Abstract Background We aimed to estimate the effect of the voluntary Norwegian Agreement on a More Inclusive Working Life (IA Agreement) on use of sickness absence (SA) and pregnancy benefits among pregnant women. Methods Pregnant women (n = 112,486) with a birth during 1.12.2003–31.12.2010 were followed from 6 to 37 gestational weeks in a continuous time multistate model with the following states: work, full SA, graded SA, pregnancy benefits, maternity leave, and other. Women working in IA companies were compared to those in non-IA companies regarding incidence and duration of SA and pregnancy benefits. Differences between groups with respect to calendar year, age, civil status, education, industry, and number of employees in the company were adjusted for using inverse probability of treatment weighting. Absolute differences in probabilities over time, expected length of stay (ELOS) in each state and differences in ELOS between IA and non-IA were calculated. 95% confidence intervals (CI) were generated using bootstrapping (1,000 repetitions). Results Adjusted analyses suggest that women working in IA companies were more likely to be in full SA in the first and last trimesters, but less likely between 14 and 28 weeks, than those in non-IA companies. The probability of being in work mirrored this, with women in IA companies tending to spend half a day more in work (ELOS difference 0.55, 95% CI -1.79, 3.02). Differences were not statistically significant. The use of graded SA was slightly higher (ELOS difference 0.46, 95% CI -0.87, 1.72) and the use of pregnancy benefits slightly lower (ELOS difference − 0.43, 95% CI -1.32, 0.42) among those in IA companies compared to non-IA companies. Conclusions Women in IA companies tended to spend more time in work and graded SA, but less time on pregnancy benefits. Differences in full SA varied during pregnancy and were most positive mid-pregnancy. This indicates that IA measures could be more effective for conditions experienced at this point. However, effects were small and not statistically significant, which may indicate the IA Agreement has not focused much on pregnant women. Key terms IA Agreement, MBRN, MoBa, multistate models, pregnancy, pregnancy benefits, sickness absence, work participation.

OBJECTIVES:This study aimed to quantify the duration of work participation and reasons for working years lost, according to gender and educational attainment, among a Norwegian population. METHODS:Register data on labor market attachment between 2000-2015 were obtained from Statistics Norway. We included five cohorts: individuals turning 20 (N=323 333), 30 (N=386 006), 40 (N=388 962), 50 (N=358 745), and 60 years (N=284 425) between 1 January 2000 and 31 December 2005. Individuals were followed for ten years. Data completeness allowed calculation of the average time spent in work and years lost to health-related absences and non-employment states per cohort. Changes in state probabilities over time were also depicted. Mean differences between genders and educational levels, and corresponding 95% confidence intervals were based on 1000 bootstrap samples. RESULTS:Both genders spent most time in work; however, per cohort, women worked approximately one year less than men. As cohorts aged, main reasons for working years lost changed from education and economic inactivity to sickness absence and disability pensioning; this trend was stronger for women than men. Individuals with a low education spent fewer years in work and more years in sickness absence and disability pensioning than highly educated peers. This difference tended to be larger for women and older cohorts. CONCLUSIONS:Per cohort, women participated one year less in work than men and, depending on age, spent more time in education, economic inactivity, sickness absence, and disability pensioning. Stronger educational gradients were seen for work and health-related absences for older cohorts and women.

Background The role of laparoscopy in the treatment of intrahepatic cholangiocarcinoma (ICC) remains unclear. This multicenter study examined the outcomes of laparoscopic liver resection for ICC. Methods Patients with ICC who had undergone laparoscopic or open liver resection between 2012 and 2019 at four European expert centers were included in the study. Laparoscopic and open approaches were compared in terms of surgical and oncological outcomes. Propensity score matching was used for minimizing treatment selection bias and adjusting for confounders (age, ASA grade, tumor size, location, number of tumors and underlying liver disease). Results Of 136 patients, 50 (36.7%) underwent laparoscopic resection, whereas 86 (63.3%) had open surgery. Median tumor size was larger (73.6 vs 55.1 mm, p = 0.01) and the incidence of bi-lobar tumors was higher (36.6 vs 6%, p < 0.01) in patients undergoing open surgery. After propensity score matching baseline characteristics were comparable although open surgery was associated with a larger fraction of major liver resections (74 vs 38%, p < 0.01), lymphadenectomy (60 vs 20%, p < 0.01) and longer operative time (294 vs 209 min, p < 0.01). Tumor characteristics were similar. Laparoscopic resection resulted in less complications (30 vs 52%, p = 0.025), fewer reoperations (4 vs 16%, p = 0.046) and shorter hospital stay (5 vs 8 days, p < 0.01). No differences were found in terms of recurrence, recurrence-free and overall survival. Conclusion Laparoscopic resection seems to be associated with improved short-term and with similar long-term outcomes compared with open surgery in patients with ICC. However, possible selection criteria for laparoscopic surgery are yet to be defined.

