Inter- and intra-rater reliability of vertebral fracture classifications in the Swedish fracture register

Reliability of thoracolumbar burst fracture classification in the Swedish Fracture Register

BACKGROUND

The Swedish Fracture Register (SFR) is a national quality register for all types of fractures in Sweden. Spine fractures have been included since 2015 and are classified using a modified AOSpine classification. The aim of this study was to determine the accuracy of the classification of thoracolumbar burst fractures in the SFR.

METHODS

Assessments of medical images were conducted in 277 consecutive patients with a thoracolumbar burst fracture (T10-L3) identified in the SFR. Two independent reviewers classified the fractures according to the AOSpine classification, with a third reviewer resolving disagreement. The combined results of the reviewers were considered the gold standard. The intra- and inter-rater reliability of the reviewers was determined with Cohen's kappa and percent agreement. The SFR classification was compared with the gold standard using positive predictive values (PPV), Cohen's kappa and percent agreement.

RESULTS

The reliability between reviewers was  high (Cohen's kappa 0.70-0.97). The PPV for correctly classifying burst fractures in the SFR was high irrespective of physician experience (76-89%), treatment (82% non-operative, 95% operative) and hospital type (83% county, 95% university). The inter-rater reliability of B-type injuries and the overall SFR classification compared with the gold standard was low (Cohen's kappa 0.16 and 0.17 respectively).

CONCLUSIONS

The SFR demonstrates a high PPV for accurately classifying burst fractures, regardless of physician experience, treatment and hospital type. However, the reliability of B-type injuries and overall classification in the SFR was found to be low. Future studies on burst fractures using SFR data where classification is important should include a review of medical images to verify the diagnosis.

Artificial Intelligence (AI) for Fracture Diagnosis: An Overview of Current Products and Considerations for Clinical Adoption, From the AJR Special Series on AI Applications

Fractures are common injuries that can be difficult to diagnose, with missed fractures accounting for most misdiagnoses in the emergency department. Artificial intelligence (AI) and, specifically, deep learning have shown a strong ability to accurately detect fractures and augment the performance of radiologists in proof-of-concept research settings. Although the number of real-world AI products available for clinical use continues to increase, guidance for practicing radiologists in the adoption of this new technology is limited. This review describes how AI and deep learning algorithms can help radiologists to better diagnose fractures. The article also provides an overview of commercially available U.S. FDA-cleared AI tools for fracture detection as well as considerations for the clinical adoption of these tools by radiology practices.

The Swedish Fracture Register - ten years of experience and 600,000 fractures collected in a National Quality Register

BACKGROUND

Before the creation of the Swedish Fracture Register (SFR), there was no national quality register that prospectively collects data regarding all types of fractures regardless of treatment in an emergency setting. Observational data on fractures registered in a sustainable way may provide invaluable tools for quality improvements in health care and research.

DESCRIPTION

Ten years after its implementation, the Swedish Fracture Register has 100% coverage among orthopaedic and trauma departments in Sweden. The completeness of registrations reached in 2020 69-96% for hip fractures at the different departments, with the majority reporting a completeness above 85%. The Swedish Fracture Register is a fully web-based national quality register created and run by orthopaedic professionals, with financial support from public healthcare providers and the government. All users have full access to both the registration platform and all aggregated statistics in real time. The web-based platform was created for use in health quality registers and it has easily gained acceptance among users. The register has gradually developed by the addition of more fracture types and skeletal parts. Research activity is high and 31 scientific publications have been published since 2016. The strategy from the start was to publish validation data and basic epidemiological data. However, over the past few years, publications on outcomes, such as re-operations and mortality, have been published and four register-based, randomised, controlled trials are ongoing.

CONCLUSION

It is possible to create a fracture register, to gain professional acceptance and to collect fracture data in a sustainable way on a national level if the platform is easy to use. Such a platform can also be used as a randomisation platform for prospective studies.

