Bayesian Multilevel Logistic Regression Models Research Articles

베이지안 다수준 모형을 활용한 고령운전자 사고의 식별

This study examines various influential factors aimed at identifying crashes involving senior drivers. We consider variables reflecting the individual characteristics of drivers, crash types, roadway conditions, and environments, along with latent effects related to space and time, as potential identifiers of senior driver-related crashes in Daejeon. To this end, we employ Bayesian multilevel logistic regression models to unveil the random influences arising from spatial dependency and temporal autocorrelation after accounting for the fixed effects by explanatory variables. The result reveals that the overall annual trend, violations, and intersections appear to be positively associated with the log odds of the probability. Conversely, snow and icy roadway surface conditions, along with several crash types, are estimated to lower the log odds. The latent spatial effects dominate overall residual fluctuation, suggesting the benefits of our adopted model. We argue that our approach lays the foundations for designing safety policies for senior drivers.

Curb the spread: Leveraging intensive longitudinal data to unravel within- and between-person processes in adherence to nonpharmaceutical interventions during COVID-19.

Adhering to behavioral recommendations and nonpharmaceutical interventions (NPIs) is a key to control COVID-19 infection rates. However, rates have decreased globally, and potentially modifiable determinants of ongoing adherence and their interaction with social and physical momentary environments are still poorly understood. Here, we comprehensively examine within-person variations and between-person differences in known behavioral determinants (capability and motivation), as well as the moderating role of situational variable environmental factors (opportunity) in predicting adherence to hygiene and social distancing behaviors. Ecological momentary assessment study over 6 months with monthly assessment bouts (4 days each and five daily assessments) in 623 German adults. Repeated daily assessments of capability, opportunity, motivation, and behavior (COM-B) model factors. Bayesian multilevel logistic regression models were estimated to examine main effects of COM-B factors and moderating effects of momentary environmental factors. Momentary adherence to NPIs was predicted by within-person changes in COM-B factors (motivation: intentions, goal conflict, and control beliefs; opportunities: regulations and norms). Between-person differences in capabilities (habit strength) and motivation (intentions and control beliefs) predicted adherence across situations. Situation-specific environmental factors moderated the motivation-behavior association (regulation measures increased; goal conflict and nonadherent others decreased the association). Individual momentary (within-person) and stable (between-person) motivation indicators predicted adherence. However, situational environmental factors such as regulations or norms have strong main effects and moderate the motivation-behavior translation. These findings have policy implications, supporting recent claims to not rely on the narrative of "personal responsibility," but instead combine health education measures to increase individual motivation with consistent regulation. (PsycInfo Database Record (c) 2023 APA, all rights reserved).

Integrating the rights-based and capability approaches in the analysis of maternal healthcare utilization in sub-Saharan Africa: A multilevel modelling study.

The rights-based and capability approaches received increased attention relative to maternal health in the aftermath of the 2015 Millennium Development Goals (MDGs). This may be in view of the sub-optimal progress gained in reducing maternal and child mortality, especially in developing countries. Despite the combined potential of these approaches, there are limited empirical studies testing their viability in aiding our understanding of maternal healthcare utilization in developing countries. This is what this study sought to accomplish. We combined several datasets, including the Demographic Health Surveys (DHS), World Development Indicators, the World Governance Indicators and Freedom House. Bayesian multilevel logistic regression models were applied on three indicators of maternal healthcare utilization (antenatal care visits, institutional delivery, and postnatal check-ups) in relation to selected variables representing right-based and capability approaches. After controlling for relevant individual and community-level factors, the results show that living in countries with high freedom status (POR = 1.19) and higher female secondary school enrolments (POR = 1.54) increases the odds of adequate antenatal care. Residence in countries with high freedom status (POR = 1.33) and higher voice and accountability (POR = 1.72) has a positive influence on institutional delivery. Similar results are reported for postnatal care where country freedom status (POR = 1.89), voice and accountability (POR = 1.25) and female school enrolment (POR = 1.41) are significant predictors. The results imply that the rights-based and capability approaches have the potential to enhance maternal healthcare utilization in sub-Saharan Africa. Therefore, policy strategies emphasizing on freedoms, accountability, and individual capability functionings should be encouraged in the pursuit of partly achieving Sustainable Development Goals (SDG) number 3.

