An introduction to multilevel regression models

Influence of Race, Ethnicity, and Gender on Clinical Performance Assessments in Graduate Medical Education

BACKGROUND

Evidence suggests disparities associated with race and ethnicity (R&E) in assessment in graduate medical education (GME). How these disparities manifest across competencies and training time and intersect with disparities associated with other factors such as gender is unclear.

OBJECTIVE

Examine the association of R&E and gender with clinical performance assessment in GME.

DESIGN

Longitudinal analysis of resident clinical performance assessments.

PARTICIPANTS

Assessment data of residents at seven internal medicine (IM) residency programs, 2014 to 2019 (9346 evaluations of 664 residents). Residents underrepresented in medicine (URiM) were identified using self-reported R&E.

MAIN MEASURES

Standardized scores were calculated for the Accreditation Council for Graduate Medical Education competencies (patient care [PC], medical knowledge [MK], practice-based learning and improvement [PBLI], systems-based practice [SBP], professionalism [PROF], and interpersonal communication and skills [ICS]). Cross-classified mixed effects regression assessed differences in standardized competency ratings with R&E over time and the interaction of R&E and gender while controlling for time of year and setting; resident gender, post-graduate year (PGY), and IM in-training examination percentile rank; and faculty gender, educational role, specialty, rank, and years of experience.

KEY RESULTS

URiM resident scores were lower than non-URIM residents across competencies (difference in standardized scores between URiM and non-URiM residents [se] PC - 0.126 [0.035], p < 0.001; MK - 0.118 [0.035], p < 0.001; PBLI - 0.122 [0.042], p = 0.004; SBP - 0.128 [0.034], p < 0.001; PROF - 0.075 [0.036], p = 0.036; ICS - 0.124 [0.039], p = 0.002). The interaction between resident R&E, gender, and PGY was significant in PBLI (estimate - 0.15 [0.06], p = 0.02) and SBP (- 0.11 [0.05], p = 0.04) indicating smaller gains over time for URiM women.

CONCLUSIONS

There were significant differences associated with R&E with lower scores for URiM residents across competencies. There was a significant interaction with resident gender and R&E over time in PBLI and SBP. Findings may reflect bias in faculty assessment, effects of non-inclusive learning environments, or structural inequities in assessment.

Social-ecological factors associated with sexual harassment across locations in US college students

OBJECTIVE

Using the social-ecological model of violence prevention, we examined the locations in which sexual harassment occurs and whether harassment differed in terms of target, harasser, relationship, and incident characteristics across locations.

PARTICIPANTS

A total of 246 mostly women, straight, White, and cisgender students at a large Midwestern university.

METHODS

Students completed an online survey and reported information about sexually-harassing experiences from the last seven days.

RESULTS

Students reported 821 instances of sexual harassment (M = 3.34 per student) over the past seven days; 74% of students reported any harassment. Using multilevel modeling, characteristics of harassers, incident characteristics, and the relationship between harassers and targets distinguished between harassment experienced across locations (i.e., housing, outside, online, public establishments, and academic buildings) more than did target characteristics.

CONCLUSIONS

These findings inform harassment prevention efforts at universities by suggesting interventions tailored to specific locations and inform debates about the proper scope of campus sexual harassment policies.

Using multi-level regression to determine associations and estimate causes and effects in clinical anesthesia due to patient, practitioner and hospital or health system practice variability

In this research methods tutorial of clinical anesthesia, we will explore techniques to estimate the influence of a myriad of factors on patient outcomes. Big data that contain information on patients, treated by individual anesthesiologists and surgical teams, at different hospitals, have an inherent multi-level data structure (Fig. 1). While researchers often attempt to determine the association between patient factors and outcomes, that does not provide clinicians with the whole story. Patient care is clustered together according to clinicians and hospitals where they receive treatment. Therefore, multi-level regression models are needed to validly estimate the influence of each factor at each level. In addition, we will explore how to estimate the influence that variability-for example, one anesthesiologist deciding to do one thing, while another takes a different approach-has on outcomes for patients, using the intra-class correlation coefficient for continuous outcomes and the median odds ratio for binary outcomes. From this tutorial, you should acquire a clearer understanding of how to perform and interpret multi-level regression modeling and estimate the influence of variable clinical practices on patient outcomes in order to answer common but complex clinical questions. Fig. 1 Infographics.

The Sustainability Spillover: Uncovering the Link Between Informal Elder Care and Eco-Conscious Behaviors Across the European Union

Background and Objectives

This study explores the association between informal caregiving for older adults and environmentally sustainable behaviors across the 27 European Union countries, aiming to identify how the gendered and domestic nature of environmentalism relates to senior care.

Research Design and Methods

Data from 41,742 respondents aged 16-74 were analyzed from the Survey of Gender Gaps in Unpaid Care, Individual and Social Activities, and conducted by a scientific consortium in 2022. Frequency of sustainable behaviors was measured across 10 indicators. Multilevel mixed-effects linear regressions examined associations between caregiving and sustainable behaviors, adjusting for demographic, socioeconomic, and attitudinal covariates.

Results

Informal caregivers engaged in eco-friendly actions significantly more frequently than noncaregivers across all 10 sustainable-behavior indicators (p < 0.001). The "caregiver effect" was strongest for sustainable-consumption choices like buying eco-friendly (β = 0.16), fair-trade (β = 0.15), and used products (β = 0.17), and weaker for household practices such as recycling (β = 0.05) and mindful resource consumption (β = 0.06). Caregivers attained higher composite environmental behavior scores (33.93 ± 8.23) than noncaregivers (31.88 ± 8.00; p < 0.001). This association remained robust after adjusting for gender, age, education, employment, household size, attitudes, and other covariates. Caregiving had the strongest association with buying used items (β = 0.20) and eco-friendly products (β = 0.14). Country-level analyses revealed consistent caregiver versus noncaregiver differences, with the largest gaps in Southern and Eastern Europe.

Discussion and Implications

This is the first large-scale cross-national study that demonstrates a consistent association between older-adult caregiving and a wide range of environmentally sustainable behaviors. Results suggest the experience of caring for a vulnerable family member is closely related to a broader sense of social and environmental responsibility. Caregivers' heightened engagement in sustainable consumption positions them as potential early adopters and change makers. Findings highlight new avenues for environmental education and caregiver support initiatives that synergistically promote interpersonal and environmental care.

Determining the influential factors of postnatal care in Bangladesh using multilevel logistic regression

BACKGROUND

Postnatal care (PNC) is the care of a newborn and mother for up to six weeks from one hour of placenta birth. The postnatal period is one of the most hazardous stages for mothers and their baby's health. The PNC is influenced by several maternal, family, biological, and socio-economic factors and it is necessary to identify the most significant factors of PNC. Therefore, the authors focus on determining the significant determinants of postnatal care in Bangladesh.

METHODS AND MATERIALS

This study is based on a secondary dataset extracted from the Bangladesh Demographic and Health Survey (BDHS)-2017/18. Descriptive statistics, chi-square test, and multilevel logistic regression have been used to determine the contributing factors of PNC.

RESULTS

The rate of postnatal care was highest in Sylhet (73.7%) and lowest in Dhaka (57.1%). Female babies had 10.1% less odds of having postnatal care than male babies. Findings depict that the children with 1-3 siblings have 1. 82 times more odds (odds ratio (OR):1.82, 95% confidence interval (CI):0.03-3.21) of PNC than babies without any siblings. Children who suffered from fever recently had 1.25 times (OR = 1.25, 95% CI:1.09-1.45) more odds of taking PNC than their counterparts. Children of working mothers had 1.33 times (OR:1.33, 95% CI:1.14-1.56) more odds of having PNC than children of non-working mothers.

CONCLUSION

The sex of a child, birth order number, place of residence, region, receiving Bacillus Calmette-Guérin (BCG) vaccine, number of antenatal care visits, having fever recently, number of household members, media exposure, and household facilities are significantly linked with PNC in Bangladesh. To ensure the good health of a child, it is necessary to focus on the targeted groups and put emphasis on the identified variables. The authors believe that the findings will be helpful to the policymakers of Bangladesh to lessen childhood morbidities which will be helpful in achieving the target of the Sustainable Development Goals (SDGs) for reducing preventable maternal and under-five deaths by 2030.

Short-term outcomes of asphyxiated neonates depending on requirement for transfer in the first 24 h of life

IMPORTANCE

In neonates with birth asphyxia (BA) and hypoxic-ischemic encephalopathy, therapeutic hypothermia (TH), initiated within six hours, is the only safe and established neuroprotective measure to prevent secondary brain injury. Infants born outside of TH centers have delayed access to cooling.

