Aye Moa

Background: Public opinion, which may be influenced by personal experiences, news, and social media, can impact compliance with public health measures (PHMs) during health emergencies. Artificial intelligence (AI) tools offer opportunities to analyze public opinion in real time during health emergencies. However, their performance in accurately identifying sentiment and themes in health-related online content remains unclear. Objective: This study aimed to evaluate the performance of natural language processing–based and large language model (LLM)–based AI tools when compared to human coding for sentiment analysis, topic modeling, and thematic analysis of public health datasets. Tools were selected to reflect those available to public health analysts and decision-makers. Methods: Data were collected via Google Alerts (GA) and social media posts from X (formerly known as Twitter) relevant to COVID-19 mitigation PHMs from December 2022 to February 2023. Following relevance screening, the sentiment of the complete datasets was analyzed by a human rater, with descriptive statistics used to summarize the overall sentiment profile. Subsets of 400 GA articles and 400 tweets were manually coded for sentiment by 2 human raters. Results were compared with outputs from 5 AI tools, including VADER (Valence Aware Dictionary and Sentiment Reasoner), SentimentGI, SentimentQDAP, Microsoft Azure, and OpenAI’s ChatGPT-4. Topic modeling of the GA and X datasets was conducted using latent Dirichlet allocation in R and zero-shot prompting in ChatGPT-4 and compared with manual topic summaries. Thematic analysis of positive and negative sentiment datasets was conducted by a human rater and ChatGPT-4, with outputs evaluated for proficiency and reasonableness. The sentiment of the entire datasets was analyzed by a human rater, and descriptive statistics were calculated. Results: Of 2227 GA results and 3484 tweets, 58% (n=1238) and 71% (n=2473), respectively, were relevant to PHMs. Human-coded sentiment analysis showed mostly neutral reporting in the news media, while social media expressed more polarized views. Across both datasets, AI tools demonstrated poor concordance with human-coded sentiment (Cohen κ <0.5 for all tools and sentiment categories). Topic modeling with ChatGPT-4 aligned more closely with human-rated topics than latent Dirichlet allocation, and of the 20 LLM-generated thematic outputs, 13 were rated proficient, and 7 were rated partially proficient. LLM outputs provided coherent, high-level summaries but lacked contextual insight. Human and LLM thematic analyses both identified themes of vaccine effectiveness, debate regarding PHMs, and public trust. Conclusions: Accessible AI tools demonstrate limited reliability for sentiment classification of health-related online text but show promise for rapid thematic exploration when combined with human oversight. These tools could complement traditional qualitative research in the context of health emergencies; however, they require human review to enhance the accuracy of interpretation. Further research is needed for non-English datasets.

Background: The reported incidence of pertussis, a vaccine-preventable disease, has been increasing in recent years. This study aimed to estimate the burden of pertussis and the vaccination rate in Australian adults in primary care. Methods: Deidentified data for participants aged ≥18 years were extracted from the MedicalDirector (MD) primary care software from 2008 to 2019. We estimated the cumulative incidence of diagnosed pertussis in adults by age and risk groups and vaccine coverage in cases and a control group (not diagnosed with pertussis or a coughing illness). We also examined the incidence of unspecified coughing illness in the study population. Results: Of the 764,864 subjects included in the study, 1788 (0.2%) were diagnosed with pertussis between 2008 and 2019, corresponding to an average annual diagnosis rate of 76.9 per 100,000 population. About 31,110 (4.1%) of adults had an unspecified coughing illness. The highest rate was observed in 2011 and higher in females (63.3%), and the diagnosis rate was stable across all age groups. Underlying chronic conditions were more prevalent among pertussis cases than controls (58.7% vs. 18.8%), with asthma or chronic obstructive pulmonary disease (COPD) being the most common. Overall, 14% of cases received a pertussis vaccination during the study period. Diagnostic testing for pertussis was performed in 34.1% of pertussis cases. Estimated conservative costs per pertussis patient ranged from AUD 473 to AUD 909, with higher costs observed in individuals with complications. Conclusions: In the outpatient setting, there was a notable burden of pertussis among adults under 65 years of age, particularly those with underlying medical conditions, such as asthma and COPD, which appear to be significant risk factors. Due to the low rate of pertussis testing among all coughing illnesses, a proportion of non-specific coughing illness may be undiagnosed pertussis. The observed low vaccination rates highlight a need for increased awareness, improved diagnostic efforts, and prevention strategies in primary care.

