In 2015 Denmark had a population of 883,909 young adults aged 19–30 years. Employment rate in this age-group increased with age from 49.3 to 69.3% (17). The government provides financial support in case of sickness or unemployment. Sickness benefits are available for a limited period or until the individual returns to work. A person must be actively job seeking to receive unemployment benefits and the benefits are reduced after prolonged sickness or unemployment. All Danish students regardless of social status are entitled to a state educational grant from the age of 18 years if they attend a youth education program (high school level) or higher education. The healthcare-system in Denmark is primarily publicly funded and all patients with the diagnoses included are exclusively treated by publicly hospitals during the acute phase.

The Danish Civil Registration System holds information on immigration, municipality at index, death, sex, and date of birth for all Danish citizens (18). The unique personal identification number assigned to all Danish citizens was used for accurate individual-level linkage between registries. The Danish National Registry of Patients (DNP) covers all contacts to Danish public hospitals. DNP holds information on all admissions since 1977 and all outpatient contacts since 1994 (19). Diagnoses have been classified according to the International Classification of Disease revision 10 (ICD-10) since 1994.

Employment status was obtained from the DREAM-registry administered by the Danish National Labor-market Authority (20), which provided weekly information of self-support as well as any public transfer payments including state education grants given to all persons aged 18–65 years with a Danish civil registration number. We used the socio-economic background of the participants' mothers. The mothers were identified in the fertility database (21) and subsequently highest attained educational level (basic school, high school, short education, higher education, or unknown) from the Danish Educational attainment Registry (22) and age-adjusted quintiles of disposable income from registries administered by Statistics Denmark were obtained. Finally, cohort members were classified as immigrants, descendants or native Danes by using information from Statistics Denmark (23).

The ABI patients were identified in DNP and were required to have a first-ever diagnosis of brain tumor, CNS infection, encephalopathy, stroke, SAH, or TBI between 1999 and 2013 (Table 1). Patients only seen at out-patient clinics were not included. In addition, we excluded cases from the period 1994–1998 to ensure that cases during 1999–2013 were first-ever cases.

The general population comparison cohort was identified using the Danish Civil Registration System by risk set sampling, i.e., controls had to be alive and at risk of first-ever ABI at the date of diagnosis (index) of the corresponding ABI patient. For each ABI patient 25 individuals matched by age (month and year), sex, and municipality at index were selected. The matching factors were chosen a priori to address confounding by age, sex, and regional differences in socio-economic position and unemployment rates. Both patients and controls were required to be residing in Denmark 1 year prior to index. Patients and controls were followed from 12 weeks after index until December 31, 2015, death or emigration, whichever came first.

For each week, individuals were defined as self-supportive if no public benefits, except for state education grant, were received or while being on leave. If an individual was self-supportive for at least 75% of 52 consecutive weeks they were considered to have a stable labor-marked attachment (sLMA) (24). This cut-off was chosen to require the individuals to be primarily self-supportive.

For the ABI cohort, time to RTW was defined as the first week of self-supportiveness after baseline (12 weeks after the diagnosis). Time to stable return to work (sRTW) was defined as the time from baseline to the first week followed by of a 52-week period with self-supportiveness in at least 75% of the period.

SLMA for the ABI-cohort was compared with the general population cohort at 1, 2, 5, and 10 years after baseline by conditional logistic regression accounting for the matching and corresponding odds ratios (OR) and 95% confidence intervals (CI) estimated. Two types of adjustment were done (1) according to the matching (simple adjustment) and (2) with further adjustment for sLMA 1-year prior to index, immigration status, mothers' income and mothers' highest achieved educational level (full adjustment). Individuals that had died or emigrated at a given time-point were excluded from that analysis. In addition, we only considered complete cases, however, treating missing values as a separate category, did not change the estimates materially (data not shown). Furthermore, we did two sensitivity analyses: (1) excluding cases identified only through emergency room contacts and (2) using alternative cut-offs for defining stable LMA, i.e., >50 and >90%, respectively.

The cumulative incidence proportions of RTW and sRTW in the ABI cohort were estimated with the Aalen-Johansen estimator with death as competing event (25). Significance level was set to 5% and all analysis were performed in “R version 3.2.3” (26). The study was approved by the Danish Data Protection Agency. Ethical approval or individual consent was not required for this type of study by Danish legislation.

Of the 8,496 patients who met the inclusion criteria, 4,645 (54.7%) had TBI, 1,333 (15.7%) had stroke, 575 (6.8%), had SAH, 829 (9.8%) had encephalopathy, 634 (7.5%) had CNS infection and 480 (5.6%) had brain tumor. Some differences were seen between the ABI cohort and controls (Table 2). Both the mother's income and educational level were lower in the ABI cohort. The proportion of immigrants and descendants were significantly higher in the ABI cohort compared to the general population cohort. During the study 393 (4.6%) died and 398 (4.7%) emigrated within the ABI cohort, whereas 1,151 (0.6%) died and 12,258 (5.9%) emigrated within the comparison cohort.