OBJECTIVE

To compare in-hospital mortality, occurrence of seizures, and functional status at discharge in newborns with BA depending on postnatal transfer for treatment to another hospital within 24 h of admission (transferred (TN) versus non-transferred neonates (NTN)).

DESIGN

Nationwide retrospective cohort study from a comprehensive hospital dataset using codes of the International Classification of Diseases, 10th modification (ICD-10). Clinical and outcome information was retrieved from diagnostic and procedural codes. Hierarchical multilevel logistic regression modeling was performed to quantify the effect of being postnatally transferred on target outcomes.

SETTING

All discharges from German hospitals from 2016 to 2021.

PARTICIPANTS

Full term neonates with birth asphyxia (ICD-10 code: P21) admitted to a pediatric department on their first day of life.

EXPOSURES

Postnatal transfer to a pediatric department within 24 h of admission to an external hospital.

MAIN OUTCOMES

In-hospital death; secondary outcomes: seizures and pediatric complex chronic conditions category (PCCC) ≥ 2.

RESULTS

Of 11,703,800 pediatric cases, 25,914 fulfilled the inclusion criteria. TNs had higher proportions of organ dysfunction, TH, organ replacement therapies, and neurological sequelae in spite of slightly lower proportions of maternal risk factors. In TNs, the adjusted odds ratios (OR) for death, seizures, and PCCC ≥ 2 were 4.08 ((95% confidence interval 3.41-4.89), 2.99 (2.65-3.38), and 1.76 (1.52-2.05), respectively. A subgroup analysis among infants receiving TH (n = 3,283) found less pronounced adjusted ORs for death (1.67 (1.29-2.17)) and seizures (1.26 (1.07-1.48)) and inverse effects for PCCC ≥ 2 (0.81 (0.64-1.02)) in TNs.

CONCLUSION AND RELEVANCE

This comprehensive nationwide study found increased odds for adverse outcomes in neonates with BA who were transferred to another facility within 24 h of hospital admission. Closely linking obstetrical units to a pediatric department and balancing geographical coverage of different levels of care facilities might help to minimize risks for postnatal emergency transfer and optimize perinatal care.

A multidimensional approach to identifying high-performing trauma centers across the United States

INTRODUCTION

The differentiators of centers performing at the highest level of quality and patient safety are likely both structural and cultural. We aimed to combine five indicators representing established domains of trauma quality and to identify and describe the structural characteristics of consistently performing centers.

METHODS

Using American College of Surgeons Trauma Quality Improvement Program data from 2017 to 2020, we evaluated five quality measures across several care domains for adult patients in levels I and II trauma centers: (1) time to operating room for patients with abdominal gunshot wounds and shock, (2) proportion of patients receiving timely venous thromboembolism prophylaxis, (3) failure to rescue (death following a complication), (4) major hospital complications, and (5) mortality. Overall performance was summarized as a composite score incorporating all measures. Centers were ranked from highest to lowest performer. Principal component analysis showed the influence of each indicator on overall performance and supported the composite score approach.

RESULTS

We identified 272 levels I and II centers, with 28 and 27 centers in the top and bottom 10%, respectively. Patients treated in high-performing centers had significant lower rates of death major complications and failure to rescue, compared with low-performing centers ( p < 0.001). The median time to operating room for gunshot wound was almost half that in high compared with low-performing centers, and rates of timely venous thromboembolism prophylaxis were over twofold greater ( p < 0.001). Top performing centers were more likely to be level I centers and cared for a higher number of severely injured patients per annum. Each indicator contributed meaningfully to the variation in scores and centers tended to perform consistently across most indicators.

CONCLUSION

The combination of multiple indicators across dimensions of quality sets a higher standard for performance evaluation and allows the discrimination of centers based on structural elements, specifically level 1 status, and trauma center volume.

LEVEL OF EVIDENCE

Therapeutic /Care Management; Level IV.

Multi-level factors associated with utilization of water, sanitation and hygiene services by mothers in Nepal

BACKGROUND

Providing improved water, sanitation, and hygiene (WASH) at a household level remains one of the major public health challenges in Nepal. Household mothers are likely to have limited access to combined WASH services, this is influenced by individual, and community factors. Individual components of an improved water source, sanitary toilet, fixed place for handwashing, and availability of soap and water were merged into one and called combined WASH. This paper aimed to identify the individual and community factors associated with combined WASH facilities and practices among mothers with children under five years in Nepal.

METHODS

A cross-sectional study was conducted using data from the Nepal Demographic and Health Survey (NDHS), 2016. The weighted sample size of this study was 4887 mothers with children under five years. The independent variables within the mothers included age, education, occupation, and caste/ethnicity in addition to education of the husband, wealth index, exposure to the newspaper, radio and television, residence, ecological zones, provinces, distance and participation in health mother groups were analyzed with the outcome variable of combined WASH. A multi-level mixed effects logistic regression model was used to assess the relationship of explanatory variables with WASH.

RESULTS

At an individual level, a rich wealth index was positively associated with combined WASH (AOR = 6.29; 95%CI: 4.63-8.54). Higher education levels and exposure to television had higher odds of having combined WASH. At the community level, the hill zone, urban residence, and Sudurpashim Provinces were positively associated with combined WASH while Madesh and Karnali Provinces and distance to water source greater than 31 minutes were associated with lower access to combined WASH.

CONCLUSION

Educated and rich household have positive association with combined WASH. It is recommended that both the health and other sectors may be instrumental in improving the combined WASH service for mothers at households.

Evidence for a role of Anopheles stephensi in the spread of drug- and diagnosis-resistant malaria in Africa

Anopheles stephensi, an Asian malaria vector, continues to expand across Africa. The vector is now firmly established in urban settings in the Horn of Africa. Its presence in areas where malaria resurged suggested a possible role in causing malaria outbreaks. Here, using a prospective case-control design, we investigated the role of An. stephensi in transmission following a malaria outbreak in Dire Dawa, Ethiopia in April-July 2022. Screening contacts of patients with malaria and febrile controls revealed spatial clustering of Plasmodium falciparum infections around patients with malaria in strong association with the presence of An. stephensi in the household vicinity. Plasmodium sporozoites were detected in these mosquitoes. This outbreak involved clonal propagation of parasites with molecular signatures of artemisinin and diagnostic resistance. To our knowledge, this study provides the strongest evidence so far for a role of An. stephensi in driving an urban malaria outbreak in Africa, highlighting the major public health threat posed by this fast-spreading mosquito.

Huangqin Qingre Chubi Capsule is Associated with Reduced Risk of Readmission in Patients with Rheumatoid Arthritis: A Real-World Retrospective Cohort Study

Purpose

The therapeutic effects of Huangqin Qingre Chubi (HQC) in rheumatoid arthritis (RA) have been documented. However, there is a lack of real-world clinical evidence supporting its efficacy.

Methods

Patients diagnosed with RA were recruited from the First Affiliated Hospital of the Anhui University of Chinese Medicine. Patient information was obtained from the hospital's database. Propensity score matching (PSM), Kaplan-Meier curve, and Cox proportional hazards model were used to control confounding factors and analyze the factors influencing readmission. Association rule analysis and random walk evaluation models were used to evaluate the correlations among HQC treatment, inflammation indicators, and self-perception of patients (SPP) scale.

Results

After PSM, 3423 patients were enrolled, with 1142 in the HQC group and 2281 in the non-HQC group. The readmission risk of the HQC group was significantly lower than that of the non-HQC group. Combined univariate and multivariate analysis results revealed that risk factors for readmission were age >60 years, female sex, hypertension, chronic gastritis, and elevated levels of laboratory indices, including anticyclic citrullinated peptide and complement component 3 (C3) and C4. HQC, disease-modifying antirheumatic drugs, nonsteroidal anti-inflammatory drugs, and glucocorticoid therapy were protective factors for readmission. HQC treatment was closely associated with improvements in many factors, including erythrocyte sedimentation rate, C-reactive protein, C3, rheumatoid factor levels, visual analog scale, depression self-assessment scale, and patient-reported activity index scores with RA.

Conclusion

HQC treatment can reduce the risk of readmission and significantly improve immune inflammatory indicators and SPP in patients with RA, with no risk of hepatorenal toxicity.

