Regression models have been in use for decades to explore and quantify the association between a dependent response and several independent variables in environmental sciences, epidemiology and public health. However, researchers often encounter situations in which some independent variables exhibit high bivariate correlation, or may even be collinear. Improper statistical handling of this situation will most certainly generate models of little or no practical use and misleading interpretations. By means of two example studies, we demonstrate how diagnostic tools for collinearity or near-collinearity may fail in guiding the analyst. Instead, the most appropriate way of handling collinearity should be driven by the research question at hand and, in particular, by the distinction between predictive or explanatory aims.

Objectives To estimate age standardised suicide rate ratios in male and female physicians compared with the general population, and to examine heterogeneity across study results. Design Systematic review and meta-analysis. Data sources Studies published between 1960 and 31 March 2024 were retrieved from Embase, Medline, and PsycINFO. There were no language restrictions. Forward and backwards reference screening was performed for selected studies using Google Scholar. Eligibility criteria for selecting studies Observational studies with directly or indirectly age standardised mortality ratios for physician deaths by suicide, or suicide rates per 100 000 person years of physicians and a reference group similar to the general population, or extractable data on physician deaths by suicide suitable for the calculation of ratios. Two independent reviewers extracted data and assessed the risk of bias using an adapted version of the Joanna Briggs Institute checklist for prevalence studies. Mean effect estimates for male and female physicians were calculated based on random effects models, with subgroup analyses for geographical region and a secondary analysis of deaths by suicide in physicians compared with other professions. Results Among 39 included studies, 38 studies for male physicians and 26 for female physicians were eligible for analyses, with a total of 3303 suicides in male physicians and 587 in female physicians (observation periods 1935-2020 and 1960-2020, respectively). Across all studies, the suicide rate ratio for male physicians was 1.05 (95% confidence interval 0.90 to 1.22). For female physicians, the rate ratio was significantly higher at 1.76 (1.40 to 2.21). Heterogeneity was high for both analyses. Meta-regression revealed a significant effect of the midpoint of study observation period, indicating decreasing effect sizes over time. The suicide rate ratio for male physicians compared with other professions was 1.81 (1.55 to 2.12). Conclusion Standardised suicide rate ratios for male and female physicians decreased over time. However, the rates remained increased for female physicians. The findings of this meta-analysis are limited by a scarcity of studies from regions outside of Europe, the United States, and Australasia. These results call for continued efforts in research and prevention of physician deaths by suicide, particularly among female physicians and at risk subgroups. Systematic review registration PROSPERO CRD42019118956.

BACKGROUND: Rates of poor sleep and hypertension are alarming worldwide. In this study, we investigate the association between sleeping difficulties and sleep duration with hypertension risk in women. METHODS: Sixty-six thousand one hundred twenty-two participants of the Nurses’ Health Study 2, who were free of hypertension at baseline (2001), were followed prospectively for 16 years and incident hypertension assessed every 2 years. Cox proportional hazards models were used to estimate hazard ratios (HRs) and 95% confidence intervals (95% CIs) for hypertension incidence associated with sleeping difficulties and sleep duration. RESULTS: During follow-up, we documented 25 987 incident cases of hypertension. After controlling for demographic and lifestyle risk factors, compared with women who slept 7 to 8 hours, women with shorter sleep duration had a significantly higher risk of hypertension (≤5 hours: HR, 1.10 [95% CI, 1.05–1.16]; 6 hours: HR, 1.07 [95% CI, 1.03–1.10]), whereas the risk for women with longer sleep duration was not statistically significant (9 hours: HR, 1.03 [95% CI, 0.97–1.10]; >9 hours: HR, 1.08 [95% CI, 0.94–1.23]). Compared with women rarely having difficulty falling or staying asleep, women sometimes or usually having these sleep difficulties had significantly higher risk of developing hypertension (HR, 1.14 [95% CI, 1.11–1.17] and 1.28 [95% CI, 1.22–1.35]; P _trend <0.001). Early morning awakening was not associated with hypertension risk ( P _trend =0.722). There was no effect modification by night work or chronotype. CONCLUSIONS: Difficulty falling or staying asleep and short sleep duration were associated with higher risk of hypertension among women in our study. Screening for poor sleep could be useful in identifying people at higher risk for hypertension.

Some evidence puts health professionals at increased risk of suicide, especially females, whereas other research suggests a lower risk in high-skilled occupations. This study investigated the suicide risk of four health professions (physicians, dentists, veterinarians, pharmacists) and three other high-skilled occupations (notaries, lawyers, tax advisors/public accountants) in Austria compared to the general population, and analyzed suicide methods across occupations. Data was collected from professional associations and Austrian cause-of-death statistics to determine suicide cases. Gender-specific standardized mortality ratios (SMRs), crude and age-adjusted suicide rates and frequencies for suicide methods were calculated for each profession (maximum time span 1986-2020). Among males, only veterinarians had a significantly elevated suicide risk compared to the general population. Physicians and tax advisors/public accountants had a significantly lower suicide risk. Among females, the veterinarians, physicians, and pharmacists had a significantly elevated suicide risk; for dentists, it was also elevated, though non-significantly. Age-adjusted suicide rates showed a smaller gap between men and women in all professions compared to the general population. Poisoning was the predominant suicide method among health professions, except dentists. These findings are consistent with some of the prior literature and call for specific suicide prevention efforts in health professions, focusing on women.

Breast cancer is highly prevalent yet a more complete understanding of the interplay between genes and probable environmental risk factors, such as night work, remains lagging. Using a discordant twin pair design, we examined the association between night shift work and breast cancer risk, controlling for familial confounding. Shift work pattern was prospectively assessed by mailed questionnaires among 5,781 female twins from the Older Finnish Twin Cohort. Over the study period (1990–2018), 407 incident breast cancer cases were recorded using the Finnish Cancer Registry. Cox proportional hazards models were used to calculate hazard ratios (HRs) and 95% confidence intervals (CIs) adjusting for potential confounders. Within-pair co-twin analyses were employed in 57 pairs to account for potential familial confounding. Compared to women who worked days only, women with shift work that included night shifts had a 1.58-fold higher risk of breast cancer (HR = 1.58; 95%CI, 1.16–2.15, highest among the youngest women i.e. born 1950–1957, HR = 2.08; 95%CI, 1.32–3.28), whereas 2-shift workers not including night shifts, did not (HR = 0.84; 95%CI, 0.59–1.21). Women with longer sleep (average sleep duration > 8 h/night) appeared at greatest risk of breast cancer if they worked night shifts (HR = 2.91; 95%CI, 1.55–5.46; P_intx=0.32). Results did not vary by chronotype (P_intx=0.74). Co-twin analyses, though with limited power, suggested that night work may be associated with breast cancer risk independent of early environmental and genetic factors. These results confirm a previously described association between night shift work and breast cancer risk. Genetic influences only partially explain these associations.