Trends In Usage Research Articles

A Review of Contact Lens Regulations in the Asia Pacific Region.

The increased usage of nonrefractive cosmetic contact lenses in Asia and a trend toward increased online purchasing of both refractive and cosmetic contact lenses poses unique challenges in the management of contact lens-related complications. However, regulations have lagged behind this increased uptake in the community. This paper reviews regulations surrounding the manufacturing, import, and distribution of contact lenses in the Asia Pacific region and compares them with international regulations from the European Union and United States. It found a need for more specific regulations and closer surveillance targeted at new trends in contact lens purchasing and usage. This paper also found a need for more effective enforcement of current regulations.

Forecasting Public Bicycle Utilization in New York City Utilizing the CitiBike Dataset

The integration of public bicycles into urban transportation systems has gained popularity due to their potential to reduce traffic congestion and air pollution, while also addressing the “last kilometer” problem. This study examines the usage patterns and predictive analysis of public bicycles in New York City using the CitiBike dataset from 2013 to 2015. We employ the Spring algorithm for networkx visualization and analyze the loan and return data to understand spatial-temporal usage dynamics. Additionally, clustering techniques, including K-means and DBSCAN, are applied to the dataset to uncover usage trends and patterns. The findings provide valuable insights for the planning and optimization of public bicycle systems, such as CitiBike, and contribute to the promotion of sustainable and environmentally friendly urban transportation.

Forecasting Public Bicycle Utilization in New York City Utilizing the CitiBike Dataset

The integration of public bicycles into urban transportation systems has gained popularity due to their potential to reduce traffic congestion and air pollution, while also addressing the “last kilometer” problem. This study examines the usage patterns and predictive analysis of public bicycles in New York City using the CitiBike dataset from 2013 to 2015. We employ the Spring algorithm for networkx visualization and analyze the loan and return data to understand spatial-temporal usage dynamics. Additionally, clustering techniques, including K-means and DBSCAN, are applied to the dataset to uncover usage trends and patterns. The findings provide valuable insights for the planning and optimization of public bicycle systems, such as CitiBike, and contribute to the promotion of sustainable and environmentally friendly urban transportation.

Optimizing Scheduled Virtual Machine Requests Placement in Cloud Environments: A Tabu Search Approach

This paper introduces a novel model for virtual machine (VM) requests with predefined start and end times, referred to as scheduled virtual machine demands (SVMs). In cloud computing environments, SVMs represent anticipated resource requirements derived from historical data, usage trends, and predictive analytics, allowing cloud providers to optimize resource allocation for maximum efficiency. Unlike traditional VMs, SVMs are not active concurrently. This allows providers to reuse physical resources such as CPU, RAM, and storage for time-disjoint requests, opening new avenues for optimizing resource distribution in data centers. To leverage this opportunity, we propose an advanced VM placement algorithm designed to maximize the number of hosted SVMs in cloud data centers. We formulate the SVM placement problem (SVMPP) as a combinatorial optimization challenge and introduce a tailored Tabu Search (TS) meta-heuristic to provide an effective solution. Our algorithm demonstrates significant improvements over existing placement methods, achieving up to a 15% increase in resource efficiency compared to baseline approaches. This advancement highlights the TS algorithm’s potential to deliver substantial scalability and optimization benefits, particularly for high-demand scenarios, albeit with a necessary consideration for computational cost.

Breaking the silence: The effects of online social movements on gender-based violence

This paper investigates the impact of online social movements against gender-based violence (GBV) on GBV-related crime. Using machine learning techniques, we construct a novel dataset tracking the prevalence of GBV-related social media movements on Twitter from 2014 to 2017. Matching this data with weekly FBI crime reports across U.S. states we estimate the effect of tweets related to GBV on GBV-related reported crime. Our econometric approach aims to mitigate concerns of spurious correlations arising from common trends in Twitter usage and crime. We estimate national level regressions with differential effects on GBV-related and non-GBV-related crimes within the same week, as well as regressions at the state-by-week level controlling for state and month fixed effects. Our main findings reveal that GBV-related tweets led to a small but significant short-term decrease in GBV-related crime reports. Further analysis suggests that the decrease is attributable to a reduction in actual crime rather than in reporting behavior.

