AI in medical research.

This review delves into the possible role of artificial intelligence (AI) in medical research, from planning to publication. AI can aid in idea generation, data analysis, and writing, with tools like chatbots and transcription systems enhancing efficiency. However, AI's limitations, including the "hallucination" problem in which it generates false information, require careful use and verification. Ensuring anonymity compliance with sensitive data is also vital. AI's transformative potential in research brings opportunities for innovation, necessitating mindful application to manage biases and data accuracy.

Improving Eye-Tracking Data Quality: A Framework for Reproducible Evaluation of Detection Algorithms.

High-quality eye-tracking data are crucial in behavioral sciences and medicine. Even with a solid understanding of the literature, selecting the most suitable algorithm for a specific research project poses a challenge. Empowering applied researchers to choose the best-fitting detector for their research needs is the primary contribution of this paper. We developed a framework to systematically assess and compare the effectiveness of 13 state-of-the-art algorithms through a unified application interface. Hence, we more than double the number of algorithms that are currently usable within a single software package and allow researchers to identify the best-suited algorithm for a given scientific setup. Our framework validation on retrospective data underscores its suitability for algorithm selection. Through a detailed and reproducible step-by-step workflow, we hope to contribute towards significantly improved data quality in scientific experiments.

Improving state medicolegal death investigation: Does system level reform improve public health data concerning drug-related mortality?

In the United States, the governance of unnatural death certification varies greatly by state. Although cross-sectional research has linked mortality data quality with variation in medicolegal death investigation systems across states-especially with regards to drug-related deaths-this relationship has not be sufficiently tested using longitudinal data. This research assesses the impact of system governance reform on the quality of drug mortality data by assessing the impact of transitioning from a coroner system to a medical examiner system on data quality. The research finds no evidence that system-level reform is associated with improved drug-related mortality data quality. These findings suggest that alternative methods should be examined for improving public health data concerning drug-related mortality. These likely include focusing on individual-level characteristics and practices of officials and offices, rather than system-level variables.

Achieving high inter-rater reliability in establishing data labels: a retrospective chart review study.

BACKGROUND: In medical research, the effectiveness of machine learning algorithms depends heavily on the accuracy of labeled data. This study aimed to assess inter-rater reliability (IRR) in a retrospective electronic medical chart review to create high quality labeled data on comorbidities and adverse events (AEs). METHODS: Six registered nurses with diverse clinical backgrounds reviewed patient charts, extracted data on 20 predefined comorbidities and 18 AEs. All reviewers underwent four iterative rounds of training aimed to enhance accuracy and foster consensus. Periodic monitoring was conducted at the beginning, middle, and end of the testing phase to ensure data quality. Weighted Kappa coefficients were calculated with their associated 95% confidence intervals (CIs). RESULTS: Seventy patient charts were reviewed. The overall agreement, measured by Conger's Kappa, was 0.80 (95% CI: 0.78-0.82). IRR scores remained consistently high (ranging from 0.70 to 0.87) throughout each phase. CONCLUSION: Our study suggests the detailed manual for chart review and structured training regimen resulted in a consistently high level of agreement among our reviewers during the chart review process. This establishes a robust foundation for generating high-quality labeled data, thereby enhancing the potential for developing accurate machine learning algorithms.

A reassessment of trends and rural-urban/regional differences in the total fertility rate in China, 2000-2020: analyses of the 2020 national census data.

The decline in the total fertility rate (TFR) is a key driver of population change and has important implications for population health and social development. However, China's TFR has been a considerable controversy due to a lack of high-quality data. Therefore, this study used the 2020 national population census of China (NPCC) data and reverse survival method to reassess temporal trends in the TFRs and to reexamine rural-urban differences and regional variations in TFRs from 2000 to 2020 in China. Overall, there were significant gaps between the estimated and reported TFRs before 2020, and the estimated TFRs based on the 2020 NPCC data remained higher than the reported TFRs from government statistics. Although TFRs rebounded shortly in the years after the two-child policy, they have shown a wavelike decline since 2010. Additionally, the estimated TFRs fluctuated below 1.5 children per woman in urban areas compared to above 1.8 in rural areas, but the rural-urban differences continued to decrease. Regarding geographic regional variations, the estimated TFRs in all regions displayed a declining trend during 2010-2020, especially in rural areas. Large decreases of over 25% in TFRs occurred in the north, east, central, and northwest regions. In addition to changing the birth policy, the government and society should adopt comprehensive strategies, including reducing the costs of marriage, childbearing, and child education, as well as promoting work-family balance, to encourage and increase fertility levels.