IMPORTANCE:. The Sequential Organ Failure Assessment (SOFA) was chosen in the definition of sepsis due to superior validity in predicting mortality. However, few studies have assessed the contributions of acute versus chronic organ failures to SOFA for mortality prediction. OBJECTIVES:. The main objective in this study was to assess the relative importance of chronic and acute organ failures in mortality prediction in patients with suspected sepsis at hospital admission. We also evaluated how the presence of infection influenced the ability of SOFA to predict 30-day mortality. DESIGN, SETTING, AND PARTICIPANTS:. Single-center prospective cohort study including 1,313 adult patients with suspected sepsis in rapid response teams in the emergency department. MAIN OUTCOMES AND MEASURES:. The main outcome was 30-day mortality. We measured the maximum total SOFA score during admission (SOFATotal), whereas preexisting chronic organ failure SOFA (SOFAChronic) score was assessed by chart review, allowing calculation of the corresponding acute SOFA (SOFAAcute) score. Likelihood of infection was determined post hoc as “No infection” or “Infection.” RESULTS:. SOFAAcute and SOFAChronic were both associated with 30-day mortality, adjusted for age and sex (adjusted odds ratios [AORs], 1.3; 95% CI, 1.3–14 and 1.3; 1.2–1.7), respectively. Presence of infection was associated with lower 30-day mortality (AOR, 0.4; 95% CI, 0.2–0.6), even when corrected for SOFA. In “No infection” patients, SOFAAcute was not associated with mortality (AOR, 1.1; 95% CI, 1.0–1.2), and in this subgroup, neither SOFAAcute greater than or equal to 2 (relative risk [RR], 1.1; 95% CI, 0.6–1.8) nor SOFATotal greater than or equal to 2 (RR, 3.6; 95% CI, 0.9–14.1) was associated with higher mortality. CONCLUSIONS AND RELEVANCE:. Chronic and acute organ failures were equally associated with 30-day mortality in suspected sepsis. A substantial part of the total SOFA score was due to chronic organ failure, calling for caution when using total SOFA in defining sepsis and as an outcome in intervention studies. SOFA’s mortality prediction ability was highly dependent on actual presence of infection.

Objectives To reduce sickness absence (SA) and increase work participation, the tripartite Agreement for a More Inclusive Working Life (IA) was established in Norway in 2001. IA companies have had access to several measures to prevent and reduce SA. Our aim in this paper was to estimate the average effect of having access to IA at the time of entering a first SA on later return-to-work (RTW) and on time spent in other work-related states. A secondary objective was to study how effects varied between women and men, and individuals with SA due to either musculoskeletal or psychological diagnoses.Design Population-based observational multistate longitudinal cohort study.Setting Individual characteristics and detailed longitudinal records of SA, work and education between 1997-2011 were obtained from population-wide registries.Participants Each individual born in Norway 1967–1976 who entered full-time SA during 2004–2011, with limited earlier SA, was included (n=187 930).Primary and secondary outcome measures Individual multistate histories containing dated periods of work, graded SA, full-time SA, non-employment and education.Methods Data were analysed in a multistate model with 500 days of follow-up. The effect of IA was assessed by estimating differences in state probabilities over time, adjusted for confounders, using inverse probability weighting.Results IA increased the probability of work after SA, with the largest difference between groups after 29 days (3.4 percentage points higher (95% CI 2.5 to 4.3)). Differences in 1-year expected length of stay were 8.4 additional days (4.9 to 11.9) in work, 7.6 (4.8 to 10.3) fewer days in full-time SA and 1.6 (-0.2 to 3.4) fewer days in non-employment. Similar trends were found within subgroups by sex, musculoskeletal and psychological diagnoses. The robustness of the findings was studied in sensitivity analyses.Conclusion Measures to prevent and reduce SA, as given through IA, were found to improve individuals’ RTW after entering SA.

