Journal of Global Health最新文献

Background: Remdesivir is a pivotal antiviral treatment introduced during the COVID-19 pandemic, and its use has changed over time. This provided an opportunity to study how drug usage evolves over the course of a pandemic. Our study aimed to examine key physiological parameters of patients hospitalised due to COVID-19, how the characteristics of patients evolved, and explore if potential differences were similar across time for patients treated with remdesivir and those not treated.

Methods: This study sourced electronic health care records from the Capital Region of Denmark. Patients aged ≥12 years hospitalised for the first time due to COVID-19 between 4 June 2020 and 1 December 2021 were included. Three time periods based on World Health Organization (WHO) treatment recommendations were used to describe temporal changes in the propensity score for remdesivir treatment as well as in individual patient characteristics.

Results: In total, 6960 patients were included. The key differences between remdesivir-treated (n = 2557) and non-treated (n = 4403) were an elevated c-reactive protein (CRP) (median of 79 vs. 35 mg/L) and increased use of glucocorticoids (41.5% vs. 10%) and antithrombotics (48.5% vs. 18.5%). When describing the temporal changes in the propensity score, there was an overall significant interaction between the time period and exposure group. From the first to the middle period for the non-treated there was a significant increase in the mean propensity score of 0.04 (95% confidence interval (CI) = 0.02-0.06). The patient characteristics that had the largest temporal variations for remdesivir-treated patients were age, alanine transaminase, mechanical ventilation, interleukin-6 inhibitors, glucocorticoids, and antithrombotics.

Conclusions: In conclusion, we found that remdesivir-treated and non-treated patients exhibited distinct sociodemographic and physiological characteristics. Also, the use of other COVID-19 treatments evolved differently between remdesivir-treated and not treated over time.

Background: Artificial intelligence (AI) tools based on large language models (LLMs) are being increasingly used by researchers and may play a role in health-related research priority-setting exercises (RPSEs). However, little is known about how these tools may differ in the types of research priorities they generate.

Methods: We examined research priorities aimed at improving treatments for four diseases: cancer, COVID-19, HIV, and Alzheimer. We compared the outputs from five AI tools (DeepSeek, ChatGPT, Claude, Perplexity, and Gemini) using SBERT-BioBERT embeddings and cosine similarity scores, and assessed the stability of differences between them by re-running identical prompts and slightly modified versions.

Results: We found that the outputs produced by Gemini were highly similar to those produced by the other tools. The two most different outputs were those produced by DeepSeek and Perplexity, whereby the former tended to emphasise technical medical issues, while the latter emphasised public health concerns. This substantive distinction between DeepSeek and Perplexity remained stable across repeated and tweaked prompts.

Conclusions: Our exploratory analysis suggests that Gemini performs well for researchers who prefer to generate health-related research priorities using a single AI model. For those planning to draw on multiple models, Perplexity and DeepSeek offer complementary perspectives.

Background: Quality research is essential to improving maternal, newborn, and child health (MNCH). Although Ethiopia has rapidly expanded academic and research institutions, duplication of studies, gaps in prioritisation and methods, and limited guidance on utilising evidence inhibit a coordinated approach to informing MNCH policy. We aim to address these challenges by characterising and prioritising the needs and opportunities of the MNCH research ecosystem in Ethiopia.

Methods: We purposively sampled experts for a three-stage Delphi study. Key informant interviews (KIIs) (n = 18) explored needs and challenges in capacity-strengthening, community engagement in research, operational infrastructure, collaborations, and funding. We thematically coded KII responses to generate 134 statements, which were then rated in an anonymous questionnaire (n = 34) on a Likert scale. We calculated average scores and percentage agreement for each statement. Finally, consensus-building discussions (n = 28) identified top priorities within each topic.

Results: Average percentage agreement across statements was 87% (range = 37-100). Highly endorsed priorities included strengthening inclusivity in research agenda-setting, prioritising research addressing key MNCH needs, enhancing research training by emphasising local experiences, cultivating intellectual curiosity, building skills in data analysis and translation, fostering research collaborations with greater multidisciplinary expertise, long-term mentorship, and capacity-building for local institutions, and engaging communities more effectively.

Conclusions: Understanding challenges in the existing research environment will enable better-informed activities and stronger research networks that address local priorities. We characterised the MNCH research ecosystem across multiple dimensions, offering actionable opportunities to strengthen research capacities, infrastructure, and innovation design and evaluation through advocacy, organisational and system strengthening efforts, curricula development, and the implementation of principles to guide partnerships and agenda-setting for a variety of stakeholders. Future efforts should prioritise fostering a culture of evidence, collaborative prioritisation of research between policymakers and researchers, and sustained commitment to scaling evidence-based practices to advance MNCH outcomes.