The influence of multidisciplinary team meetings on treatment decisions in advanced bladder cancer

OBJECTIVES

To investigate the role of specialised genitourinary multidisciplinary team meetings (MDTMs) in decision-making and identify factors that influence the probability of receiving a treatment plan with curative intent for patients with muscle invasive bladder cancer (MIBC).

PATIENTS AND METHODS

Data relating to patients with cT2-4aN0/X-1 M0 urothelial cell carcinoma, diagnosed between November 2017 and October 2019, were selected from the nationwide, population-based Netherlands Cancer Registry ('BlaZIB study'). Curative treatment options were defined as radical cystectomy (RC) with or without neoadjuvant chemotherapy, chemoradiation or brachytherapy. Multilevel logistic regression analyses were used to examine the association between MDTM factors and curative treatment advice and how this advice was followed.

RESULTS

Of the 2321 patients, 2048 (88.2%) were discussed in a genitourinary MDTM. Advanced age (>80 years) and poorer World Health Organization performance status (score 1-2 vs 0) were associated with no discussion (P < 0.001). Being discussed was associated with undergoing treatment with curative intent (odds ratio [OR] 3.0, 95% confidence interval [CI] 1.9-4.9), as was the involvement of a RC hospital (OR 1.70, 95% CI 1.09-2.65). Involvement of an academic centre was associated with higher rates of bladder-sparing treatment (OR 2.05, 95% CI 1.31-3.21). Patient preference was the main reason for non-adherence to treatment advice.

CONCLUSIONS

For patients with MIBC, the probability of being discussed in a MDTM was associated with age, performance status and receiving treatment with curative intent, especially if a representative of a RC hospital was present. Future studies should focus on the impact of MDTM advice on survival data.

A survey of the knowledge, perceptions of and attitudes to digital health of healthcare professionals in 14 Bulgarian hospitals: First large-scale study on digital health in Bulgarian inpatient facilities

Objective

To explore the knowledge, perceptions of and attitudes to digital health of Bulgarian hospital professionals in the first study of digital health in this professional group.

Methods

A paper-based questionnaire was administered to doctors, trainee doctors, nurses, midwives, and laboratory assistants working in multiprofile or specialized hospitals. Topics included the following: state, objectives, benefits, and future of digital health; data storage, access, security, and sharing; main software used; patient-held Personal Information System (PIS); and telemedicine. A total of 1187 participants from 14 hospitals completed the survey in two phases: September 2013-April 2014 and May 2015-April 2017. Data were analyzed through descriptive statistics and multilevel logistic regression.

Results

Three-quarters of participants evaluated the state of development of digital health in Bulgaria as subpar (36.0% negative; 38.9% passable; 24.5% positive). 27.2% (323) endorsed patients having unconditional access to their data. In contrast, 89.5% (1062) of participants considered it appropriate to have full access to patient data recorded by colleagues. Doctors were more likely to endorse patients having access to their data than healthcare specialists (OR = 1.79 at facility, OR = 1.77 at location).

Conclusion

The largely negative or lukewarm attitudes toward the state of development of digital health in Bulgaria are likely to result from the high number of failed projects, unmet expectations, misunderstood benefits, and unforeseen challenges. This study provides a much-needed stimulus and baseline for researching the ways in which the digital health landscape in Bulgaria has matured-or not.

Advancing translational research in neuroscience through multi-task learning

Translational research in neuroscience is increasingly focusing on the analysis of multi-modal data, in order to account for the biological complexity of suspected disease mechanisms. Recent advances in machine learning have the potential to substantially advance such translational research through the simultaneous analysis of different data modalities. This review focuses on one of such approaches, the so-called "multi-task learning" (MTL), and describes its potential utility for multi-modal data analyses in neuroscience. We summarize the methodological development of MTL starting from conventional machine learning, and present several scenarios that appear particularly suitable for its application. For these scenarios, we highlight different types of MTL algorithms, discuss emerging technological adaptations, and provide a step-by-step guide for readers to apply the MTL approach in their own studies. With its ability to simultaneously analyze multiple data modalities, MTL may become an important element of the analytics repertoire used in future neuroscience research and beyond.

Impact of National Pandemic Lockdowns on Perceived Threat of Immigrants: A Natural Quasi-Experiment Across 23 Countries

Xenophobia and anti-immigrant attacks rose during the COVID-19 pandemic, yet this may not be solely due to the disease threat. According to theories of frustration and scapegoating, situational obstructions and deprivation can motivate prejudice against outgroups. Using a global natural quasi-experimental design, this study tests whether the restrictiveness of national lockdowns can explain higher individual-level perceptions of immigrant threat. Data of 45,894 participants from 23 countries were analyzed. Both lockdown duration and lockdown severity were positively associated with individuals’ perceived threat of immigrants. The lockdown effects were independent of objective and subjective measures of disease threat, and there was no evidence that disease threat drives people’s prejudice toward immigrants. Subgroup analysis suggested the lockdown effects were reliable in Europe and the Americas, but not in Asia. These findings suggest a need to mitigate frustration and scapegoating when implementing lockdowns, and to distinguish the influence of societal restrictions from disease threat.

Association Between Resident Race and Ethnicity and Clinical Performance Assessment Scores in Graduate Medical Education

PURPOSE

To assess the association between internal medicine (IM) residents' race/ethnicity and clinical performance assessments.

METHOD

The authors conducted a cross-sectional analysis of clinical performance assessment scores at 6 U.S. IM residency programs from 2016 to 2017. Residents underrepresented in medicine (URiM) were identified using self-reported race/ethnicity. Standardized scores were calculated for Accreditation Council for Graduate Medical Education core competencies. Cross-classified mixed-effects regression assessed the association between race/ethnicity and competency scores, adjusting for rotation time of year and setting; resident gender, postgraduate year, and IM In-Training Examination percentile rank; and faculty gender, rank, and specialty.

RESULTS

Data included 3,600 evaluations by 605 faculty of 703 residents, including 94 (13.4%) URiM residents. Resident race/ethnicity was associated with competency scores, with lower scores for URiM residents (difference in adjusted standardized scores between URiM and non-URiM residents, mean [standard error]) in medical knowledge (-0.123 [0.05], P = .021), systems-based practice (-0.179 [0.05], P = .005), practice-based learning and improvement (-0.112 [0.05], P = .032), professionalism (-0.116 [0.06], P = .036), and interpersonal and communication skills (-0.113 [0.06], P = .044). Translating this to a 1 to 5 scale in 0.5 increments, URiM resident ratings were 0.07 to 0.12 points lower than non-URiM resident ratings in these 5 competencies. The interaction with faculty gender was notable in professionalism (difference between URiM and non-URiM for men faculty -0.199 [0.06] vs women faculty -0.014 [0.07], P = .01) with men more than women faculty rating URiM residents lower than non-URiM residents. Using the 1 to 5 scale, men faculty rated URiM residents 0.13 points lower than non-URiM residents in professionalism.

CONCLUSIONS

Resident race/ethnicity was associated with assessment scores to the disadvantage of URiM residents. This may reflect bias in faculty assessment, effects of a noninclusive learning environment, or structural inequities in assessment.

Surgeon Thyroidectomy Case Volume Impacts Disease-free Survival in the Management of Thyroid Cancer

OBJECTIVES

To assess the association between surgeons thyroidectomy case volume and disease-free survival (DFS) for patients with well-differentiated thyroid cancer (WDTC). A secondary objective was to assess a surgeon volume cutoff to optimize outcomes in those with WDTC. We hypothesized that surgeon volume will be an important predictor of DFS in patients with WDTC after adjusting for hospital volume and sociodemographic and clinical factors.

METHODS

In this retrospective population-based cohort study, we identified WDTC patients in Ontario, Canada, who underwent thyroidectomy confirmed by both hospital-level and surgeon-level administrative data between 1993 and 2017 (N = 37,233). Surgeon and hospital volumes were calculated based on number of cases performed in the year prior by the physician and at an institution performing each case, respectively and divided into quartiles. A multilevel hierarchical Cox regression model was used to estimate the effect of volume on DFS.

RESULTS

A crude model without patient or treatment characteristics demonstrated that both higher surgeon volume quartiles (p < 0.001) and higher hospital volume quartiles (p < 0.001) were associated with DFS. After controlling for clustering and patient/treatment covariates and hospital volume, moderately low (18-39/year) and low (0-17/year) volume surgeons (hazard ratios [HR]: 1.23, 95% confidence interval [CI]: 1.09-1.39 and HR: 1.34, 95% CI: 1.17-1.53 respectively) remained an independent statistically significant negative predictor of DFS.

