A multicentre, observational, snapshot study was performed from August to October 2020. Twelve hospitals in the Netherlands participated including one academic hospital, 10 teaching hospitals and one non-teaching hospital.

The study was reported according to the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) statement [9]. Local investigators were trained by the coordinating investigator to identify eligible patients and to collect data. Castor Electronic Data Capture software was used to store this data. Ethical approval was obtained from the local ethics committee of all participating hospitals. A waiver of informed consent was granted for this study due to the study design, the expected large study population and associated burden to obtain informed consent for this type of injury, and the minimal risk to human subjects.

1718 consecutive patients of 18 years and older who presented with fractures or dislocations of the metacarpals and phalanges were included in the database. A selected group of patients from this database were included for this study. Inclusion criteria were patient presentation at the emergency department (ED) within 14 days after injury, no additional injury and a single MC fracture treated non-operatively. The study flow chart is presented in Fig. 1.

Flow-diagram of included patients

Patient demographics were retrieved from the electronic patient file. The following patient and injury characteristics were documented: age, sex, smoking, diabetes mellitus, work status, presence of hand comorbidity, mechanism of injury, dominant hand is affected hand, days between injury and presentation, affected digit, presence of soft tissue injuries (including open fractures, significant lacerations, ligament-, tendon- and nail bed injuries), clinically observed angulation observed in the ED (including volar, dorsal, ulnar or radial angulation), rotational deformity observed in the ED, fracture displacement of more than two millimetres on the radiograph (i.e. an abnormal position of the distal fracture fragment in relation to the proximal bone), closed reduction of the fracture, treating physician (emergency physician trauma, orthopaedic, plastic or surgeon) and referral for hand therapy. Patients who were treated by an emergency physician received definitive treatment at the ED and were not assessed by another medical specialist. Complications were recorded until one month after first presentation.

Fractures were classified as intra-articular base, extra-articular base, shaft, neck and intra-articular head fractures. The classification of fractures was in accordance with the standard radiology report. Uncertainties in the radiologist’s report were addressed through consultation with the local surgeon participating in the collaborative study group.

Non-operative treatment was classified as functional treatment or immobilisation. Functional treatment included all patients who received no immobilisation, a buddy tape, a lucerne cast (a metacarpal cast, leaving the wrist joint and the palmar metacarpophalangeal joint (MCPJ) exposed, and dorsally fixating the proximal interphalangeal joints (PIP) and the MCPJs in 70°–90° of flexion to allow motion of the PIPJs), a removable wrist brace, a removable cast or a pressure bandage. Immobilisation included immobilisation with a cast in neutral position, a cast in intrinsic position or an immobilizing rigid digit splint.

Analyses of treatment method and duration of immobilisation were performed separately for different groups. Specifically, intra-articular and extra-articular base fractures were analysed separately for digits 2 and 3, and for digits 4 and 5. For neck fractures, the results were analysed for digits 2–4, and separately for digit 5. Shaft fractures and intra-articular head fractures were analysed collectively across all digits. Consequently, the analyses were performed for eight distinct groups.

The primary outcome was the applied non-operative treatment method, i.e. the number of patients who received immobilisation versus functional treatment, and in case of immobilisation the duration of immobilisation. The applied non-operative treatment method was determined at the ED. If variation was observed, a bivariate analysis was performed of the fracture characteristics between functional treatment and immobilisation. This was done for extra-articular base fractures of digit 4 and 5, shaft fractures, and fifth MC neck fractures.

A sub-analysis was performed to compare the percentage of immobilisation between the 12 participating hospitals. This analysis was feasible only for shaft fractures and fifth MC neck fractures. The small number of fractures in the other fracture categories (less than 24 fractures per fracture category) and the limited variation in treatment precluded similar comparative analyses.

Of patients who were eligible for evaluation of follow-up at the out-patient clinic, a change in treatment from immobilisation to functional treatment, and the duration of immobilisation was assessed. Patients who received follow-up in a different hospital and patients who did not attend to the scheduled follow-up appointment were excluded from these analyses.

The secondary outcome was the Michigan Hand Outcomes Questionnaire (MHQ) score [10]. This is a hand specific outcome instrument that contains six distinct scales: (1) overall hand function, (2) activities of daily living, (3) pain, (4) work performance, (5) aesthetics and (6) patient satisfaction with hand function. The maximum score is 100 points, which is the best possible score. For shaft fractures and fifth MC neck fractures, differences in MHQ scores between functional treatment and immobilisation and differences in MHQ scores between an immobilisation period of less than 21 days and more than 21 days were analysed.

Baseline characteristics and outcome measures were described using descriptive statistics. For continuous data, the mean and standard deviation (normally distributed data) or the median and percentiles (non-normally distributed data) were reported. For categorical data numbers and frequencies were reported. Visual inspection of the distribution was used to assess normality. Associations between duration of immobilisation and hospital and injury characteristics were analysed using multivariable linear regression analysis, in order to quantify and distinguish the roles of patient characteristics from hospital variation. Expert-based variable selection was used to select independent variables for the analysis. Missing values were imputed by assuming the absence of the outcome (four cases had on clinically observed angulation and three on clinically observed rotational deformity). A p value of < 0.05 was taken as a threshold of statistical significance and all tests were two-sided.