A logic framework for addressing medical racism in academic medicine: an analysis of qualitative data.

BACKGROUND: Despite decades of anti-racism and equity, diversity, and inclusion (EDI) interventions in academic medicine, medical racism continues to harm patients and healthcare providers. We sought to deeply explore experiences and beliefs about medical racism among academic clinicians to understand the drivers of persistent medical racism and to inform intervention design. METHODS: We interviewed academically-affiliated clinicians with any racial identity from the Departments of Family Medicine, Cardiac Sciences, Emergency Medicine, and Medicine to understand their experiences and perceptions of medical racism. We performed thematic content analysis of semi-structured interview data to understand the barriers and facilitators of ongoing medical racism. Based on participant narratives, we developed a logic framework that demonstrates the necessary steps in the process of addressing racism using if/then logic. This framework was then applied to all narratives and the barriers to addressing medical racism were aligned with each step in the logic framework. Proposed interventions, as suggested by participants or study team members and/or identified in the literature, were matched to these identified barriers to addressing racism. RESULTS: Participant narratives of their experiences of medical racism demonstrated multiple barriers to addressing racism, such as a perceived lack of empathy from white colleagues. Few potential facilitators to addressing racism were also identified, including shared language to understand racism. The logic framework suggested that addressing racism requires individuals to understand, recognize, name, and confront medical racism. CONCLUSIONS: Organizations can use this logic framework to understand their local context and select targeted anti-racism or EDI interventions. Theory-informed approaches to medical racism may be more effective than interventions that do not address local barriers or facilitators for persistent medical racism.

Factors impacting clinical data and documentation quality in Australian aged care and disability services: a user-centred perspective.

BACKGROUND: Research has highlighted a need to improve the quality of clinical documentation and data within aged care and disability services in Australia to support improved regulatory reporting and ensure quality and safety of services. However, the specific causes of data quality issues within aged care and disability services and solutions for optimisation are not well understood. OBJECTIVES: This study explored aged care and disability workforce (referred to as 'data-users') experiences and perceived root causes of clinical data quality issues at a large aged care and disability services provider in Western Australia, to inform optimisation solutions. METHODS: A purposive sample of n = 135 aged care and disability staff (including community-based and residential-based) in clinical, care, administrative and/or management roles participated in semi-structured interviews and web-based surveys. Data were analysed using an inductive thematic analysis method, where themes and subthemes were derived. RESULTS: Eight overarching causes of data and documentation quality issues were identified: (1) staff-related challenges, (2) education and training, (3) external barriers, (4) operational guidelines and procedures, (5) organisational practices and culture, (6) technological infrastructure, (7) systems design limitations, and (8) systems configuration-related challenges. CONCLUSION: The quality of clinical data and documentation within aged care and disability services is influenced by a complex interplay of internal and external factors. Coordinated and collaborative effort is required between service providers and the wider sector to identify behavioural and technical optimisation solutions to support safe and high-quality care and improved regulatory reporting.

A Comparison Study of Deep Learning Methodologies for Music Emotion Recognition.

Classical machine learning techniques have dominated Music Emotion Recognition. However, improvements have slowed down due to the complex and time-consuming task of handcrafting new emotionally relevant audio features. Deep learning methods have recently gained popularity in the field because of their ability to automatically learn relevant features from spectral representations of songs, eliminating such necessity. Nonetheless, there are limitations, such as the need for large amounts of quality labeled data, a common problem in MER research. To understand the effectiveness of these techniques, a comparison study using various classical machine learning and deep learning methods was conducted. The results showed that using an ensemble of a Dense Neural Network and a Convolutional Neural Network architecture resulted in a state-of-the-art 80.20% F1 score, an improvement of around 5% considering the best baseline results, concluding that future research should take advantage of both paradigms, that is, combining handcrafted features with feature learning.

Improved data quality and statistical power of trial-level event-related potentials with Bayesian random-shift Gaussian processes.

Studies of cognitive processes via electroencephalogram (EEG) recordings often analyze group-level event-related potentials (ERPs) averaged over multiple subjects and trials. This averaging procedure can obscure scientifically relevant variability across subjects and trials, but has been necessary due to the difficulties posed by inference of trial-level ERPs. We introduce the Bayesian Random Phase-Amplitude Gaussian Process (RPAGP) model, for inference of trial-level amplitude, latency, and ERP waveforms. We apply RPAGP to data from a study of ERP responses to emotionally arousing images. The model estimates of trial-specific signals are shown to greatly improve statistical power in detecting significant differences in experimental conditions compared to existing methods. Our results suggest that replacing the observed data with the de-noised RPAGP predictions can potentially improve the sensitivity and accuracy of many of the existing ERP analysis pipelines.

