Underrepresented Subgroups Research Articles

Prioritizing cases from a multi-institutional cohort for a dataset of pathologist annotations

ObjectiveWith the increasing energy surrounding the development of artificial intelligence and machine learning (AI/ML) models, the use of the same external validation dataset by various developers allows for a direct comparison of model performance. Through our High Throughput Truthing project, we are creating a validation dataset for AI/ML models trained in the assessment of stromal tumor-infiltrating lymphocytes (sTILs) in triple negative breast cancer (TNBC). Materials and methodsWe obtained clinical metadata for hematoxylin and eosin-stained glass slides and corresponding scanned whole slide images (WSIs) of TNBC core biopsies from two US academic medical centers. We selected regions of interest (ROIs) from the WSIs to target regions with various tissue morphologies and sTILs densities. Given the selected ROIs, we implemented a hierarchical rank-sort method for case prioritization. ResultsWe received 122 glass slides and clinical metadata on 105 unique patients with TNBC. All received cases were female, and the mean age was 63.44 years. 60% of all cases were White patients, and 38.1% were Black or African American. After case prioritization, the skewness of the sTILs density distribution improved from 0.60 to 0.46 with a corresponding increase in the entropy of the sTILs density bins from 1.20 to 1.24. We retained cases with less prevalent metadata elements. ConclusionThis method allows us to prioritize underrepresented subgroups based on important clinical factors. In this manuscript, we discuss how we sourced the clinical metadata, selected ROIs, and developed our approach to prioritizing cases for inclusion in our pivotal study.

Safety of treating acute pulmonary embolism at home: an individual patient data meta-analysis.

Home treatment is considered safe in acute pulmonary embolism (PE) patients selected by a validated triage tool (e.g. simplified PE severity index score or Hestia rule), but there is uncertainty regarding the applicability in underrepresented subgroups. The aim was to evaluate the safety of home treatment by performing an individual patient-level data meta-analysis. Ten prospective cohort studies or randomized controlled trials were identified in a systematic search, totalling 2694 PE patients treated at home (discharged within 24 h) and identified by a predefined triage tool. The 14- and 30-day incidences of all-cause mortality and adverse events (combined endpoint of recurrent venous thromboembolism, major bleeding, and/or all-cause mortality) were evaluated. The relative risk (RR) for 14- and 30-day mortalities and adverse events is calculated in subgroups using a random effects model. The 14- and 30-day mortalities were 0.11% [95% confidence interval (CI) 0.0-0.24, I2 = 0) and 0.30% (95% CI 0.09-0.51, I2 = 0). The 14- and 30-day incidences of adverse events were 0.56% (95% CI 0.28-0.84, I2 = 0) and 1.2% (95% CI 0.79-1.6, I2 = 0). Cancer was associated with increased 30-day mortality [RR 4.9; 95% prediction interval (PI) 2.7-9.1; I2 = 0]. Pre-existing cardiopulmonary disease, abnormal troponin, and abnormal (N-terminal pro-)B-type natriuretic peptide [(NT-pro)BNP] at presentation were associated with an increased incidence of 14-day adverse events [RR 3.5 (95% PI 1.5-7.9, I2 = 0), 2.5 (95% PI 1.3-4.9, I2 = 0), and 3.9 (95% PI 1.6-9.8, I2 = 0), respectively], but not mortality. At 30 days, cancer, abnormal troponin, and abnormal (NT-pro)BNP were associated with an increased incidence of adverse events [RR 2.7 (95% PI 1.4-5.2, I2 = 0), 2.9 (95% PI 1.5-5.7, I2 = 0), and 3.3 (95% PI 1.6-7.1, I2 = 0), respectively]. The incidence of adverse events in home-treated PE patients, selected by a validated triage tool, was very low. Patients with cancer had a three- to five-fold higher incidence of adverse events and death. Patients with increased troponin or (NT-pro)BNP had a three-fold higher risk of adverse events, driven by recurrent venous thromboembolism and bleeding.

Exploring biorepository donation patterns, experiences, and recommendations: a mixed-methods study among Appalachian adults enrolled in a sugary drink reduction program.