Socioeconomic disparities associated with mortality in patients hospitalized for COVID-19 in Colombia.

Socioeconomic disparities play an important role in the development of severe clinical outcomes including deaths from COVID-19. However, the current scientific evidence in regard the association between measures of poverty and COVID-19 mortality in hospitalized patients is scant. The objective of this study was to investigate whether there is an association between the Colombian Multidimensional Poverty Index (CMPI) and mortality from COVID-19 in hospitalized patients in Colombia from May 1, 2020 to August 15, 2021. This was an ecological study using individual data on hospitalized patients from the National Institute of Health of Colombia (INS), and municipal level data from the High-Cost Account and the National Administrative Department of Statistics. The main outcome variable was mortality due to COVID-19. The main exposure variable was the CMPI that ranges from 0 to 100% and was categorized into five levels: (i) level I (0%-20%), (ii) level II (20%-40%), (iii) level III (40%-60%), (iv) level IV (60%-80%); and (v) level V (80%-100%). The higher the level, the higher the level of multidimensional poverty. A Bayesian multilevel logistic regression model was applied to estimate Odds Ratio (OR) and their corresponding 95% credible intervals (CI). In addition, a subgroup analysis was performed according to the epidemiological COVID-19 waves using the same model. The odds for dying from COVID-19 was 1.46 (95% CI 1.4-1.53) for level II, 1.41 (95% CI 1.33-1.49) for level III and 1.70 (95% CI 1.54-1.89) for level IV hospitalized COVID-19 patients compared with the least poor patients (CMPI level I). In addition, age and male sex also increased mortality in COVID-19 hospitalized patients. Patients between 26 and 50 years-of-age had 4.17-fold increased odds (95% CI 4.07-4.3) of death compared with younger than 26-years-old patients. The corresponding for 51-75 years-old patients and those above the age of 75 years were 9.17 (95% CI 8.93-9.41) and 17.1 (95% CI 16.63-17.56), respectively. Finally, the odds of death from COVID-19 in hospitalized patients gradually decreased as the pandemic evolved. In conclusion, socioeconomic disparities were a major risk factor for mortality in patients hospitalized for COVID-19 in Colombia.

Bayesian multilevel logistic regression models: a case study applied to the results of two questionnaires administered to university students

Bayesian multilevel models—also known as hierarchical or mixed models—are used in situations in which the aim is to model the random effect of groups or levels. In this paper, we conduct a simulation study to compare the predictive ability of 1-level Bayesian multilevel logistic regression models with that of 2-level Bayesian multilevel logistic regression models by using the prior Scaled Beta2 and inverse-gamma distributions to model the standard deviation in the 2-level. Then, these models are employed to estimate the correct answers in two questionnaires administered to university students throughout the first academic semester of 2018. The results show that 2-level models have a better predictive ability and provide more precise probability intervals than 1-level models, particularly when the prior Scaled Beta2 distribution is used to model the standard deviation in the second level. Moreover, the probability intervals of 1-level Bayesian multilevel logistic regression models proved to be more precise when Scaled Beta2 distributions, rather than an inverse-gamma distribution, are employed to model the standard deviation or when 1-level Bayesian multilevel logistic regression models, are used.

Lack of ownership of mobile phones could hinder the rollout of mHealth interventions in Africa.