Background: The burden of age-related macular degeneration (AMD) has steadily increased in recent decades. We aimed to estimate the prevalence of AMD, including its subtypes, among individuals aged 40-89 years in China.

Methods: We conducted an updated literature search in the CNKI, Wanfang, Chinese Science and Technology Journal Database, PubMed, Embase, and MEDLINE for studies published between 27 June 2016 and 30 July 2024 that reported on the prevalence of AMD in China. We also included data from the 2017 China AMD Study. We utilised a multi-level mixed-effects meta-regression model to estimate age- and sex-specific prevalence of any AMD and its subtypes at the national level. For any AMD, we additionally conducted random-effects meta-analyses to pool odds ratios for associated factors, after which we incorporated these estimates into an associated factor-based model to estimate prevalence at regional and provincial levels.

Results: We included 40 articles, of which 24 contributed data for modelling analysis. The estimated national prevalence in China in 2020 was 4.70% (95% CI = 3.40, 6.46) for any AMD, 4.06% (95% CI = 2.92, 5.60) for early AMD, and 0.64% (95% CI = 0.48, 0.86) for late AMD, including 0.30% (95% CI = 0.25, 0.37) for geographic atrophy and 0.34% (95% CI = 0.23, 0.49) for neovascular AMD. These corresponded to 32.42 million cases (95% CI = 23.43, 44.54) with any AMD, 28.00 million (95% CI = 20.15, 38.61) with early AMD, 4.42 million (95% CI = 3.28, 5.93) with late AMD, 2.09 million (95% CI = 1.71, 2.52) with geographic atrophy, and 2.33 million (95% CI = 1.57, 3.41) with neovascular AMD. Regionally, the highest prevalence and number of cases was observed in Southwest China (5.95%; 95% CI = 4.48, 7.81) and South Central China (10.68 million; 95% CI = 7.60, 14.82), respectively. At the provincial level, Hainan and Guangdong exhibited the highest prevalence (7.64%; 95% CI = 4.61, 12.22) and the largest number of individuals affected (3.50 million; 95% CI = 2.34, 5.13), respectively.

Conclusions: We observed a substantial burden of AMD in Mainland China, with variations across subtypes, regions, and provinces. These findings underscore a need for targeted public health strategies to address AMD in the context of ageing.

Registration: PROSPERO: CRD420251080685.

Background: The World Health Organization revised its pneumonia guideline for managing children with chest indrawing and/or fast breathing, classifying them as 'pneumonia' and treating them with oral antibiotics at home; the Integrated Management of Childhood Illness (IMCI) protocol was revised accordingly. Evidence is needed on outcomes and treatment practices with the application of revised guidelines in programme settings.

Methods: This prospective observational cohort study was conducted in seven selected health centres in Uganda from November 2022 to May 2023. The IMCI-trained health workers identified and enrolled children aged 2-59 months presenting with cough and/or difficult breathing and chest indrawing without any general danger signs. The primary outcome was vital status at day 15 after the initial assessment. Secondary outcomes were prevalence of and adherence to antibiotic use, and hospitalisation. Enrolled children were given age-appropriate treatment.

Results: Of the 316 children who were enrolled, 68.4% (n/N = 216/316) were aged 12-59 months, and 93.7% (n/N = 296/316) had comorbidities, primarily malaria and diarrhoea. All were prescribed oral amoxicillin. Two children were lost to follow-up; thus, we followed 314 children on day 15. In children aged 2-11 months, 93.9% (n/N = 93/99) received a correct prescription compared to 20% (n/N = 43/215) among 12-59-month-olds. Adherence to five days of treatment was reported for 64.3% (n/N = 202/314) of children. According to the mothers' self report, no deaths were reported, 95.2% (n/N = 299/314) were cured; 2.2% (n/N = 7/314) were worse with six of seven hospitalised, and 2.5% (n/N = 8/314) were the same as the condition at time of enrolment. Most children were well-nourished; 3.8% had a weight-for-height (WHZ) z-score<-3, 6.7% had a weight-for-age (WAZ) z-score<-3, and 0.3% had a mid-upper arm circumference (MUAC)<115 mm. At follow-up on day 15, of 16 children hospitalised at any time after enrolment, 10 (62.5%) had recovered and were discharged, while six (37.5%) were still hospitalised. The presence of any severe malnutrition was associated with a 4-fold increased risk of hospitalisation. In contrast, a longer duration of oral amoxicillin treatment was associated with a 66% decrease in risk of hospitalisation during the follow-up period.

