Confounder Adjustment Using the Disease Risk Score: A Proposal for Weighting Methods.

Propensity score analysis is a common approach to addressing confounding in nonrandomized studies. Its implementation, however, requires important assumptions (e.g., positivity). The disease risk score (DRS) is an alternative confounding score that can relax some of these assumptions. Like the propensity score, the DRS summarizes multiple confounders into a single score, on which conditioning by matching allows the estimation of causal effects. However, matching relies on arbitrary choices for pruning out data (e.g., matching ratio, algorithm, and caliper width) and may be computationally demanding. Alternatively, weighting methods, common in propensity score analysis, are easy to implement and may entail fewer choices, yet none have been developed for the DRS. Here we present 2 weighting approaches: One derives directly from inverse probability weighting; the other, named target distribution weighting, relates to importance sampling. We empirically show that inverse probability weighting and target distribution weighting display performance comparable to matching techniques in terms of bias but outperform them in terms of efficiency (mean squared error) and computational speed (up to >870 times faster in an illustrative study). We illustrate implementation of the methods in 2 case studies where we investigate placebo treatments for multiple sclerosis and administration of aspirin in stroke patients.

Intergenerational transmission of tobacco smoking: The role of the child's behavioral difficulties. Data from the Danish National Birth cohort (DNBC).

AIMS: This study explores the role of offspring behavioral difficulties in the intergeneration transmission of tobacco smoking. METHODS: This longitudinal cohort study is based on children born in Denmark in 1996-2003 participating in the Danish National Birth Cohort (DNBC), followed-up until 18years of age. We included mother-child pairs with complete data regarding the exposure (4 trajectories of maternal daily smoking quantity during pregnancy: low, intermediate/stable, intermediate/decreasing and high), outcome (offspring daily smoking status at 18 years) and mediator (offspring symptoms of hyperactivity-inattention at 11 years), that is 24,588 mother-child pairs. RESULTS: In our study population, during pregnancy respectively 86.2%, 6.80%, 4.08% and 2.97% mothers belonged to the low, intermediate/stable, intermediate/decreasing and high smoking trajectory groups. After controlling for covariates using propensity scores, the direct effect of maternal smoking in pregnancy on offspring smoking in adolescence was statistically significant, especially when the mother belonged to the intermediate/stable smoking trajectory group (ORIPW = 2.09, 95% CI: 1.70 - 2.61) or to the high smoking trajectory group (ORIPW = 2.08, 95% CI: 1.52 - 3.11) compared to the low smoking trajectory group. None of the indirect effects of maternal smoking in pregnancy were statistically significant, and neither were the proportions mediated. CONCLUSION: Maternal pregnancy smoking seems to have an influence on offspring smoking in early adulthood, which does not appear to be mediated by offspring behavioral difficulties. Women should be strongly encouraged to quit smoking in pregnancy to reduce both short and long-term health risks among their offspring.

Use of systemic and biological therapy in patients with moderate-to-severe psoriasis.

BACKGROUND: Patients with moderate-to-severe psoriasis are candidates for systemic treatment, but it is unknown how many receive such therapy at a national level in Denmark. OBJECTIVES: We aimed to determine the prevalence of conventional systemic therapy use in patients with moderate-to-severe psoriasis and, further, to investigate the time to discontinuation of conventional systemic therapy and initiation of biological therapy among biologic-naïve patients. METHODS: This registry-based study identified a cohort of patients with psoriasis in Denmark. We estimated the prevalence of moderate-to-severe psoriasis at a national level using registry data. Inverse probability weighting was used to mitigate potential selection bias in the prevalence estimate of moderate-to-severe psoriasis. Analyses were then performed on the weighted cohort. RESULTS: Of patients with psoriasis in Denmark, 10.9% were estimated to have moderate-to-severe psoriasis, of whom 62.3% received either conventional systemic or biological therapy, meaning 37.7% who were considered candidates for systemic therapy did not receive any systemic treatment. The study demonstrated that, comparing previous time periods with more recent years: (i) time on conventional systemic therapy for patients with moderate-to-severe psoriasis has become shorter, with a median (interquartile range) of 3.0 years (0.6-10.0) in 1985-1994 vs. 0.6 years (0.3-2.0) in 2014-2018; (ii) more patients initiated biologics as second-line therapy, with 69.5% in 2010-2013 vs. 71.2% in 2014-2018; and (iii) the median time from initiation of systemic therapy to initiation of biological therapy decreased from 13.3 years (11.5-16.8) in 2010-2013 to 1.9 years (1.7-2.4) in 2014-2018. CONCLUSIONS: This study found that nearly 37.7% of Danish patients with moderate-to-severe psoriasis do not receive systemic treatment even though they would qualify for this. Furthermore, for patients treated with conventional systemics, drug survival decreased during the observation period.