Mobile health (mHealth) interventions, which require ownership of mobile phones, are being investigated throughout Africa. We estimate the percentage of individuals who own mobile phones in 33 African countries, identify a relationship between ownership and proximity to a health clinic (HC), and quantify inequities in ownership. We investigate basic mobile phones (BPs) and smartphones (SPs): SPs can connect to the internet, BPs cannot. We use nationally representative data collected in 2017-2018 from 44,224 individuals in Round 7 of the Afrobarometer surveys. We use Bayesian multilevel logistic regression models for our analyses. We find 82% of individuals in 33 countries own mobile phones: 42% BPs and 40% SPs. Individuals who live close to an HC have higher odds of ownership than those who do not (aOR: 1.31, Bayesian 95% highest posterior density [HPD] region: 1.24-1.39). Men, compared with women, have over twice the odds of ownership (aOR: 2.37, 95% HPD region: 1.96-2.84). Urban residents, compared with rural residents, have almost three times the odds (aOR: 2.66, 95% HPD region: 2.22-3.18) and, amongst mobile phone owners, nearly three times the odds of owning an SP (aOR: 2.67, 95% HPD region: 2.33-3.10). Ownership increases with age, peaks in 26-40 year olds, then decreases. Individuals under 30 are more likely to own an SP than a BP, older individuals more likely to own a BP than an SP. Probability of ownership decreases with the Lived Poverty Index; however, some of the poorest individuals own SPs. If the digital devices needed for mHealth interventions are not equally available within the population (which we have found is the current situation), rolling out mHealth interventions in Africa is likely to propagate already existing inequities in access to healthcare.

Human papillomavirus vaccination disparities among U.S. college students: An intersectional Multilevel Analysis of Individual Heterogeneity and Discriminatory Accuracy (MAIHDA)

We investigated how gender identity, sexual orientation, and race/ethnicity intersect to shape the social epidemiology of HPV vaccination initiation among U.S. college students. Cross-sectional survey data were from the National College Health Assessment (Fall, 2019-Spring, 2020; N = 65,047). We conducted an intersectional Multilevel Analysis of Individual Heterogeneity and Discriminatory Accuracy by nesting participants within 36 social strata defined using gender identity, sexual orientation, and race/ethnicity. Bayesian multilevel logistic regression models with random intercepts for social strata were fit for HPV vaccination initiation. Intersectional models adjusted for the additive main effects to isolate intersectional interactions, controlling for age and geographic region. Social strata that included cisgender men, transgender women, and non-binary assigned-male-at-birth individuals and strata that included racial/ethnic minorities had a significantly lower likelihood of HPV vaccination initiation relative to strata including cisgender women and non-Hispanic White individuals, respectively, while strata including lesbian/gay and bisexual/pansexual/queer individuals had a significantly higher likelihood of HPV vaccination initiation relative to strata including heterosexual individuals. We also observed substantial between-stratum inequities in the predicted prevalence of HPV vaccination initiation, with estimates ranging from 59.2% for heterosexual, racial/ethnic minority, cisgender men to 87.1% for bisexual/pansexual/queer, racial/ethnic minority, non-binary assigned-female-at-birth individuals. That being said, the majority of the observed between-stratum variance was driven by additive rather than intersectional interaction effects and the discriminatory accuracy of intersectional stratification with respect to predicting HPV vaccination initiation was low. Collectively, our findings point to a need for more universal guidelines and clinician recommendations that promote HPV vaccine uptake for all adolescents, regardless of race/ethnicity, gender identity, sex-assigned-at-birth, or sexual orientation; however, utilizing an intersectional lens will ensure that resulting public health interventions address inequities and center the needs and experiences of multiply marginalized adolescents.

Use of troponins in the classification of myocardial infarction from electronic health records. The Atherosclerosis Risk in Communities (ARIC) Study

ObjectiveElectronic health record (EHR) data are underutilized for abstracting classification criteria for heart disease. We compared extraction of EHR data on troponin I and T levels with human abstraction. MethodsUsing EHR for hospitalizations identified through the Atherosclerosis Risk in Communities (ARIC) Study in four US hospitals, we compared blood levels of troponins I and T extracted from EHR structured data elements with levels obtained through data abstraction by human abstractors to 3 decimal places. Observations were divided randomly 50/50 into training and validation sets. Bayesian multilevel logistic regression models were used to estimate agreement by hospital in first and maximum troponin levels, troponin assessment date, troponin upper limit of normal (ULN), and classification of troponin levels as normal (< ULN), equivocal (1-2× ULN), abnormal (>2× ULN), or missing. ResultsEstimated overall agreement in first measured troponin level in the validation data was 88.2% (95% credible interval: 65.0%-97.5%) and 95.5% (91.2-98.2%) for the maximum troponin level observed during hospitalization. The largest variation in probability of agreement was for first troponin measured, which ranged from 66.4% to 95.8% among hospitals. ConclusionExtraction of maximum troponin values during a hospitalization from EHR structured data is feasible and accurate.