Under-represented subgroups in biomarker research linked to behavioral health trials may impact the promise of precision health. This mixed methods study examines biorepository donations across an Appalachian sample enrolled in a sugary drink reduction intervention trial. Participants enrolled in the behavioral trial were asked to join an optional biomarker study and were tracked for enrollment and biospecimen returns (stool and/or buccal sample). At 6 months, participants completed a summative interview on decision-making process, experiences collecting samples, and recommendations to encourage biospecimen donation. Return rates were analyzed across demographics (i.e., age, gender, race, education, income, health literacy status, and rurality status) using chi-squares. Qualitative data were content coded with differences compared by biomarker study enrollment and donation choices. Of the 249 invited participants, 171 (61%) enrolled, and 63% (n = 157) returned buccal samples and 49% (n = 122) returned stool samples. Metro residing participants were significantly more likely (56%) to return stool samples compared to non-metro (39%) counterparts [x2(1) = 6.61; p = 0.01]. Buccal sample return had a similar trend, 67 and 57%, respectively for metro vs. non-metro [x2(1) = 2.84; p = 0.09]. An additional trend indicated that older (≥40 years) participants were more likely (55%) to donate stool samples than younger (43%) participants [x2(1) = 3.39; p = 0.07]. No other demographics were significantly associated with biospecimen return. Qualitative data indicated that societal (66-81%) and personal (41-51%) benefits were the most reported reasons for deciding to donate one or both samples, whereas mistrust (3-11%) and negative perceptions of the collection process (44-71%) were cited the most by those who declined one or both samples. Clear instructions (60%) and simple collection kits (73%) were donation facilitators while challenges included difficult stool collection kits (16%) and inconveniently located FedEx centers (16%). Recommendations to encourage future biorepository donation were to clarify benefits to science and others (58%), provide commensurate incentives (25%), explain purpose (19%) and privacy protections (20%), and assure ease in sample collection (19%). Study findings suggest the need for biomarker research awareness campaigns. Researchers planning for future biomarker studies in medically underserved regions, like Appalachia, may be able to apply findings to optimize enrollment.

“A sincere ‘how are you?’ is already a sign of acknowledgement that you're there too.” - Interview study on the support needs of adolescents and young adults (AYAs) living with a parent with cancer

PurposeParental cancer brings changes and challenges which affect the whole family. Evidence shows heightened psychosocial risk among the offspring. Research among adolescents and young adults (AYAs) facing parental cancer has mainly focused on these psychosocial problems. As a reaction, there has been an increase in research on the needs of this population, since this knowledge is crucial for developing appropriate support. Despite this increase, some AYAs are still underrepresented, namely those over eighteen years old and those who have a parent with incurable cancer. Moreover, the factors hampering or fostering access to and receipt of support addressing their needs are under-investigated. Increasing our understanding of these issues is crucial to developing more effective support interventions for these young people. Our aim was therefore to gain insight into the perceived support needs of AYAs who have a parent with cancer, including the mentioned underrepresented subgroups, and the perceived factors facilitating or hindering their use of support. MethodsWe conducted semi-structured interviews with 17 AYAs who have a parent with cancer of any type and stage, which were analyzed using conventional Qualitative Content Analysis (QCA). ResultsSix main categories emerged from the data, encompassing various subcategories: 1) Needs surrounding the time of disclosure, 2) Information needs, 3) Informal support needs, 4) Need for effective personal coping strategies, 5) Formal support needs, and 6) Contextual and medical factors affecting support needs. AYAs perceived family members, friends and teachers as important sources of informal support, but lacked contact with peers who also have a parent with cancer. They expressed a need for formal informational and emotional support from their parent's medical team, (onco)psychologists, and their general practitioner. AYAs faced barriers in accessing formal support. A proactive attitude from healthcare providers and information about available support would help them meet their needs. Furthermore, we shed new light on needs and issues in post-secondary education. ConclusionsAYAs who are living with their parent with cancer experience informational, emotional, and practical needs in multiple contexts, such as the home, school, leisure and hospital environment. They identify their informal and formal network as important key figures in addressing these needs.

Transfer learning under the Cox model with interval‐censored data

AbstractTransfer learning, focusing on information borrowing to address limited sample size issues, has gained increasing attention in recent years. Our method aims to utilize data from other population groups as a complement to enhance risk factor discernment and failure time prediction among underrepresented subgroups. However, a literature gap exists in effective knowledge transfer from the source to the target for risk assessment with interval‐censored data while accommodating population incomparability and privacy constraints. Our objective is to bridge this gap by developing a transfer learning approach under the Cox proportional hazards model. We introduce the tuning‐free Trans‐Cox‐MIC algorithm, enabling adaptable information sharing in regression coefficients and baseline hazards, while ensuring computational efficiency. Our approach accommodates covariate distribution shifts, coefficient variations, and baseline hazard discrepancies. Extensive simulations showcase the method's accuracy, robustness, and efficiency. Application to the prostate cancer screening data demonstrates enhanced risk estimation precision and predictive performance in the African American population.