Propensity-based standardization to enhance the validation and interpretation of prediction model discrimination for a target population.

External validation of the discriminative ability of prediction models is of key importance. However, the interpretation of such evaluations is challenging, as the ability to discriminate depends on both the sample characteristics (ie, case-mix) and the generalizability of predictor coefficients, but most discrimination indices do not provide any insight into their respective contributions. To disentangle differences in discriminative ability across external validation samples due to a lack of model generalizability from differences in sample characteristics, we propose propensity-weighted measures of discrimination. These weighted metrics, which are derived from propensity scores for sample membership, are standardized for case-mix differences between the model development and validation samples, allowing for a fair comparison of discriminative ability in terms of model characteristics in a target population of interest. We illustrate our methods with the validation of eight prediction models for deep vein thrombosis in 12 external validation data sets and assess our methods in a simulation study. In the illustrative example, propensity score standardization reduced between-study heterogeneity of discrimination, indicating that between-study variability was partially attributable to case-mix. The simulation study showed that only flexible propensity-score methods (allowing for non-linear effects) produced unbiased estimates of model discrimination in the target population, and only when the positivity assumption was met. Propensity score-based standardization may facilitate the interpretation of (heterogeneity in) discriminative ability of a prediction model as observed across multiple studies, and may guide model updating strategies for a particular target population. Careful propensity score modeling with attention for non-linear relations is recommended.

Saving time and money in biomedical publishing: the case for free-format submissions with minimal requirements.

BACKGROUND: Manuscript preparation and the (re)submission of articles can create a significant workload in academic jobs. In this exploratory analysis, we estimate the time and costs needed to meet the diverse formatting requirements for manuscript submissions in biomedical publishing. METHODS: We reviewed 302 leading biomedical journals' submission guidelines and extracted information on the components that tend to vary the most among submission guidelines (the length of the title, the running title, the abstract, and the manuscript; the structure of the abstract and the manuscript, number of items and references allowed, whether the journal has a template). We estimated annual research funding lost due to manuscript formatting by calculating hourly academic salaries, the time lost to reformatting articles, and quantifying the total number of resubmissions per year. We interviewed several researchers and senior journal editors and editors-in-chief to contextualize our findings and develop guidelines that could help both biomedical journals and researchers work more efficiently. RESULTS: Among the analyzed journals, we found a huge diversity in submission requirements. By calculating average researcher salaries in the European Union and the USA, and the time spent on reformatting articles, we estimated that ~ 230 million USD were lost in 2021 alone due to reformatting articles. Should the current practice remain unchanged within this decade, we estimate ~ 2.5 billion USD could be lost between 2022 and 2030-solely due to reformatting articles after a first editorial desk rejection. In our interviews, we found alignment between researchers and editors; researchers would like the submission process alignment between researchers and editors; researchers would like the submission process to be as straightforward and simple as possible, and editors want to easily identify strong, suitable articles and not waste researchers' time. CONCLUSIONS: Based on the findings from our quantitative analysis and contextualized by the qualitative interviews, we conclude that free-format submission guidelines would benefit both researchers and editors. However, a minimum set of requirements is necessary to avoid manuscript submissions that lack structure. We developed our guidelines to improve the status quo, and we urge the publishers and the editorial-advisory boards of biomedical journals to adopt them. This may also require support from publishers and major international organizations that govern the work of editors.

Assessing racial bias in type 2 diabetes risk prediction algorithms.