Spatial Distribution and Determinants of Nonautonomy on Decision Regarding Contraceptive Utilization among Married Reproductive-Age Women in Ethiopia: Spatial and Bayesian Multilevel Analysis.

Background Studies conducted to date in Ethiopia did not explore the spatial distribution, individual-level, and community-level factors affecting women's nonautonomy on decision to use contraceptives. Hence, this study aimed to assess the spatial distribution of women's nonautonomy on decision regarding contraceptive utilization and its determinants in Ethiopia. Methods Data were accessed from the Demographic Health Survey program official database website (https://dhsprogram.com). A weighted sample of 3,668 married reproductive-age women currently using contraceptives was included in this analysis. Bayesian multilevel logistic regression models were fitted to identify the determinants of women's nonautonomy on contraceptive utilization. Adjusted odds ratio with 95% credible interval was used to select variables that have a significant effect on nonautonomy on contraceptive utilization. Results A high proportion of women with nonautonomy on decision regarding contraceptive utilization was found in northern parts of Southern Nations, Nationalities, and People's Region, Southern parts of Oromia, and Benishangul-Gumuz regions of the country. Overall, 2876 (78.40% (95% CI: 77.0%, 79.7%)) women were nonautonomous on decision regarding contraceptive utilization. In the final model, age from 35–49 (AOR (95% CI) = 0.63 (0.54, 0.72)), living in the richer households (AOR (95% CI) = 0.12 (0.03, 0.26)), being married at 18 years or above (AOR (95% CI) = 0.33 (0.19, 0.57)), and residing in an rural areas (AOR (95% CI) = 1.34 (1.01, 1.71)) and metropolitan regions (AOR (95% CI) = 0.71(0.54, 0.91)) were associated with women's nonautonomy on decision regarding contraceptive utilization. Conclusions In Ethiopia, the spatial distribution of women's nonautonomy on decision about contraceptive utilization was nonrandom. More than three-fourths of married reproductive-age women in Ethiopia are nonautonomous on decision regarding contraceptive utilization. Region, residence, current age, age at marriage, and wealth index were statistically associated with women's nonautonomy on decision regarding contraceptive utilization.

Who Joins a Forum—And Who Does Not?—Evaluating Drivers of Forum Participation in Polycentric Governance Systems

AbstractIn polycentric governance systems, decisions that influence a given policy issue are often made across a series of forums: venues where actors meet to resolve collective action problems. Here, we examine who does and does not participate in forums, and the factors driving that participation. We analyze forum participation patterns of 307 actors involved in Swiss water governance, who could participate in water governance forums. We find that the majority of actors do not participate in any forums. Results from a Bayesian multilevel logistic regression model show that especially those concerned with a broader range of policy issues and those that have more organizational resources at their disposal are more likely to participate. To a lesser extent, this also holds for organizations that represent policy beliefs consistent with median beliefs in the system. A belief that increased cross‐sectoral coordination is needed to promote more effective governance does not have a discernible impact on participation. These results question the integrative characteristics often attributed to forums in polycentric governance more generally.