CONCLUSION

Both high-volume surgeons and hospitals are predictors of better DFS in patients with WDTC. DFS is higher among surgeons performing more than 40 thyroidectomies a year.

LEVEL OF EVIDENCE

3 Laryngoscope, 2022.

Long-Term Ultrasonic Benchmarking for Microstructure Characterization with Bayesian Updating

Ultrasonic non-destructive characterization is an appealing technique for identifying the microstructures of materials in place of destructive testing. However, the existing ultrasonic characterization techniques do not have sufficient long-term gage repeatability and reproducibility (GR&R), since benchmarking data are not updated. In this study, a hierarchical Bayesian regression model was utilized to provide a long-term ultrasonic benchmarking method for microstructure characterization, suitable for analyzing the impacts of experimental setups, human factors, and environmental factors on microstructure characterization. The priori distributions of regression parameters and hyperparameters of the hierarchical model were assumed and the Hamilton Monte Carlo (HMC) algorithm was used to calculate the posterior distributions. Characterizing the nodularity of cast iron was used as an example, and the benchmarking experiments were conducted over a 13-week transition period. The results show that updating a hierarchical model can increase its performance and robustness. The outcome of this study is expected to pave the way for the industrial uptake of ultrasonic microstructure characterization techniques by organizing a gradual transition from destructive sampling inspection to non-destructive one-hundred-percent inspection.

Clinical Outcomes, Health Care Costs and Prognostic Factors for Total Knee Arthroplasty: A Multilevel Analysis of a National Cohort Study Using Administrative Claims Data

Total knee arthroplasty (TKA) is one of the most frequent orthopaedic surgeries. The purpose of this study was to determine the 30-day postoperative mortality rate, total episode-of-care costs, and prognostic factors associated with these outcomes, for adult patients who underwent TKA in Colombia's contributory health care system. A retrospective cohort study of all adult patients enrolled in Colombia's contributory health system, who underwent TKA between January 1, 2012 and November 30, 2015 was performed. Thirty-day postoperative mortality rates, 30-day ICU admissions rates, 30-day hospital readmission rates, 1-year arthroplasty revision rates, and total episode-of-care costs were estimated. Multilevel, generalized linear models were generated, to determine the prognostic factors associated with outcomes presented. A total of 12,453 patients were included. The 30-day mortality rate was 0.13 per 100 surgeries and the ICU admissions rate at 30 days postoperative was 4.44 per 100 surgeries. The 30-day hospital readmission rate was 4.28 per 100 surgeries and the 1-year arthroplasty revision rate was 1.22 per 100 surgeries. The prognostic factors associated with mortality were age, Charlson Index, and type of insurer. The prognostic factors associated with hospital readmission were age category, Charlson Index, and geographic region; younger age and higher Charlson Index were found to be associated with a higher 1-year arthroplasty revision rate. The median of total episode-of-care costs was USD$ 6,190.07 (interquartile range: 2,299-7,282). The multivariate model found that age, the Charlson Index, the Atlantic region, and type of insurer were associated with the costs incurred by the health system. For patients undergoing TKA in Colombia, age, the Charlson Index, insurers, and geographic region are associated with mortality, ICU admissions, 30-day hospital readmissions, 1-year arthroplasty revisions, and total costs incurred by the health system.

Health insurance coverage and timely antenatal care attendance in sub-Saharan Africa

Introduction

Out-of-pocket payments for healthcare remain a significant health financing challenge in sub-Saharan Africa (SSA), preventing women from using maternal health services. There is a paucity of empirical literature on the influence of health insurance coverage on the timeliness of antenatal care (ANC) attendance in low- and middle-income countries. In this study, we examined the association between health insurance coverage and timely ANC attendance among pregnant women in SSA.

Methods

Secondary data from Demographic and Health Surveys conducted between 2015 and 2020 in sixteen (16) sub-Saharan African countries with 113,918 women aged 15-49 years were included in the analysis. The outcome variable was the timing of antenatal care (ANC). A multilevel binary logistic regression analysis was carried out to determine the association between health insurance coverage and timely ANC.

Results

The overall coverage of health insurance and timely antenatal attendance among pregnant women in SSA were 4.4% and 39.0% respectively. At the country level, the highest coverage of health insurance was found in Burundi (24.3%) and the lowest was in Benin (0.9%). For timely ANC attendance, the highest prevalence was in Liberia (72.4%) and the lowest was in Nigeria (24.2%). The results in the model showed that women who were covered by health insurance were more likely to have timely ANC attendance compared to those who were not covered by health insurance (aOR = 1.21, 95% CI = 1.11-1.31).

Conclusion

Our findings show that that being covered under health insurance is associated with higher likelihood of seeking timely ANC attendance. To accelerate progress towards achievement of the Sustainable Development Goal targets by the year 2030, we recommend that governments and health insurance authorities across the sub-Saharan African countries actively implement health insurance policies as well as roll out health educational programmes that facilitate and ensure increased coverage of health insurance.

Evaluation of a Sexual Transmitted Infection Prevention Program Among University Students in Beira City Central Mozambique: A Study Protocol

Background: Unhealthy sexual behaviors, such as unprotected sexual intercourse and lack of using screening services increase cyclical transmission of sexually transmitted infections including Human Immunodeficiency Virus (HIV), especially among young adults. Hence health promotion programs can contribute to reduce the consequences, by changing (determinants of) these behaviors. Such interventions need to embrace a comprehensive approach and apply theory-and evidence-based methods. This article describes the protocol for a process and effect evaluation study of a sexually transmitted infection prevention program among university students in Beira city, central Mozambique.

Methods: The on-going program at Universidade Católica de Moçambique is described following the six steps of Intervention Mapping (IM), with a focus on the evaluation plan (i.e., the final step in IM). The details regarding previous steps in the protocol are briefly described as well, as they lay the foundation for the final step. The overall study will apply a hybrid type 1 approach by assessing the effectiveness of the intervention while gettering implementation. The process evaluation will apply qualitative and quantitative methods to gain insight in the context, reach, dose delivered, dose received and recruitment. Interviews with closed and open-ended questions will be conducted with program implementers and users. A quasi-experimental non-equivalent control group design is used to evaluate the effectiveness. A cohort of university students will be followed for 6 months. Self-administrated questionnaires will be used to collect data every 3 months.

Discussion: A combination of process and effect evaluation is proposed. This is a useful and fruitful procedure, since concurrent process evaluation can allow researchers to better interpret findings from the effect evaluation and understand how the intervention might replicate in similar contexts. We decided to follow the IM approach since, it is a theory-and evidence-based, systematic and detailed guide regarding what to do at every steps. A quasi-experimental non-equivalent control group design was chosen to fit the context of the study and generate outcomes with high external validity.

Study Registration: 004/CIBS/2020.

The effect of short term exposure to outdoor air pollution on fertility

BACKGROUND

There is evidence to suggest that long term exposure to air pollution could be associated with decreased levels of fertility, although there is controversy as to how short term exposure may compromise fertility in IVF patients and what windows of exposure during the IVF process patients could be most vulnerable.

METHODS

This prospective cohort study aimed to evaluate the impact of acute exposure that air pollution have on reproductive outcomes in different moments of the IVF process. Women undergoing IVF living in Barcelona were recruited. Individual air pollution exposures were modelled at their home address 15 and 3 days before embryo transfer (15D and 3D, respectively), the same day of transfer (D0), and 7 days after (D7). The pollutants modelled were: PM2.5 [particulate matter (PM) ≤2.5 μm], PMcoarse (PM between 2.5 and 10μm), PM10 (PM≤10 μm), PM2.5 abs, and NO2 and NOx. Outcomes were analyzed using multi-level regression models, with adjustment for co-pollutants and confouding factors. Two sensitivity analyses were performed. First, the model was adjusted for subacute exposure (received 15 days before ET). The second analysis was based on the first transfer performed on each patient aiming to exclude patients who failed previous transfers.

RESULTS

One hundred ninety-four women were recruited, contributing with data for 486 embryo transfers. Acute and subacute exposure to PMs showed a tendency in increasing miscarriage rate and reducing clinical pregnancy rate, although results were not statistically significant. The first sensitivity analysis, showed a significant risk of miscarriage for PM2.5 exposure on 3D after adjusting for subacute exposure, and an increased risk of achieving no pregnancy for PM2.5, PMcoarse and PM10 on 3D. The second sensitivity analysis showed a significant risk of miscarriage for PM2.5 exposure on 3D, and a significant risk of achieving no pregnancy for PM2.5, PMcoarse and PM10 particularly on 3D. No association was observed for nitrogen dioxides on reproductive outcomes.