Risk prediction models for type 2 diabetes can be useful for the early detection of individuals at high risk. However, models may also bias clinical decision-making processes, for instance by differential risk miscalibration across racial groups. We investigated whether the Prediabetes Risk Test (PRT) issued by the National Diabetes Prevention Program, and two prognostic models, the Framingham Offspring Risk Score, and the ARIC Model, demonstrate racial bias between non-Hispanic Whites and non-Hispanic Blacks. We used National Health and Nutrition Examination Survey (NHANES) data, sampled in six independent two-year batches between 1999 and 2010. A total of 9,987 adults without a prior diagnosis of diabetes and with fasting blood samples available were included. We calculated race- and year-specific average predicted risks of type 2 diabetes according to the risk models. We compared the predicted risks with observed ones extracted from the US Diabetes Surveillance System across racial groups (summary calibration). All investigated models were found to be miscalibrated with regard to race, consistently across the survey years. The Framingham Offspring Risk Score overestimated type 2 diabetes risk for non-Hispanic Whites and underestimated risk for non-Hispanic Blacks. The PRT and the ARIC models overestimated risk for both races, but more so for non-Hispanic Whites. These landmark models overestimated the risk of type 2 diabetes for non-Hispanic Whites more severely than for non-Hispanic Blacks. This may result in a larger proportion of non-Hispanic Whites being prioritized for preventive interventions, but it also increases the risk of overdiagnosis and overtreatment in this group. On the other hand, a larger proportion of non-Hispanic Blacks may be potentially underprioritized and undertreated.

Development and internal validation of a diagnostic prediction model for psoriasis severity.

BACKGROUND: While administrative health records such as national registries may be useful data sources to study the epidemiology of psoriasis, they do not generally contain information on disease severity. OBJECTIVES: To develop a diagnostic model to distinguish psoriasis severity based on administrative register data. METHOD: We conducted a retrospective registry-based cohort study using the Danish Skin Cohort linked with the Danish national registries. We developed a diagnostic model using a gradient boosting machine learning technique to predict moderate-to-severe psoriasis. We performed an internal validation of the model by bootstrapping to account for any optimism. RESULTS: Among 4016 adult psoriasis patients (55.8% women, mean age 59 years) included in this study, 1212 (30.2%) patients were identified as having moderate-to-severe psoriasis. The diagnostic prediction model yielded a bootstrap-corrected discrimination performance: c-statistic equal to 0.73 [95% CI: 0.71-0.74]. The internal validation by bootstrap correction showed no substantial optimism in the results with a c-statistic of 0.72 [95% CI: 0.70-0.74]. A bootstrap-corrected slope of 1.10 [95% CI: 1.07-1.13] indicated a slight under-fitting. CONCLUSION: Based on register data, we developed a gradient boosting diagnostic model returning acceptable prediction of patients with moderate-to-severe psoriasis.

Out-of-home care placement and mortality rate in early adulthood: identifying vulnerable subgroups in a nationwide cohort study.

Out-of-home care has been linked to excess mortality across the lifespan. We examined whether this association is modified by the age at first out-of-home care placement and the number of placements. In this population-based cohort study, we used register data covering all children born in Denmark between 1 and 1980 and 31 December 1999, totalling 1,111,193 individuals followed until 31 December 2018. We divided participants according to sex, out-of-home care status, age at first placement, and the number of placements. We estimated adjusted hazard ratios and hazard differences per 10,000 person-years for all-cause mortality and mortality due to suicide, accidents, and cancer between ages 18 and 39. 53,015 (4.8%) of the participants were placed in out-of-home care before age 18. The adjusted hazard ratio for all-cause mortality was 3.4 (95% CI 3.1-3.7) for males and 4.7 (4.0-5.4) for females, corresponding to 20.6 (19.0-22.2) and 10.3 (9.1-11.5) additional deaths per 10,000 individuals annually among males and females, respectively. Associations did not vary substantially according to age at first placement or the number of placements. Both males and females with a history of out-of-home care were more likely to die from suicide, accidents, and cancer compared with their peers. We show a markedly higher all-cause and cause-specific mortality among children who have been placed in out-of-home care, but contrary to our hypothesis, age at first placement and the number of placements did not modify this relation. These results warrant further investigation into potential target points for interventions that may prevent premature mortality in this group of disadvantaged individuals.