Controls of outbursts of moraine-dammed lakes in the greater Himalayan region

Abstract. Glacial lakes in the Hindu Kush–Karakoram–Himalayas–Nyainqentanglha (HKKHN) region have grown rapidly in number and area in past decades, and some dozens have drained in catastrophic glacial lake outburst floods (GLOFs). Estimating regional susceptibility of glacial lakes has largely relied on qualitative assessments by experts, thus motivating a more systematic and quantitative appraisal. Before the backdrop of current climate-change projections and the potential of elevation-dependent warming, an objective and regionally consistent assessment is urgently needed. We use an inventory of 3390 moraine-dammed lakes and their documented outburst history in the past four decades to test whether elevation, lake area and its rate of change, glacier-mass balance, and monsoonality are useful inputs to a probabilistic classification model. We implement these candidate predictors in four Bayesian multi-level logistic regression models to estimate the posterior susceptibility to GLOFs. We find that mostly larger lakes have been more prone to GLOFs in the past four decades regardless of the elevation band in which they occurred. We also find that including the regional average glacier-mass balance improves the model classification. In contrast, changes in lake area and monsoonality play ambiguous roles. Our study provides first quantitative evidence that GLOF susceptibility in the HKKHN scales with lake area, though less so with its dynamics. Our probabilistic prognoses offer improvement compared to a random classification based on average GLOF frequency. Yet they also reveal some major uncertainties that have remained largely unquantified previously and that challenge the applicability of single models. Ensembles of multiple models could be a viable alternative for more accurately classifying the susceptibility of moraine-dammed lakes to GLOFs.

Clinical, health systems and neighbourhood determinants of tuberculosis case fatality in urban Blantyre, Malawi: a multilevel epidemiological analysis of enhanced surveillance data

Abstract We investigated whether household to clinic distance was a risk factor for death on tuberculosis (TB) treatment in Malawi. Using enhanced TB surveillance data, we recorded all TB treatment initiations and outcomes between 2015 and 2018. Household locations were geolocated, and distances were measured by a straight line or shortest road network. We constructed Bayesian multi-level logistic regression models to investigate associations between distance and case fatality. A total of 479/4397 (10.9%) TB patients died. Greater distance was associated with higher (odds ratio (OR) 1.07 per kilometre (km) increase, 95% credible interval (CI) 0.99–1.16) odds of death in TB patients registered at the referral hospital, but not among TB patients registered at primary clinics (OR 0.98 per km increase, 95% CI 0.92–1.03). Age (OR 1.02 per year increase, 95% CI 1.01–1.02) and HIV-positive status (OR 2.21, 95% CI 1.73–2.85) were also associated with higher odds of death. Model estimates were similar for both distance measures. Distance was a risk factor for death among patients at the main referral hospital, likely due to delayed diagnosis and suboptimal healthcare access. To reduce mortality, targeted community TB screening interventions for TB disease and HIV, and expansion of novel sensitive diagnostic tests are required.

Determinants of malaria among under-five children in Ethiopia: Bayesian multilevel analysis

BackgroundIn Ethiopia, malaria is one of the public health problems, and it is still among the ten top leading causes of morbidity and mortality among under-five children.However, the studies conducted in the country have been inconclusive and inconsistent. Thus, this study aimed to assess factors associated with malaria among under-five children in Ethiopia.MethodsWe retrieved secondary data from the malaria indicator survey data collected from September 30 to December 10, 2015, in Ethiopia. A total of 8301 under-five-year-old children who had microscopy test results were included in the study. Bayesian multilevel logistic regression models were fitted and Markov chain Monte Carlo simulation was used to estimate the model parameters using Gibbs sampling. Adjusted Odd Ratio with 95% credible interval in the multivariable model was used to select variables that have a significant association with malaria.ResultsIn this study, sleeping under the insecticide-treated bed nets during bed time (ITN) [AOR 0.58,95% CI, 0.31–0.97)], having 2 and more ITN for the household [AOR 0.43, (95% CI, 0.17–0.88)], have radio [AOR 0.41, (95% CI, 0.19–0.78)], have television [AOR 0.19, (95% CI, 0.01–0.89)] and altitude [AOR 0.05, (95% CI, 0.01–0.13)] were the predictors of malaria among under-five children.ConclusionsThe study revealed that sleeping under ITN, having two and more ITN for the household, altitude, availability of radio, and television were the predictors of malaria among under-five children in Ethiopia. Thus, the government should strengthen the availability and utilization of ITN to halt under-five mortality due to malaria.

