Omicron, a severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) variant, entered Taiwan at the end of 2021. The Taiwanese government ended its "zero-COVID" policy in March 2022. Multiple coronavirus disease 2019 (COVID-19) outbreaks began in April 2022. We monitored the replacement of Omicron subvariants after BA.1/BA.2 and analyzed their correlation with COVID-19 outbreaks. We collected SARS-CoV-2 real-time qRTPCR-positive nasopharyngeal swabs from Kaohsiung Medical University Hospital (KMUH), Kaohsiung City, Taiwan, and performed sequencing for specimens exhibiting a cytopathic effect in Vero E6 cells to determine their clades and lineages. We analyzed the medical records of COVID-19 patients and identified hospitalization risk factor(s). We retrieved SARS-CoV-2 sequences identified in Taiwan from GISAID and analyzed their correlation with COVID-19 data from the Taiwan Centers for Disease Control. We analyzed the phylogenesis of KMUH-47 to KMUH-104 (SARS-CoV-2 isolates identified herein) and all of the Omicron subvariants from BA.5 to XBB.1 (n = 1930). Age and comorbidities were hospitalization risk factors. Men generally exhibited a greater fatality rate than women. COVID-19-related deaths predominantly occurred in individuals over 70 years old. The COVID-19-related case fatality rate increased as nucleotide (NT) and amino acid (AA) substitutions increased. The number of COVID-19-related cases and deaths progressively decreased with each outbreak between August 2022 and October 2023. Hospitalization was associated with age and the presence of comorbidities. COVID-19-related fatality was linked to sex, age, and the accumulation of NT and AA substitutions in emerging Omicron subvariants.
6 August 2025 in Journal of Infection and Public Health
Dengue virus (DENV) is a mosquito-borne flavivirus that remains a global public health threat, particularly in tropical regions where it is endemic. Accurate and timely diagnosis is critical for optimal patient care and effective outbreak containment. While reverse transcription-polymerase chain reaction (RT-PCR) remains the gold standard for dengue diagnosis, nonstructural protein 1 (NS1) rapid diagnostic tests (RDTs) offer a more accessible alternative. However, their performance may vary with infection type, serotype, and the timing of sample collection. We evaluated the diagnostic performance of two NS1 RDTs, Abbott NS1 and PBF NS1, against RT-PCR as the reference standard. A total of 183 dengue-positive and 40 dengue-negative samples were obtained during the 2023 dengue outbreak in Kaohsiung, Taiwan. Sensitivity, specificity, and predictive values were comprehensively assessed based on infection type, serotype, and post-symptom onset (PSO) intervals. Additionally, we analyzed the diagnostic benefits of combining NS1 detection with anti-DENV IgM serology. Abbott NS1 exhibited notably higher sensitivity (87.4 % vs. 83.1 %) and specificity (100 % vs. 97.5 %) compared to PBF NS1, though the differences were not statistically significant. Sensitivity varied by PSO, with Abbott NS1 achieving significantly higher sensitivity beyond three days (97.4 % vs. 76.9 %, P = 0.034). In secondary infections, Abbott NS1 outperformed PBF NS1 (86.8 % vs. 79.2 %). When combined with anti-DENV IgM serology, both RDTs demonstrated improved diagnostic sensitivity, particularly in secondary infections. Our findings underscore the importance of context-specific diagnostic tools. Combining NS1 antigen detection with IgM serology enhances diagnostic accuracy. While Abbott NS1 demonstrated slightly better performance, PBF NS1 remains a viable and non-inferior alternative.
Background: Multimodal large language models (MLLMs) show promise in medical imaging, but their performance is highly dependent on prompt engineering. This study systematically evaluates how different prompting strategies affect diagnostic accuracy in clinical laboratory image interpretation. Methods: We evaluated five MLLMs (ChatGPT-4o, Gemini 2.0 Flash, Claude 3.5 Sonnet, Grok-2, and Perplexity Pro (Claude 3.5 Sonnet)) using 177 proficiency testing images across three domains: blood smears (n = 78), urinalysis (n = 50), and parasitology (n = 49). Three prompting approaches were compared: (1) complex multi-choice prompts with 20 diagnostic options, (2) zero-shot open-ended prompts, and (3) two-step descriptive-reasoning prompts. Images were sourced from the Taiwan Society of Laboratory Medicine external quality assurance archives with expert consensus diagnoses. Results: Zero-shot prompting significantly outperformed complex multi-choice prompts across all models and domains (p < 0.001). With zero-shot prompts, Gemini achieved 78.5% overall accuracy (urinalysis: 92.0%; parasitology: 75.5%; blood smears: 64.1%), representing a 17% improvement over complex prompts. Two-step descriptive-reasoning prompts further improved blood smear accuracy by 8–12% for top-performing models, but showed minimal benefit in urinalysis and parasitology. The re-query mechanism (“please reconsider”) improved urinalysis accuracy by 7.6% but had a negligible effect on blood smears and parasitology. Conclusions: Prompting strategy critically determines MLLM diagnostic performance. Zero-shot approaches with minimal constraints consistently outperform complex multi-choice formats. The remarkable performance of general-purpose models in structured domains like urinalysis (>90% accuracy) demonstrates the considerable progress of multimodal AI. However, complex morphological tasks like blood smear interpretation require either specialized prompting techniques or domain-specific fine-tuning. These findings provide evidence-based guidance for optimizing AI integration in clinical laboratories.
Individuals with asymptomatic SARS-CoV-2 infection can unknowingly transmit the virus, yet identifying such infections in vaccinated populations remains challenging. We conducted a longitudinal study of 129 vaccine recipients immunized with various combinations of SARS-CoV-2 spike (S) protein vaccine platforms. Sera were collected before the first dose (v1), at 2 weeks (v7) and 6 months (v8) after the third dose. Taiwan’s first major COVID-19 outbreak occurred between v7 and v8. We measured anti-nucleocapsid (anti-N) and anti-S IgG antibody titers by ELISA and assessed virus-neutralizing activity using live virus and pseudovirus assays. By developing an iterative serial screening method, we identified asymptomatic infections among unconfirmed cases. Our v7-v8 paired cohort resolved into three distinct groups: confirmed cases (21%), asymptomatic infections (17%), and uninfected cases (62%). In normalized v8 sera, confirmed cases exhibited an anti-S+++/anti-N+++ phenotype, while uninfected cases showed an anti-S+/anti-N+ phenotype. Statistical analysis validated a distinct asymptomatic group characterized by intermediate anti-S but baseline anti-N antibody levels (anti-S++/anti-N+). This approach enables more accurate estimates of infection prevalence and vaccine efficacy.