Short summary

The Danish Fracture Database contains vast amounts of data on 75,000 patients who have undergone fracture-related surgery. These data represents a unique source of knowledge on fracture-related surgery, however, no sufficient examination of validity concerning the Danish Fracture Database have been made. As such, data can not be considered reliable until validation of the database has been carried out. 

Rationale

In Denmark, the lifetime prevalence of fractures is 38% and it is estimated that around 20,000 fracture-related surgical procedures are carried out annually. Being an area of such magnitude, systematic quality-assessment of fracture related surgery is paramount in order to identifying risk factors for failures and prognosis, and monitoring and upholding high quality treatment. For this purpose, The Danish Fracture Database (DFDB) was established in 2012 and was nationally implemented in 2016. It contains substantial amounts of routinely collected data on patients who have undergone fracture-related surgery (not including spinal- , and cranial fractures). The DFDB has the potential to act as an important part of the research infrastructure for this field of surgery. But one of the concerns with the usage of clinical databases for research and systematic quality-assessment is the risk of poor data validity and completeness. This is also the case for the DFDB. 

Completeness of a database is measured as sensitivity and specificity. In Table 1, a schematic representation of possible outcomes illustrates these concepts. Sensitivity can be seen as the percentage of correctly registered patients out of all the patients that should have been registered. As a measurement for all the patients that should have been registered it is common practice to use an independent register, which has already been proven to have a high completeness and validity - the gold standard. Validity of a database is measured as the positive predictive value (PPV) and negative predictive value (NPV) of variables in the database. In the context of the DFDB, validity is measured for individual variables. As such, the PPV and NPV would be calculated for each individual variable registered in the DFDB. The PPV and NPV can be seen as the likelihood that a patient has or has not the registered diagnosis or received treatment that is registered in the database 

When assessing completeness and validity of databases, a suitable gold standard is required. For a number of validation studies on Danish registers, the gold standard has been the Danish National Patient Registry (DNPR). The DNPR is an independent database founded in 1977 and has been found to record 99.4% of all hospital discharges in Denmark. But the literature is sparse on validation studies regarding the registration of orthopedic surgery diagnostic and procedure codes in the DNPR. For orthopedic diagnosis codes, the Danish Health and Medicines Authority published an evaluation of the DNPR in 1993. According to this publication, the highest PPV associated with diagnosis codes for orthopedic surgery was 83%. In 2006, Lass et al. found that the positive predictive value for orthopedic procedure codes in the DNPR was 63.2% (59.1-67.1). A review of the DNPR, by Schmidt et al. concluded that the usage of the DNPR as gold standard for validation studies is not without problems and should ideally be preceded by a validation study of the DNPR for the relevant field of research. 

It is clear that the DFDB has the potential to be an important part of the research infrastructure in the field of fracture-related surgery assuming the data within is valid. The database was assessed in regards to completeness and validity following its implementation in 2012 by Gromov et al. in which 322 patients from two orthopedic surgery departments in Denmark was included. The study concluded that the completeness of data was 83% and the validity for individually assessed variables was between 90% and 100%. The DFDB has since then expanded to receive data from 21 orthopedic departments across the nation. It now contains data on more than 75,000 fracture-related surgery procedures. Therefore, new studies regarding completeness and validity of the DFDB is strongly warranted and of the greatest interest for the utilization of the DFDB in research and systematic qualityassessment. To do this a reliable gold standard for comparison of data with the DFDB is required.

The aim of this project is to assess the completeness and validity of the DFDB as well as the validity of the DNPR. 

Description of the cohort

The cohort will consist of all patients who have undergone fracture-related surgery in 2016 on one of the 21 Danish orthopedic surgery departments who report to the DFDB. A 2-year follow up period will include patients who underwent reoperation following fracture-related surgery performed in 2016. 

Data and biological material

Data included will consist of register data, radiology and data from patient medical records. 

Collaborating researchers and departments

Department of Orthopedic Surgery and Traumatology, Kolding Hospital - part of Hospital Lillebaelt

• Associate Professor Bjarke Viberg, Head of Research, Ph.D., MD. 
• Per Hviid Gundtoft, Ph.D., MD

Department of Orthopedic Surgery and Traumatology, Hvidovre Hospital

• Peter Toft Tengberg, Ph.D., MD.