Effect of unintended pregnancy on skilled antenatal care uptake in Bangladesh: analysis of national survey data

BackgroundAround 48% of all pregnancies in low- and middle-income countries are unintended. Unintended pregnancy may contribute to lower use of antenatal care (ANC); however, current research in the area is largely inconclusive due to the methodological approaches applied.MethodsResponses from 4493 women extracted from the 2014 Bangladesh Demographic and Health Survey (BDHS) were used to assess the association between unintended pregnancy and subsequent uptake of at least one and at least four skilled ANC visits. For this, Bayesian multilevel logistic regression models with informative priors (representing a range of values within which the researcher is certain the true effect of the parameters included lies) were used, adjusting for other factors that affect ANC uptake. Informative priors were selected from the BDHS data collected in 2004, 2007, and 2011.ResultsAround 64% of women in Bangladesh who had at least one pregnancy within 3 years prior to the survey (that ended in a live birth) received ANC at least once, and of these around 32% used ANC at least four times. Mistimed (aOR, 0.73, 95% Cred I, 0.66–0.81) and unwanted (aOR, 0.69, 95% Cred I, 0.64–0.75) pregnancy were associated with reduced odds of attending the recommended minimum of four skilled ANC visits compared with wanted pregnancy. These likelihoods were even lower for at least one skilled ANC visit among women with a mistimed (aOR, 0.59, 95% Cred I, 0.53–0.65) or an unwanted pregnancy (aOR, 0.67, 95% Cred I, 0.61–0.74) than women with a wanted pregnancy.ConclusionsIn Bangladesh, more than one-quarter of women who report an unintended pregnancy at conception and do not terminate the pregnancy are at high risk of not using ANC. It is important for policies to include women with unintended pregnancy in mainstream healthcare services. This will increase the use of ANC and reduce associated adverse consequences.

Bayesian Multilevel Analysis of Utilization of Antenatal Care Services in Ethiopia.

In sub-Saharan Africa, 72% of pregnant women received an antenatal care visit at least once in their pregnancy period. Ethiopia has one of the highest rates of maternal mortality in sub-Saharan African countries. So, this high maternal mortality levels remain a major public health problem. According to EDHS, 2016, the antenatal care (ANC), delivery care (DC), and postnatal care (PNC) were 62%, 73%, and 13%, respectively, indicating that ANC is in a low level. The main objective of this study was to examine the factors that affect the utilization of antenatal care services in Ethiopia using Bayesian multilevel logistic regression models. The data used for this study comes from the 2016 Ethiopian Demographic and Health Survey which was conducted by the Central Statistical Agency (CSA). The statistical method of data analysis used for this study is the Bayesian multilevel binary logistic regression model in general and the Bayesian multilevel logistic regression for the random coefficient model in particular. The convergences of parameters are estimated by using Markov chain Monte-Carlo (MCMC) using SPSS and MLwiN software. The descriptive result revealed that out of the 7171 women who are supposed to use ANC services, 2479 (34.6%) women were not receiving ANC services, while 4692 (65.4%) women were receiving ANC services. Moreover, women in the Somali and Afar regions are the least users of ANC. Using the Bayesian multilevel binary logistic regression of random coefficient model factors, place of residence, religion, educational attainment of women, husband educational level, employment status of husband, beat, household wealth index, and birth order were found to be the significant factors for usage of ANC. Regional variation in the usage of ANC was significant.

Spatio-temporal variations in unmet need for family planning in Ghana: 2003\u20132014

Ghana has long prioritized family planning as a key strategy for improving health and socioeconomic development. However, despite the heavy investments in the sector over the last decade, the family planning program has not successfully improved the country’s family planning indicators. In this study, we describe the spatial and temporal patterns of unmet need for family planning from 2003 to 2014 and mainly estimate the socioeconomic factors affecting it. Using data from the 2003, 2008, and 2014 Ghana Demographic and Health Surveys, we map the regional and temporal trends in unmet need for family planning and used Bayesian multilevel logistic regression models to estimate the effects of individual-level socioeconomic characteristics on unmet need for family planning. The results show that there are considerable regional disparities in unmet need for family planning for the study period. These disparities show diverse trends among the ten regions in the country over the study period. The patterns of unmet need for family planning are considerably affected by socioeconomic factors such as educational attainment, household wealth, marital, and work status, as well as age, ethnicity, parity, and time (year). Both socioeconomic and demographic factors play a significant role in the patterns of unmet need for family planning in Ghana. There is the need to formulate consistently effective family planning policies focusing mainly on the higher risk groups of women in the country.

