Search

Background: Tobacco industry interference has been identified as the greatest obstacle to the implementation of evidence-based measures to reduce tobacco use. Understanding and addressing industry interference in public health policy-making is therefore crucial. Existing conceptualisations of corporate political activity (CPA) are embedded in a business perspective and do not attend to CPA's social and public health costs; most have not drawn on the unique resource represented by internal tobacco industry documents. Building on this literature, including systematic reviews, we develop a critically informed conceptual model of tobacco industry political activity.Methods and Findings: We thematically analysed published papers included in two systematic reviews examining tobacco industry influence on taxation and marketing of tobacco; we included 45 of 46 papers in the former category and 20 of 48 papers in the latter (n = 65). We used a grounded theory approach to build taxonomies of "discursive" (argument-based) and "instrumental" (action-based) industry strategies and from these devised the Policy Dystopia Model, which shows that the industry, working through different constituencies, constructs a metanarrative to argue that proposed policies will lead to a dysfunctional future of policy failure and widely dispersed adverse social and economic consequences. Simultaneously, it uses diverse, interlocking insider and outsider instrumental strategies to disseminate this narrative and enhance its persuasiveness in order to secure its preferred policy outcomes. Limitations are that many papers were historical (some dating back to the 1970s) and focused on high-income regions.Conclusions: The model provides an evidence-based, accessible way of understanding diverse corporate political strategies. It should enable public health actors and officials to preempt these strategies and develop realistic assessments of the industry's claims. [ABSTRACT FROM AUTHOR]

In a Perspective linked to the Research Article by Haar and colleagues, Michael Spagat discusses the challenges and importance of conducting research on mortality in regions affected by violent conflicts. [ABSTRACT FROM AUTHOR]

Elaine Mardis and Marc Ladanyi discuss how large-scale genomics has driven advances in cancer translational medicine, with a focus on publications in this month's Special Issue. [ABSTRACT FROM AUTHOR]

Background: Preterm infants are a group at high risk of having experienced placental insufficiency. It is unclear which growth charts perform best in identifying infants at increased risk of stillbirth and other adverse perinatal outcomes. We compared 2 birthweight charts (population centiles and INTERGROWTH-21st birthweight centiles) and 3 fetal growth charts (INTERGROWTH-21st fetal growth charts, World Health Organization fetal growth charts, and Gestation Related Optimal Weight [GROW] customised growth charts) to identify which chart performed best in identifying infants at increased risk of adverse perinatal outcome in a preterm population.Methods and Findings: We conducted a retrospective cohort study of all preterm infants born at 24.0 to 36.9 weeks gestation in Victoria, Australia, from 2005 to 2015 (28,968 records available for analysis). All above growth charts were applied to the population. Proportions classified as <5th centile and <10th centile by each chart were compared, as were proportions of stillborn infants considered small for gestational age (SGA, <10th centile) by each chart. We then compared the relative performance of non-overlapping SGA cohorts by each chart to our low-risk reference population (infants born appropriate size for gestational age [>10th and <90th centile] by all intrauterine charts [AGAall]) for the following perinatal outcomes: stillbirth, perinatal mortality (stillbirth or neonatal death), Apgar <4 or <7 at 5 minutes, neonatal intensive care unit admissions, suspicion of poor fetal growth leading to expedited delivery, and cesarean section. All intrauterine charts classified a greater proportion of infants as <5th or <10th centile than birthweight charts. The magnitude of the difference between birthweight and fetal charts was greater at more preterm gestations. Of the fetal charts, GROW customised charts classified the greatest number of infants as SGA (22.3%) and the greatest number of stillborn infants as SGA (57%). INTERGROWTH classified almost no additional infants as SGA that were not already considered SGA on GROW or WHO charts; however, those infants classified as SGA by INTERGROWTH had the greatest risk of both stillbirth and total perinatal mortality. GROW customised charts classified a larger proportion of infants as SGA, and these infants were still at increased risk of mortality and adverse perinatal outcomes compared to the AGAall population. Consistent with similar studies in this field, our study was limited in comparing growth charts by the degree of overlap, with many infants classified as SGA by multiple charts. We attempted to overcome this by examining and comparing sub-populations classified as SGA by only 1 growth chart.Conclusions: In this study, fetal charts classified greater proportions of preterm and stillborn infants as SGA, which more accurately reflected true fetal growth restriction. Of the intrauterine charts, INTERGROWTH classified the smallest number of preterm infants as SGA, although it identified a particularly high-risk cohort, and GROW customised charts classified the greatest number at increased risk of perinatal mortality. [ABSTRACT FROM AUTHOR]