Impact of hypothetical improvements in the psychosocial work environment on sickness absence rates: a simulation study.

BACKGROUND: The association between psychosocial working environments and sickness absence is well-known. However, the potential for reducing sickness absences of different lengths through improvements in psychosocial work factors is not fully understood. We aim to quantify the potential for reducing short-, intermediate- and long-term sickness absence rates, respectively, through hypothetical improvements in several psychosocial work factors. METHODS: This longitudinal study includes 24 990 public hospital employees from the 2014 wave of the Well-being in Hospital Employees study. The 1-year sickness absence rate was divided into short- (1-3 days), intermediate- (4-28 days) and long-term (29 days or more) periods. We simulated hypothetical scenarios with improvements in 17 psychosocial work factors using the parametric g-formula and estimated resulting changes in sickness absence rate ratios (RRs) with 95% confidence intervals (95% CIs). RESULTS: Setting all 17 psychosocial work factors to their most desirable levels (vs. least desirable levels) was associated with an overall 54% lower rate of sickness absence (95% CI: 48-60%). Reducing bullying (no vs. yes RR: 0.86, 95% CI: 0.83-0.90) and perceived stress (low vs. high RR: 0.90, 95% CI: 0.87-0.92), and increasing skill discretion (high vs. low RR: 0.91, 95% CI: 0.89-0.94) held the largest potential for reducing the total sickness absence rate. Overall, associations were similar for short-, intermediate- and long-term sickness absence. CONCLUSIONS: The psychosocial working environment was strongly associated with sickness absence. Improving the working environment may have a great impact on short-, intermediate- and long-term sickness absence rates.

Clinical prediction models for mortality in patients with covid-19: external validation and individual participant data meta-analysis.

OBJECTIVE: To externally validate various prognostic models and scoring rules for predicting short term mortality in patients admitted to hospital for covid-19. DESIGN: Two stage individual participant data meta-analysis. SETTING: Secondary and tertiary care. PARTICIPANTS: 46 914 patients across 18 countries, admitted to a hospital with polymerase chain reaction confirmed covid-19 from November 2019 to April 2021. DATA SOURCES: Multiple (clustered) cohorts in Brazil, Belgium, China, Czech Republic, Egypt, France, Iran, Israel, Italy, Mexico, Netherlands, Portugal, Russia, Saudi Arabia, Spain, Sweden, United Kingdom, and United States previously identified by a living systematic review of covid-19 prediction models published in The BMJ, and through PROSPERO, reference checking, and expert knowledge. MODEL SELECTION AND ELIGIBILITY CRITERIA: Prognostic models identified by the living systematic review and through contacting experts. A priori models were excluded that had a high risk of bias in the participant domain of PROBAST (prediction model study risk of bias assessment tool) or for which the applicability was deemed poor. METHODS: Eight prognostic models with diverse predictors were identified and validated. A two stage individual participant data meta-analysis was performed of the estimated model concordance (C) statistic, calibration slope, calibration-in-the-large, and observed to expected ratio (O:E) across the included clusters. MAIN OUTCOME MEASURES: 30 day mortality or in-hospital mortality. RESULTS: Datasets included 27 clusters from 18 different countries and contained data on 46 914patients. The pooled estimates ranged from 0.67 to 0.80 (C statistic), 0.22 to 1.22 (calibration slope), and 0.18 to 2.59 (O:E ratio) and were prone to substantial between study heterogeneity. The 4C Mortality Score by Knight et al (pooled C statistic 0.80, 95% confidence interval 0.75 to 0.84, 95% prediction interval 0.72 to 0.86) and clinical model by Wang et al (0.77, 0.73 to 0.80, 0.63 to 0.87) had the highest discriminative ability. On average, 29% fewer deaths were observed than predicted by the 4C Mortality Score (pooled O:E 0.71, 95% confidence interval 0.45 to 1.11, 95% prediction interval 0.21 to 2.39), 35% fewer than predicted by the Wang clinical model (0.65, 0.52 to 0.82, 0.23 to 1.89), and 4% fewer than predicted by Xie et al's model (0.96, 0.59 to 1.55, 0.21 to 4.28). CONCLUSION: The prognostic value of the included models varied greatly between the data sources. Although the Knight 4C Mortality Score and Wang clinical model appeared most promising, recalibration (intercept and slope updates) is needed before implementation in routine care.