Evaluation of antibiotic consumption using WHO -antimicrobial consumption tool and AWaRe classification

ABSTRACT Introduction: Antimicrobial resistance is a public health concern with global ramifications. Antibiotic misuse and overuse, are rampant in our country but more alarming is the data on the use of antibiotics primarily because of lack of access is another threat. A majority of the data on drug sales and consumption in India comes from the private sector and is typically gathered from private commercial organization. Because data on antibiotic usage in India is rarely reported, the study’s objectives were to estimate antibiotic consumption in ICU patients and also to assess antibiotic usage trends using the WHO AWaRE classification. Materials and Methods: A prospective observational study was carried out in the ICU department for six months after obtaining institutional ethics committee approval (P14/2021/14). In-patient records and details on antibiotic prescriptions were collected from the day of admission until they were discharged from the intensive care unit. Data on antibiotic consumption were retrieved from the hospital pharmacy. Descriptive statistics such as frequency, percentages, and means were used to summarize the data. Microsoft Excel 2016 was used for all statistical analyses. Results: The results of the study showed that female patients (54%), made up a significant majority of those receiving antimicrobial treatment. The age group between 40–49 years (21.68%) was the most common for those receiving antimicrobial treatment. Surgical prophylaxis (37.34%) emerged as the most common indication. Among the different antimicrobials studied, ceftriaxone 1 gm (108.5 DDD) emerged as the most prescribed antimicrobial. The access group of antibiotics accounted for 55% of the total usage. Conclusion: Appropriate use of antibiotics is important to minimize the risk of adverse events and antimicrobial resistance.

Impact of pharmacist-led aminoglycoside stewardship: a 10-year observational study

BackgroundAminoglycosides are crucial for treating multidrug-resistant gram-negative infections and endocarditis. However, aminoglycosides are associated with significant risks of nephrotoxicity, necessitating careful dose selection and therapeutic drug monitoring. Therapeutic drug monitoring is essential for minimizing risk; however, few institutions routinely perform it. This study aimed to assess the impact of a pharmacist-driven therapeutic drug monitoring intervention on aminoglycoside usage trends and clinical outcomes.MethodsThis retrospective cohort study included 263 patients treated with aminoglycosides between 2014 and 2023. A pharmacist-led therapeutic drug monitoring intervention began in 2017, focusing on monitoring renal function, documenting patient weight, and closely managing aminoglycoside concentrations. Trends in aminoglycoside use and renal outcomes were analyzed.ResultsOver the study period, appropriate use of aminoglycosides at the time of initial prescription increased from 49 to 82% (P < 0.01). Pharmacist dosing design at initial prescription increased significantly from 21% pre-intervention to 60% post-intervention (P < 0.01). The proportion of pharmacist intervention in initial dosing design increased over time. The proportion of patients with measured aminoglycoside blood concentrations significantly increased from 53% pre-intervention to 72% post-intervention (P < 0.01). The proportion of patients who were able to manage target blood concentrations from the initial aminoglycoside dose without dose adjustments increased from 31% pre-intervention to 42% post-intervention, although the results were not significantly different (P = 0.07). The incidence rate of renal impairment remained similar (11% vs. 12%; P = 0.85), although the annual average number of cases decreased from 4.3 before the intervention to 2.5 after. Similarly, there were no significant differences in clinical efficacy before and after the intervention (65% vs. 71%; P = 0.35). Furthermore, aminoglycoside stewardship led to a 56% cost saving.ConclusionsPharmacist-led aminoglycoside stewardship significantly improved the appropriate use of aminoglycosides and decreased the associated costs. Thus, pharmacist involvement is essential for the proper use of aminoglycosides. However, many patients required aminoglycoside dose reductions despite the pharmacist’s guideline-based dosing design. Therefore, further accumulation of information on the management of aminoglycoside blood concentration may be necessary for the revision of these guidelines.