Pregnancy health in a multi-state U.S. population of systemically underserved patients and their children: PROMISE cohort design and baseline characteristics

BackgroundGestational weight gain (GWG) is a routinely monitored aspect of pregnancy health, yet critical gaps remain about optimal GWG in pregnant people from socially marginalized groups, or with pre-pregnancy body mass index (BMI) in the lower or upper extremes. The PROMISE study aims to determine overall and trimester-specific GWG associated with the lowest risk of adverse birth outcomes and detrimental infant and child growth in these underrepresented subgroups. This paper presents methods used to construct the PROMISE cohort using electronic health record data from a network of community-based healthcare organizations and characterize the cohort with respect to baseline characteristics, longitudinal data availability, and GWG.MethodsWe developed an algorithm to identify and date pregnancies based on outpatient clinical data for patients 15 years or older. The cohort included pregnancies delivered in 2005–2020 with gestational age between 20 weeks, 0 days and 42 weeks, 6 days; and with known height and adequate weight measures needed to examine GWG patterns. We linked offspring data from birth records and clinical records. We defined study variables with attention to timing relative to pregnancy and clinical data collection processes. Descriptive analyses characterize the sociodemographic, baseline, and longitudinal data characteristics of the cohort, overall and within BMI categories.ResultsThe cohort includes 77,599 pregnancies: 53% had incomes below the federal poverty level, 82% had public insurance, and the largest race and ethnicity groups were Hispanic (56%), non-Hispanic White (23%) and non-Hispanic Black (12%). Pre-pregnancy BMI groups included 2% underweight, 34% normal weight, 31% overweight, and 19%, 8%, and 5% Class I, II, and III obesity. Longitudinal data enable the calculation of trimester-specific GWG; e.g., a median of 2, 4, and 6 valid weight measures were available in the first, second, and third trimesters, respectively. Weekly rate of GWG was 0.00, 0.46, and 0.51 kg per week in the first, second, and third trimesters; differences in GWG between BMI groups were greatest in the second trimester.ConclusionsThe PROMISE cohort enables characterization of GWG patterns and estimation of effects on child growth in underrepresented subgroups, ultimately improving the representativeness of GWG evidence and corresponding guidelines.

Demographic diversity of genetic databases used in Alzheimer's disease research.

For several years, experts have warned about the lack of diversity in genetic research databases, and researchers have devoted time and resources to recruiting subjects from underrepresented subgroups. In this study, we review published reports in academic journals of genetic studies of Alzheimer's disease to note whether demographic diversity was indicated in the reports and, if so, the extent of representation of non-European subjects over the period from 1997 to 2022. We use multivariate regression analysis to analyze changes over time and to explain variation across studies. Our analysis indicates that reported diversity has not changed over time. Rather, it appears to have remained relatively constant, since Genome-Wide Association Studies (GWASs) were first used in the 1990s. We find most variation to be across journals rather than within journals, suggesting that characteristics of journals are an important influence on the dissemination of research with diverse samples. Lack of racial diversity in genetic databases used to develop clinical applications could lead to disparities in the effectiveness of those applications for underrepresented groups.

Higher performance for women than men in MRI-based Alzheimer’s disease detection

IntroductionAlthough machine learning classifiers have been frequently used to detect Alzheimer’s disease (AD) based on structural brain MRI data, potential bias with respect to sex and age has not yet been addressed. Here, we examine a state-of-the-art AD classifier for potential sex and age bias even in the case of balanced training data.MethodsBased on an age- and sex-balanced cohort of 432 subjects (306 healthy controls, 126 subjects with AD) extracted from the ADNI data base, we trained a convolutional neural network to detect AD in MRI brain scans and performed ten different random training-validation-test splits to increase robustness of the results. Classifier decisions for single subjects were explained using layer-wise relevance propagation.ResultsThe classifier performed significantly better for women (balanced accuracy 87.58pm 1.14%) than for men (79.05pm 1.27%). No significant differences were found in clinical AD scores, ruling out a disparity in disease severity as a cause for the performance difference. Analysis of the explanations revealed a larger variance in regional brain areas for male subjects compared to female subjects.DiscussionThe identified sex differences cannot be attributed to an imbalanced training dataset and therefore point to the importance of examining and reporting classifier performance across population subgroups to increase transparency and algorithmic fairness. Collecting more data especially among underrepresented subgroups and balancing the dataset are important but do not always guarantee a fair outcome.