Conclusions: Children aged 2-59 months with chest indrawing pneumonia without danger signs can be successfully managed at home with a five-day course of oral amoxicillin, highlighting the importance of the new policy and approach.

Registration: ISRCTN12687253.

Background: Bangladesh observed a sudden massive influx of Rohingya refugees in August 2017. This large migrant population relative to a smaller host community placed a burden and threat on the public health sector. Due to the lack of pathogen-specific predicting factors and the influence of seasonal variation on childhood diarrhoeal pathogens in a densely populated area, we aimed to explore the same among Rohingya refugees and the host population.

Methods: We collected data from under-five children of Rohingya refugees and hosts between 2018 and 2023 from the Diarrhea Treatment Center (DTC)-based surveillance system that served our study population. We collected and tested stool samples to detect enteric pathogens. We performed a multiple logistic regression analysis to identify factors associated with individual pathogens.

Results: Out of 3534 children, 1479 (41.9%) were Rohingya refugees, and 2055 (58.1%) were host children who visited DTCs. Bacterial pathogens were identified in 15% (n/N = 533/3534) of children, and rotavirus in 58% (n/N = 1492/2564). We found higher odds of Vibrio cholerae (adjusted odds ratio (aOR) = 2.12; 95% confidence interval (CI) = 1.21-3.74), non-typhoidal Salmonella (NTS) (aOR = 4.45; 95% CI = 2.04-9.68), and lower odds of rotavirus infection (aOR = 0.72; 95% CI = 0.59-0.89) during the wet season compared to the cold season. Lack of handwashing with soap before feeding the child increased the risk of Aeromonas infection (aOR = 1.85; 95% CI = 1.21-2.81). Drinking tube well water lowers the risk of Vibrio cholerae (95% CI = 0.24-0.71), rotavirus (95% CI = 0.57-0.86), and Aeromonas (95% CI = 0.36-0.75) infection. We found that the recent intake of vitamin A was a protective factor for Vibrio cholerae (95% CI = 0.26-0.76), Aeromonas (95% CI = 0.44-0.89), and NTS (95% CI = 0.12-0.56) enteric infections.

Conclusions: Our results underscore the necessity of reinforcing routine diarrhoea surveillance for early detection of epidemics, vitamin A supplementation for children under five, and health education to prevent diarrhoea in vulnerable areas such as refugee camps.

Background: Iron deficiency anaemia (IDA) can lead to impairment of immunity, cognitive function, and poorer academic performance. Current health policies worldwide focus primarily on IDA prevention among preschoolers and women, overlooking school-aged children (aged 5-12 years) as a susceptible group. Through this systematic review and meta-analysis, we aimed to determine the prevalence of IDA among this population.

Methods: We searched PubMed, Embase, CINAHL, and EBSCO Open Dissertation from inception until July 2023 for English-language observational studies reporting on the prevalence of IDA children aged 5-12. We calculated the pooled prevalence using a random-effect model and performed subgroup analyses by regions, countries' income, and diagnostic criteria. We assessed the study quality using Hoy's risk of bias tool.

Results: We included 55 studies involving over 2.1 million children. None of these studies had a high risk of bias. The pooled global prevalence of IDA among children aged 5-12 years in community settings was 9.4% (95% confidence interval = 6.5%, 12.7%, I² = 99.6%). Subgroup analyses indicated moderate public health concerns among sub-Saharan Africa (21.9%) and South Asia (15.8%), or among low-income (29.7%) and lower-middle-income (24.5%) countries.

Conclusions: IDA is an important public health issue among children aged 5-12 years globally which even poses a significant concern in some populations or regions. Our findings could guide the development of national detection strategies and health prevention programmes targeted at improving children's health and educational outcomes.

Registration: PROSPERO (CRD42022335700).

Background: Chronic obstructive pulmonary disease (COPD), the third leading cause of death worldwide, demands prompt and precise identification and phenotyping for effective management. This study aims to develop a multimodal multi-task learning framework that concurrently performs automated detection and classification of COPD.

Methods: Retrospective multi-task model fusing computed tomography (CT) and clinical data (n = 2320) at a tertiary hospital. Predictive performance for lung-function metrics was assessed using the concordance correlation coefficient (CCC) and mean absolute error (MAE). Classification efficacy was evaluated via the area under the receiver operating characteristic curve (AUC), accuracy (ACC), precision, recall, and F1-score. Generalisability was further verified by replicating the experiments on three distinct backbone networks.