Enhancing Energy Demand Prediction Using Elman Neural Network and Support Vector Machine Model: A Case Study in Lagos State, Nigeria

Energy and power resources must be managed well to guarantee their best use. To ensure that the required demand for energy generation is met, it is necessary to estimate Energy Demand (ED) accurately. Regression analysis and time series analysis were the mainstays of ED prediction in the past. But thanks to recent developments, accuracy has been increased by utilizing machine learning (ML) techniques to identify trends in data on electricity consumption. The purpose of this study is to offer insightful information on the relationships between various variables and how those relationships affect trends in energy usage. The goal is to create a prediction model that can reliably predict ED in a certain study area. In Lagos State, Nigeria, the Elman neural network (ELNN) and support vector machine (SVM) model are used for ED forecasting. The performance of SVM is optimized by selecting suitable kernel functions. The research region's 24-hour dataset was used to train the SVM model. Metrics including the correlation coefficient (R), Pearson correlation coefficient (PCC), mean square error (MSE), mean absolute error (MAE), and Root mean square error (RMSE) were used to evaluate the ML model. The findings show that the ELNN-M3 model fits the data well (R &gt; 0.9). It is noteworthy that the study considered temperature, wind speed, and sun radiation as predicting variables. This study adds to the ongoing attempts to improve the accuracy of energy demand prediction, especially in dynamic contexts such as Lagos State.

Trends and Determinants of Palliative Care Utilization Among Patients With Metastatic Upper Tract Urothelial Carcinoma in the National Cancer Database

Objective To identify patterns of palliative care usage and identify determinants of palliative intervention using a large contemporary cohort of patients with metastatic upper tract urothelial carcinoma (mUTUC). Methods: The National Cancer Database was queried from 2004 to 2020 for patients with mUTUC. Patients with a prior malignancy, non-mUTUC, and missing follow up or palliative care information were excluded. Demographics and baseline characteristics were compared between patients with mUTUC who received palliative care and those that did not. Trends in annual palliative care usage were assessed via logistic regression. Univariate and multivariate logistic regression models were used to identify predictors of receipt of palliative care. Results: Four thousand and forty-four patients with mUTUC were included in the final cohort, among which 908 received palliative care (22.5%) and 3136 did not (77.5%). We found that the utilization of palliative care increased significantly from 2004 (15.0%) to 2019 (23.1%), P &lt; .001. Additionally, on multivariate analysis we found that a recent year of diagnosis, receipt of a prior nonsurgical treatment paradigm, and an overall survival &lt;6 months were independent predictors of palliative intervention, all P &lt; .001. On the other hand, undergoing treatment at a minority serving hospital and older age were associated with lower likelihood of receiving palliative care, both P &lt; .001. Conclusions: There is a low but increasing trend of utilization of palliative care among patients with mUTUC. Expansion of palliative care services, particularly among older patients and those at minority-serving hospitals, remains a key opportunity to improve quality of life and enhance patient-centered care among those with mUTUC.

Real-world research on beta-blocker usage trends in China and safety exploration based on the FDA Adverse Event Reporting System (FAERS)

BackgroundBeta-blockers are widely used, with continuously updated clinical recommendations. However, their application faces challenges in personalized treatment and safety. The study aimed to investigate the frequency and patterns of prescribing beta-blockers in China and to explore potential adverse event risk signals associated with beta-blockers, providing reference for rational medication use in clinical settings.MethodsPrescription data for beta-blockers from January 2018 to June 2023 were extracted through the Hospital Prescription Analysis Collaborative Project in China to analyze clinical usage trends. While adverse drug reaction reports for beta-blockers were obtained from the FDA Adverse Event Reporting System (FAERS) database. The classification and standardization of adverse drug event (ADE) reports were based on the preferred terms (PT) and corresponding system organ classes (SOC) from the Medical Dictionary for Regulatory Activities (MedDRA). Signal detection utilized a proportion imbalance method.ResultsIn clinical practice, metoprolol dominated beta-blocker prescriptions in China, accounting for 62.2%. Beta-blockers were primarily prescribed to the elderly (65.7%) and male patients (57.0%). However, off-label use of beta-blockers was relatively widespread. For instance, sotalol was prescribed for hypertension at 18.25%, while esmolol was used for angina and heart failure at rates of 12.94% and 14.98%, respectively. In addition, we identified newly discovered adverse reactions associated with beta-blockers, such as BRASH syndrome (metoprolol: n = 186, ROR = 391.285; carvedilol: n = 72, ROR = 256.459), acute kidney injury (bisoprolol: n = 247, ROR = 5.641), premature baby (labetalol: n = 110, ROR = 91.385), and sleep disorder (propranolol: n = 254, ROR = 10.98).ConclusionsMetoprolol led the beta-blocker market in China. Attention was warranted regarding the newly discovered adverse reactions, such as the risk of acute kidney injury with bisoprolol and the potential for BRASH syndrome with metoprolol and carvedilol.