Asian New Zealanders' experiences of racism during the COVID-19 pandemic and its association with life satisfaction.

Racism is an important social determinant of wellbeing. This study describes New Zealand Asians' experience of racism and the association between their racism experiences and their impacts on life satisfaction during the COVID-19 pandemic. This study collected 1,452 responses by the cross-sectional online survey conducted in 2021. Descriptive and logistic regression analyses yielded the prevalent types and settings of racism, identified subgroups exposed more to racism and its association with life satisfaction. Results show that nearly 40% (37.7-42.9%) of participants experienced racism, mainly in public places, social media and mainstream media. Verbal attacks and microaggressions were predominant types of racism. Younger, student, temporary visa holding, and rural area participants were more likely to experience racism. Associations between not experiencing racism and high life satisfaction were significant. This study identified two under-represented subgroups, students and migrants, who were disproportionately exposed to racism. It also revealed that experiences of COVID-era racism are associated with life satisfaction. These findings inform us where anti-racism interventions are most needed, and that such interventions are needed to ensure the wellbeing of Asian communities in a COVID-19 world.

Under-representation of key demographic groups in opioid use disorder trials

BackgroundThe extent to which clinical trials of medications for opioid use disorder (MOUD) are representative or not is unknown. Some patient characteristics modify MOUD effectiveness; if these same characteristics differ in distribution between the trial population and usual-care population, this could contribute to lack of generalizability—a discrepancy between trial and usual-care effectiveness. Our objective was to identify interpretable, multidimensional subgroups who were prescribed MOUD in substance use treatment programs in the US but who were not represented or under-represented by clinical trial participants. MethodsThis was a secondary descriptive analysis of trial and real-world data. The trial data included twenty-seven US opioid treatment programs in the National Drug Abuse Treatment Clinical Trials Network, N=2,199 patients. The real-world data included US substance use treatment programs that receive public funding, N=740,015 patients. We characterized real-world patient populations who were non-represented and under-represented in the trial data in terms of sociodemographic and clinical characteristics that could modify MOUD effectiveness. ResultsWe found that 10.7% of MOUD patients in TEDS-A were not represented in the three clinical trials. As expected, pregnant MOUD patients (n=19,490) were not represented. Excluding pregnancy, education and marital status from the characteristics, 2.6% of MOUD patients were not represented. Patients aged 65 years and older (n=11,204), and those 50-64 years who identified as other (non-White, non-Black, and non-Hispanic) race/ethnicity or multi-racial (n=7,281) were under-represented. ConclusionsQuantifying and characterizing non- or under-represented subgroups in trials can provide the data necessary to improve representation in future trials and address research-to-practice gaps.

External Validation of Two Established Clinical Risk Scores Predicting Outcome after Local Treatment of Colorectal Liver Metastases in a Nationwide Cohort.

Simple SummaryPatients with colorectal liver metastases (CRLM) are able to achieve long-term survival when they receive local treatment of CRLM (resection or tumor ablation). Existing clinical risk scores (CRSs) predicting prognosis of patients after resection of colorectal liver metastases were developed in highly specialized centers and thus may not function in the general population. We validated the Fong and GAME CRSs in a large population-based cohort, including two important subgroups: young/elderly and with/without perioperative chemotherapy. Both CRSs showed predictive ability. However, they were not able to discriminate preoperative risk sufficiently for clinical decision-making and, thus, require improvement.Optimized surgical techniques and systemic therapy have increased the number of patients with colorectal liver metastases (CRLM) eligible for local treatment. To increase postoperative survival, we need to stratify patients to customize therapy. Most clinical risk scores (CRSs) which predict prognosis after CRLM resection were based on the outcome of studies in specialized centers, and this may hamper the generalizability of these CRSs in unselected populations and underrepresented subgroups. We aimed to externally validate two CRSs in a population-based cohort of patients with CRLM. A total of 1105 patients with local treatment of CRLM, diagnosed in 2015/2016, were included from a nationwide population-based database. Survival outcomes were analyzed. The Fong and more recently developed GAME CRS were externally validated, including in pre-specified subgroups (≤70/>70 years and with/without perioperative systemic therapy). The three-year DFS was 22.8%, and the median OS in the GAME risk groups (high/moderate/low) was 32.4, 46.7, and 68.1 months, respectively (p < 0.005). The median OS for patients with versus without perioperative therapy was 47.6 (95%CI [39.8, 56.2]) and 54.9 months (95%CI [48.8, 63.7]), respectively (p = 0.152), and for below/above 70 years, it was 54.9 (95%CI [49.3–64.1]) and 44.2 months (95%CI [37.1–54.3]), respectively (p < 0.005). The discriminative ability for OS of Fong CRS was 0.577 (95%CI [0.554, 0.601]), and for GAME, it was 0.596 (95%CI [0.572, 0.621]), and was comparable in the subgroups. In conclusion, both CRSs showed predictive ability in a population-based cohort and in predefined subgroups. However, the limited discriminative ability of these CRSs results in insufficient preoperative risk stratification for clinical decision-making.