Background: There has been a growing interest in understanding the effects of social networks on health-related behaviour, with a particular backdrop being the emerging prominence of complexity or systems science in public health. Social network interventions specifically use or alter the characteristics of social networks to generate, accelerate, or maintain health behaviours. We conducted a systematic review and meta-analysis to investigate health behaviour outcomes of social network interventions.Methods and Findings: We searched eight databases and two trial registries from 1990 to May 28, 2019, for English-language reports of randomised controlled trials (RCTs) and before-and-after studies investigating social network interventions for health behaviours and outcomes. Trials that did not specifically use social networks or that did not include a comparator group were excluded. We screened studies and extracted data from published reports independently. The primary outcome of health behaviours or outcomes at ≤6 months was assessed by random-effects meta-analysis. Secondary outcomes included those measures at >6-12 months and >12 months. This study is registered with the International Prospective Register of Systematic Reviews, PROSPERO: CRD42015023541. We identified 26,503 reports; after exclusion, 37 studies, conducted between 1996 and 2018 from 11 countries, were eligible for analysis, with a total of 53,891 participants (mean age 32.4 years [SD 12.7]; 45.5% females). A range of study designs were included: 27 used RCT/cluster RCT designs, and 10 used other study designs. Eligible studies addressed a variety of health outcomes, in particular sexual health and substance use. Social network interventions showed a significant intervention effect compared with comparator groups for sexual health outcomes. The pooled odds ratio (OR) was 1.46 (95% confidence interval [CI] 1.01-2.11; I2 = 76%) for sexual health outcomes at ≤6 months and OR 1.51 (95% CI 1.27-1.81; I2 = 40%) for sexual health outcomes at >6-12 months. Intervention effects for drug risk outcomes at each time point were not significant. There were also significant intervention effects for some other health outcomes including alcohol misuse, well-being, change in haemoglobin A1c (HbA1c), and smoking cessation. Because of clinical and measurement heterogeneity, it was not appropriate to pool data on these other behaviours in a meta-analysis. For sexual health outcomes, prespecified subgroup analyses were significant for intervention approach (p < 0.001), mean age of participants (p = 0.002), and intervention length (p = 0.05). Overall, 22 of the 37 studies demonstrated a high risk of bias, as measured by the Cochrane Risk of Bias tool. The main study limitations identified were the inclusion of studies of variable quality; difficulty in isolating the effects of specific social network intervention components on health outcomes, as interventions included other active components; and reliance on self-reported outcomes, which have inherent recall and desirability biases.Conclusions: Our findings suggest that social network interventions can be effective in the short term (<6 months) and longer term (>6 months) for sexual health outcomes. Intervention effects for drug risk outcomes at each time point were not significant. There were also significant intervention effects for some other health outcomes including alcohol misuse, well-being, change in HbA1c, and smoking cessation. [ABSTRACT FROM AUTHOR]

Background: In the UK, approximately 4,200 men who have sex with men (MSM) are living with HIV but remain undiagnosed. Maximising the number of high-risk people testing for HIV is key to ensuring prompt treatment and preventing onward infection. This study assessed how different HIV test characteristics affect the choice of testing option, including remote testing (HIV self-testing or HIV self-sampling), in the UK, a country with universal access to healthcare.Methods and Findings: Between 3 April and 11 May 2017, a cross-sectional online-questionnaire-based discrete choice experiment (DCE) was conducted in which respondents who expressed an interest in online material used by MSM were asked to imagine that they were at risk of HIV infection and to choose between different hypothetical HIV testing options, including the option not to test. A variety of different testing options with different defining characteristics were described so that the independent preference for each characteristic could be valued. The characteristics included where each test is taken, the sampling method, how the test is obtained, whether infections other than HIV are tested for, test accuracy, the cost of the test, the infection window period, and how long it takes to receive the test result. Participants were recruited and completed the instrument online, in order to include those not currently engaged with healthcare services. The main analysis was conducted using a latent class model (LCM), with results displayed as odds ratios (ORs) and probabilities. The ORs indicate the strength of preference for one characteristic relative to another (base) characteristic. In total, 620 respondents answered the DCE questions. Most respondents reported that they were white (93%) and were either gay or bisexual (99%). The LCM showed that there were 2 classes within the respondent sample that appeared to have different preferences for the testing options. The first group, which was likely to contain 86% of respondents, had a strong preference for face-to-face tests by healthcare professionals (HCPs) compared to remote testing (OR 6.4; 95% CI 5.6, 7.4) and viewed not testing as less preferable than remote testing (OR 0.10; 95% CI 0.09, 0.11). In the second group, which was likely to include 14% of participants, not testing was viewed as less desirable than remote testing (OR 0.56; 95% CI 0.53, 0.59) as were tests by HCPs compared to remote testing (OR 0.23; 95% CI 0.15, 0.36). In both classes, free remote tests instead of each test costing £30 was the test characteristic with the largest impact on the choice of testing option. Participants in the second group were more likely to have never previously tested and to be non-white than participants in the first group. The main study limitations were that the sample was recruited solely via social media, the study advert was viewed only by people expressing an interest in online material used by MSM, and the choices in the experiment were hypothetical rather than observed in the real world.Conclusions: Our results suggest that preferences in the context we examined are broadly dichotomous. One group, containing the majority of MSM, appears comfortable testing for HIV but prefers face-to-face testing by HCPs rather than remote testing. The other group is much smaller, but contains MSM who are more likely to be at high infection risk. For these people, the availability of remote testing has the potential to significantly increase net testing rates, particularly if provided for free. [ABSTRACT FROM AUTHOR]