The Effects of COVID-19 on Antifungal Prescribing in the UK-Lessons to Learn.

Fungal infections are increasingly prevalent; however, antifungal stewardship attracts little funding or attention. Previous studies have shown that knowledge of guidelines and scientific evidence regarding antifungals is poor, leading to prescribing based on personal experiences and the inherent biases this entails. We carried out a retrospective study of inpatient antifungal usage at two major hospitals. We assessed the longitudinal trends in antifungal usage and the effect of COVID-19 on antifungal prescription, alongside levels of empirical and diagnostically targeted antifungal usage. Our results showed that the longitudinal patterns of total systemic antifungal usage within the trusts were similar to national prescribing trends; however, the composition of antifungals varied considerably, even when looking exclusively at the more homogenous group of COVID-19 patients. We showed a high level of empirical antifungal use in COVID-19 patients, with neither trust adhering to international recommendations and instead appearing to follow prior prescribing habits. This study highlights the significant challenges to optimise antifungal use with prescribing behaviour largely dictated by habit, a lack of adherence to guidelines, and high rates of empirical non-diagnostic-based prescribing. Further research and resources are required to understand the impact of antifungal stewardship on improving antifungal prescribing behaviours in this setting and the effects on outcome.

Abstract 4112853: Trends in Statin Use for the Secondary Prevention of Atherosclerotic Cardiovascular Disease among U.S. Adults by Age, 1999-2020

Background: Current research focusing on trends in statin medication usage, particularly for different age groups, is limited. This study explored the change trends in statin usage from 1999 to 2020, with a specific focus on different age groups. Methods: Utilizing data from the National Health and Nutrition Examination Survey spanning the period from 1999 to 2020, our study included individuals aged 18 years and older diagnosed with atherosclerotic cardiovascular disease (ASCVD). Results: In the 18-44 years age group, usage changed from 18.34% (95% CI, 0.18 to 36.50) in 1999 to 16.63% (95% CI, 0.36 to 32.91) in 2020 (linear P value = 0.540, quadratic P value = 0.610) (Figure 1). For individuals aged 45-64 years, usage increased from 39.33% (95% CI, 29.62 to 49.05) in 1999 to 54.97% (95% CI, 47.10 to 662.83) in 2020 (linear P value &lt; 0.001, quadratic P value = 0.777) (Figure 1). In the group aged 65 years and older, usage increased from 35.17% (95% CI, 29.57 to 40.77) in 1999 to 60.62% (95% CI, 55.73 to 65.51) in 2010, and further to 68.62% (95% CI, 63.65 to 73.60) in 2020 (quadratic P value &lt; 0.001) (Figure 1). There was an interaction effect between statin usage trends and age (P for interaction &lt;0.001) (Figure 1). Conclusion: Overall, current health policies are effective for the use of statins for the secondary prevention of ASCVD. However, there was almost no improvement in statin usage within the 18-44 years age group. This highlights the need for greater focus on young ASCVD patients. Therefore, additional research is needed on statin use among young people.