30-day and 1-year mortality after skeletal fractures: a register study of 295,713 fractures at different locations

Background and purpose - Few studies have reported the mortality rate after skeletal fractures involving different locations, within the same population. We analyzed the 30-day and 1-year mortality rates following different fractures.Patients and methods - We included 295,713 fractures encountered in patients 16-108 years of age, registered in the Swedish Fracture Register (SFR) from 2012 to 2018. Mortality rates were obtained by linkage of the SFR to the Swedish Tax Agency population register. The standardized mortality ratios (SMR) at 30 days and 1 year were calculated for fractures in any location and for each of 27 fracture locations, using age- and sex-life tables from Statistics Sweden (www.scb.se).Results - The overall SMR at 30 days was 6.8 (95% CI 6.7-7.0) and at 1 year 2.2 (CI 2.2-2.2). The SMR was > 2 for 19/27 and 13/27 of the fracture locations at 30 days and 1 year, respectively. Humerus, femur, and tibial diaphysis fractures were all associated with high SMR, at both 30 days and 1 year.Interpretation - Patients sustaining a fracture had approximately a 7-fold increased mortality at 30 days and over 2-fold increased mortality at 1 year as compared with what would be expected in the general population. High mortality rates were seen for patients with axial skeletal and proximal extremity fractures, indicating frailty in these patient groups.

Is the AO spine thoracolumbar injury classification system reliable and practical? a systematic review

Controversy surrounding the classification of thoracolumbar injuries has given rise to various classification systems over the years, including the most recent AOSpine Thoracolumbar Injury Classification System (ATLICS). This systematic review aims to provide an up-to-date evaluation of the literature, including assessment of a further three studies not analysed in previous reviews. In doing so, this is the first systematic review to include the reliability among non-spine subspecialty professionals and to document the wide variety between reliability across studies, particularly with regard to sub-type classification. Relevant studies were found via a systematic search of PubMed, EBESCO, Cochrane and Web of Science. Data extraction and quality assessment were conducted in line with Cochrane Collaboration guidelines. Twelve articles assessing the reliability of ATLICS were included in this review. The overall inter-observer reliability varied from fair to substantial, but the three additional studies in this review, compared to previous reviews, presented on average only fair reliability. The greatest variation of results was seen in A1 and B3 subtypes. Least reliably classified on average was A4 subtype. This systematic review concludes that ATLICS is reliable for the majority of injuries, but the variability within subtypes suggests the need for further research in assessing the needs of users in order to increase familiarity with ATLICS or perhaps the necessity to include more subtype-specific criteria into the system. Further research is also recommended on the reliability of modifiers, neurological classification and the application of ATLICS in a paediatric context.

Treatment and re-operation rates in one thousand and three hundred tibial fractures from the Swedish Fracture Register

PURPOSE

Approximately, 50 persons per 100,000 per year sustain a tibial fracture. There is, however, a lack of large cohort studies that describe the treatment and re-operation frequencies of tibial fractures. The aim of this study was to describe the treatment and re-operation rates of tibial fractures in all segments of the tibia.

METHODS

Data related to all patients aged 16 and above treated for tibial fractures (ICD-10 S82.10-31) at Sahlgrenska University Hospital in 2011-2015 were extracted from the Swedish Fracture Register. To make sure all re-operations were included in the study, the operation planning system was checked for all patients included in the study.

RESULTS

The study comprised 1371 tibial fractures - 712 proximal, 417 diaphyseal and 242 distal tibial fractures. Among the proximal and distal tibial fractures, plate fixation was the most commonly used surgical method, whereas among tibial shaft fractures, an intramedullary nail was the most commonly used surgical method. Almost 30% (29.8%) of all surgically treated tibial fractures underwent re-operation. Among proximal tibial fractures, 24.0% underwent re-operation; tibial shaft fractures 37.0% and distal tibial fractures 26.8%. Re-operations due to infection were more or less equally common in all segments (3.9-5.4%).

CONCLUSION

This study describes the treatment and re-operation rates after tibial fractures in a cohort of 1371 tibial fractures at Sahlgrenska University Hospital during a period of 5 years. The study shows an overall re-operation rate of 29.8% for fractures in all segments of the tibia.