CONCLUSIONS

Exposure to particulate matter has a negative impact on reproductive outcomes in IVF patients. Subacute exposure seems to increase the harmful effect of the acute exposure on miscarriage and pregnancy rates. Nitrogen dioxides do not modify significantly the reproductive success.

Passport to a Mighty Nation: Exploring Sociocultural Foundation of Chinese Public's Attitude to COVID-19 Vaccine Certificates

Vaccination against COVID-19 is essential against the pandemic. There are broad discussions on adopting certificates for vaccination and the immunity obtained after infection. Based on a national sample of over 2000 participants administered in April 2021, the current study examines the Chinese public's attitudes to the so-called COVID-19 vaccination passport and factors contributing to their viewpoints. Generally, the Chinese people had favorable opinions on the passport. Among possible contributing factors, income, personal benefit perception, the subjective norm of COVID-19 vaccination, and nationalism were significantly associated with the public's positive attitude. At the same time, general vaccine knowledge and scientific literacy had an inconstant effect. Echoing recent studies, these findings reveal a collectivism-oriented attitude of the Chinese public towards the proposal to certify vaccination publicly. Theoretical and practical implications of the results were discussed.

Modeling Energy Expenditure Estimation in Occupational Context by Actigraphy: A Multi Regression Mixed-Effects Model

The accurate prediction of energy requirements for healthy individuals has many useful applications. The occupational perspective has also been proven to be of great utility for improving workers' ergonomics, safety, and health. This work proposes a statistical regression model based on actigraphy and personal characteristics to estimate energy expenditure and cross-validate the results with reference standardized methods. The model was developed by hierarchical mixed-effects regression modeling based on the multitask protocol data. Measurements combined actigraphy, indirect calorimetry, and other personal and lifestyle information from healthy individuals (n = 50) within the age of 29.8 ± 5 years old. Results showed a significant influence of the variables related to movements, heart rate and anthropometric variables of body composition for energy expenditure estimation. Overall, the proposed model showed good agreement with energy expenditure measured by indirect calorimetry and evidenced a better performance than the methods presented in the international guidelines for metabolic rate assessment proving to be a reliable alternative to normative guidelines. Furthermore, a statistically significant relationship was found between daily activity and energy expenditure, which raised the possibility of further studies including other variables, namely those related to the subject's lifestyle.

Variation in end-of-life care and hospital palliative care among hospitals and local authorities: A preliminary contribution of big data

BACKGROUND

Many studies explore the clinical and ethical dimensions of care at the end-of-life, but fewer use administrative data to examine individual and geographic differences, including the use of palliative care.

AIM

Provide a population-based perspective on end-of-life and hospital palliative care among local authorities and hospitals in France.

DESIGN

Retrospective cohort study of care received by 17,928 decedents 65 and over (last 6 months of life), using the French national health insurance database.

RESULTS

55.7% of decedents died in acute-care hospitals; 79% were hospitalized in them at least once; 11.7% were admitted at least once for hospital palliative care. Among 31 academic medical centers, intensive care unit admissions ranged from 12% to 67.4%; hospital palliative care admissions, from 2% to 30.6%. Across local authorities, for intensive care unit days and hospital palliative care admissions, the ratios between the values at the third and the first quartile were 2.4 and 1.5. The odds of admission for hospital palliative care or to an intensive care unit for more than 7 days were more than twice as high among people ⩽85 years (aOR = 2.11 (1.84-2.43) and aOR = 2.59 (2.12-3.17), respectively). The odds of admission for hospital palliative care were about 25% lower (p = 0.04) among decedents living in local authorities with the lowest levels of education than those with the highest levels.

CONCLUSION

The variation we document in end-of-life and hospital palliative care across different categories of hospitals and 95 local authorities raises important questions as to what constitutes appropriate hospital use and intensity at the end-of-life.

Consequences of ignoring clustering in linear regression

Background

Clustering of observations is a common phenomenon in epidemiological and clinical research. Previous studies have highlighted the importance of using multilevel analysis to account for such clustering, but in practice, methods ignoring clustering are often employed. We used simulated data to explore the circumstances in which failure to account for clustering in linear regression could lead to importantly erroneous conclusions.

Methods

We simulated data following the random-intercept model specification under different scenarios of clustering of a continuous outcome and a single continuous or binary explanatory variable. We fitted random-intercept (RI) and ordinary least squares (OLS) models and compared effect estimates with the “true” value that had been used in simulation. We also assessed the relative precision of effect estimates, and explored the extent to which coverage by 95% confidence intervals and Type I error rates were appropriate.

Results

We found that effect estimates from both types of regression model were on average unbiased. However, deviations from the “true” value were greater when the outcome variable was more clustered. For a continuous explanatory variable, they tended also to be greater for the OLS than the RI model, and when the explanatory variable was less clustered. The precision of effect estimates from the OLS model was overestimated when the explanatory variable varied more between than within clusters, and was somewhat underestimated when the explanatory variable was less clustered. The cluster-unadjusted model gave poor coverage rates by 95% confidence intervals and high Type I error rates when the explanatory variable was continuous. With a binary explanatory variable, coverage rates by 95% confidence intervals and Type I error rates deviated from nominal values when the outcome variable was more clustered, but the direction of the deviation varied according to the overall prevalence of the explanatory variable, and the extent to which it was clustered.

Conclusions

In this study we identified circumstances in which application of an OLS regression model to clustered data is more likely to mislead statistical inference. The potential for error is greatest when the explanatory variable is continuous, and the outcome variable more clustered (intraclass correlation coefficient is ≥ 0.01).

Supplementary Information

The online version contains supplementary material available at 10.1186/s12874-021-01333-7.

Hospital-specific probability of cystectomy affects survival from muscle-invasive bladder cancer

OBJECTIVES

Radical cystectomies (RCs) are increasingly centralized, but bladder cancer can be diagnosed in every hospital The aim of this study is to assess the variation between hospitals of diagnosis in a patient's chance to undergo a RC before and after the volume criteria for RCs, to identify factors associated with this variation and to assess its effect on survival.

METHODS AND MATERIALS

Patients diagnosed with muscle-invasive bladder cancer (cT2-4a,N0/X,M0/X) without nodal or distant metastases between 2008 and 2016 were identified through the Netherlands Cancer Registry. Multilevel logistic regression analysis was used to investigate the hospital specific probability of undergoing a cystectomy. Cox proportional hazard regression analysis was used to assess the case-mix adjusted effect of hospital-specific probabilities on survival.

RESULTS

Of the 9,215 included patients, 4,513 (49%) underwent a RC. The percentage of RCs varied between 7% and 83% by hospital of diagnosis before the introduction of the first volume criteria (i.e., 2008-2009; minimum of 10 RCs). This variation decreased slightly to 17%-77% after establishment of the second volume criteria (i.e., 2015-2016; minimum of 20 RCs). Age, cT-stage and comorbidity were inversely and socioeconomic status was positively associated with RC. Both being diagnosed in a community hospital and/or being diagnosed in a hospital fulfilling the RC volume criteria were associated with increased use of RC compared to academic hospitals and hospitals not fulfilling the volume criteria. For each 10% increase in the percentage of RC in the hospital of diagnosis, 2-year case-mix adjusted survival increased 4% (hazard ratio 0.96, 95% confidence interval 0.94-0.98).

CONCLUSION

Probability of RC varied between hospitals of diagnosis and affected 2-year overall survival. Undergoing a RC was associated with age, cT-stage, socioeconomic status, type of hospital, and whether the hospital of diagnosis fulfilled the RC volume criteria. Future research is needed to identify patient, tumor, and hospital characteristics affecting utilization of curative treatment as this may benefit overall survival.

Determining Associations and Estimating Effects with Regression Models in Clinical Anesthesia

There are an increasing number of "big data" studies in anesthesia that seek to answer clinical questions by observing the care and outcomes of many patients across a variety of care settings. This Readers' Toolbox will explain how to estimate the influence of patient factors on clinical outcome, addressing bias and confounding. One approach to limit the influence of confounding is to perform a clinical trial. When such a trial is infeasible, observational studies using robust regression techniques may be able to advance knowledge. Logistic regression is used when the outcome is binary (e.g., intracranial hemorrhage: yes or no), by modeling the natural log for the odds of an outcome. Because outcomes are influenced by many factors, we commonly use multivariable logistic regression to estimate the unique influence of each factor. From this tutorial, one should acquire a clearer understanding of how to perform and assess multivariable logistic regression.