Background: Face-to-face brief interventions for problem drinking are effective, but they have found limited implementation in routine care and the community. Internet-based interventions could overcome this treatment gap. We investigated effectiveness and moderators of treatment outcomes in internet-based interventions for adult problem drinking (iAIs).Methods and Findings: Systematic searches were performed in medical and psychological databases to 31 December 2016. A one-stage individual patient data meta-analysis (IPDMA) was conducted with a linear mixed model complete-case approach, using baseline and first follow-up data. The primary outcome measure was mean weekly alcohol consumption in standard units (SUs, 10 grams of ethanol). Secondary outcome was treatment response (TR), defined as less than 14/21 SUs for women/men weekly. Putative participant, intervention, and study moderators were included. Robustness was verified in three sensitivity analyses: a two-stage IPDMA, a one-stage IPDMA using multiple imputation, and a missing-not-at-random (MNAR) analysis. We obtained baseline data for 14,198 adult participants (19 randomised controlled trials [RCTs], mean age 40.7 [SD = 13.2], 47.6% women). Their baseline mean weekly alcohol consumption was 38.1 SUs (SD = 26.9). Most were regular problem drinkers (80.1%, SUs 44.7, SD = 26.4) and 19.9% (SUs 11.9, SD = 4.1) were binge-only drinkers. About one third were heavy drinkers, meaning that women/men consumed, respectively, more than 35/50 SUs of alcohol at baseline (34.2%, SUs 65.9, SD = 27.1). Post-intervention data were available for 8,095 participants. Compared with controls, iAI participants showed a greater mean weekly decrease at follow-up of 5.02 SUs (95% CI -7.57 to -2.48, p < 0.001) and a higher rate of TR (odds ratio [OR] 2.20, 95% CI 1.63-2.95, p < 0.001, number needed to treat [NNT] = 4.15, 95% CI 3.06-6.62). Persons above age 55 showed higher TR than their younger counterparts (OR = 1.66, 95% CI 1.21-2.27, p = 0.002). Drinking profiles were not significantly associated with treatment outcomes. Human-supported interventions were superior to fully automated ones on both outcome measures (comparative reduction: -6.78 SUs, 95% CI -12.11 to -1.45, p = 0.013; TR: OR = 2.23, 95% CI 1.22-4.08, p = 0.009). Participants treated in iAIs based on personalised normative feedback (PNF) alone were significantly less likely to sustain low-risk drinking at follow-up than those in iAIs based on integrated therapeutic principles (OR = 0.52, 95% CI 0.29-0.93, p = 0.029). The use of waitlist control in RCTs was associated with significantly better treatment outcomes than the use of other types of control (comparative reduction: -9.27 SUs, 95% CI -13.97 to -4.57, p < 0.001; TR: OR = 3.74, 95% CI 2.13-6.53, p < 0.001). The overall quality of the RCTs was high; a major limitation included high study dropout (43%). Sensitivity analyses confirmed the robustness of our primary analyses.Conclusion: To our knowledge, this is the first IPDMA on internet-based interventions that has shown them to be effective in curbing various patterns of adult problem drinking in both community and healthcare settings. Waitlist control may be conducive to inflation of treatment outcomes. [ABSTRACT FROM AUTHOR]

Background: Non-small-cell lung cancer (NSCLC) patients often demonstrate varying clinical courses and outcomes, even within the same tumor stage. This study explores deep learning applications in medical imaging allowing for the automated quantification of radiographic characteristics and potentially improving patient stratification.Methods and Findings: We performed an integrative analysis on 7 independent datasets across 5 institutions totaling 1,194 NSCLC patients (age median = 68.3 years [range 32.5-93.3], survival median = 1.7 years [range 0.0-11.7]). Using external validation in computed tomography (CT) data, we identified prognostic signatures using a 3D convolutional neural network (CNN) for patients treated with radiotherapy (n = 771, age median = 68.0 years [range 32.5-93.3], survival median = 1.3 years [range 0.0-11.7]). We then employed a transfer learning approach to achieve the same for surgery patients (n = 391, age median = 69.1 years [range 37.2-88.0], survival median = 3.1 years [range 0.0-8.8]). We found that the CNN predictions were significantly associated with 2-year overall survival from the start of respective treatment for radiotherapy (area under the receiver operating characteristic curve [AUC] = 0.70 [95% CI 0.63-0.78], p < 0.001) and surgery (AUC = 0.71 [95% CI 0.60-0.82], p < 0.001) patients. The CNN was also able to significantly stratify patients into low and high mortality risk groups in both the radiotherapy (p < 0.001) and surgery (p = 0.03) datasets. Additionally, the CNN was found to significantly outperform random forest models built on clinical parameters-including age, sex, and tumor node metastasis stage-as well as demonstrate high robustness against test-retest (intraclass correlation coefficient = 0.91) and inter-reader (Spearman's rank-order correlation = 0.88) variations. To gain a better understanding of the characteristics captured by the CNN, we identified regions with the most contribution towards predictions and highlighted the importance of tumor-surrounding tissue in patient stratification. We also present preliminary findings on the biological basis of the captured phenotypes as being linked to cell cycle and transcriptional processes. Limitations include the retrospective nature of this study as well as the opaque black box nature of deep learning networks.Conclusions: Our results provide evidence that deep learning networks may be used for mortality risk stratification based on standard-of-care CT images from NSCLC patients. This evidence motivates future research into better deciphering the clinical and biological basis of deep learning networks as well as validation in prospective data. [ABSTRACT FROM AUTHOR]