"Exploring Social Media Usage Trends Among Adolescents in a Public Senior Secondary School in Abuja, Nigeria"

This study investigates social media usage trends and prevalence among 260 adolescents in a public senior secondary school in Abuja, Nigeria. A cross-sectional study design was employed, with students aged 13 to 18 years. Data collected revealed that a significant majority of respondents, 63.1% possess a smartphone and indicated a high level of access to social media platforms. Their engagement with social media daily is primarily for social interaction, social comparison, entertainment, and academic support. Furthermore, the findings highlight patterns of usage that vary by gender, with female students displaying a higher, 60% tendency for interaction through visual content platforms, while male students favored gaming and sports-related content. The study also identifies potential concerns regarding the implications of excessive social media use, including mental health issues, cyberbullying, and academic distraction. These results underscore the need for targeted educational programs and parental guidance to promote safe and responsible social media engagement. Overall, this research contributes to a deeper understanding of adolescents' social media behaviors and offers recommendations for fostering a balanced approach to digital engagement in educational settings.

Adoption of Artificial Intelligence in Public and Private Libraries of China: Determinants, Challenges, and Perceived Benefits

Artificial intelligence (AI) has much significance in different industries, including education. Due to increasing technological advancement, universities show their growing concerns about adopting several facilities linked with AI, specifically in their library management systems. Motivated by such an emerging trend, this research was conducted to examine AI adoption in both public and private libraries in China. The study explored the existing landscape of AI powered services in libraries, the perception of AI benefits, challenges related to AI adoption, and key determinants of such adoption for the targeted libraries. Using a questionnaire technique, the study collected data from different public and private libraries for which a sample of 154 respondents was finalized over the time span of 4.5 weeks. The analysis of the collected data shows that both the situations of launching and planning to launch different AI-related services were observed in Chinese libraries. Moreover, participants also confirmed several benefits of AI, including the enhancement of library operations by improving search accuracy, automating repetitive tasks, offering personalized resource recommendations, revealing usage trends through analytics, curating digital collections, improving cataloging, digitizing rare materials, and providing quick responses via chatbots, thereby allowing librarians to conduct their other more complex tasks. Additionally, the study confirmed several challenges associated with AI adoption. The results also revealed that along with other factors, the support from university administration, maturity and reliability of AI applications, and training and human resource management are positive and significant determinants of AI adoption. On the contrary, funding/cost associated with implementation and innovative services/alignment with technological trends reflected negative coefficients for AI adoption in Chinese libraries.

Shifting Trends of Antimicrobial Resistance Patterns Among Uropathogenic Bacteria Before and During the COVID-19 Pandemic.

Urinary tract infections (UTIs) represent a significant global health concern, primarily caused by uropathogenic bacteria and their rising trend of antimicrobial resistance (AMR). This study investigates the prevalence and pattern of AMR among uropathogenic bacteria during the COVID-19 pandemic, highlighting its impact on antimicrobial usage and resistance trends. This retrospective cross-sectional study, conducted at King Fahad Medical City, Riyadh, Saudi Arabia, from January 1, 2018, to December 31, 2022, analyzed 10,031 patients with positive urine cultures for resistance patterns pre-COVID-19 and during COVID-19. Bacterial identification and antimicrobial susceptibility testing were performed using the BD Phoenix system according to Clinical Laboratory Standard Institute guidelines. This study recorded an overall decrease in AMR during the pandemic among the most prevalent uropathogens (Escherichia coli,Klebsiella pneumoniae, andPseudomonas aeruginosa); however, notable increases in AMR were observed for specific antibiotics like cefoxitin, imipenem, and colistin. resistance. The multidrug resistance (MDR) amongP. aeruginosastrains significantly decreased from 22.9% pre-pandemic to 9.2% during the pandemic. The decline in AMR patterns during the COVID-19 pandemic likely resulted from altered antibiotic usage and healthcare practices, emphasizing the importance of ongoing monitoring and targeted antimicrobial management in response to changing AMR dynamics during global health emergencies.