Intersectoral Partnerships Between Local Governments and Health Organisations in High-Income Contexts: A Scoping Review.

BACKGROUND: Local governments are the closest level of government to the communities they serve. Traditionally providing roads, rates and garbage services, they are also responsible for policy and regulation, particularly land use planning and community facilities and services that have direct and indirect impacts on (equitable) health and well-being. Partnerships between health agencies and local government are therefore an attractive proposition to progress actions that positively impact community health and well-being. Yet, the factors underpinning these partnerships across different contexts are underdeveloped, as mechanisms to improve population health and well-being. METHODS: A scoping review was conducted to gain insight into the concepts, theories, sources, and knowledge gaps that shape partnerships between health and local governments. The search strategy followed the Preferred Reporting Items for Systematic Reviews and Meta-Analyses extension for Scoping Reviews (PRISMA-ScR) guidelines and was informed by a critical realist approach that identifies necessary, contingent and contextual factors in the literature. MEDLINE, Scopus, Web of Science, and ProQuest Central databases were searched for studies published between January 2005 and July 2021. RESULTS: The search yielded 3472 studies, after deleting duplicates and initial title and abstract screening, 188 papers underwent full text review. Twenty-nine papers were included in the review. Key themes shaping partnerships included funding and resources; partnership qualities; governance and policy; and evaluation and measures of success. The functional, organisational and individual aspects of these themes are explored and presented in a framework. CONCLUSION: Given that local government are the closest level of government to community, this paper provides a sophisticated roadmap that can underpin partnerships between local government and health agencies aiming to influence population health outcomes. By identifying key themes across contexts, we provide a framework that may assist in designing and evaluating evidence-informed health and local government partnerships.

Quality of routine health facility data for monitoring maternal, newborn and child health indicators: A desk review of DHIS2 data in Lumbini Province, Nepal.

INTRODUCTION: Health-facility data serves as a primary source for monitoring service provision and guiding the attainment of health targets. District Health Information Software (DHIS2) is a free open software predominantly used in low and middle-income countries to manage the facility-based data and monitor program wise service delivery. Evidence suggests the lack of quality in the routine maternal and child health information, however there is no robust analysis to evaluate the extent of its inaccuracy. We aim to bridge this gap by accessing the quality of DHIS2 data reported by health facilities to monitor priority maternal, newborn and child health indicators in Lumbini Province, Nepal. METHODS: A facility-based descriptive study design involving desk review of Maternal, Neonatal and Child Health (MNCH) data was used. In 2021/22, DHIS2 contained a total of 12873 reports in safe motherhood, 12182 reports in immunization, 12673 reports in nutrition and 12568 reports in IMNCI program in Lumbini Province. Of those, monthly aggregated DHIS2 data were downloaded at one time and included 23 priority maternal and child health related data items. Of these 23 items, nine were chosen to assess consistency over time and identify outliers in reference years. Twelve items were selected to examine consistency between related data, while five items were chosen to assess the external consistency of coverage rates. We reviewed the completeness, timeliness and consistency of these data items and considered the prospects for improvement. RESULTS: The overall completeness of facility reporting was found within 98% to 100% while timeliness of facility reporting ranged from 94% to 96% in each Maternal, Newborn and Child Health (MNCH) datasets. DHIS2 reported data for all 9 MNCH data items are consistent over time in 4 of 12 districts as all the selected data items are within ±33% difference from the provincial ratio. Of the eight MNCH data items assessed, four districts reported ≥5% monthly values that were moderate outliers in a reference year with no extreme outliers in any districts. Consistency between six-pairs of data items that are expected to show similar patterns are compared and found that three pairs are within ±10% of each other in all 12 districts. Comparison between the coverage rates of selected tracer indicators fall within ±33% of the DHS survey result. CONCLUSION: Given the WHO data quality guidance and national benchmark, facilities in the Lumbini province well maintained the completeness and timeliness of MNCH datasets. Nevertheless, there is room for improvement in maintaining consistency over time, plausibility and predicted relationship of reported data. Encouraging the promotion of data review through the data management committee, strengthening the system inbuilt data validation mechanism in DHIS2, and promoting routine data quality assessment systems should be greatly encouraged.