Nationwide Heterogeneity in Hospital-Specific Probabilities of Rectal Cancer Understaging and Its Effects on Outcomes.

Rectal cancer patients who are understaged may not be offered the highest quality treatment modalities, which are based on an accurate assessment of preoperative staging. The objective of this study was to evaluate heterogeneity in the probability of being understaged at Commission on Cancer hospitals in the United States and to assess how this variation affects outcomes. The 2006-2013 National Cancer Data Base was queried for clinical stage I-III rectal cancer patients who underwent resection. The initial clinical stage was compared with the "gold standard," pathological stage. A Bayesian multilevel logistic regression model was used to characterize variation in hospital-specific probabilities of being understaged (clinical stage < pathologic stage). Separate analyses assessed the impact of being understaged on positive circumferential resection margins (CRM), receipt of adjuvant chemotherapy, and 5-year overall survival. Among 12,684 patients who did not receive neoadjuvant chemoradiation and treated at 1176 hospitals, 3044 (24%) were understaged. After patient level risk-adjustment, a 24-fold difference in the probability of being understaged was observed between hospitals (range 3-72%, median = 15%). Understaging was independently associated with positive CRM [odds ratio (OR) 1.59, 95% confidence interval (CI) 1.39, 1.92] and receipt of adjuvant chemotherapy (OR 14.22, 95% CI 13.55, 18.88). Despite an increase in the delivery of systemic therapy after surgical resection, understaging was associated with worse survival (hazard ratio = 1.61, 95% CI 1.48, 1.95). Deficiencies in high-quality rectal cancer management begin with incorrect clinical staging. The risk-adjusted probability of understaging varied widely between hospitals. This institutional failure to provide optimal oncological management at the start of care was associated with worse long-term survival.

Surgical site infection after gastrointestinal surgery in high-income, middle-income, and low-income countries: a prospective, international, multicentre cohort study

SummaryBackgroundSurgical site infection (SSI) is one of the most common infections associated with health care, but its importance as a global health priority is not fully understood. We quantified the burden of SSI after gastrointestinal surgery in countries in all parts of the world.MethodsThis international, prospective, multicentre cohort study included consecutive patients undergoing elective or emergency gastrointestinal resection within 2-week time periods at any health-care facility in any country. Countries with participating centres were stratified into high-income, middle-income, and low-income groups according to the UN's Human Development Index (HDI). Data variables from the GlobalSurg 1 study and other studies that have been found to affect the likelihood of SSI were entered into risk adjustment models. The primary outcome measure was the 30-day SSI incidence (defined by US Centers for Disease Control and Prevention criteria for superficial and deep incisional SSI). Relationships with explanatory variables were examined using Bayesian multilevel logistic regression models. This trial is registered with ClinicalTrials.gov, number NCT02662231.FindingsBetween Jan 4, 2016, and July 31, 2016, 13 265 records were submitted for analysis. 12 539 patients from 343 hospitals in 66 countries were included. 7339 (58·5%) patient were from high-HDI countries (193 hospitals in 30 countries), 3918 (31·2%) patients were from middle-HDI countries (82 hospitals in 18 countries), and 1282 (10·2%) patients were from low-HDI countries (68 hospitals in 18 countries). In total, 1538 (12·3%) patients had SSI within 30 days of surgery. The incidence of SSI varied between countries with high (691 [9·4%] of 7339 patients), middle (549 [14·0%] of 3918 patients), and low (298 [23·2%] of 1282) HDI (p<0·001). The highest SSI incidence in each HDI group was after dirty surgery (102 [17·8%] of 574 patients in high-HDI countries; 74 [31·4%] of 236 patients in middle-HDI countries; 72 [39·8%] of 181 patients in low-HDI countries). Following risk factor adjustment, patients in low-HDI countries were at greatest risk of SSI (adjusted odds ratio 1·60, 95% credible interval 1·05–2·37; p=0·030). 132 (21·6%) of 610 patients with an SSI and a microbiology culture result had an infection that was resistant to the prophylactic antibiotic used. Resistant infections were detected in 49 (16·6%) of 295 patients in high-HDI countries, in 37 (19·8%) of 187 patients in middle-HDI countries, and in 46 (35·9%) of 128 patients in low-HDI countries (p<0·001).InterpretationCountries with a low HDI carry a disproportionately greater burden of SSI than countries with a middle or high HDI and might have higher rates of antibiotic resistance. In view of WHO recommendations on SSI prevention that highlight the absence of high-quality interventional research, urgent, pragmatic, randomised trials based in LMICs are needed to assess measures aiming to reduce this preventable complication.FundingDFID-MRC-Wellcome Trust Joint Global Health Trial Development Grant, National Institute of Health Research Global Health Research Unit Grant.

