The number of primary surgery in the adult population (age 18 years or older) will be collected from patient information systems using the NOMESCO procedure code NHG80 regardless of diagnosis. These patients will be further screened for diabetes and arthritic diseases (the International Classification of Diseases 10th Revision (ICD-10) diagnostic codes E10.\*\*-E14.\*\* and M05.\*\*-M14.\*\*). As reoperations will be considered surgery under the NOMESCO procedure codes NH\*\*\*, QDB\*\*\*, QDG\*\*\*, ZZH\*\*\*. Comorbidities were collected from the previous patient health records by diagnosis codes for diabetes type 1 (E10.\*\*), other diabetes (E11.\*\*-14.\*\*), seropositive rheumatoid arthritis (M05.9), seronegative rheumatoid arthritis (M06.0), psoriasis (M07.\*\*), gout (M10.\*\*), other arthritis (M09.\*\*, M11-M14).
The data will be divided and analyzed in two separate phases. First we will gather all the information that we can reliably utilize from the register. The data will include the primary operation, all reoperations, experience level of the surgeon, patients' age, sex, and whether the patient has been diagnosed with other comorbidities stated above.
In the second phase, we will collect matched patients for the patients who underwent reoperations using propensity score matching with ratio of 1:3 with nearest neighbour method. The variables used for matching will be age, sex and all other comorbidities (diabetes type 1, other diabetes, seropositive rheumatoid arthritis, seronegative rheumatoid arthritis, psoriasis, gout, other arthritis) We will further screen the medical records of the reoperated patients as well the matched controls. From pre- and post-operative x-rays we will determine the degree of arthrosis and hallux valgus and intermetatarsal angles. Operation reports will be screened for surgical technique (i.e. method of cartilage removal, fixation and use of bone transport). These data will be used to analyze the association of perioperative factors and reoperations. These models will not be including variables that were used in matching.
During the both phases, we will use logistic regression to analyze association between assumed risk factors and reoperations by using the data from the register (first phase) or from the patients' health records and x-rays (second phase). We will evaluate causality with directed acyclic graphs (DAG), and develop logistic models separately for each of the assumed risk factors. The variables used in the DAGs will be selected based on previous knowledge and hypothesized causal relationships. Results from both analyses will interpreted with adjusted odds ratios (OR) with 95 % confidence intervals (CI). Model fit will be assessed by McFadden R2 -index, multicollinearity by evaluating Variance Inflation Factor (VIF) and heteroscedasticity by inspecting the fitted values vs. residual plots. Statistical analyses will be performed with R version 4.0.5.