Abstract P122: Addressed-based Sampling Of People Living In Rural Southeast Usa

Introduction: The RURAL cohort study addresses critical knowledge gaps regarding heart and lung disorders in rural Southeastern USA. Using CDC mortality data, RURAL selected ten counties from four states (Alabama, Mississippi, Louisiana, and Kentucky) that are ecologically paired for their five-year low and high CVD mortality rates. RURAL will recruit a representative sample of 4600 participants using addressed-based sampling and an intensive community engagement, yet representative sampling from rural areas is challenging. Hypothesis: Address-based sampling using dedicated mailings and telephone calls will yield representative samples from the ten target counties. Methods: A dynamic sampling strategy was implemented to obtain balanced response rates for sex, age groups and race. An initial wave of random addresses was used to estimate response rates. A second wave with Bernoulli probabilities was used to reduce overrepresented subgroups. Finally, a third wave with requests for referrals (network sampling) will be used to increase underrepresented subgroups. Eligible participants collected from households and referrals are rostered, complete a baseline exam, undergo an mHealth assessment and are scheduled for a medical examination in a mobile examination unit (MEU). Results: Sampling of 693 participants from 19,460 inhabitants in age range 25 to 64 years started in Dallas County, Alabama. Address sizes, rostered participants and sample size imbalances are listed in the Table. Bernoulli probabilities [55-64 men: 80%; 55-65 women: 55%; 45-55 women:76%] corrected the imbalance in response rates for sex and age groups 45-54 and 55-64 (see Table). Referrals in wave 2 will further balance the sample. Conclusion: Our initial sampling experiences in rural Alabama showed lower than expected response rates and imbalances across subgroups. These results emphasize the importance of network sampling with well-adjusted Bernoulli probabilities for representative sampling in rural areas.

Abstract P207: Dietary Diversity And Quality And Cvd Risk Factors In The Reasons For Geographic And Racial Differences In Stroke (REGARDS)

Background: Diet diversity and quality are each recommended to promote a healthy diet and prevent chronic disease. Yet, little is known on how diet diversity relates to diet quality or cardiometabolic health, particularly in underrepresented subgroups disproportionally affected by poor nutrition and CVD. Aims: To determine relationships of 1) dietary diversity with diet quality, and 2) diet diversity and quality with CVD risk factors (blood pressure, plasma LDL-C, HDL-C, triglycerides, total:HDL ratio, CRP, glucose) in a diverse population of U.S. adults. Methods: We evaluated 21,350 Black and White adults aged ≥45 years in REGARDS. A validated dietary questionnaire at baseline was used to estimate two distinct aspects of diet diversity: food count (no. of food items eaten at least once a week) and dissimilarity (Jaccard distance, capturing diversity of attributes of foods eaten). Diet quality was estimated using the AHA-2020 and AHEI-2010 scores. Cross-sectional associations with CVD risk factors were examined using multivariable linear regression. Results: Food count and dissimilarity each inversely correlated with diet quality scores (r ranging from -0.04 to -0.11). In multivariable models, diet diversity had mixed or mostly null associations with CVD risk factors (See Table). For instance, food count inversely associated with HDL-C and LDL-C; while dissimilarity inversely associated with plasma glucose but positively with triglycerides. In contrast, AHA-2020 and AHEI-2010 were consistently favorably associated with nearly all CVD risk factors. Associations of diet diversity were similar in Blacks and Whites, while most associations of diet quality were stronger among Whites ( P -interaction &lt; 0.0015). Conclusion: Diet diversity metrics were not associated with diet healthfulness or most CVD risk factors among U.S. adults. Our findings support a focus on diet quality, with adequate intake of healthful food groups, rather than diversity, to promote cardiometabolic health.