How the Plants for Joints multidisciplinary lifestyle intervention achieved its effects: a mixed methods process evaluation.

BACKGROUND: Plants for Joints (PFJ) is a multidisciplinary intervention centered around a whole-food plant-based diet, physical activity, and sleep and stress management. The PFJ intervention successfully improved disease activity and symptoms in people with rheumatoid arthritis (RA) or osteoarthritis (OA), respectively, and metabolic health. To investigate how these effects were achieved a mixed methods process evaluation was conducted to understand the context, implementation, and mechanism of impact of the PFJ intervention. Also, the relationship between degree of implementation and lifestyle changes was explored. METHODS: Quantitative and qualitative data were collected across the evaluation domains context (i.e. reach), implementation (i.e. recruitment and delivery), and mechanism of impact (i.e. responsiveness) of both the participants and coaches (incl. dietitians, sport coaches) according to the UK MRC guidelines for process evaluations. Data was collected from the participants via focus groups and questionnaires after the intervention, and interviews with coaches. Qualitative data were analyzed thematically, and quantitative data were assessed with descriptive statistics and linear regression analyses. Degree of implementation was quantified using a theory-driven implementation index score composed of different process evaluation constructs. RESULTS: Of the 155 participants who participated in the PFJ intervention, 106 (68%) took part in the questionnaire and 34 (22%) attended a focus group. Participants felt the intervention was complete, coherent, and would recommend the intervention to others (mean score 9.2 (SD 1.4) out of 10). Participants felt heard and empowered to take control of their lifestyle and health outcomes. Components perceived as most useful were self-monitoring, social support, practical and theoretical information, and (individual) guidance by the multidisciplinary team. Participants perceived the intervention as feasible, and many indicated it effectively improved their health outcomes. In an explorative analysis there was no significant difference in healthy lifestyle changes across implementation index score groups. CONCLUSION: This process evaluation offers important insights into why the PFJ intervention works and how the intervention can be optimized for future implementation. Results indicating the intervention's high satisfaction, feasibility, and perceived effectiveness, further support the use of plant-based lifestyle interventions as an additional treatment option for patients with RA, OA, or other chronic diseases. TRIAL REGISTRATION: International Clinical Trial Registry Platform numbers: NL7800, NL7801, and NL7802, all registered 17-06-2019.

Can cognitive dissonance explain beliefs regarding meritocracy?

Why do economically disadvantaged people often regard inequality as fair? The literature on deliberative justice suggests that people regard inequality as fair when it is proportional to inequality in effort or other inputs - i.e. when it is meritocratic. But in the real-world there is substantial uncertainty over the distribution of income and merit - so what compels disadvantaged people to legitimate their own disadvantage? This paper suggests it is a reaction to cognitive dissonance. When inequality is high, and when people lack control, their only way to reduce dissonance is to convince themselves the distribution is fair. I implement an online experiment to test this theory. Results do not support a cognitive dissonance mechanism behind meritocracy. But they do indicate that disadvantaged individuals are more likely to regard inequality as fair when they lack control. Analysis of qualitative data indicates that deprivation of control engenders a fatalistic response to inequality.

Professional Experiences and Career Trajectories of Mid- to Senior-Career Women Clinician-Scientists: A Qualitative Study.

Importance: Despite increasing evidence and recognition of persistent gender disparities in academic medicine, qualitative data detailing the association of gender-based experiences with career progression remain sparse, particularly at the mid- to senior-career stage. Objective: To investigate the role gender has played in everyday professional experiences of mid- to senior-career women clinician-scientists and their perceptions of gender-related barriers experienced across their careers. Design, Setting, and Participants: In this qualitative study, a total of 60 of 159 invited clinician-scientists who received National Institutes of Health K08 or K23 awards between 2006 and 2009 and responded to a survey in 2021 agreed to participate. Invitees were selected using random, purposive sampling to support sample heterogeneity. Semistructured in-depth interviews were conducted January to May 2022. For this study, interviews from 31 women were analyzed using the framework approach to thematic analysis. Data analyses were performed between August and October 2023. Main Outcomes and Measures: Descriptive themes of participant experiences of gender and gender-based barriers in academic medicine. Results: A total of 31 women clinician-scientists (8 identifying as Asian [25.8%], 14 identifying as White [45.2%], and 9 identifying as members of a minority group underrepresented in medicine [29.0%]; 14 aged 40-49 years [45.2%] and 14 aged 50-59 years [45.2%]) were included. Among them, 17 participants (54.8%) had children who required adult supervision or care, 7 participants (22.6%) had children who did not require supervision or care, and 6 participants (19.4%) did not have children. There were 4 dominant themes identified within participant experiences in academic medicine: the mental burden of gendered expectations at work and home, inequitable treatment of women in bureaucratic processes, subtle and less subtle professional exclusion of women, and value of communities built on shared identities, experiences, and solidarity. Conclusions and Relevance: This study found that women perceived the institution of academic medicine as a male-centric system misaligned with the needs of women, with associated feelings of exclusion, disillusionment, and loss of trust in their institutions. Findings suggest that the confluence of domestic obligations and unaccommodating institutional environments may make it difficult for women clinician-scientists to achieve established timelines of career progression and productivity; these findings may have long-term implications for the well-being and retention of women in academic medicine.