Background: Resuscitated cardiac arrest is associated with high mortality; however, the ability to estimate risk of adverse outcomes using existing illness severity scores is limited. Using in-hospital data available within the first 24 hours of admission, we aimed to develop more accurate models of risk prediction using both logistic regression (LR) and machine learning (ML) techniques, with a combination of demographic, physiologic, and biochemical information.Methods and Findings: Patient-level data were extracted from the Australian and New Zealand Intensive Care Society (ANZICS) Adult Patient Database for patients who had experienced a cardiac arrest within 24 hours prior to admission to an intensive care unit (ICU) during the period January 2006 to December 2016. The primary outcome was in-hospital mortality. The models were trained and tested on a dataset (split 90:10) including age, lowest and highest physiologic variables during the first 24 hours, and key past medical history. LR and 5 ML approaches (gradient boosting machine [GBM], support vector classifier [SVC], random forest [RF], artificial neural network [ANN], and an ensemble) were compared to the APACHE III and Australian and New Zealand Risk of Death (ANZROD) predictions. In all, 39,566 patients from 186 ICUs were analysed. Mean (±SD) age was 61 ± 17 years; 65% were male. Overall in-hospital mortality was 45.5%. Models were evaluated in the test set. The APACHE III and ANZROD scores demonstrated good discrimination (area under the receiver operating characteristic curve [AUROC] = 0.80 [95% CI 0.79-0.82] and 0.81 [95% CI 0.8-0.82], respectively) and modest calibration (Brier score 0.19 for both), which was slightly improved by LR (AUROC = 0.82 [95% CI 0.81-0.83], DeLong test, p < 0.001). Discrimination was significantly improved using ML models (ensemble and GBM AUROCs = 0.87 [95% CI 0.86-0.88], DeLong test, p < 0.001), with an improvement in performance (Brier score reduction of 22%). Explainability models were created to assist in identifying the physiologic features that most contributed to an individual patient's survival. Key limitations include the absence of pre-hospital data and absence of external validation.Conclusions: ML approaches significantly enhance predictive discrimination for mortality following cardiac arrest compared to existing illness severity scores and LR, without the use of pre-hospital data. The discriminative ability of these ML models requires validation in external cohorts to establish generalisability. [ABSTRACT FROM AUTHOR]

Background: The current acute kidney injury (AKI) risk prediction model for patients undergoing percutaneous coronary intervention (PCI) from the American College of Cardiology (ACC) National Cardiovascular Data Registry (NCDR) employed regression techniques. This study aimed to evaluate whether models using machine learning techniques could significantly improve AKI risk prediction after PCI.Methods and Findings: We used the same cohort and candidate variables used to develop the current NCDR CathPCI Registry AKI model, including 947,091 patients who underwent PCI procedures between June 1, 2009, and June 30, 2011. The mean age of these patients was 64.8 years, and 32.8% were women, with a total of 69,826 (7.4%) AKI events. We replicated the current AKI model as the baseline model and compared it with a series of new models. Temporal validation was performed using data from 970,869 patients undergoing PCIs between July 1, 2016, and March 31, 2017, with a mean age of 65.7 years; 31.9% were women, and 72,954 (7.5%) had AKI events. Each model was derived by implementing one of two strategies for preprocessing candidate variables (preselecting and transforming candidate variables or using all candidate variables in their original forms), one of three variable-selection methods (stepwise backward selection, lasso regularization, or permutation-based selection), and one of two methods to model the relationship between variables and outcome (logistic regression or gradient descent boosting). The cohort was divided into different training (70%) and test (30%) sets using 100 different random splits, and the performance of the models was evaluated internally in the test sets. The best model, according to the internal evaluation, was derived by using all available candidate variables in their original form, permutation-based variable selection, and gradient descent boosting. Compared with the baseline model that uses 11 variables, the best model used 13 variables and achieved a significantly better area under the receiver operating characteristic curve (AUC) of 0.752 (95% confidence interval [CI] 0.749-0.754) versus 0.711 (95% CI 0.708-0.714), a significantly better Brier score of 0.0617 (95% CI 0.0615-0.0618) versus 0.0636 (95% CI 0.0634-0.0638), and a better calibration slope of observed versus predicted rate of 1.008 (95% CI 0.988-1.028) versus 1.036 (95% CI 1.015-1.056). The best model also had a significantly wider predictive range (25.3% versus 21.6%, p < 0.001) and was more accurate in stratifying AKI risk for patients. Evaluated on a more contemporary CathPCI cohort (July 1, 2015-March 31, 2017), the best model consistently achieved significantly better performance than the baseline model in AUC (0.785 versus 0.753), Brier score (0.0610 versus 0.0627), calibration slope (1.003 versus 1.062), and predictive range (29.4% versus 26.2%). The current study does not address implementation for risk calculation at the point of care, and potential challenges include the availability and accessibility of the predictors.Conclusions: Machine learning techniques and data-driven approaches resulted in improved prediction of AKI risk after PCI. The results support the potential of these techniques for improving risk prediction models and identification of patients who may benefit from risk-mitigation strategies. [ABSTRACT FROM AUTHOR]

