University of Hargeisa

Robot-assisted distal pancreatectomy (RDP) was developed to overcome technical limitations of laparoscopic distal pancreatectomy (LDP), yet uncertainty persists regarding oncologic adequacy, learning-curve effects, and outcomes in high-risk subgroups. We synthesized current evidence to address these gaps. We systematically searched PubMed and EMBASE, in accordance with PRISMA guidelines, from inception to 2025 to identify comparative studies of RDP versus LDP. Using random-effects models, we calculated weighted mean differences (WMDs) for continuous outcomes and risk ratios (RRs) for dichotomous outcomes, and performed subgroup analyses, including pancreatic ductal adenocarcinoma (PDAC), along with meta-regression to explore heterogeneity sources. Sixty-four studies comprising 15,790 patients (5,723 RDP; 10,067 LDP; mean age 60.5 years; BMI 26.1 kg/m²) were included. RDP resulted in lower blood loss (WMD − 52.0 mL; p < 0.00001), fewer conversions (RR 0.49; p < 0.00001), and fewer unplanned splenectomies (RR 0.59; p < 0.0001). Operative time was longer (WMD + 24.06 min; p < 0.00001). Postoperative morbidity, POPF, PPH, infection, reintervention, and mortality were comparable. Length of stay was shorter with RDP (WMD − 0.57 days; p < 0.00001). Although lymph node yield appeared higher with LDP in the overall and PDAC cohorts, this difference was no longer significant in a sensitivity analysis, and R0 resection rates remained comparable. Costs were higher with RDP, with substantial heterogeneity. RDP and LDP demonstrate comparable safety and oncologic outcomes. RDP reduces blood loss, conversions, and splenectomy but increases operative time and cost. The operative time disadvantage likely reflects learning-curve. Selective use in high-risk and complex resections is supported; cost-effectiveness warrants further study.

Biological invasions, driven by the spread of non-native species, have become a critical global issue because of their far-reaching ecological and socioeconomic impacts. Effective communication of the risks of biological invasions is essential for implementing robust policy and legislation and gaining public support for conservation efforts. However, current policies often suffer from fragmentation and ineffectiveness, largely due to inadequate risk communication and complex multi-level governance. To address this challenge, we develop a global framework designed to enhance clearer communication about biological invasion risks. The framework contextualizes key terms across three domains in invasion science: species invasiveness, risk analysis, and decision support tools. Using both diffusion-of-English and ecology-of-language paradigms, and following a three-step process involving preliminary consensus, AI querying, and ground-truthing with final consensus, we validate the framework in 70 non-English languages which, together with English, have official status in at least one country and collectively cover all 195 countries worldwide. Our findings reveal that while terminology for risk analysis is well established, terminology for species invasiveness and, especially, for decision support tools remains underdeveloped in many languages, hindering effective communication and policy implementation. Our framework underscores the importance of cultural and political neutrality. By promoting clearer risk communication among scientists, policymakers, and the public globally, we aim to reduce policy fragmentation and foster enhanced collaboration in risk mitigation. We recommend expanding multilingual decision support tools to include the full risk analysis process: risk identification, risk assessment, and risk management. This will support intergovernmental mitigation efforts and promote a unified global response to biological invasions.

Multimorbidity is a growing public health concern globally, yet remains under-researched in fragile and conflict-affected settings, including Somalia, a country in the Horn of Africa with limited health infrastructure. This study investigates the prevalence, regional patterns, and individual- and contextual-level determinants of multimorbidity among women aged 15–49 years in Somalia. A cross-sectional analytical study design was applied using nationally representative data from the 2020 Somalia Demographic and Health Survey (DHS). The final sample included 8,172 women after excluding cases with missing or incomplete information. Multimorbidity was defined as the presence of ≥ 2 self-reported chronic conditions. Bayesian multilevel logistic regression was employed to account for hierarchical data structure— individuals (Level 1) nested within regions and residence types (Level 2). Random intercept and slope models were compared using Deviance Information Criterion (DIC), with posterior distributions estimated via Markov Chain Monte Carlo (MCMC) simulation. Associations were assessed using odds ratios (ORs) and 95

Malnutrition and household food insecurity are major public health concerns among under-five children in Somaliland, often sharing overlapping risk factors. Using data from the 2020 Somaliland Health and Demographic Survey (n = 3961), this study assessed their prevalence, spatial distribution, and shared determinants through descriptive statistics, spatial autocorrelation, and a multivariate Bayesian spatial model. The prevalence of malnutrition and food insecurity was 42.7% and 26.7%, respectively, with higher risk of malnutrition among children in food-insecure households (OR = 1.17). Both outcomes were most prevalent in eastern regions such as Sool and Sanaag, with significant clustering of food insecurity and their joint occurrence. Risk factors for food insecurity included large family size, lack of radio and agricultural land, and longer water-fetching time, while higher wealth, electricity access, and more sleeping rooms were protective. Malnutrition was associated with female sex, absence of cooling facilities, and longer water-fetching time. The joint burden was influenced by nomadic residence, household wealth, family size, sleeping rooms, and electricity. These findings highlight the geographic concentration and interconnectedness of malnutrition and food insecurity and underscore the need for multi-sectoral interventions targeting poverty reduction, water access, and infrastructure, with a particular focus on eastern regions. Strengthening nutrition-sensitive policies and integrated programs that address both household resources and structural barriers could reduce the overlapping burden of child malnutrition and food insecurity, providing critical evidence for policymakers and development partners in Somaliland and similar fragile contexts.

Offensive language presents significant challenges on the internet and requires robust moderation. However, the efficacy of such moderation often depends on providing clear and interpretable justifications for each classification. Unfortunately, many existing datasets lack annotated rationales, and most detection models offer limited interpretability and transparency. These limitations hinder the development of trustworthy systems and the implementation of effective content moderation strategies. In this paper, we introduce SomOffXplain, an interpretable framework for detecting offensive language in Somali, which generates human-understandable explanations for its predictions. SomOffXplain performs span-level rationale extraction at both the word and phrase levels, enabling it to highlight text segments that support its predictions. Given that Somali is a low-resource language, we first construct a new benchmark dataset of 10,175 samples, each annotated with human-provided rationales. We evaluate our method against five fine-tuned pre-trained models using Local Interpretable Model-Agnostic Explanations (LIME). Additionally, we adapt four large language models (LLMs) through few-shot and zero-shot prompting to assess their ability to understand and produce rationales in Somali. Our proposed model demonstrates superiority in terms of explainability and predictive accuracy, exhibiting higher plausibility and faithfulness compared to the baselines. Furthermore, our results reveal that half of the state-of-the-art LLMs evaluated fail to generate high-quality rationales that align with human-annotated ground truth rationales, whereas LIME-based methods also prove to be weak explainers for Somali text. We believe our contributions support online safety, help prevent harassment in under-resourced language communities, enhance the trustworthiness of language models, and promote transparency in artificial intelligence systems.