Chronic disorders, work-unit leadership quality and long-term sickness absence among 33 025 public hospital employees.

OBJECTIVE: This study aimed to examine the association between work-unit level leadership quality and individual-level long-term sickness absence (LTSA) in the hospital sector and effect modification by chronic disorders. METHODS: This longitudinal analysis included 33 025 Danish public hospital employees who were followed-up for one year after baseline in March 2014. Leadership quality was assessed by questionnaire with mean responses aggregated by work-unit and characterized in tertiles. LTSA during follow-up was determined from employer records. Chronic disorders at baseline was assessed from the Danish hospital and prescription registers. We performed multilevel logistic regression to estimate odds ratios (OR) and 95% confidence intervals (CI) adjusting for potential confounders. We evaluated interaction between chronic illness and low leadership quality on multiplicative and additive scales. RESULTS: We identified employees as healthy (60.8%) or with somatic (31.6%), mental (3.3%), or both somatic and mental (4.3%) disorders. During follow-up, 6% of employees registered a LTSA. Medium and high leadership quality were associated with lower risk of LTSA with OR of 0.84 (95% CI 0.76-0.94) and 0.73 (95% CI 0.65-0.82) respectively, compared to low leadership quality. Associations were similar for healthy employees and employees with only somatic disorders, whereas no association was observed for employees with mental disorders (in presence or absence of somatic disorders). No statistically significant (α=0.05) interactions between leadership quality and chronic disorders on LTSA were observed. CONCLUSION: The findings suggest that the quality of leadership in work units is associated with risk of long-term sick leave in the Danish public hospital sector and that strong leadership protects employees against LTSA.

Impact of housing conditions on changes in youth's mental health following the initial national COVID-19 lockdown: a cohort study.

We aimed to investigate if declines in youth's mental health during lockdown were dependent on housing condition among 7445 youth (median age ~ 20 years) from the Danish National Birth Cohort (DNBC), with data collected at 18 years of age and again three weeks into the first national lockdown (April 2020). We examined associations between housing conditions (access to outdoor spaces, urbanicity, household density, and household composition) and changes in mental health (mental well-being, Quality of Life (QoL) and loneliness). We report results from multivariate linear and logistic regression models. Youth without access to outdoor spaces experienced greater declines in mental well-being (vs. garden; mean difference: - 0·75 (95% CI - 1·14, - 0·36)), and correspondingly greater odds of onset of low mental well-being (vs. garden; OR: 1·72 (95% CI 1·20, 2·48)). Youth in higher density households vs. below median or living alone vs. with parents only also had greater odds of onset of low mental well-being (OR: 1·26 (95% CI 1·08, 1·46) and OR: 1·62 (95% CI 1·17, 2·23), respectively). Living in denser households (vs. below median; OR: 1·18 (95% CI 1·06, 1·33), as well as living alone (vs. with parents; OR: 1·38 (95% CI 1·04, 1·82) was associated with onset of low QoL. Living alone more than doubled odds of onset of loneliness compared to living with parents, OR: 2·12 (95% CI 1·59, 2·82). Youth living alone, in denser households, and without direct access to outdoor spaces may be especially vulnerable to mental health declines.

Mediation of the parental education gradient in early adult mortality by childhood adversity: a population-based cohort study of more than 1 million children.