Resilience-aware MLOps for AI-based medical diagnostic system.

Background: The healthcare sector demands a higher degree of responsibility, trustworthiness, and accountability when implementing Artificial Intelligence (AI) systems. Machine learning operations (MLOps) for AI-based medical diagnostic systems are primarily focused on aspects such as data quality and confidentiality, bias reduction, model deployment, performance monitoring, and continuous improvement. However, so far, MLOps techniques do not take into account the need to provide resilience to disturbances such as adversarial attacks, including fault injections, and drift, including out-of-distribution. This article is concerned with the MLOps methodology that incorporates the steps necessary to increase the resilience of an AI-based medical diagnostic system against various kinds of disruptive influences. Methods: Post-hoc resilience optimization, post-hoc predictive uncertainty calibration, uncertainty monitoring, and graceful degradation are incorporated as additional stages in MLOps. To optimize the resilience of the AI based medical diagnostic system, additional components in the form of adapters and meta-adapters are utilized. These components are fine-tuned during meta-training based on the results of adaptation to synthetic disturbances. Furthermore, an additional model is introduced for post-hoc calibration of predictive uncertainty. This model is trained using both in-distribution and out-of-distribution data to refine predictive confidence during the inference mode. Results: The structure of resilience-aware MLOps for medical diagnostic systems has been proposed. Experimentally confirmed increase of robustness and speed of adaptation for medical image recognition system during several intervals of the system's life cycle due to the use of resilience optimization and uncertainty calibration stages. The experiments were performed on the DermaMNIST dataset, BloodMNIST and PathMNIST. ResNet-18 as a representative of convolutional networks and MedViT-T as a representative of visual transformers are considered. It is worth noting that transformers exhibited lower resilience than convolutional networks, although this observation may be attributed to potential imperfections in the architecture of adapters and meta-adapters. Сonclusion: The main novelty of the suggested resilience-aware MLOps methodology and structure lie in the separating possibilities and activities on creating a basic model for normal operating conditions and ensuring its resilience and trustworthiness. This is significant for the medical applications as the developer of the basic model should devote more time to comprehending medical field and the diagnostic task at hand, rather than specializing in system resilience. Resilience optimization increases robustness to disturbances and speed of adaptation. Calibrated confidences ensure the recognition of a portion of unabsorbed disturbances to mitigate their impact, thereby enhancing trustworthiness.

Academic Detailing as a Health Information Technology Implementation Method: Supporting the Design and Implementation of an Emergency Department-Based Clinical Decision Support Tool to Prevent Future Falls.