Development and Validation of a Risk Score to Predict Low Birthweight Using Characteristics of the Mother: Analysis from BUNMAP Cohort in Ethiopia

At least one ultrasound is recommended to predict fetal growth restriction and low birthweight earlier in pregnancy. However, in low-income countries, imaging equipment and trained manpower are scarce. Hence, we developed and validated a model and risk score to predict low birthweight using maternal characteristics during pregnancy, for use in resource limited settings. We developed the model using a prospective cohort of 379 pregnant women in South Ethiopia. A stepwise multivariable analysis was done to develop the prediction model. To improve the clinical utility, we developed a simplified risk score to classify pregnant women at high- or low-risk of low birthweight. The accuracy of the model was evaluated using the area under the receiver operating characteristic curve (AUC) and calibration plot. All accuracy measures were internally validated using the bootstrapping technique. We evaluated the clinical impact of the model using a decision curve analysis across various threshold probabilities. Age at pregnancy, underweight, anemia, height, gravidity, and presence of comorbidity remained in the final multivariable prediction model. The AUC of the model was 0.83 (95% confidence interval: 0.78 to 0.88). The decision curve analysis indicated the model provides a higher net benefit across ranges of threshold probabilities. In general, this study showed the possibility of predicting low birthweight using maternal characteristics during pregnancy. The model could help to identify pregnant women at higher risk of having a low birthweight baby. This feasible prediction model would offer an opportunity to reduce obstetric-related complications, thus improving the overall maternal and child healthcare in low- and middle-income countries.

Variability in intensive care unit admission among pregnant and postpartum women in Canada: a nationwide population-based observational study

Background

Pregnancy-related critical illness results in approximately 300,000 deaths globally each year. The objective was to describe the variation in ICU admission and the contribution of patient- and hospital-based factors in ICU admission among acute care hospitals for pregnant and postpartum women in Canada.

Methods

A nationwide cohort study between 2004 and 2015, comprising all pregnant or postpartum women admitted to Canadian hospitals. The primary outcome was ICU admission. Secondary outcomes were severe maternal morbidity (a potentially life-threatening condition) and maternal death (during and within 6 weeks after pregnancy). The proportion of total variability in ICU admission rates due to the differences among hospitals was described using the median odds ratio from multi-level logistic regression models, adjusting for individual hospital clusters.

Results

There were 3,157,248 identifiable pregnancies among women admitted to 342 Canadian hospitals. The overall ICU admission rate was 3.2 per 1000 pregnancies. The rate of severe maternal morbidity was 15.8 per 1000 pregnancies, of which 10% of women were admitted to an ICU. The most common severe maternal morbidity events included postpartum hemorrhage (n = 16,364, 0.52%) and sepsis (n = 11,557, 0.37%). Of the 195 maternal deaths (6.2 per 100,000 pregnancies), only 130 (67%) were admitted to ICUs. Patients dying in hospital, without admission to ICU, included those with cardiovascular compromise, hemorrhage, and sepsis. For 2 pregnant women with similar characteristics at different hospitals, the average (median) odds of being admitted to ICU was 1.92 in 1 hospital compared to another. Hospitals admitting the fewest number of pregnant patients had the highest incidence of severe maternal morbidity and mortality. Patient-level factors associated with ICU admission were maternal comorbidity index (OR 1.88 per 1 unit increase, 95%CI 1.86–1.99), urban residence (OR 1.09, 95%CI 1.02–1.16), and residing at the lowest income quintile (OR 1.44, 95%CI 1.34–1.55).

Conclusions

Most women who experience severe maternal morbidity are not admitted to an ICU. There exists a wide hospital-level variability in ICU admission, with patients living in urban locations and patients of lowest income levels most likely to be admitted to ICU. Cardiovascular compromise, hemorrhage, and sepsis represent an opportunity for improved patient care and outcomes.

Association of Maternal Age With Severe Maternal Morbidity and Mortality in Canada

IMPORTANCE

Over the past 2 decades, there has been a trend toward increasing maternal age in many high-income countries. Maternal age may lead to greater attendant morbidity and mortality for Canadian mothers.

OBJECTIVE

To investigate the association of maternal age, adjusting for patient-level and hospital-level factors, with severe maternal morbidity (SMM) and maternal death in Canada.

DESIGN, SETTING, AND PARTICIPANTS

A nationwide population-based cohort study of all antepartum, peripartum, and postpartum women and adolescents seen at Canadian acute care hospitals from April 1, 2004, to March 31, 2015. All analyses were completed on September 13, 2018.

EXPOSURES

Maternal age at the index delivery.

MAIN OUTCOMES AND MEASURES

Severe maternal morbidity and maternal death during pregnancy and within 6 weeks after termination of pregnancy.

RESULTS

During the study period, there were 3 162 303 new pregnancies (mean [SD] maternal age, 29.5 [5.6] years) and 3 533 259 related hospital admissions. There were 54 219 episodes of SMM (17.7 cases per 1000 deliveries) in the entire study period, with a 9.8% relative increase from 2004-2005 to 2014-2015, in addition to an increasing proportion of pregnancies to older mothers. Independent patient-level factors associated with SMM included increasing Maternal Comorbidity Index; maternal age 19 years or younger and 30 years or older, with the greatest risk experienced by women 45 years or older (odds ratio [OR], 2.69; 95% CI, 2.34-3.06 compared with maternal age 20-24 years); and lowest income quintile (OR, 1.19; 95% CI, 1.14-1.22 compared with highest income quintile). Hospital-level factors associated with SMM included specific provinces. Independent patient-level factors associated with maternal mortality included increasing Maternal Comorbidity Index, age 40 to 44 years (OR, 3.39; 95% CI, 1.68-6.82 compared with age 20-24 years), age 45 years or older (OR, 4.39; 95% CI, 1.01-19.10 compared with age 20-24 years), and lowest income quintile (OR, 4.14; 95% CI, 2.03-8.50 compared with highest income quintile). Hospital-level factors associated with maternal mortality included lowest hospital pregnancy volume.

CONCLUSIONS AND RELEVANCE

In Canada, maternal age and SMM have increased over the past decade. Results of this study suggest that province of residence, maternal comorbidity, residence income quintile, and extremes of maternal age, especially those 45 years or older, were associated with SMM and mortality. These findings are relevant to prospective parents, their health care team, and public health planning.

Association of maternal disease and medication use with the risk of congenital heart defects in offspring: a case-control study using logistic regression with a random-effects model

Objective To examine the association between maternal diseases and congenital heart defects (CHDs) and to evaluate whether those associations vary with corresponding medication use. Methods A multi-hospital case-control study conducted from February 2010 to December 2014 analysed 916 controls and 1236 cases. Participating mothers were asked whether they suffered from influenza, common cold, herpes and threatened abortion or had used corresponding medication during the periconception period or the early pregnancy period. We used a random-effects logistic regression model to compute the odds ratios (ORs), adjusted odds ratios (AORs) and 95% confidence intervals (CIs) while controlling for potential confounders. Results Compared with the results for mothers with no exposure, there were significant associations between maternal diseases with medication non-use and CHDs in the aggregate, including influenza (AOR, 1.83; 95% CI, 1.13-2.95), common cold (AOR, 2.05; 95% CI, 1.60-2.64) and herpes (AOR, 7.00; 95% CI, 2.15-22.84). There was no significant association between medication users and offspring with any subtype of CHDs, except that maternal common cold with medication use slightly increased the risk of the specific subtype, namely, isolated cardiac defects. However, an association was observed between maternal threatened abortion and medication and isolated cardiac defects (AOR, 1.33; 95% CI, 1.01-1.75). Conclusion Maternal influenza, common cold, herpes and threatened abortion from 3 months before pregnancy through the first trimester were associated with an increased risk of congenital heart disease in offspring. The teratogenic effect of these conditions may be attenuated by medication use, except for threatened abortion.