Background: Chest radiograph interpretation is critical for the detection of thoracic diseases, including tuberculosis and lung cancer, which affect millions of people worldwide each year. This time-consuming task typically requires expert radiologists to read the images, leading to fatigue-based diagnostic error and lack of diagnostic expertise in areas of the world where radiologists are not available. Recently, deep learning approaches have been able to achieve expert-level performance in medical image interpretation tasks, powered by large network architectures and fueled by the emergence of large labeled datasets. The purpose of this study is to investigate the performance of a deep learning algorithm on the detection of pathologies in chest radiographs compared with practicing radiologists.Methods and Findings: We developed CheXNeXt, a convolutional neural network to concurrently detect the presence of 14 different pathologies, including pneumonia, pleural effusion, pulmonary masses, and nodules in frontal-view chest radiographs. CheXNeXt was trained and internally validated on the ChestX-ray8 dataset, with a held-out validation set consisting of 420 images, sampled to contain at least 50 cases of each of the original pathology labels. On this validation set, the majority vote of a panel of 3 board-certified cardiothoracic specialist radiologists served as reference standard. We compared CheXNeXt's discriminative performance on the validation set to the performance of 9 radiologists using the area under the receiver operating characteristic curve (AUC). The radiologists included 6 board-certified radiologists (average experience 12 years, range 4-28 years) and 3 senior radiology residents, from 3 academic institutions. We found that CheXNeXt achieved radiologist-level performance on 11 pathologies and did not achieve radiologist-level performance on 3 pathologies. The radiologists achieved statistically significantly higher AUC performance on cardiomegaly, emphysema, and hiatal hernia, with AUCs of 0.888 (95% confidence interval [CI] 0.863-0.910), 0.911 (95% CI 0.866-0.947), and 0.985 (95% CI 0.974-0.991), respectively, whereas CheXNeXt's AUCs were 0.831 (95% CI 0.790-0.870), 0.704 (95% CI 0.567-0.833), and 0.851 (95% CI 0.785-0.909), respectively. CheXNeXt performed better than radiologists in detecting atelectasis, with an AUC of 0.862 (95% CI 0.825-0.895), statistically significantly higher than radiologists' AUC of 0.808 (95% CI 0.777-0.838); there were no statistically significant differences in AUCs for the other 10 pathologies. The average time to interpret the 420 images in the validation set was substantially longer for the radiologists (240 minutes) than for CheXNeXt (1.5 minutes). The main limitations of our study are that neither CheXNeXt nor the radiologists were permitted to use patient history or review prior examinations and that evaluation was limited to a dataset from a single institution.Conclusions: In this study, we developed and validated a deep learning algorithm that classified clinically important abnormalities in chest radiographs at a performance level comparable to practicing radiologists. Once tested prospectively in clinical settings, the algorithm could have the potential to expand patient access to chest radiograph diagnostics. [ABSTRACT FROM AUTHOR]

Background: There is interest in using convolutional neural networks (CNNs) to analyze medical imaging to provide computer-aided diagnosis (CAD). Recent work has suggested that image classification CNNs may not generalize to new data as well as previously believed. We assessed how well CNNs generalized across three hospital systems for a simulated pneumonia screening task. Methods and findings: A cross-sectional design with multiple model training cohorts was used to evaluate model generalizability to external sites using split-sample validation. A total of 158,323 chest radiographs were drawn from three institutions: National Institutes of Health Clinical Center (NIH; 112,120 from 30,805 patients), Mount Sinai Hospital (MSH; 42,396 from 12,904 patients), and Indiana University Network for Patient Care (IU; 3,807 from 3,683 patients). These patient populations had an age mean (SD) of 46.9 years (16.6), 63.2 years (16.5), and 49.6 years (17) with a female percentage of 43.5%, 44.8%, and 57.3%, respectively. We assessed individual models using the area under the receiver operating characteristic curve (AUC) for radiographic findings consistent with pneumonia and compared performance on different test sets with DeLong’s test. The prevalence of pneumonia was high enough at MSH (34.2%) relative to NIH and IU (1.2% and 1.0%) that merely sorting by hospital system achieved an AUC of 0.861 (95% CI 0.855–0.866) on the joint MSH–NIH dataset. Models trained on data from either NIH or MSH had equivalent performance on IU (P values 0.580 and 0.273, respectively) and inferior performance on data from each other relative to an internal test set (i.e., new data from within the hospital system used for training data; P values both <0.001). The highest internal performance was achieved by combining training and test data from MSH and NIH (AUC 0.931, 95% CI 0.927–0.936), but this model demonstrated significantly lower external performance at IU (AUC 0.815, 95% CI 0.745–0.885, P = 0.001). To test the effect of pooling data from sites with disparate pneumonia prevalence, we used stratified subsampling to generate MSH–NIH cohorts that only differed in disease prevalence between training data sites. When both training data sites had the same pneumonia prevalence, the model performed consistently on external IU data (P = 0.88). When a 10-fold difference in pneumonia rate was introduced between sites, internal test performance improved compared to the balanced model (10× MSH risk P < 0.001; 10× NIH P = 0.002), but this outperformance failed to generalize to IU (MSH 10× P < 0.001; NIH 10× P = 0.027). CNNs were able to directly detect hospital system of a radiograph for 99.95% NIH (22,050/22,062) and 99.98% MSH (8,386/8,388) radiographs. The primary limitation of our approach and the available public data is that we cannot fully assess what other factors might be contributing to hospital system–specific biases. Conclusion: Pneumonia-screening CNNs achieved better internal than external performance in 3 out of 5 natural comparisons. When models were trained on pooled data from sites with different pneumonia prevalence, they performed better on new pooled data from these sites but not on external data. CNNs robustly identified hospital system and department within a hospital, which can have large differences in disease burden and may confound predictions. [ABSTRACT FROM AUTHOR]