Using Quotas as a Remedy for Structural Injustice

We analyze a frequent but undertheorized form of structural injustice, one that arises due to the difficulty of reaching numerically equitable representation of underrepresented subgroups within a larger group. This form of structural injustice is significant because it could occur even if it were possible to completely eliminate bias and overt discrimination from hiring and recruitment practices. The conceptual toolkit we develop can be used to analyze such situations and propose remedies. Specifically, based on a simple mathematical model, we offer a new argument in favour of quotas, explore implications for policy-making, and consider the wider philosophical significance of the problem. We show that in order to reach more equitable representations, quota-based recruitment may often be practically unavoidable. Assuming that members of groups in statistical minority are more likely to quit due to their marginalization, their proportions can stabilize at a low level, preventing a shift towards more equal representation and conserving the minority status of the subgroup. We show that this argument has important implications for addressing, preventing, and remediating the structural injustice of unfair representation.

Gender-biased evaluation or actual differences? Fairness in the evaluation of faculty teaching.

How do we know if a faculty teaching evaluation is biased? Biasing factors studies are an influential source of evidence for arguing about biased teaching evaluations. These studies examine existing evaluation data and compare the results by gender, race, or ethnicity, interpreting a significant difference between subgroups as evidence of bias. However, only a difference explained by irrelevant aspects embedded in the evaluation would compromise its fairness. The study aims to amend how practitioners and researchers address gender bias concerns in faculty teaching evaluations by defining fairness, disparate impact, and statistical bias from an educational measurement standpoint. The study illustrates the use of differential item functioning (DIF) analysis, a strategy to examine whether the meaning of an item changes depending on the gender of the instructor. The study examines instructor’s gender bias using responses to a course evaluation questionnaire from education graduate students from two academic departments within the same institution. In one of the departments, the analysis suggested a fair evaluation and no gender gap. In the other department, four of the eight items in the rating scale were easier for women than men with similar teaching ability, and women achieved better evaluations than men. The discussion addresses the conceptual and methodological advantages of adopting an educational measurement perspective on fairness in faculty teaching evaluation. Findings encourage practitioners and administrators to use the best available tools to strengthen the credibility of faculty teaching evaluations and prevent unfair personnel decisions affecting underrepresented subgroups in academia by gender, race, or ethnicity.

Quantifying representativeness in randomized clinical trials using machine learning fairness metrics.

ObjectiveWe help identify subpopulations underrepresented in randomized clinical trials (RCTs) cohorts with respect to national, community-based or health system target populations by formulating population representativeness of RCTs as a machine learning (ML) fairness problem, deriving new representation metrics, and deploying them in easy-to-understand interactive visualization tools.Materials and MethodsWe represent RCT cohort enrollment as random binary classification fairness problems, and then show how ML fairness metrics based on enrollment fraction can be efficiently calculated using easily computed rates of subpopulations in RCT cohorts and target populations. We propose standardized versions of these metrics and deploy them in an interactive tool to analyze 3 RCTs with respect to type 2 diabetes and hypertension target populations in the National Health and Nutrition Examination Survey.ResultsWe demonstrate how the proposed metrics and associated statistics enable users to rapidly examine representativeness of all subpopulations in the RCT defined by a set of categorical traits (eg, gender, race, ethnicity, smoking status, and blood pressure) with respect to target populations.DiscussionThe normalized metrics provide an intuitive standardized scale for evaluating representation across subgroups, which may have vastly different enrollment fractions and rates in RCT study cohorts. The metrics are beneficial complements to other approaches (eg, enrollment fractions) used to identify generalizability and health equity of RCTs.ConclusionBy quantifying the gaps between RCT and target populations, the proposed methods can support generalizability evaluation of existing RCT cohorts. The interactive visualization tool can be readily applied to identified underrepresented subgroups with respect to any desired source or target populations.

Trial eligibility in advanced hepatocellular carcinoma: Does it support clinical practice in underrepresented subgroups?