Aims: The study objectives were to provide a quantitative description of work participation among young adults, and to outline the relations between work participation and social, educational and health-related characteristics throughout the life-course. Methods: We collected data in several national registries for all 318,705 individuals born in Norway 1967–1971 who were national residents on 1 January 1993. The criterion for work was annual occupational income above the boundary which identifies the core workforce. We analysed associations between social, educational and health-related characteristics, and the number of years at work and the risk of never working during 19 years of follow-up (1993–2011; age 22–44 years). Results: The overall work participation was high, with a median of 14 years and a 0.074 risk of never working. Women worked fewer years than men (medians 11 v. 16 years) and had higher risk of never working (0.103 v. 0.047). Combined educational and health problems before 1993 had a strong influence on subsequent work participation. The educational gradient in risks of never working was considerably stronger for women than for men. Diagnostic groups of mental disorders had high risks of never working, ranging from affective (risk 0.150) and stress-related disorders (risk 0.163) to intellectual disability (risk 0.933). Conclusions: The complex problems characterising individuals with low work participation suggest that preventive measures should take sex into account and be targeted at social, educational and mental issues in early life, and focusing on identified vulnerable groups.

Multi-state models are increasingly being used to model complex epidemiological and clinical outcomes over time. It is common to assume that the models are Markov, but the assumption can often be unrealistic. The Markov assumption is seldomly checked and violations can lead to biased estimation of many parameters of interest. This is a well known problem for the standard Aalen-Johansen estimator of transition probabilities and several alternative estimators, not relying on the Markov assumption, have been suggested. A particularly simple approach known as landmarking have resulted in the Landmark-Aalen-Johansen estimator. Since landmarking is a stratification method a disadvantage of landmarking is data reduction, leading to a loss of power. This is problematic for “less traveled” transitions, and undesirable when such transitions indeed exhibit Markov behaviour. Introducing the concept of partially non-Markov multi-state models, we suggest a hybrid landmark Aalen-Johansen estimator for transition probabilities. We also show how non-Markov transitions can be identified using a testing procedure. The proposed estimator is a compromise between regular Aalen-Johansen and landmark estimation, using transition specific landmarking, and can drastically improve statistical power. We show that the proposed estimator is consistent, but that the traditional variance estimator can underestimate the variance of both the hybrid and landmark estimator. Bootstrapping is therefore recommended. The methods are compared in a simulation study and in a real data application using registry data to model individual transitions for a birth cohort of 184 951 Norwegian men between states of sick leave, disability, education, work and unemployment.

In non-Markov multi-state models, the traditional Aalen–Johansen (AJ) estimator for state transition probabilities is generally not valid. An alternative, suggested by Putter and Spitioni, is to analyse a subsample of the full data, consisting of the individuals present in a specific state at a given landmark time-point. The AJ estimator of occupation probabilities is then applied to the landmark subsample. Exploiting the result by Datta and Satten, that the AJ estimator is consistent for state occupation probabilities even in non-Markov models given that censoring is independent of state occupancy and times of transition between states, the landmark Aalen–Johansen (LMAJ) estimator provides consistent estimates of transition probabilities. So far, this approach has only been studied for non-parametric estimation without covariates. In this paper, we show how semi-parametric regression models and inverse probability weights can be used in combination with the LMAJ estimator to perform covariate adjusted analyses. The methods are illustrated by a simulation study and an application to population-wide registry data on work, education and health-related absence in Norway. Results using the traditional AJ estimator and the LMAJ estimator are compared, and show large differences in estimated transition probabilities for highly non-Markov multi-state models.

When comparing time varying treatments in a non-randomized setting, one must often correct for time-dependent confounders that influence treatment choice over time and that are themselves influenced by treatment. We present a new two-step procedure, based on additive hazard regression and linear increments models, for handling such confounding when estimating average treatment effects on the treated. The approach can also be used for mediation analysis. The method is applied to data from the Swiss HIV Cohort Study, estimating the effect of antiretroviral treatment on time to acquired immune deficiency syndrome or death. Compared with other methods for estimating the average treatment effects on the treated the method proposed is easy to implement by using available software packages in R.