Increasing Utilization of Extended Oral Antibiotic Prophylaxis Following Total Knee Arthroplasty from 2010 to 2022

Extended oral antibiotic (EOA) prophylaxis has been shown to reduce rates of periprosthetic joint infection in high-risk patients following total knee arthroplasty (TKA). Although national societies' recommendations against their use and clinical efficacy remain controversial, the increase in the literature surrounding EOA prophylaxis suggests a potential change in practice patterns that may warrant the creation of national guidelines. The purpose of this study was to investigate the trends in the utilization of EOA prophylaxis following TKA from 2010 to2022. Patients who underwent primary or aseptic revision TKA between 2010 and 2022 were identified in a national administrative claims database. Temporal trends in EOA usage following primary or revision TKA were calculated, and secondary analysis examined rates of utilization across demographics, including patients considered at high risk for infection. In total, 1,258,759 primary and 91,530 aseptic revision TKA patients were included in this study. From 2010 to 2022, EOA prescriptions increased by 321 and 368% following primary and revision TKA, respectively. Rates of utilization among high-risk individuals were similar to those of the general population. The use of EOA prophylaxis after TKA has increased significantly since 2010. Given their increasing usage and variation in prescription patterns among patients at high risk of infection, guidelines surrounding their use are necessary to promote antibiotic stewardship while preventing rates of infection.

Real-world study on disease burden and current clinical practice of hospital-acquired pneumonia in Japan

BackgroundHospital-acquired pneumonia (HAP) is a common nosocomial infection and is associated with high mortality. Despite advances in the understanding of the causes and prevention of HAP, it continues to be a frequent complication associated with hospital care. Presently, there are no large retrospective cohort studies on HAP in Japan. MethodsA retrospective cohort study was conducted using the Medical Data Vision Co. Ltd. database for the study period (1 April 2015 to 31 May 2018). The study population was defined based on ICD-10 codes for bacterial pneumonia, characteristics of hospitalization, and prescription of injection-only antibiotics. The study included patients ≥ 18 years of age with at least one episode of HAP during the identification period, where the episode was defined as hospitalization with HAP within the study identification period. ResultsA total of 2,968 patients were included in this study contributing to 2,979 HAP episodes. Patients with HAP were more likely to be male (64.9%) and older than age 65 (86.5%). The top three frequently prescribed antibiotics were sulbactam-ampicillin (39.7%; 1,183 episodes), tazobactam-piperacillin (28.4%; 846 episodes) and ceftriaxone (23.2%; 690 episodes). The mean (± SD) length of hospital stay during overall hospitalization and the HAP period were 49.9 (± 34.2) days and 11.3 ± 7.3 days respectively. The HAP patient mortality at discharge was 22.0%. ConclusionThe present study provided insights regarding the characteristics, treatment patterns of HAP patients in Japan. Further, the study provided noteworthy information regarding antibiotic usage trends in the aging Japanese population.

Analyzing Antibiotic Usage Trends in Genitourinary Infections and Procedures: Insights From the TriNetX Database

ObjectiveTo analyze antibiotic prescribing trends, particularly the decline in fluoroquinolone use for various GUI and procedures, and the growing preference for alternative treatments. MethodsThe TriNetX database was queried for antibiotic utilization rates from 2016 to 2022. We analyzed cases of cystitis, pyelonephritis, prostatitis, and other GUIs as well as periprocedure utilization for benign prostatic hyperplasia (BPH), urethral stricture disease, pelvic organ prolapse/stress urinary incontinence (POP/SUI), and overactive bladder/interstitial cystitis (OAB/IC). We determined the top four to six antibiotics prescribed within five days of diagnosis or procedure. Statistical analysis included linear regression modeling. ResultsFrom 2016-2022, fluoroquinolone use decreased in cystitis (p<0.001), pyelonephritis (p < 0.001), and prostatitis (p < 0.01). These infections were increasingly treated with beta-lactams (p<0.0001, p<0.01, p<0.01, respectively). Additionally, we found a decrease in fluoroquinolone utilization in procedures for BPH (p<0.001), urethral stricture (p<0.01), POP/SUI (p<0.001), and OAB/IC (p < 0.001). All cohorts except for OAB/IC showed a significant increase in beta-lactam use (p<0.001, p<0.01, p<0.01, respectively). ConclusionsWe noted a significant decline in fluoroquinolone use for GUI and procedures, accompanied by increased beta-lactam use, reflecting changes in prescribing practices. Judicious use is essential to promote antimicrobial stewardship and mitigate severe adverse effects.

