Trial design, participants
This was a pragmatic randomised controlled trial (RCT) conducted in close collaboration with NAV. It included all employees aged 18–65 years, sick-listed (fully or partially) in 2015/16 with an International Classification of Primary Care 2nd edition (ICPC-2) diagnosis (indicating being sick listed by a GP), living in Hordaland County in Norway, and whose sickness absence spell had lasted for 26 weeks. Trial candidates were drawn from a central NAV register at 22 weeks of sick leave. Crossing 22 weeks lead to an automatic notification at the desk of a local NAV counsellor who was trained to identify eligible participants based on inclusion/exclusion criteria. The trial excluded pregnant women, individuals with secret address, NAV employees, people with International Classification of Diseases 10th edition (ICD-10) diagnoses (i.e., diagnosed and sick-listed by the specialist health services), ICPC-2 cancer or dementia diagnoses. No changes were made to the methods after trial commencement.
Representatives from NAV had an active role in all phases of setting up this pragmatic RCT. The main reason for this was that the intervention had to be temporarily implemented at NAV in Hordaland County for this trial to be conducted, while participants maintained all their legal rights. Secondly, the intervention had to be designed in such a way that NAV could implement it nationwide if the results had been in favour of the IME consultations and the politicians had decided to make IME part of NAV’s regular follow up of sick listed employees in Norway.
Ethics and consent
The Norwegian Regional Ethical Committee (REK nr. 2015/560) assesses research proposals covered by the Health Research act and declared the project as exempt from review. The reason was that the main aim of the study was to evaluate the effect of IME on RTW. Work participation is not viewed as a health outcome by the Norwegian Health research act. As a result, the study was instead covered by the Personal Data Act. This trial was performed as part of NAV’s follow up of sick listed employees and therefore regulated by The Insurance Law Act 1997-02-28-19-§25–13 § 4 and a specific regulation for testing IME in Norway [12]. This Act authorizes NAV to summon sick listed employees receiving sickness benefits from the national social security insurance scheme to consultations. This enabled the study to be performed without informed consent from the participants. Importantly, as this was a trial and not implemented practises for all sick listed employees in Norway, there were no economical or treatment sanctions for not attending.
The study was first registered in the international register ClinicalsTirals.gov on 14.08.2015, trial number NCT02524392.
All methods were carried out in accordance with relevant guidelines and regulations.
Interventions, randomisation
After identifying eligible participants, the NAV counsellor sent an e-mail with the participant’s ID-number, year of birth and gender to a research technician (not part of the research team conducting the trial) at NORCE for randomisation. Allocation was returned by e-mail and participants were consecutively randomised to either the IME group or the TAU group. We performed simple randomisation stratified per the 13 NAV offices in Hordaland County (1:1) where the randomisation allocation sequence was decided by use of a computer-generated randomisation list. Participants randomised to receive one IME consultation received a letter from the local NAV office with information about the new initiative and time and place for the consultation. The letter stated that the consultation was mandatory, and they were asked to prioritise the consultation. Participants randomised to TAU did not receive any information and were merely followed up as usual by their regular GP and NAV.
The RCT was carried out as part of NAV’s follow up of sick-listed employees. To test this new intervention, nine GPs were employed as IME physicians by NAV in the trial period performing the IMEs. They had offices in different parts of Hordaland County. The intervention included a consultation between the IME physician and the sick-listed employee, in a regular GP clinic, lasting up to 1 h. The focus in the consultation was assessment of previous treatment, return to work-related actions and the employee’s and the IME physician’s evaluation of work ability and RTW possibilities. The IME physician ended the consultation by writing a report (based on a predefined report scheme) to NAV on his/her findings and making a recommendation for further follow-up and sick leave grade. This report was also sent to the treating GP who assessed the need and potential for further actions. The IME physician did not have authority to change the sick leave status and according to Norwegian law the regular GPs remained responsible for the follow-up of all sick listed employees on his/her list. Thus, all judicial rights and obligations for the sick-listed employees also remained unaffected and the only factor that divided the intervention group from the TAU group was the IME summoning and, for those who met with the IME physician, the IME consultation. For simplicity, we label the treatment “IME” rather than “IME + TAU”. For full description of the IME physicians’ training program, time and form of the preparation and consultation, content of the IME report and how the IME physicians evaluated the employee’s RTW barriers and potential, please see our protocol paper [10]. No changes were made to methods after trial commencement.
Outcomes
The primary outcome was number of days on sickness benefits, weighted and not weighted for grading, after inclusion to the trial, based on complete and objective data from NAV’s national social insurance register. There was no loss to follow-up at 7, 9, and 12 months.
Data collected throughout the project was saved in NAV’s secure online database. All directly identifiable personal data (name, social security number, address etc.) was replaced by a unique ID-number before data analyses, when data were kept in NORCE’s secure database. The main outcome, number of days on sick leave after randomisation, was measured in calendar days based on data from GPs’ sick leave certificates registered in NAV (unweighted). However, the most informative way of describing the outcome is days on sick leave after randomisation weighted with the actual sick leave grade, e.g., 1 day with 100% sick leave counts as 1 day, while 1 day with 50% sick leave counts as 0.5 day, etc.
Sample size
Sample size calculations were based on data from Hordaland County showing that 15% of employees who had been sick-listed for 6 months in 2008 were not receiving sickness benefits at 7 months follow-up (Statistics Norway’s events database (FD-Trygd). We estimated that a 3% effect of the IME compared to the TAU group, i.e., 18% vs 15% of employees who had been sick listed for 6 months would not be sick listed at 7 months, could be discovered if 1892 individuals in two groups participate in the study. The calculations were performed in STATA 14 (StataCorp LP, USA) by the command “Power”. Statistical significance level was set to 5% with expected positive effect and power of 80%.
Blinding
The sick-listed employees in the TAU group were not aware of their participation in the trial. The researchers assessing the outcomes (THH and KM) were blinded for intervention assignment.
Statistical methods
For the main effect analysis, we observed mean days of sick leave after randomisation for participants in the two groups. In separate multiple linear regression analyses, we also examined effects of the intervention including adjustment for group differences in age, gender, country of birth, marital status, number of children under the age of 7 or 18, number of sick days 2 years before randomisation, calendar year and month, and geographical site indicators. We estimated the effect for all who were offered an IME consultation, i.e., an analysis adhering to the “intention-to-treat” (ITT) principle, using the ordinary least squares method. We also estimated the effect for the group who took part in the IME consultation, i.e., an effect of “treatment on the treated” (TT). In the latter analysis, randomisation status was utilized as instrument in an instrumental-variable approach. This approach solves the problem of causal inference in an RCT with partial compliance, in this case, where the group who received the IME consultation is self-selected [13].
To test for potential subgroup effects of IME, we conducted analyses of subsamples defined by gender, age, and ICPC-2 diagnosis.