Results: This study included 1624 patients for model training, 348 patients for the validation set, and an additional 348 patients for the independent test set. The optimal model achieved a maximum CCC of 0.75 for forced vital capacity (FVC), corresponding to an MAE of 0.37, and a maximum CCC of 0.77 for forced expiratory volume in one second (FEV1), corresponding to an MAE of 0.33. For the binary classification task (COPD/Non-COPD), the highest AUC achieved through multi-task learning was 0.88, with a maximum ACC of 0.83. In the ternary classification task (COPD/preserved ratio impaired spirometry (PRISm)/Normal), the highest AUC reached 0.87, with a maximum ACC of 0.79.

Conclusions: Multitask-learning models that integrate chest CT images with basic clinical variables outperform their single-task counterparts in both the identification and classification of COPD. This approach supports evidence-based clinical decision-making and advances the delivery of precision medicine.

Background: The multifactorial mechanisms driving childhood obesity, a global public health challenge, are yet to be fully elucidated. We aimed to develop and externally validate three widely applied machine learning models alongside logistic regression in 2-18-year-old children and adolescents in Beijing and Tangshan to predict obesity risk. As a further step, we wanted to interpret the optimised model and translate it into a web-based tool to inform clinical decision-making.

Methods: We analysed data of 19 024 (training/testing) and 2410 (external validation) children and adolescents from Beijing and Tangshan, respectively. Using a set of factors including demographic, familial, socioeconomic, lifestyle, and perinatal variables, we developed four models (light gradient boosting machine, random forest, eXtreme gradient boosting (XGBoost), and logistic regression) and compared their predictive performance. After validation, we selected an optimised model and interpreted it using SHapley Additive exPlanations (SHAP) analysis. Then, we developed an online calculator with interpretable visualisations to enable real-time risk assessment.

Results: The XGBoost model exhibited superior performance, with an area under the receiver operating characteristic curve (AUROC) of 0.875 on the external validation set, significantly outperforming the logistic regression model (AUROC = 0.718). To identify the minimal feature subset that maintained model efficacy, we incrementally incorporated predictors in the descending order of SHAP importance values while assessing key performance metrics (accuracy, AUROC, and F-beta score). This SHAP-based analysis identified nine key predictors of childhood obesity: birth length, paternal body mass index (BMI), maternal BMI, sleep duration, physical activity, birth weight, maternal age at delivery, delivery mode, and gestational age. The deployed online tool provides individualised risk probabilities and SHAP-derived explanations.

Conclusions: The XGBoost model in our study was the superior ensemble learning method for predicting childhood obesity. The digital tool integrates this model and can help clinical practitioners determine individuals' risk of childhood obesity.

Background: Bangladesh faces significant challenges in accurately documenting causes of death (COD), largely due to incomplete vital registration systems, which lack COD reporting. A substantial number of deaths occur outside health facilities, often without medical certification, leading to further gaps in mortality data. Verbal autopsy (VA) has emerged as a viable method in low-resource settings to bridge this gap. We aimed to explore the feasibility and appropriateness of using VA by tracking burial records' contact information to enhance mortality documentation and inform health policies in the graveyards of urban Bangladesh.

Methods: We employed an exploratory design using both quantitative and qualitative methods. We conducted VAs using the contact details from six graveyards' burial records of Dhaka North City Corporation in Bangladesh, identifying participants through random sampling. In-depth interviews with data collectors, graveyard managers, and study participants provided insights into the feasibility and challenges of this process. We collected the data using the World Health Organization VA tool and assigned CODs using the InSilicoVA algorithm, applying thematic analysis to qualitative findings. We compared mortality trends with national data sets.

Results: We conducted 531 VAs using the contact information from burial site records in Dhaka North City Corporation graveyards, with sub-optimal consent rates varying by location. The leading CODs were acute respiratory infections (21%) and cardiac disease (19%), demonstrating the practicality of obtaining COD from the VA, and the feasibility of collecting burial records and contact details, if consent rates could be improved. Qualitative findings indicated that using burial records for such data collection faces obstacles, including low response rates, socioeconomic disparities in participation, difficulty finding contacts, and sampling inconsistencies.

Conclusions: We are the first to explore VA using contact information from burial records in urban Bangladesh. While the approach shows promise, the current feasibility results are of limited value without substantially improving consent coverage, representativeness, and standardisation. Only with these improvements can this method meaningfully strengthen COD documentation and provide reliable insights into population-level mortality trends.