Background: HIV testing rates are suboptimal among at-risk men. Crowdsourcing may be a useful tool for designing innovative, community-based HIV testing strategies to increase HIV testing. The purpose of this study was to use a stepped wedge cluster randomized controlled trial (RCT) to evaluate the effect of a crowdsourced HIV intervention on HIV testing uptake among men who have sex with men (MSM) in eight Chinese cities.Methods and Findings: An HIV testing intervention was developed through a national image contest, a regional strategy designathon, and local message contests. The final intervention included a multimedia HIV testing campaign, an online HIV testing service, and local testing promotion campaigns tailored for MSM. This intervention was evaluated using a closed cohort stepped wedge cluster RCT in eight Chinese cities (Guangzhou, Shenzhen, Zhuhai, and Jiangmen in Guangdong province; Jinan, Qingdao, Yantai, and Jining in Shandong province) from August 2016 to August 2017. MSM were recruited through Blued, a social networking mobile application for MSM, from July 29 to August 21 of 2016. The primary outcome was self-reported HIV testing in the past 3 months. Secondary outcomes included HIV self-testing, facility-based HIV testing, condom use, and syphilis testing. Generalized linear mixed models (GLMMs) were used to analyze primary and secondary outcomes. We enrolled a total of 1,381 MSM. Most were ≤30 years old (82%), unmarried (86%), and had a college degree or higher (65%). The proportion of individuals receiving an HIV test during the intervention periods within a city was 8.9% (95% confidence interval [CI] 2.2-15.5) greater than during the control periods. In addition, the intention-to-treat analysis showed a higher probability of receiving an HIV test during the intervention periods as compared to the control periods (estimated risk ratio [RR] = 1.43, 95% CI 1.19-1.73). The intervention also increased HIV self-testing (RR = 1.89, 95% CI 1.50-2.38). There was no effect on facility-based HIV testing (RR = 1.00, 95% CI 0.79-1.26), condom use (RR = 1.00, 95% CI 0.86-1.17), or syphilis testing (RR = 0.92, 95% CI 0.70-1.21). A total of 48.6% (593/1,219) of participants reported that they received HIV self-testing. Among men who received two HIV tests, 32 individuals seroconverted during the 1-year study period. Study limitations include the use of self-reported HIV testing data among a subset of men and non-completion of the final survey by 23% of participants. Our study population was a young online group in urban China and the relevance of our findings to other populations will require further investigation.Conclusions: In this setting, crowdsourcing was effective for developing and strengthening community-based HIV testing services for MSM. Crowdsourced interventions may be an important tool for the scale-up of HIV testing services among MSM in low- and middle-income countries (LMIC).Trial Registration: ClinicalTrials.gov NCT02796963. [ABSTRACT FROM AUTHOR]

In a Perspective, David Peiris and Dorairaj Prabhakaran discuss implications and challenges of cardiovascular disease risk assessments in the population of India. [ABSTRACT FROM AUTHOR]

Background: Verbal autopsy (VA) is a practical method for determining probable causes of death at the population level in places where systems for medical certification of cause of death are weak. VA methods suitable for use in routine settings, such as civil registration and vital statistics (CRVS) systems, have developed rapidly in the last decade. These developments have been part of a growing global momentum to strengthen CRVS systems in low-income countries. With this momentum have come pressure for continued research and development of VA methods and the need for a single standard VA instrument on which multiple automated diagnostic methods can be developed.Methods and Findings: In 2016, partners harmonized a WHO VA standard instrument that fully incorporates the indicators necessary to run currently available automated diagnostic algorithms. The WHO 2016 VA instrument, together with validated approaches to analyzing VA data, offers countries solutions to improving information about patterns of cause-specific mortality. This VA instrument offers the opportunity to harmonize the automated diagnostic algorithms in the future.Conclusions: Despite all improvements in design and technology, VA is only recommended where medical certification of cause of death is not possible. The method can nevertheless provide sufficient information to guide public health priorities in communities in which physician certification of deaths is largely unavailable. The WHO 2016 VA instrument, together with validated approaches to analyzing VA data, offers countries solutions to improving information about patterns of cause-specific mortality. [ABSTRACT FROM AUTHOR]

In an Essay, Matthew Todd and colleagues discuss an open source approach to drug development. [ABSTRACT FROM AUTHOR]

In January's Guest Editorial, Sarah Stock and Jenny Myers discuss approaches to fetal and neonatal growth assessment. [ABSTRACT FROM AUTHOR]