의사결정나무 분석에 대한 국내 연구 동향 탐색

This study explores the basic system and applicability of decision tree techniques by examining domestic research trends in this field. We analyzed 1,075 academic papers using domestic decision tree analysis from 2001 to 2023, categorized by year, journal, topic field, author, paper frequency, and frequency of algorithms used. Key findings reveal that the frequency of papers began with one in 2001, increasing to 87 in 2023. The distribution across fields was as follows: social science (377 papers, 35.07%), natural science (233 papers, 21.67%), and engineering (230 papers, 21.40%). Among the algorithms, CHAID was used most frequently (315 times, 36.42%), followed by CART (310 times, 35.84%), ensemble techniques like random forest (124 times, 14.34%), and C5.0 (94 times, 10.87%). The Quest algorithm was rarely used (10 times, 1.16%). The annual usage trends indicate a growing preference for ensemble techniques to enhance decision tree prediction rates. The paper's keywords included “decision trees” (687), “data mining” (239), “machine learning” (105), “logistic” (98), and “neural networks” (68). Notably, keywords related to health, such as “suicide,” “depression,” “high blood pressure,” and “health,” emerged prominently, signifying the extensive use of decision trees in medical research. Based on these findings, we propose implications and directions for follow-up studies.

Dynamic recommendation for anticoagulant treatment in patients with atrial fibrillation using deep reinforcement learning method

Abstract Background Recent evidence suggests that the guideline-directed anticoagulant therapy for atrial fibrillation (AF) remains controversial. Widely-used CHA2DS2-VASc score is solely based on limited traditional cardiovascular risk factors, omitting AF characteristics and other markers of thromboembolic risk. A more efficient, safer and more personalized anticoagulant approach is warranted. Purpose To develop a data-driven deep reinforcement learning (DRL) model for guiding dynamic anticoagulant treatment in AF patients to improve cardiovascular outcomes. Methods Participants of this study were enrolled from the multicentred China Atrial Fibrillation (China-AF) Registry between August 2011 and December 2022, who were with regular follow-up every 6 months. We excluded patients on warfarin at baseline due to its declining usage trend in non-valvular AF patients in China. The DRL model was trained in 70% randomly selected patients for optimal dynamic decision-making, and then subsequently tested in the remaining 30%. Data of sociodemographic characteristics, AF characteristics, medical history, lifestyle factors, laboratory examination, and medications were input for model training. Concordance rate between DRL model’s recommendations and physicians’ actual decisions of non-vitamin-K-antagonist oral anticoagulant (NOAC) prescription among all visits before censoring was calculated for each patient. Primary outcome was the composite of cardiovascular death, ischemic stroke, transient ischemic attack or systemic embolism (SSE), and major bleeding. Shapley additive explanation analysis ranked the most important factors affecting decision-making of the DRL model. Results A total of 20068 patients (mean age: 63.0±12.0 years; 36.2% female) were randomly divided into a training cohort of 14050 patients and a testing cohort of 6018 patients. The model’s NOAC recommendations were mostly affected by age, prior NOAC prescription, body mass index, hypertension history and prior statin prescription (Figure 1). Patients with concordance rates of 50.1%-75% and 75.1%-100% had significant risk reductions for the primary outcome (adjusted HR =0.63; 95% CI, 0.46-0.85; P = 0.003 and adjusted HR =0.59; 95% CI, 0.46-0.75; P &amp;lt;0.001, respectively), compared to those with a concordance rate of 0-25%. Similar results were observed for all-cause death, cardiovascular death and SSE outcomes, that patients with the highest concordance rate had a significant lower risk, compared to those with the lowest concordance rate. There was a nonsignificant but similar trend with regard to major bleeding events (Figure 2). Conclusions This modelling study suggests that a data-driven DRL model might provide more efficient, safer and more personalized anticoagulant suggestions, potentially assisting physicians in clinical practice.