Although hepatocellular carcinoma is considered a highly lethal malignancy, recent therapeutic advances have been achieved during the last 10 years. This scenario resulted in an unprecedented improvement in survival for patients with advanced hepatocellular carcinoma, almost reaching 20-26 mo of overall survival after first-second line sequential treatment. The advent of the combination of atezolizumab with bevacizumab showed, for the first time, superiority over sorafenib with improvement in overall survival. However, first and second-line trials were correctly based on the premise that a strict selection of patients enhances the power to capture the positive effect of treatment by excluding competing risks for mortality such as liver failure, decompensated cirrhosis or other underlying medical conditions. As a result, the inclusion criteria used in clinical trials do not support the use of novel therapies in several real-world scenarios involving underrepresented subgroups, such as patients with unpreserved liver function, other comorbid conditions, a history of solid-organ transplantation, autoimmune disorders and those with a high risk of bleeding. The present text aims at discussing treatment strategies in these subgroups.

Assessing Nonresponse Bias in Farm Injury Surveillance Data.

Despite a response rate of about 18%, there was scant evidence of nonresponse bias. Farm and ranch characteristics of operations responding to a mailed survey were similar to those not responding. Responders were significantly more likely to be married but only slightly older and more educated than nonresponders. Earlier survey responders reported more injuries and greater severity when injured compared to later responders. Nonresponse bias in a survey can result in misleading estimates of agricultural injuries and can misdirect prevention efforts aimed at reducing the burden of injuries on farmers. Responders (n = 2,977) and nonresponders (n = 13,849) were compared based on demographics and agricultural production characteristics to identify underrepresented subgroups. Injury characteristics were compared between early (n = 1,667) and late (n = 1,309) responders. Methods accounted for correlated data, sample size inflation of p-values, and assessment of meaningful differences. Few differences were identified between responders and nonresponders. Responders differed from nonresponders by state of residence, and responders were more likely to be married. Other characteristics (age, gender, education, farm size, crops grown, animals raised) were similar across groups. Early responders reported more injuries and more often sought medical care for an injury than late responders. The differences identified between responders and nonresponders were minimal and not likely to create bias. Differential reporting of injury and injury severity between early and late responders is worthy of further investigation.

Clinical Trial Generalizability Assessment in the Big Data Era: A Review.

Clinical studies, especially randomized, controlled trials, are essential for generating evidence for clinical practice. However, generalizability is a long‐standing concern when applying trial results to real‐world patients. Generalizability assessment is thus important, nevertheless, not consistently practiced. We performed a systematic review to understand the practice of generalizability assessment. We identified 187 relevant articles and systematically organized these studies in a taxonomy with three dimensions: (i) data availability (i.e., before or after trial (a priori vs. a posteriori generalizability)); (ii) result outputs (i.e., score vs. nonscore); and (iii) populations of interest. We further reported disease areas, underrepresented subgroups, and types of data used to profile target populations. We observed an increasing trend of generalizability assessments, but < 30% of studies reported positive generalizability results. As a priori generalizability can be assessed using only study design information (primarily eligibility criteria), it gives investigators a golden opportunity to adjust the study design before the trial starts. Nevertheless, < 40% of the studies in our review assessed a priori generalizability. With the wide adoption of electronic health records systems, rich real‐world patient databases are increasingly available for generalizability assessment; however, informatics tools are lacking to support the adoption of generalizability assessment practice.

Intersectionality in the narratives of black women in computing through the education and workforce pipeline

PurposeThe purpose of this study is to investigate the narratives of 93 Black women in computing in the USA to identify salient themes that are at the intersection of race and gender in the field of computer science.Design/methodology/approachThe study uses a multi-method approach with a survey to describe the sample and a series of focus groups for in-depth analysis of themes. The qualitative methodology uses a grounded theory and consensual qualitative research approach with a research team that includes computer scientists and social scientists to collect and analyze data. Given the highly technical field of computer science and the intersectional experiences of the participants, this approach was optimal to capture and code data through the lens of Black women in computing.FindingsThe authors found four main themes that represented specific needs for Black women in the computing community. The first is the importance of linking Black women in computing (i.e. their recruitment, retention and career growth) to the bottom line of organizational and personal accountability. The second is effective cultural and educational supports for Black women in computing across pathways, starting in middle school. The third is to provide leadership development as a part of their educational and workplace experience. The fourth is a collection of empirical research and scholarship about and for Black women as a part of the computing literature.Originality/valueBlack women comprise one of the most underrepresented subgroups in the area of computer science in the USA. There is very little research about Black women in computing. To promote broadened participation in computing, there is a critical need to understand the narratives of successful Black women in the space.