Growth scaling for the early dynamics of HIV/AIDS epidemics in Brazil and the influence of socio-demographic factors

The early dynamics of an infectious disease outbreak can be affected by various factors including the transmission mode of the disease and host-specific factors. While recent works have highlighted the presence of sub-exponential growth patterns during the early phase of epidemics, empirical studies examining the contribution of different factors to early epidemic growth dynamics are lacking. Here we aim to characterize and explain the early incidence growth patterns of local HIV/AIDS epidemics in Brazil as a function of socio-demographic factors. For this purpose, we accessed annual AIDS incidence series and state-level socio-demographic variables from publicly available databases. To characterize the early growth dynamics of the HIV/AIDS epidemic, we employed the generalized-growth model to estimate with quantified uncertainty the scaling of growth parameter (p) which captures growth patterns ranging from constant incidence (p=0) to sub-exponential (0 < p < 1) and exponential growth dynamics (p=1) at three spatial scales: national, regional, and state levels. We evaluated the relationship between socio-demographic variables and epidemic growth patterns across 27 Brazilian states using mixed-effect regression analyses. We found wide variation in the early dynamics of the AIDS epidemic in Brazil, displaying sub-exponential growth patterns with the p parameter estimated substantially below 1.0. The mean p was estimated to be 0.81 at the national level, with a range of 0.72-0.85 at the regional level, and a range of 0.28-0.96 at the state level. Our findings support the notion that socio-demographic factors contribute to shaping the early growth dynamics of the epidemic at the local level. Gini index and socio-demographic index were negatively associated with the parameter p, whereas urbanicity was positively associated with p_. The results could have theoretical significance in understanding differences in growth scaling across different sexually transmitted disease systems, and have public health implications to guide control.

Centralization of ovarian cancer in the Netherlands: Hospital of diagnosis no longer determines patients' probability of undergoing surgery

OBJECTIVE

Surgical care for advanced stage epithelial ovarian cancer (EOC) patients has been centralized in the Netherlands since 2012. We evaluated whether the likelihood for patients to undergo surgery depends on the hospital of initial diagnosis before and after centralization of surgical care.

METHODS

Patients with EOC FIGO stage IIB-IV, diagnosed in the Netherlands between 2000 and 2015, were identified from the Netherlands Cancer Registry. Multilevel multivariate logistic regression was used to study the association between hospital of diagnosis and patients' likelihood of undergoing surgery in subsequent time periods. Furthermore, changes in overall survival were analyzed by multivariable Cox regression models.

RESULTS

15,314 EOC patients were selected from the NCR. Hospital of diagnosis was identified as a significant level for patients' likelihood of undergoing surgery in 2000-2005 (LR test p<0.001), as well as in 2006-2011 (LR test p=0.002) but not in 2012-2015 (LR test p=0.127). Patients who underwent surgery in 2012-2015 had a better survival when compared to 2006-2011 (HR 0.90(0.84-0.96)).

CONCLUSION

This study shows that centralization of surgical care resolved the variation between hospitals in the probability to undergo cytoreductive surgery for patients with advanced EOC. Since centralization was established in 2012, the decision to operate patients seems solely attributable to patient and tumor characteristics. This supports the growing evidence in favor of centralizing (surgical) treatment for complex and heterogeneous diseases such as EOC.

Associations between the workplace-effort in psychosocial risk management and the employee-rating of the psychosocial work environment - a multilevel study of 7565 employees in 1013 workplaces

Aims: This study examined the association between the workplace-effort in psychosocial risk management and later employee-rating of the psychosocial work environment. Method: The study is based on data from two questionnaire surveys – one including 1013 workplaces and one including 7565 employees from these workplaces. The association was analyzed using multi-level linear regression. The association for five different trade-groups and for five different psychosocial work environment domains was examined. Results: Limited but statistically significant better employee-ratings of the psychosocial work environment in the respective domains were observed among Danish workplaces that prioritized “development possibilities for employees,” “recognition of employees,” “employees influence on own work tasks,” good “communication at the workplace,” and “help to prevent work overload.” Conclusion: Danish workplaces with a high effort in psychosocial risk management in the preceding year had a small but significantly more positive rating of the psychosocial work environment by the employees. However, future studies are needed to establish the causality of the associations.

The Use of Fixed-and Random-Effects Models for Classifying Hospitals as Mortality Outliers: A Monte Carlo Assessment

Background. There is an increasing movement towards the release of hospital “report-cards.” However, there is a paucity of research into the abilities of the different methods to correctly classify hospitals as performance outliers.Objective.To examine the ability of risk-adjusted mortality rates computed using conventional logistic regression and random-effects logistic regression models to correctly identify hospitals that have higher than acceptable mortality.Research Design.Monte Carlo simulations.Measures.Sensitivity, specificity, and positive predictive value of a classification as a high-outlier for identifying hospitals with higher than acceptable mortality rates.Results.When the distribution of hospital-specific log-odds of death was normal, random-effects models had greater specificity and positive predictive value than fixed-effects models. However, fixed-effects models had greater sensitivity than random-effects models.Conclusions.Researchers and policy makers need to carefully consider the balance between false positives and false negatives when choosing statistical models for determining which hospitals have higher than acceptablemortality in performance profiling.

Abuse of Older Men in Seven European Countries: A Multilevel Approach in the Framework of an Ecological Model

BACKGROUND

Several studies on elder abuse indicate that a large number of victims are women, but others report that men in later life are also significantly abused, especially when they show symptoms of disability and poor health, and require help for their daily activities as a result. This study focused on the prevalence of different types of abuse experienced by men and on a comparison of male victims and non-victims concerning demographic/socio-economic characteristics, lifestyle/health variables, social support and quality of life. Additionally, the study identified factors associated with different types of abuse experienced by men and characteristics associated with the victims.

METHODS

The cross-sectional data concerning abuse in the past 12 months were collected by means of interviews and self-response during January-July 2009, from a sample of 4,467 not demented individuals aged between 60-84 years living in seven European countries (Germany, Greece, Italy, Lithuania, Portugal, Spain and Sweden). We used a multilevel approach, within the framework of an Ecological Model, to explore the phenomenon of abuse against males as the complex result of factors from multiple levels: individual, relational, community and societal.

RESULTS

Multivariate analyses showed that older men educated to higher levels, blue-collar workers and men living in a rented accommodation were more often victims than those educated to lower levels, low-rank white-collar workers and home owners, respectively. In addition, high scores for factors such as somatic and anxiety symptoms seemed linked with an increased probability of being abused. Conversely, factors such as increased age, worries about daily expenses (financial strain) and greater social support seemed linked with a decreased probability of being abused.

CONCLUSIONS

Male elder abuse is under-recognized, under-detected and under-reported, mainly due to the vulnerability of older men and to social/cultural norms supporting traditional male characteristics of stoicism and strength. Further specific research on the topic is necessary in the light of the present findings. Such research should focus, in particular, on societal/community aspects, as well as individual and family ones, as allowed by the framework of the Ecological Model, which in turn could represent a useful method also for developing prevention strategies for elder abuse.

Concordance of performance metrics among U.S. trauma centers caring for injured children

BACKGROUND

Several indicators of quality pediatric trauma care have been proposed including low in-hospital mortality, nonoperative management of blunt splenic injury, use of intracranial pressure monitors after severe traumatic brain injury, and craniotomy for children with severe subdural or epidural hematomas. It is not known if center-level performance is consistent in each of these metrics. We evaluated whether center performance in one area of quality predicted similar performance in other areas of quality.

METHODS

We reviewed patients 18 years or younger who were hospitalized with an injury Abbreviated Injury Scale (AIS) score of 2 or greater from 2010 to 2011 at trauma centers (n = 150) participating in the Trauma Quality Improvement Program. Random-intercept multilevel modeling was used to generate center-specific adjusted odds ratios for each quality indicator. We evaluated correlations between center-specific adjusted odds ratios of each quality indicator and mortality using Pearson correlation coefficients. Weighted κ statistics were used to test multiple pairwise agreements between indicators and the overall agreement across all four indicators.

RESULTS

Among 84,880 children identified for analysis, 3,603 had blunt splenic injury, 3,503 had severe traumatic brain injury, and 1,286 had an epidural or subdural hematoma. A negative correlation between center-specific odds of mortality and craniotomy was present (Pearson correlation coefficient, -0.18; p = 0.03). There were no significant correlations between other indicators. Although κ statistics showed slight agreement for the pairwise comparison of odds of mortality and craniotomy (0.17, 0.02-0.32), there was no agreement for all other pairwise comparisons or the overall comparison of all four indicators (-0.01, -0.07 to 0.06).

CONCLUSION

Our findings demonstrate a lack of concordance in center-level performance across the four pediatric trauma quality indicators we evaluated. These findings should be considered by pediatric trauma quality improvement initiatives to allow for comprehensive measurement of hospital quality as opposed to benchmarking using a single indicator.