Context and Objective Vitamin D status may affect cardiovascular disease (CVD) development and survival. We studied the relationship between concentrations of the circulating biomarker 25-hydroxyvitamin D (25OHD) and all-cause and cardiovascular mortality risk. Design, Setting, Participants, and Main Outcome Measures 25OHD, the sum of 25-hydroxyvitamin D3 and 25-hydroxyvitamin D2, was analyzed in plasma samples from 4114 white patients suspected of having stable angina pectoris and was adjusted for seasonal variation. Hazard ratios (HRs) for all-cause and cardiovascular mortality were estimated by using multivariable Cox models with 25OHD as the main exposure variable, with adjustment for study site, age, sex, smoking, body mass index, estimated glomerular filtration rate, and systolic blood pressure. Results A total of 895 (21.8%) deaths, including 407 (9.9%) from CVD causes, occurred during a mean ± standard deviation follow-up of 11.9 ± 3.0 years. Compared with the first 25OHD quartile, HRs in the second, third, and fourth quartiles were 0.64 [95% confidence interval (CI), 0.54 to 0.77], 0.56 (95% CI, 0.46 to 0.67), and 0.56 (95% CI, 0.46 to 0.67) for all-cause mortality and 0.70 (95% CI, 0.53 to 0.91), 0.60 (95% CI, 0.45 to 0.79), and 0.57 (95% CI, 0.43 to 0.75) for cardiovascular mortality, respectively. Threshold analysis demonstrated increased all-cause and CVD mortality in patients with 25OHD concentrations below ∼42.5 nmol/L. Moreover, analysis suggested increased all-cause mortality at concentrations >100 nmol/L. Conclusion Plasma 25OHD concentrations were inversely associated with cardiovascular mortality and nonlinearly (U-shaped) associated with all-cause mortality.

When comparing time varying treatments in a non-randomized setting, one must often correct for time-dependent confounders that influence treatment choice over time and that are themselves influenced by treatment. We present a new two-step procedure, based on additive hazard regression and linear increments models, for handling such confounding when estimating average treatment effects on the treated. The approach can also be used for mediation analysis. The method is applied to data from the Swiss HIV Cohort Study, estimating the effect of antiretroviral treatment on time to acquired immune deficiency syndrome or death. Compared with other methods for estimating the average treatment effects on the treated the method proposed is easy to implement by using available software packages in R.

Background Completing upper secondary education is associated with higher work participation and less health-related absence from work. Although these outcomes are closely interrelated, most studies focus on single outcomes, using cross-sectional designs or short follow-up periods. As such, there is limited knowledge of the long-term outcomes, and how paths for completers and non-completers unfold over time. In this paper, we use multi-state models for time-to-event data to assess the long-term effects of completing upper secondary education on employment, tertiary education, sick leave, and disability pension over twelve and a half years for young men. Methods Baseline covariates and twelve and a half years of follow-up data on employment, tertiary education, sick leave and disability pension were obtained from national registries for all males born in Norway between 1971 and 1976 (n =184951). The effects of completing upper secondary education (by age 23) were analysed in a multi-state framework, adjusting for both individual and family level confounders. All analyses were done separately for general studies and vocational tracks. Results Completers do better on a range of outcomes compared to non-completers, for both fields of upper secondary education, but effects of completion change over time. The largest changes are for tertiary education and work, with the probability of work increasing reciprocally to the probability of education. Vocational students are quicker to transfer to the labour market, but tend to have more unemployment, sick leave and disability, and the absolute effects of completion on these outcomes are largest for vocational tracks. However, the relative effects of completion are larger for general studies. Conclusion Completing upper secondary education increases long-term work participation and lowers health-related absence for young men, but effects diminish over time. Studies that have used shorter follow-up periods could be overstating the negative effects of dropout on labour market participation. Multi-state models are well suited to analyse data on work, education and health-related absence, and can be useful in understanding the dynamic aspects of these outcomes.