BACKGROUND: Children born into disadvantaged socioeconomic circumstances are more likely to experience both adversity during childhood and premature mortality. The aim of this study was to investigate how much of the parental education gradient in early adult mortality is explained by exposure to childhood adversity. METHODS: We used data from the nationwide register-based Danish Life Course cohort study. Our sample consisted of all individuals born between Jan 1, 1980, and Dec 31, 2001, who did not emigrate or die before age 16 years, and for whom information on parental education level was available. These individuals were followed up for mortality from age 16 years until Dec 31, 2018. Highest attained parental education level at birth was divided into low (≤9 years), medium (10-12 years), and high (>12 years) according to years in education. Individuals were assigned to one of five childhood adversity trajectory groups based on their annual exposure between age 0 and 16 years to a broad selection of adversities in three dimensions: material deprivation, loss or threat of loss in the family, and family dynamics. Childhood abuse was not included. The association between parental education level and mortality was assessed with a Cox proportional hazards model. To assess the magnitude of mediation of this association by childhood adversity, we used counterfactual mediation analysis and an Aalen additive hazards model. Analyses were unadjusted and adjusted for parental origin and parental ages at birth. FINDINGS: Our sample consisted of 1 278 156 individuals followed up from birth until age 16-38 years. The sample comprised 655 633 (51·3%) men and 622 523 (48·7%) women, and 1 243 981 (97·3%) participants were of European descent. During follow-up, 5387 deaths were registered. Compared with the high parental education group, we calculated a total effect equal to 8·7 additional deaths (95% CI 6·6-10·9) per 100 000 person-years in the medium parental education group and 31·9 (28·5 to 35·2) per 100 000 person-years in the low parental education group. Mediation through childhood adversity trajectories accounted for 41·5% (95% CI 8·0-67·5) of the additional deaths in the medium parental education group and 46·4% (32·9-58·8) of the additional deaths in the low parental education group. The results were similar when adjusting the analyses for sociodemographic factors. INTERPRETATION: The experience of childhood adversity seems to be an important mediator of the association between parental education and mortality in early adulthood. Interventions reducing the exposure to childhood adversity might thus reduce the parental education gradient in early adult mortality. FUNDING: Netherlands Organisation for Health Research and Development.

Reducing employee turnover in hospitals: estimating the effects of hypothetical improvements in the psychosocial work environment.

OBJECTIVES: Poor psychosocial work environments in hospitals are associated with higher employee turnover. In this prospective cohort study, we aimed to identify and quantify which aspects of the psychosocial work environment have the greatest impact on one-year employee turnover rates within a hospital setting, both overall and within occupational groups. METHODS: The study population included 24 385 public hospital employees enrolled in the Danish Well-being in Hospital Employees cohort in 2014. We followed the participants for one year and registered if they permanently left their workplace. Using baseline sociodemographic, workplace, and psychosocial work environment characteristics, we applied the parametric g-formula to simulate hypothetical improvements in the psychosocial work environment and estimated turnover rate differences (RD) per 10 000 employees per year and 95% confidence intervals (95% CI). RESULTS: Of the 24 385 participants, 2552 (10.5%) left the workplace during the one-year follow-up. Up to 44% of this turnover was potentially preventable through hypothetical improvements in the psychosocial work environment. The specific hypothetical improvements with the largest effects were in satisfaction with work prospects (RD -522 turnovers per 10 000 person-years, 95% CI -536- -508), general job satisfaction (RD -339, 95% CI -353- -325) and bullying (RD -200, 95% CI -214- -186). The potential for preventing turnover was larger for nurses than for physicians and other healthcare employees. CONCLUSIONS: Improvements in the psychosocial work environment may have great potential for reducing turnover among hospital staff, particularly among nurses.

Internal-external cross-validation helped to evaluate the generalizability of prediction models in large clustered datasets.

OBJECTIVE: To illustrate how to evaluate the need of complex strategies for developing generalizable prediction models in large clustered datasets. STUDY DESIGN AND SETTING: We developed eight Cox regression models to estimate the risk of heart failure using a large population-level dataset. These models differed in the number of predictors, the functional form of the predictor effects (non-linear effects and interaction) and the estimation method (maximum likelihood and penalization). Internal-external cross-validation was used to evaluate the models' generalizability across the included general practices. RESULTS: Among 871,687 individuals from 225 general practices, 43,987 (5.5%) developed heart failure during a median follow-up time of 5.8 years. For discrimination, the simplest prediction model yielded a good concordance statistic, which was not much improved by adopting complex strategies. Between-practice heterogeneity in discrimination was similar in all models. For calibration, the simplest model performed satisfactorily. Although accounting for non-linear effects and interaction slightly improved the calibration slope, it also led to more heterogeneity in the observed/expected ratio. Similar results were found in a second case study involving patients with stroke. CONCLUSION: In large clustered datasets, prediction model studies may adopt internal-external cross-validation to evaluate the generalizability of competing models, and to identify promising modelling strategies.