Within- and between-group regression for improving the robustness of causal claims in cross-sectional analysis

BACKGROUND

A major objective of environmental epidemiology is to elucidate exposure-health outcome associations. To increase the variance of observed exposure concentrations, researchers recruit individuals from different geographic areas. The common analytical approach uses multilevel analysis to estimate individual-level associations adjusted for individual and area covariates. However, in cross-sectional data this approach does not differentiate between residual confounding at the individual level and at the area level. An approach allowing researchers to distinguish between within-group effects and between-group effects would improve the robustness of causal claims.

METHODS

We applied an extended multilevel approach to a large cross-sectional study aimed to elucidate the hypothesized link between drinking water pollution from perfluoroctanoic acid (PFOA) and plasma levels of C-reactive protein (CRP) or lymphocyte counts. Using within- and between-group regression of the individual PFOA serum concentrations, we partitioned the total effect into a within- and between-group effect by including the aggregated group average of the individual exposure concentrations as an additional predictor variable.

RESULTS

For both biomarkers, we observed a strong overall association with PFOA blood levels. However, for lymphocyte counts the extended multilevel approach revealed the absence of a between-group effect, suggesting that most of the observed total effect was due to individual level confounding. In contrast, for CRP we found consistent between- and within-group effects, which corroborates the causal claim for the association between PFOA blood levels and CRP.

CONCLUSION

Between- and within-group regression modelling augments cross-sectional analysis of epidemiological data by supporting the unmasking of non-causal associations arising from hidden confounding at different levels. In the application example presented in this paper, the approach suggested individual confounding as a probable explanation for the first observed association and strengthened the robustness of the causal claim for the second one.

Prevalence and factors associated with depressive symptoms among young women and men in the Eastern Cape Province, South Africa

PURPOSE

There is little research on prevalence of depressive symptoms and associated factors among youth in sub-Saharan Africa. This paper explores factors associated with depressive symptomatology in South Africa.

METHODS

A cross-sectional analysis of interviews with 1 415 women and 1 368 men aged 15-26 was undertaken. The Centre for Epidemiological Studies on Depression Scale (CESD Scale) was used to establish depressive symptomatology.

RESULTS

The prevalence of depressive symptoms was 20.5% in women and 13.5% in men. For women, depressive symptoms were associated with increased childhood adversity (aOR 1.34 95% CI 1.116, 1.55); drug use (aOR 1.98 CI 1.17, 3.35); experience of intimate partner violence (aOR 2.21 CI 1.16, 3.00); sexual violence before the age of 18 years (aOR 1.45 CI 1.02, 2.02) and lower perceptions of community cohesion (aOR 1.23 CI 1.07, 1.40). For men, depressive symptoms were associated with a mother's death (aOR 2.24 CI 1.25, 4.00); childhood adversity (aOR 1.61 CI 1.38, 1.88); alcohol abuse (aOR 1.63 CI 1.13, 2.35), sexual coercion by a woman (aOR 2.36 CI 1.47, 3.80) and relationship conflict (aOR 1.07 CI 1.01, 1.12).

CONCLUSIONS

Depressive symptoms were more highly prevalent in women than in men. Depressed mood was associated with childhood adversity, sexual violence and substance misuse in both women and men. This study further suggests gender differences in that for women, depressive symptoms were associated with intimate partner violence and lower perceptions of community cohesion, while for men the associations were with a mother's death and relationship conflict.

Epidemiology of congenital diaphragmatic hernia in Europe: a register-based study

INTRODUCTION

Published prevalence rates of congenital diaphragmatic hernia (CDH) vary. This study aims to describe the epidemiology of CDH using data from high-quality, population-based registers belonging to the European Surveillance of Congenital Anomalies (EUROCAT).

METHODS

Cases of CDH delivered between 1980 and 2009 notified to 31 EUROCAT registers formed the population-based case series. Prevalence over time was estimated using multilevel Poisson regression, and heterogeneity between registers was evaluated from the random component of the intercept.

RESULTS

There were 3373 CDH cases reported among 12 155 491 registered births. Of 3131 singleton cases, 353 (10.4%) were associated with a chromosomal anomaly, genetic syndrome or microdeletion, 784 (28.2%) were associated with other major structural anomalies. The male to female ratio of CDH cases overall was 1:0.69. Total prevalence was 2.3 (95% CI 2.2 to 2.4) per 10 000 births and 1.6 (95% CI 1.6 to 1.7) for isolated CDH cases. There was a small but significant increase (relative risk (per year)=1.01, 95% credible interval 1.00-1.01; p=0.030) in the prevalence of total CDH over time but there was no significant increase for isolated cases (ie, CDH cases that did not occur with any other congenital anomaly). There was significant variation in total and isolated CDH prevalence between registers. The proportion of cases that survived to 1 week was 69.3% (1392 cases) for total CDH cases and 72.7% (1107) for isolated cases.

CONCLUSIONS

This large population-based study found an increase in total CDH prevalence over time. CDH prevalence also varied significantly according to geographical location. No significant association was found with maternal age.

Physicians declining patient enrollment in a critical care trial: a case study in thromboprophylaxis

PURPOSE

To analyze the frequency, rationale and determinants of attending physicians requesting that their eligible patients not be approached for participation in a thromboprophylaxis trial.

METHODS

Research personnel in 67 centers prospectively documented eligible non-randomized patients due to physicians declining to allow their patients to be approached.

RESULTS

In 67 centers, 3,764 patients were enrolled, but 1,460 eligible patients had no consent encounter. For 218 (14.9 %) of these, attending physicians requested that their patients not be approached. The most common reasons included a high risk of bleeding (31.2 %) related to fear of heparin bioaccumulation in renal failure, the presence of an epidural catheter, peri-operative status or other factors; specific preferences for thromboprophylaxis (12.4 %); morbid obesity (9.6 %); uncertain prognosis (6.4 %); general discomfort with research (3.7 %) and unclear reasons (17.0 %). Physicians were more likely to decline when approached by less experienced research personnel; considering those with[10 years of experience as the reference category, the odds ratios (OR) for physician refusals to personnel without trial experience was 10.47 [95 % confidence interval (CI) 2.19-50.02] and those with less than 10 years experience was 1.72 (95 % CI 0.61-4.84). Physicians in open rather than closed units were more likely to decline (OR 4.26; 95 % CI 1.27-14.34). Refusals decreased each year of enrollment compared to the pilot phase.

CONCLUSIONS

Tracking, analyzing, interpreting and reporting the rates and reasons for physicians declining to allow their patients to be approached for enrollment provides insights into clinicians' concerns and attitudes to trials. This information can encourage physician communication and education, and potentially enhance efficient recruitment.

Hirschsprung's disease prevalence in Europe: a register based study

BACKGROUND

Hirschsprung's disease is a congenital gut motility disorder, characterised by the absence of the enteric ganglion cells along the distal gut. The aim of this study was to describe the epidemiology of Hirschsprung's disease, including additional congenital anomalies, total prevalence, trends, and association with maternal age.

METHODS

Cases of Hirschsprung's disease delivered during 1980 to 2009 notified to 31 European Surveillance of Congenital Anomaly registers formed the population-based case-series. Prevalence rates and 95% confidence intervals were calculated as the number of cases per 10,000 births. Multilevel Poisson regression was performed to investigate trends in prevalence, geographical variation and the association with maternal age.

RESULTS

There were 1,322 cases of Hirschsprung's disease among 12,146,210 births. The total prevalence was 1.09 (95% confidence interval, 1.03-1.15) per 10,000 births and there was a small but significant increase in prevalence over time (relative risk = 1.01; 95% credible interval, 1.00-1.02; p = 0.004). There was evidence of geographical heterogeneity in prevalence (p < 0.001). Excluding 146 (11.0%) cases with chromosomal anomalies or genetic syndromes, there were 1,176 cases (prevalence = 0.97; 95% confidence interval, 0.91-1.03 per 10,000 births), of which 137 (11.6%) had major structural anomalies. There was no evidence of a significant increased risk of Hirschsprung's disease in cases born to women aged ≥35 years compared with those aged 25 to 29 (relative risk = 1.09; 95% credible interval, 0.91-1.31; p = 0.355).

CONCLUSION

This large population-based study found evidence of a small increasing trend in Hirschsprung's disease and differences in prevalence by geographic location. There was also no evidence of an association with maternal age.