Background: Perinatal mortality and morbidity continue to be major global health challenges strongly associated with prematurity and reduced fetal growth, an issue of further interest given the mounting evidence that fetal growth in general is linked to degrees of risk of common noncommunicable diseases in adulthood. Against this background, WHO made it a high priority to provide the present fetal growth charts for estimated fetal weight (EFW) and common ultrasound biometric measurements intended for worldwide use.Methods and Findings: We conducted a multinational prospective observational longitudinal study of fetal growth in low-risk singleton pregnancies of women of high or middle socioeconomic status and without known environmental constraints on fetal growth. Centers in ten countries (Argentina, Brazil, Democratic Republic of the Congo, Denmark, Egypt, France, Germany, India, Norway, and Thailand) recruited participants who had reliable information on last menstrual period and gestational age confirmed by crown-rump length measured at 8-13 wk of gestation. Participants had anthropometric and nutritional assessments and seven scheduled ultrasound examinations during pregnancy. Fifty-two participants withdrew consent, and 1,387 participated in the study. At study entry, median maternal age was 28 y (interquartile range [IQR] 25-31), median height was 162 cm (IQR 157-168), median weight was 61 kg (IQR 55-68), 58% of the women were nulliparous, and median daily caloric intake was 1,840 cal (IQR 1,487-2,222). The median pregnancy duration was 39 wk (IQR 38-40) although there were significant differences between countries, the largest difference being 12 d (95% CI 8-16). The median birthweight was 3,300 g (IQR 2,980-3,615). There were differences in birthweight between countries, e.g., India had significantly smaller neonates than the other countries, even after adjusting for gestational age. Thirty-one women had a miscarriage, and three fetuses had intrauterine death. The 8,203 sets of ultrasound measurements were scrutinized for outliers and leverage points, and those measurements taken at 14 to 40 wk were selected for analysis. A total of 7,924 sets of ultrasound measurements were analyzed by quantile regression to establish longitudinal reference intervals for fetal head circumference, biparietal diameter, humerus length, abdominal circumference, femur length and its ratio with head circumference and with biparietal diameter, and EFW. There was asymmetric distribution of growth of EFW: a slightly wider distribution among the lower percentiles during early weeks shifted to a notably expanded distribution of the higher percentiles in late pregnancy. Male fetuses were larger than female fetuses as measured by EFW, but the disparity was smaller in the lower quantiles of the distribution (3.5%) and larger in the upper quantiles (4.5%). Maternal age and maternal height were associated with a positive effect on EFW, particularly in the lower tail of the distribution, of the order of 2% to 3% for each additional 10 y of age of the mother and 1% to 2% for each additional 10 cm of height. Maternal weight was associated with a small positive effect on EFW, especially in the higher tail of the distribution, of the order of 1.0% to 1.5% for each additional 10 kg of bodyweight of the mother. Parous women had heavier fetuses than nulliparous women, with the disparity being greater in the lower quantiles of the distribution, of the order of 1% to 1.5%, and diminishing in the upper quantiles. There were also significant differences in growth of EFW between countries. In spite of the multinational nature of the study, sample size is a limiting factor for generalization of the charts.Conclusions: This study provides WHO fetal growth charts for EFW and common ultrasound biometric measurements, and shows variation between different parts of the world. [ABSTRACT FROM AUTHOR]

In a Perspective, Gordon Schiff discusses the importance of appropriately analyzing adverse event reports. [ABSTRACT FROM AUTHOR]

Background: In February 2011, an Ohio law took effect mandating use of the United States Food and Drug Administration (FDA)-approved protocol for mifepristone, which is used with misoprostol for medication abortion. Other state legislatures have passed or enacted similar laws requiring use of the FDA-approved protocol for medication abortion. The objective of this study is to examine the association of this legal change with medication abortion outcomes and utilization.Methods and Findings: We used a retrospective cohort design, comparing outcomes of medication abortion patients in the prelaw period to those in the postlaw period. Sociodemographic and clinical chart data were abstracted from all medication abortion patients from 1 y prior to the law's implementation (January 2010-January 2011) to 3 y post implementation (February 2011-October 2014) at four abortion-providing health care facilities in Ohio. Outcome data were analyzed for all women undergoing abortion at ≤49 d gestation during the study period. The main outcomes were as follows: need for additional intervention following medication abortion (such as aspiration, repeat misoprostol, and blood transfusion), frequency of continuing pregnancy, reports of side effects, and the proportion of abortions that were medication abortions (versus other abortion procedures). Among the 2,783 medication abortions ≤49 d gestation, 4.9% (95% CI: 3.7%-6.2%) in the prelaw and 14.3% (95% CI: 12.6%-16.0%) in the postlaw period required one or more additional interventions. Women obtaining a medication abortion in the postlaw period had three times the odds of requiring an additional intervention as women in the prelaw period (adjusted odds ratio [AOR] = 3.11, 95% CI: 2.27-4.27). In a mixed effects multivariable model that uses facility-months as the unit of analysis to account for lack of independence by site, we found that the law change was associated with a 9.4% (95% CI: 4.0%-18.4%) absolute increase in the rate of requiring an additional intervention. The most common subsequent intervention in both periods was an additional misoprostol dose and was most commonly administered to treat incomplete abortion. The percentage of women requiring two or more follow-up visits increased from 4.2% (95% CI: 3.0%-5.3%) in the prelaw period to 6.2% (95% CI: 5.5%-8.0%) in the postlaw period (p = 0.003). Continuing pregnancy was rare (0.3%). Overall, 12.6% of women reported at least one side effect during their medication abortion: 8.4% (95% CI: 6.8%-10.0%) in the prelaw period and 15.6% (95% CI: 13.8%-17.3%) in the postlaw period (p < 0.001). Medication abortions fell from 22% (95% CI: 20.8%-22.3%) of all abortions the year before the law went into effect (2010) to 5% (95% CI: 4.8%-5.6%) 3 y after (2014) (p < 0.001). The average patient charge increased from US$426 in 2010 to US$551 in 2014, representing a 16% increase after adjusting for inflation in medical prices. The primary limitation to the study is that it was a pre/post-observational study with no control group that was not exposed to the law.Conclusions: Ohio law required use of a medication abortion protocol that is associated with a greater need for additional intervention, more visits, more side effects, and higher costs for women relative to the evidence-based protocol. There is no evidence that the change in law led to improved abortion outcomes. Indeed, our findings suggest the opposite. In March 2016, the FDA-protocol was updated, so Ohio providers may now legally provide current evidence-based protocols. However, this law is still in place and bans physicians from using mifepristone based on any new developments in clinical research as best practices continue to be updated. [ABSTRACT FROM AUTHOR]