'Standing together - at a distance': Documenting changes in mental-health indicators in Denmark during the COVID-19 pandemic.

Aims: There is a need to document the mental-health effects of the COVID-19 pandemic and its associated societal lockdowns. We initiated a large mixed-methods data collection, focusing on crisis-specific worries and mental-health indicators during the lockdown in Denmark. Methods: The study incorporated five data sources, including quantitative surveys and qualitative interviews. The surveys included a time series of cross-sectional online questionnaires starting on 20 March 2020, in which 300 (3×100) Danish residents were drawn every three days from three population groups: the general population (N=1046), families with children (N=1032) and older people (N=1059). These data were analysed by trend analysis. Semi-structured interviews were conducted with 32 people aged 24-83 throughout Denmark to provide context to the survey results and to gain insight into people's experiences of the lockdown. Results: Absolute level of worries, quality of life and social isolation were relatively stable across all population groups during the lockdown, although there was a slight deterioration in older people's overall mental health. Many respondents were worried about their loved ones' health (74-76%) and the potential long-term economic consequences of the pandemic (61-66%). The qualitative interviews documented significant variation in people's experiences, suggesting that the lockdown's effect on everyday life had not been altogether negative. Conclusions: People in Denmark seem to have managed the lockdown without alarming changes in their mental health. However, it is important to continue investigating the effects of the pandemic and various public-health measures on mental health over time and across national contexts.

Incomparability of treatment groups is often blindly ignored in randomised controlled trials - a post hoc analysis of baseline characteristic tables.

OBJECTIVES: Randomisation is often believed to lead to baseline comparability of treatment groups in controlled trials. This study aims to challenge this popular belief, which is relevant in expectation- but not necessarily in realisation. STUDY DESIGN AND SETTING: After presenting an overview of methods for assessing baseline comparability of treatment groups in randomised controlled trials (RCTs), we reviewed RCTs published over 1 year in three high-impact medical journals. We extracted data regarding the methods used to evaluate baseline comparability. To quantify baseline balance, we calculated post hoc standardised mean differences (SMDs) in baseline characteristics reported in these trials. RESULTS: Amongst 142 RCTs, 120 (84.5%) claimed that baseline comparability was achieved. However, 81 RCTs (57%) did not report how they assessed this balance. The rest (61 RCTs, 43%) used traditional statistical tests, which are deemed inappropriate for balance checking. Our post hoc calculation of SMDs showed that 49 (34.5%) RCTs had at least one baseline variable, which might have been strongly unbalanced (i.e., SMD ≥25%) across treatment groups. CONCLUSION: Baseline incomparability of treatment groups in RCTs is often blindly ignored. We suggest it be thoroughly evaluated and transparently reported, using the standardised mean difference or other proper balance metrics.

Counterfactual clinical prediction models could help to infer individualized treatment effects in randomized controlled trials-An illustration with the International Stroke Trial.

OBJECTIVE: Causal treatment effects are estimated at the population level in randomized controlled trials, while clinical decision is often to be made at the individual level in practice. We aim to show how clinical prediction models used under a counterfactual framework may help to infer individualized treatment effects. STUDY DESIGN AND SETTING: As an illustrative example, we reanalyze the International Stroke Trial. This large, multicenter trial enrolled 19,435 adult patients with suspected acute ischemic stroke from 36 countries, and reported a modest average benefit of aspirin (vs. no aspirin) on a composite outcome of death or dependency at 6 months. We derive and validate multivariable logistic regression models that predict the patient counterfactual risks of outcome with and without aspirin, conditionally on 23 predictors. RESULTS: The counterfactual prediction models display good performance in terms of calibration and discrimination (validation c-statistics: 0.798 and 0.794). Comparing the counterfactual predicted risks on an absolute difference scale, we show that aspirin-despite an average benefit-may increase the risk of death or dependency at 6 months (compared with the control) in a quarter of stroke patients. CONCLUSIONS: Counterfactual prediction models could help researchers and clinicians (i) infer individualized treatment effects and (ii) better target patients who may benefit from treatments.