Patterns and Yearly Time Trends in the Use of Radiation Therapy During the Last 30 Days of Life Among Patients With Metastatic Rectal Cancer in the United States From 2004 to 2012

Although radiation therapy (RT) can provide palliative benefits for patients with metastatic rectal cancer, its role at the end of life remains unclear. The objective of this study was to assess sociodemographic and clinical factors associated with the use of RT during the last 30 days of life and to evaluate yearly time trends in RT utilization among stage IV patients with rectal cancer. The 2004 to 2012 National Cancer DataBase was queried for patients with metastatic rectal cancer who had a documented death during follow-up. A Bayesian multilevel logistic regression model was used to characterize predictive factors and yearly time trends associated with RT use in the last 30 days of life. Among 10 431 patients who met inclusion criteria, 345 (3%) received RT during the last 30 days of life. Factors independently associated with RT use included older age, female sex, African American race, nonprivate insurance, higher comorbidity burden, and worse grade. The odds of RT use at the end of life decreased by 28% between 2007 and 2009 (odds ratio [OR] = 0.72, 95% Credible Interval (CI) = 0.58-0.93), but then increased by 16% from 2010 to 2012 (OR = 1.16, 95% CI = 1.13-1.33), relative to 2004 to 2006. Radiation therapy use for patients with metastatic rectal cancer is beneficial, and efforts to optimize its appropriate use are important. Several factors associated with RT use during the last 30 days of life included disparities in sociodemographic and clinical subgroups. Research is needed to understand the underlying causes of these inequalities and the role of predictive models in clinical decision-making.

Variation in Hospital-Specific Rates of Suboptimal Lymphadenectomy and Survival in Colon Cancer: Evidence from the National Cancer Data Base

Little is known about between-hospital differences in the rate of suboptimal lymphadenectomy. This study characterizes variation in hospital-specific rates of suboptimal lymphadenectomy and its effect on overall survival in a national hospital-based registry. Stage I-III colon cancer patients were identified from the 2003-2012 National Cancer Data Base. Bayesian multilevel logistic regression models were used to assess the impact of patient- and hospital-level factors on hospital-specific rates of suboptimal lymphadenectomy (<12 lymph nodes), and multilevel Cox models were used to estimate the effect of suboptimal lymphadenectomy at the patient (yes vs. no) and hospital level (quartiles of hospital-specific rates) on overall survival. A total of 360,846 patients across 1345 hospitals in the US met the inclusion criteria, of which 25% had a suboptimal lymphadenectomy. Wide variation was observed in hospital-specific rates of suboptimal lymphadenectomy (range 0-82%, median 44%). Older age, male sex, comorbidity score, no insurance, positive margins, lower tumor grade, lower T and N stage, and sigmoid and left colectomy were associated with higher odds of suboptimal lymphadenectomy. Patients treated at lower-volume and non-academic hospitals had higher odds of suboptimal lymphadenectomy. Patient- and hospital-level factors explained 5% of the between-hospital variability in suboptimal lymphadenectomy, leaving 95% unexplained. Higher suboptimal lymphadenectomy rates were associated with worse survival (quartile 4 vs. quartile 1: hazard ratio 1.19, 95% confidence interval 1.16-1.22). Large differences in hospital-specific rates of suboptimal lymphadenectomy were observed, and this variation was associated with survival. Quality improvement initiatives targeting hospital-level adherence to the national standard may improve overall survival among resected colon cancer patients.