Fracture incidence in adults in relation to age and gender: A study of 27,169 fractures in the Swedish Fracture Register in a well-defined catchment area

Studies on fracture incidence have mostly been based on retrospectively registered data from local hospital databases. The Swedish Fracture Register (SFR) is a national quality register collecting data prospectively on fractures, at the time of care-seeking. In the present study the incidence of all different fractures, regardless of location, in adults’ ≥ 16 years treated at the only care provider for patients with fractures within a catchment area of approximately 550,000 inhabitants, during 2015‒2018 are described. Age, gender, and fracture location (according to AO/OTA classification) was used for the analyses and presentation of fracture incidences. During the 4-year study period, 23,917 individuals sustained 27,169 fractures. The mean age at fracture was 57.9 years (range 16‒105 years) and 64.5% of the fractures occurred in women. The five most common fractures accounted for more than 50% of all fractures: distal radius, proximal femur, ankle, proximal humerus, and metacarpal fractures. Seven fracture incidence distribution groups were created based on age- and gender-specific incidence curves, providing visual and easily accessible information on fracture distribution. This paper reports on incidence of all fracture locations based on prospectively collected data in a quality register. The knowledge on fracture incidence related to age and gender may be of importance for the planning of orthopaedic care, involving both in- and out-patients as well as allocating surgical resources. Further, this might be useful for organizing preventive measures, especially in countries with similar socioeconomic structure and fracture burden.

Development of the Italian fractures registry (RIFra): A call for action to improve quality and safety

In recent years, there has been an increasing interest in the development of arthroplasty registries, therefore, in our country, the Italian Arthroplasty Registry (RIAP), was issued by the National Law No. 221/2012. In the last decade, however, some European countries -namely Sweden, Denmark, Norway, and Germany (in development)- have introduced another nationwide orthopaedic registry than arthroplasty registers: the fracture registry. The development of this new tool aims to improve quality and safety in fracture management, thus trying to provide a better postoperative quality of life in trauma patients. Based on these findings, the AO-Trauma Italy Council encouraged the development of a national fracture registry in Italy. The present study aims to (1) provide an overview of the fracture registries in Europe and (2) to develop, for the first time, a pilot Italian Fracture Registry (RIFra). Thirteen AO-Trauma Italy members, chairmen of Level-I orthopaedic and trauma centres, diffused throughout Italy, were involved in the RIFra project. The RIFra form, developed between November 2019 and March 2020, consists of 5 main sections, namely: epidemiologic data, previous surgical procedure (if any), patient and fracture features, surgical procedure, surgical implant details. This study constitutes the first step to start, in future years, the bureaucratic procedure leading to the final establishment of a RIAP-like fracture registry in Italy.

Interobserver reliability of the Gehweiler classification and treatment strategies of isolated atlas fractures: an internet-based multicenter survey among spine surgeons

Purpose

Atlas (C1) fractures are commonly rated according to the Gehweiler classification, but literature on its reliability is scarce. In addition, evaluation of fracture stability and choosing the most appropriate treatment regime for C1-injuries are challenging. This study aimed to investigate the interobserver reliability of the Gehweiler classification and to identify whether evaluation of fracture stability as well as the treatment of C1-fractures are consistent among spine surgeons.

Methods

Computed tomography images of 34 C1-fractures and case-specific information were presented to six experienced spine surgeons. C1-fractures were graded according to the Gehweiler classification, and the suggested treatment regime was recorded in a questionnaire. For data analyses, SPSS was used, and interobserver reliability was calculated using Fleiss’ kappa (κ) statistics.

Results

We observed a moderate reliability for the Gehweiler classification (κ = 0.50), the evaluation of fracture stability (κ = 0.50), and whether a surgical or non-surgical therapy was indicated (κ = 0.53). Type 1, 2, 3a, and 5 fractures were rated stable and treated non-surgically. Type 3b fractures were rated unstable in 86.7% of cases and treated by surgery in 90% of cases. Atlas osteosynthesis was most frequently recommended (65.4%). Overall, 25.8% of type 4 fractures were rated unstable, and surgery was favoured in 25.8%.

Conclusion

We found a moderate reliability for the Gehweiler classification and for the evaluation of fracture stability. In particular, diverging treatment strategies for type 3b fractures emphasise the necessity of further clinical and biomechanical investigations to determine the optimal treatment of unstable C1-fractures.