The cumulative incidences of RTW and stable RTW are shown in Figure 1 and in Supplementary Table 1. Across diagnoses 50–75% had returned to work at baseline, the corresponding numbers for sRTW were 30–60%. Overall, the cumulative incidence of sRTW mainly increased during the first 5 years, while RTW and sRTW reached levels between 88–95 and 55–80% after 10 years for RTW and sRTW, respectively. Brain tumors had the lowest cumulative incidence of stable RTW followed by encephalopathy, and stroke, whereas CNS infection had the highest proportion of RTW and stable RTW followed by SAH and TBI. Persons with sLMA 1-year prior to index had a higher likelihood of stable RTW (data not shown). Excluding emergency room cases gave slightly lower RTW and sRTW however with similar patterns as in the main analysis (data not shown).

Simple and fully adjusted odds ratios for sLMA for the ABI cohort compared to the matched controls are given in Table 3. The fully adjusted odds ratio for 1 year (aOR_1YR), 2 years (aOR_2YR), 5 years (aOR_5yr), and 10 years (aOR_10yr) are shown in Figure 2. The proportion of the control cohort with sLMA was 77.4–85.7% throughout the follow-up period (Table 3). In the ABI cohort 52.9% had sLMA 1 year after baseline, which was significantly lower compared to the control cohort (aOR_1yr: 0.25 [95% CI 0.24–0.27]). Ten years after baseline the proportion of sLMA in the ABI cohort had increased slightly to 53.2% but remained significantly lower compared to the controls (aOR_10YR 0.35 [95% CI: 0.33–0.38]). Despite significant variations in the proportion of patients with sLMA between diagnoses, sLMA remained significantly lower compared to the control cohort for all diagnoses. Most of the diagnoses (i.e., TBI, SAH, stroke, encephalopathy, and CNS infections) showed no improvements in LMA from 2 to 5 years after baseline. However, after 2 years CNS infections had the highest OR for sLMA compared to control cohort (aOR_2YR: 0.53 [95% CI: 0.43–0.66]), followed by SAH (aOR_2YR: 0.47 [95% CI: 0.37–0.58]), TBI (aOR_2YR: 0.36[95% CI: 0.33–0.39]), stroke (aOR_2YR: 0.31 [95% CI: 0.27–0.35]), encephalopathy (aOR_2YR: 0.30 [95% CI: 0.25–0.37]), and brain tumor (aOR_2YR: 0.17 [95% CI: 0.13–0.21]). In the sensitivity analyses of defining sLMA using 50 and 90% cut-offs, respectively, no significant changes were seen (data not shown). Similarly, excluding emergency room contacts yielded similar conclusions (data not shown). Only SAH had lower ORs after excluding emergency room contacts (aOR_5YR: 0.32 [95% CI: 0.23–0.45]) and when defining sLMA using 50% as cut-off (aOR_5YR: 0.39 [95% CI: 0.30–0.50]), respectively.

The main strengths of our study are the nation-wide design, the use of a comparison cohort and the long-term data of public benefits. Furthermore, the unique Danish personal identification number enabled us to link accurately between medical data and employment data. Nonetheless, registry studies are at risk of misclassification and in the current study we relied on the accuracy of discharge-diagnoses and DREAM-data. The discharge data are generally of high quality, however only the validity of stroke and specific CNS infections are known (19). For both diagnoses some misclassification may have occurred in up to 15–20% of the diagnoses and for emergency room contacts even higher (34). However, excluding emergency room contact gave similar patterns although slightly lower LMA. Employment data were retrieved from DREAM, which can elucidate the dynamics of RTW because of the weekly individual registrations (24). Data in the registry have a high positive predictive value (20, 35), however, only sick leaves longer than 2 weeks are registered if the employer claims a reimbursement fee (20). Also, misclassification may have occurred for persons living on own assets or support from family, but this is rare in Denmark. Preexisting comorbidity may also decrease RTW, but to overcome this bias, we adjusted for sLMA prior to index to account for pre-injury differences in LMA between the cohorts. Another limitation to this study is that we did not have information on the severity of ABI. Largely, because there is no uniform measure of injury severity across different types of ABI and such information was not systematically registered for all the included diagnoses in Denmark. In addition, structural factors such as economic recession and structural changes in the Danish health care sector might also have influenced the results. But these factors were not likely to influence the results because the OR for sLMA stratified by year of diagnosis showed no significant differences (data not shown). Furthermore, we might have overestimated the 5- and 10-year estimates for LMA among brain tumor patients, since in contrast to the rest of the ABI cohort, the mortality after brain tumor increases even 5–10 years after index leaving primarily individuals with less severe impairments. A significant proportion of the diagnoses included in this study is associated with low social status (36, 37), which also influences the ability to RTW and subsequent sLMA. Thus, one may overestimate the impact of brain injury if this selection bias is ignored. This problem was addressed by matching and adjusting for social status and consequently we believe this bias is small. Furthermore, socio-economic background of the participants' mothers was used since the socio-economic position of a young adult may be better reflected by the mother's data than by the person's own data in this age-group. However, we cannot rule out some degree of residual confounding, but it is unlikely that it explains the difference in sLMA in ABI patients and their population controls. Finally, the cut-off for sLMA of 75% favors a conservative estimate of the impact of ABI on the ability to assume work.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.