Background: Dentists prescribe approximately 10% of antibiotics dispensed in UK community pharmacies. Despite clear clinical guidance, dentists often prescribe antibiotics inappropriately. This cluster-randomised controlled trial used routinely collected National Health Service (NHS) dental prescribing and treatment claim data to compare the impact of individualised audit and feedback (A&F) interventions on dentists' antibiotic prescribing rates.Methods and Findings: All 795 antibiotic prescribing NHS general dental practices in Scotland were included. Practices were randomised to the control (practices = 163; dentists = 567) or A&F intervention group (practices = 632; dentists = 1,999). A&F intervention practices were allocated to one of two A&F groups: (1) individualised graphical A&F comprising a line graph plotting an individual dentist's monthly antibiotic prescribing rate (practices = 316; dentists = 1,001); or (2) individualised graphical A&F plus a written behaviour change message synthesising and reiterating national guidance recommendations for dental antibiotic prescribing (practices = 316; dentists = 998). Intervention practices were also simultaneously randomised to receive A&F: (i) with or without a health board comparator comprising the addition of a line to the graphical A&F plotting the monthly antibiotic prescribing rate of all dentists in the health board; and (ii) delivered at 0 and 6 mo or at 0, 6, and 9 mo, giving a total of eight intervention groups. The primary outcome, measured by the trial statistician who was blinded to allocation, was the total number of antibiotic items dispensed per 100 NHS treatment claims over the 12 mo post-delivery of the baseline A&F. Primary outcome data was available for 152 control practices (dentists = 438) and 609 intervention practices (dentists = 1,550). At baseline, the number of antibiotic items prescribed per 100 NHS treatment claims was 8.3 in the control group and 8.5 in the intervention group. At follow-up, antibiotic prescribing had decreased by 0.4 antibiotic items per 100 NHS treatment claims in control practices and by 1.0 in intervention practices. This represents a significant reduction (-5.7%; 95% CI -10.2% to -1.1%; p = 0.01) in dentists' prescribing rate in the intervention group relative to the control group. Intervention subgroup analyses found a 6.1% reduction in the antibiotic prescribing rate of dentists who had received the written behaviour change message relative to dentists who had not (95% CI -10.4% to -1.9%; p = 0.01). There was no significant between-group difference in the prescribing rate of dentists who received a health board comparator relative to those who did not (-4.3%; 95% CI -8.6% to 0.1%; p = 0.06), nor between dentists who received A&F at 0 and 6 mo relative to those who received A&F at 0, 6, and 9 mo (0.02%; 95% CI -4.2% to 4.2%; p = 0.99). The key limitations relate to the use of routinely collected datasets which did not allow evaluation of any effects on inappropriate prescribing.Conclusions: A&F derived from routinely collected datasets led to a significant reduction in the antibiotic prescribing rate of dentists.Trial Registration: Current Controlled Trials ISRCTN49204710. [ABSTRACT FROM AUTHOR]

Background: The intra-tumor diversity of cancer cells is under intense investigation; however, little is known about the heterogeneity of the tumor microenvironment that is key to cancer progression and evolution. We aimed to assess the degree of microenvironmental heterogeneity in breast cancer and correlate this with genomic and clinical parameters.Methods and Findings: We developed a quantitative measure of microenvironmental heterogeneity along three spatial dimensions (3-D) in solid tumors, termed the tumor ecosystem diversity index (EDI), using fully automated histology image analysis coupled with statistical measures commonly used in ecology. This measure was compared with disease-specific survival, key mutations, genome-wide copy number, and expression profiling data in a retrospective study of 510 breast cancer patients as a test set and 516 breast cancer patients as an independent validation set. In high-grade (grade 3) breast cancers, we uncovered a striking link between high microenvironmental heterogeneity measured by EDI and a poor prognosis that cannot be explained by tumor size, genomics, or any other data types. However, this association was not observed in low-grade (grade 1 and 2) breast cancers. The prognostic value of EDI was superior to known prognostic factors and was enhanced with the addition of TP53 mutation status (multivariate analysis test set, p = 9 × 10-4, hazard ratio = 1.47, 95% CI 1.17-1.84; validation set, p = 0.0011, hazard ratio = 1.78, 95% CI 1.26-2.52). Integration with genome-wide profiling data identified losses of specific genes on 4p14 and 5q13 that were enriched in grade 3 tumors with high microenvironmental diversity that also substratified patients into poor prognostic groups. Limitations of this study include the number of cell types included in the model, that EDI has prognostic value only in grade 3 tumors, and that our spatial heterogeneity measure was dependent on spatial scale and tumor size.Conclusions: To our knowledge, this is the first study to couple unbiased measures of microenvironmental heterogeneity with genomic alterations to predict breast cancer clinical outcome. We propose a clinically relevant role of microenvironmental heterogeneity for advanced breast tumors, and highlight that ecological statistics can be translated into medical advances for identifying a new type of biomarker and, furthermore, for understanding the synergistic interplay of microenvironmental heterogeneity with genomic alterations in cancer cells. [ABSTRACT FROM AUTHOR]

: Minnie Sarwal and colleagues developed a gene expression assay using peripheral blood samples to detect patients with renal transplant at high risk for acute rejection. Please see later in the article for the Editors' Summary [ABSTRACT FROM AUTHOR]