of bony rib	N/A	2 cm off the start of bony rib—lateral border of latissimus	lateral border of latissimus—2 cm from TP	Head of rib—2 cm past the TP
11–12	N/A	N/A	N/A	lateral border of latissimus—2 cm from TP	Head of rib—2 cm past the TP

A, anterior; CC, costochondral; L, lateral; N/A, not assessed; P, posterior; TP, transverse process.

Figure 1. Rib anatomy. Sketch of the axial view of ribs 3–7. A, anterior; CC, costochondral; L, lateral; P, posterior; TP, transverse process.

Fracture severity

The fracture severity is described as incomplete (type 1), bicortical non-displaced (type 2), or displaced/distracted/comminuted (type 3). The numbers are used in shorthand notation when documenting (“mapping”) rib fractures (figure 2). Incomplete fractures are defined as fractures with only a single cortex that has a clear fracture (and no clear fracture line that crosses the second cortex) or that have no clear fracture on either cortex but have at least one cortex with a “buckle” fracture. Bicortical non-displaced fractures are fractures with a clear fracture line through both cortices, but at least one cortex, or both, remains aligned. Displaced fractures are fractures where both cortices are completely off alignment, where there is comminution, or where there is a 3 mm or greater distraction. Additionally, a “Z-like” fracture pattern, where there is often not a clear cortical fracture line, is considered displaced (often seen in the higher anterior ribs).

Figure 2. Fracture types. Sketch of different patterns for Type 1, 2, and 3 fractures.

Example use of taxonomy

Verbose description of fractures

Left-sided rib fractures: third rib, single cortical fracture at the CC sector, fourth rib bicortical non-displaced fracture at the anterior sector and a displaced fracture at the lateral sector, and fifth rib displaced fracture at the posterior sector with a single cortical fracture in the TP sector.

Shorthand description of fractures (AKA MAP)

Left 3CC1, 4A2L3, 5P3TP1.

Results

The initial baseline agreement between the two lead surgeons was 36.4% for Set 1. The lead surgeons scored 86.3% agreement on Set 2. This served as the baseline agreement for the rest of the study. The radiologist scored 83.9% agreement for Set 1 (meeting the baseline agreement from Set 2 of the lead surgeons) and was excluded from further review. The four surgeons and one chief resident scored a mean of 41.9% agreement for Set 1 (42.6%, 31.3%, 50.5%, 49.4%, and 35.5%, respectively). On Set 2, they scored a mean of 72% (77.8%, 69.8%, 72.2%, 72%, and 68.3%, respectively).

Discussion

Chest wall trauma is ubiquitous, with an average of five chest rib fracture patients seen every day across our three trauma centers. Although other rib fracture taxonomies have been described, our TARF was developed to address the need for a rib fracture taxonomy that provides detailed information and is simple enough to drive clinical decision-making. As such, we combined all displaced, distracted, and comminuted fractures into a single category. Additionally, we added the CC and TP sectors to provide optimal clinical information. Finally, we added single cortical fracture classification, which we think is an important component of the proposed taxonomy. When single cortical fractures are identified, they can lead to the overtriage of patients. For example, many patients whose CT images revealed “multiple bilateral anterior rib fractures” experienced incomplete rib fractures that were stable. Although these can cause significant pain, they do not necessarily require hospitalization. On the other hand, we think that many single cortical fractures are unrecognized, leading to undertreatment and patient dissatisfaction. Furthermore, missed rib fractures on CT images may result in an under-reporting of the total burden of rib fractures seen in trauma centers, and we think that adding the single cortical fracture classification can help alleviate this underreporting and better approach the true incidence.

In 2019, Bemelman et al published their IOA based on the Muller AO classification system.⁹ Although instructive, this classification system is overly complex in our assessment owing to the description of the “type” and “subtype” of fractures. They achieved 62% agreement in the interpretation of a limited number of images, perhaps due to the complexity of their classification system. The CWIS has made important changes in creating a taxonomy² and showing IOA.¹⁰ However, the CWIS IOA was based on email responses from study participants who were shown a fracture and asked for the observer’s interpretation of severity and sector. In contrast, our study required the observers to read 100 CT images in their entirety and identify the rib fractures, the position of the fractures on each rib, and the severity of the fractures. The lead surgeons developed the taxonomy using basic parameters and aimed to assess their IOA rate. Through this process, they realized that the definition of rib fracture types needed greater clarity. They also observed that many fractures occurred near the borders of the defined sectors, prompting them to be more mindful of these boundaries and to strictly follow the established criteria. Once these issues were addressed and the two lead authors reached a consensus on how to proceed, they reviewed the second set of CT scans using the same approach as the first. This resulted in a significantly improved agreement rate. The study was subsequently extended to include surgeons from all three participating centers, who showed comparable results. Participants received the key map for Set 1 before proceeding to analyze Set 2, allowing them to better understand accurate and inaccurate readings. With this training, the participants achieved 72% agreement on Set 2. Likely, the substantial improvement noted in Set 2 stems from enhanced pattern recognition through this simple training method. This excellent rate of agreement is comparable to that reported in the Bemelman and CWIS studies.^{9 10} As such, we think that this TARF can be effectively employed in clinical practice and provide the best description of the rib fracture burden for both clinical practice and optimizing comparisons in ongoing research. We intend to publish further studies with this TARF to provide a more descriptive update on the epidemiology and outcome of rib fracture patients with and without surgical rib fixation. Limitations of the study include limited generalizability as findings were applicable to three trauma medical centers. The sample is restricted to specific locations, patient demographics, or healthcare settings. The study population is limited as well however the scans were randomly selected. Future research will include larger and multi-regional samples to improve reliability and generalizability. Additionally, analysis of fracture subtypes will be conducted to identify which specific fracture types are most susceptible to interpretive disagreement.¹¹

Conclusions

Our experience suggests that this taxonomy can be used among surgical and radiology experts to better describe the true rib fracture burden, allowing for highly detailed information in a concise clinically relevant format. Furthermore, this TARF may allow for optimal comparison of patients with rib fractures and potentially provide improved insight into fracture patterns and patient outcomes with or without surgical rib fixation. Further investigations utilizing this TARF are needed.

Data availability statement

Data are available upon reasonable request.