BACKGROUND: Clinical decision support (CDS) tools that incorporate machine learning-derived content have the potential to transform clinical care by augmenting clinicians' expertise. To realize this potential, such tools must be designed to fit the dynamic work systems of the clinicians who use them. We propose the use of academic detailing-personal visits to clinicians by an expert in a specific health IT tool-as a method for both ensuring the correct understanding of that tool and its evidence base and identifying factors influencing the tool's implementation. OBJECTIVE: This study aimed to assess academic detailing as a method for simultaneously ensuring the correct understanding of an emergency department-based CDS tool to prevent future falls and identifying factors impacting clinicians' use of the tool through an analysis of the resultant qualitative data. METHODS: Previously, our team designed a CDS tool to identify patients aged 65 years and older who are at the highest risk of future falls and prompt an interruptive alert to clinicians, suggesting the patient be referred to a mobility and falls clinic for an evidence-based preventative intervention. We conducted 10-minute academic detailing interviews (n=16) with resident emergency medicine physicians and advanced practice providers who had encountered our CDS tool in practice. We conducted an inductive, team-based content analysis to identify factors that influenced clinicians' use of the CDS tool. RESULTS: The following categories of factors that impacted clinicians' use of the CDS were identified: (1) aspects of the CDS tool's design (2) clinicians' understanding (or misunderstanding) of the CDS or referral process, (3) the busy nature of the emergency department environment, (4) clinicians' perceptions of the patient and their associated fall risk, and (5) the opacity of the referral process. Additionally, clinician education was done to address any misconceptions about the CDS tool or referral process, for example, demonstrating how simple it is to place a referral via the CDS and clarifying which clinic the referral goes to. CONCLUSIONS: Our study demonstrates the use of academic detailing for supporting the implementation of health information technologies, allowing us to identify factors that impacted clinicians' use of the CDS while concurrently educating clinicians to ensure the correct understanding of the CDS tool and intervention. Thus, academic detailing can inform both real-time adjustments of a tool's implementation, for example, refinement of the language used to introduce the tool, and larger scale redesign of the CDS tool to better fit the dynamic work environment of clinicians.

An open-source, high-resolution, automated fluorescence microscope.

Fluorescence microscopy is a fundamental tool in the life sciences, but the availability of sophisticated equipment required to yield high-quality, quantitative data is a major bottleneck in data production in many laboratories worldwide. This problem has long been recognized and the abundancy of low-cost electronics and the simplification of fabrication through 3D-printing have led to the emergence of open-source scientific hardware as a research field. Cost effective fluorescence microscopes can be assembled from cheaply mass-produced components, but lag behind commercial solutions in image quality. On the other hand, blueprints of sophisticated microscopes such as light-sheet or super-resolution systems, custom-assembled from high quality parts, are available, but require a high level of expertise from the user. Here, we combine the UC2 microscopy toolbox with high-quality components and integrated electronics and software to assemble an automated high-resolution fluorescence microscope. Using this microscope, we demonstrate high resolution fluorescence imaging for fixed and live samples. When operated inside an incubator, long-term live-cell imaging over several days was possible. Our microscope reaches single molecule sensitivity, and we performed single particle tracking and SMLM super-resolution microscopy experiments in cells. Our setup costs a fraction of its commercially available counterparts but still provides a maximum of capabilities and image quality. We thus provide a proof of concept that high quality scientific data can be generated by lay users with a low-budget system and open-source software. Our system can be used for routine imaging in laboratories that do not have the means to acquire commercial systems and through its affordability can serve as teaching material to students.

Characteristics, Progression, and Output of Randomized Platform Trials: A Systematic Review.

Importance: Platform trials have become increasingly common, and evidence is needed to determine how this trial design is actually applied in current research practice. Objective: To determine the characteristics, progression, and output of randomized platform trials. Evidence Review: In this systematic review of randomized platform trials, Medline, Embase, Scopus, trial registries, gray literature, and preprint servers were searched, and citation tracking was performed in July 2022. Investigators were contacted in February 2023 to confirm data accuracy and to provide updated information on the status of platform trial arms. Randomized platform trials were eligible if they explicitly planned to add or drop arms. Data were extracted in duplicate from protocols, publications, websites, and registry entries. For each platform trial, design features such as the use of a common control arm, use of nonconcurrent control data, statistical framework, adjustment for multiplicity, and use of additional adaptive design features were collected. Progression and output of each platform trial were determined by the recruitment status of individual arms, the number of arms added or dropped, and the availability of results for each intervention arm. Findings: The search identified 127 randomized platform trials with a total of 823 arms; most trials were conducted in the field of oncology (57 [44.9%]) and COVID-19 (45 [35.4%]). After a more than twofold increase in the initiation of new platform trials at the beginning of the COVID-19 pandemic, the number of platform trials has since declined. Platform trial features were often not reported (not reported: nonconcurrent control, 61 of 127 [48.0%]; multiplicity adjustment for arms, 98 of 127 [77.2%]; statistical framework, 37 of 127 [29.1%]). Adaptive design features were only used by half the studies (63 of 127 [49.6%]). Results were available for 65.2% of closed arms (230 of 353). Premature closure of platform trial arms due to recruitment problems was infrequent (5 of 353 [1.4%]). Conclusions and Relevance: This systematic review found that platform trials were initiated most frequently during the COVID-19 pandemic and declined thereafter. The reporting of platform features and the availability of results were insufficient. Premature arm closure for poor recruitment was rare.