Objectives High school dropout has unfavourable consequences. We examine the association between dropout and mortality and estimate the mediating role of an unfavourable work trajectory. Methods Subjects born in Norway 1967–1976 were followed up in several national registries. The association between dropping out of high school by age 23 and all-cause and cause-specific mortality 10 years onwards were estimated as hazard ratios (HR), applying Cox regression in confounder-adjusted models. Characteristics relating to different trajectories during follow-up were considered as mediators of dropout effects on survival. Results Thirty-nine percent out of 396 373 participants were classified as dropouts. All-cause mortality during follow-up was 69 per 1 00 000 person-years (2689 cases). The crude dropout HR was 2.1 (1.9–2.3), being reduced to 1.8 (1.6–2.0) in the confounder-adjusted model. After including the potential mediators in the model, dropouts no longer had an excess mortality compared to non-dropouts (HR 1.0, CI (0.9–1.1)). Mediators relating to financial situation, social benefits and employment accounted for the largest reduction in the association. Deaths caused by overdose and drug dependence were strongly associated with dropout (454 cases; rate 12; confounder-adjusted HR 5.0 (3.8–6.5)). Inclusion of work trajectory factors attenuated the association by 63% (women) and 48% (men). Conclusions Mortality was clearly associated with high school dropout, strongest for drug-related causes. The results suggest that an unfavourable work career following high school dropout could be an important mechanism for this association.

Background Increasing work participation is an important political objective in many countries. In Norway, a voluntary national intervention program aimed at increasing work participation (the IA Agreement) was implemented in 2001, and is still ongoing. One of the main goals of the IA Agreement is to reduce sickness absence (SA) and disability pensions (DP). Organisations that sign the agreement commit themselves to take active measures in order to reduce SA and DP, and in return, they are given special rights such as grants for workplace adjustments and job training. Scientific evaluations of the IA Agreement have been limited. We aim to estimate the impact of the IA Agreement on SA and DP in Norway during 2001–2009. Methods The source population is a national cohort of all 626 928 individuals born 1967–1976. Individual-level data on SA, DP, employment, and background characteristics were obtained from national registries. We apply a quasi-experimental design using difference-in-difference analysis, comparing employees in IA organisations (intervention group) with employees in non-IA organisations (control group), with respect to pre-post differences in SA and DP. Results The IA Agreement was signed by 13 760 organisations by March 2003 and covered 43% of Norwegian employees. Compared to the general working population, IA organisations had a higher proportion of women and workers in the healthcare sector. We will analyse the nationwide impact of the IA Agreement on SA and DP and explore differences by gender, industry and diagnostic category.

The issue of time-dependent confounding is central in causal inference. When comparing treatments in a non-randomised setting, one will often need to correct for confounders that influence treatment choice over time, and that are themselves influenced by treatment. We present a new method for handling such confounding when estimating treatment effects of the treated. Using Farewell's linear increments model, we estimate what the values of possible confounding processes would have been, counterfactually, if treated individuals had not started treatment. These estimated counterfactuals are used to find a causally valid estimate of the treatment effect. The approach also yields a definition of a natural direct and indirect effect as functions of time. The validity of the approach is demonstrated in a simulation study. The method is also applied to data from the Swiss HIV Cohort Study, analysing the effect of antiretroviral treatment in HIV patients on time to AIDS or death. Compared to other methods that are used to adjust for time-dependent confounding our approach makes more assumptions about the covariate processes themselves rather than the treatment initiation. R code for carrying out the analyses is available.

Seasonal variation may reduce the validity of 25-hydroxyvitamin D (25OHD) as a biomarker of vitamin D status. Here we aimed to identify potential determinants of seasonal variation in 25OHD concentrations and to evaluate cosinor modelling as a method to adjust single 25OHD measurements for seasonal variation. In Caucasian cardiovascular patients (1999–2004), we measured 25OHD by liquid chromatography tandem mass spectrometry in 4116 baseline and 528 follow-up samples. To baseline values, we fitted a cosinor model for monthly concentrations of 25OHD. Using the model, we estimated each patient's adjusted annual 25OHD value. Further, we studied how covariates affected the annual mean 25OHD concentration and seasonal variation of the study cohort. To evaluate the model, we predicted follow-up measurements with and without covariates and compared accuracy with carrying forward baseline values and linear regression adjusting for season, common approaches in research and clinical practice, respectively. The annual mean (59.6 nmol/l) was associated with participants' age, gender, smoking status, body mass, physical activity level, diabetes diagnosis, vitamin D supplement use and study site (adjusted models, P<0.05). Seasonal 25OHD variation was 15.8 nmol/l, and older age (>62 years) was associated with less variation (adjusted model, P=0.025). Prediction of follow-up measurements was more accurate with the cosinor model compared with the other approaches (P<0.05). Adding covariates to cosinor models did not improve prediction (P>0.05). We find cosinor models suitable and flexible for analysing and adjusting for seasonal variation in 25OHD concentrations, which is influenced by age.