Background: The reported incidence of pertussis, a vaccine-preventable disease, has been increasing in recent years. This study aimed to estimate the burden of pertussis and vaccination rate in Australian adults in primary care. Methods: Deidentified data for subjects ≥18 years were extracted from Medical Director (MD) primary care software from 2008-2019. We estimated the cumulative incidence of diagnosed pertussis in adults by age and risk groups, and vaccine coverage in cases and a control group (not diagnosed with pertussis or a coughing illness). We also examined the incidence of undiagnosed coughing illness in the study population. Results: Of the 764,864 subjects included in the study, 1,788 (0.2%) were diagnosed with pertussis between 2008 and 2019, corresponding to an average annual diagnosis rate of 76.9 per 100,000 population. About 31,110 (4.1%) of adults had an undiagnosed coughing illness. The highest rate was observed in 2011 and higher in females (63.3%), and the diagnosis rate was stable across all age groups. Underlying chronic conditions were more prevalent among pertussis cases than controls (58.7% vs 18.8%), with asthma or chronic obstructive pulmonary disease (COPD) being the most common. Overall, 14% of cases received a pertussis vaccination during the study period. Diagnostic testing for pertussis was performed in 34.1% of pertussis cases. Estimated conservative costs per pertussis patient ranged from AUD $473 to $909, with higher costs observed in individuals with complications. Conclusion: In the outpatient setting, there was a notable burden of pertussis among adults under 65 years of age, particularly those with underlying medical conditions, such as asthma and COPD which appear to be significant risk factors. Due to low rate of pertussis testing among all coughing illness, a proportion of non-specific coughing illness may be undiagnosed pertussis. The observed low vaccination rates highlight a need for increased awareness, improved diagnostic efforts and prevention strategies in primary care.

Background: In Australia, 2017 was a severe A/H3N2 season and, therefore, we estimated the effectiveness of standard-dose quadrivalent influenza vaccine in preventing hospitalization for cardiovascular disease (CVD) among New South Wales (NSW) residents aged ≥50 years. Methods: We conducted a nested, matched case–control study within the 45 and Up study, linking data from the Australian Immunization Register, NSW Admitted Patient Data Collection and Pharmaceutical Benefits Schedule. Cases were individuals hospitalized for CVD and controls were those who were hospitalized for gastrointestinal diseases. The two groups were balanced using 1:1 propensity score matching based on age group (50–64, 65–74, 75–84, and ≥85 years) and sex. After adjusting for confounders (smoking, body mass index and income), we calculated the adjusted odds ratio (aOR) for vaccination during the season using multivariable logistic regression. E-values were estimated to assess residual confounding. Vaccine effectiveness (VE) was calculated as (1 − aOR) × 100. Results: There were 10,445 (4452 cases and 5993 controls) study participants. After matching, 8904 (85.2%) were retained with a mean age of 76.4 ± 10.4 years and 58.3% men. Following adjustment for confounders, the aOR of averting a CVD hospitalization was 0.15 (95% CI: 0.13 to 0.17; p < 0.001). The estimated VE against CVD hospitalization was 85% (95% CI: 83 to 87). We found an E-value of 12.82, indicating strong evidence of minimal residual confounding. Conclusions: In the severe 2017 influenza A/H3N2 season in Australia, we observed a high VE in preventing cardiovascular hospitalization despite a low VE against influenza infection prevention. Improving vaccine uptake may reduce cardiovascular burden.

Background: The cause of acute paediatric hepatitis of unknown aetiology (2022) has not been established despite extensive investigation.Objective: To summarise the evidence for and against a causal role for human adenovirus (HAdv), adeno-associated virus 2 (AAV-2) and SARS-CoV-2 in outbreaks of paediatric hepatitis in 2022.Methods: We appraised and summarised relevant evidence for each of the Bradford Hill criteria for causality using quantitative (statistical modelling) and qualitative (narrative coherence) approaches. Each team member scored the evidence base for each criterion separately for HAdv, AAV-2 and SARS-CoV-2; differences were resolved by discussion. We additionally examined criteria of strength and temporality by examining the lagged association between SARS-CoV-2 positivity, respiratory HAdv positivity, positive faecal HAdv specimens and excess A&E attendances in 1–4 years for liver conditions in England.Results: Assessing criteria using the published literature and our modelling: for HAdv three Bradford Hill criteria (strength, consistency and temporality) were partially met; and five criteria (consistency, coherence, experimental manipulation, analogy and temporality) were minimally met. For AAV-2, the strength of association criterion was fully met, five criteria (consistency, temporality, specificity, biological gradient and plausibility) were partially met and three (coherence, analogy and experimental manipulation) were minimally met. For SARS-CoV-2, five criteria (strength of association, plausibility, temporality, coherence and analogy) were fully met; one (consistency) was partially met and three (specificity, biological gradient and experimental manipulation) were minimally met.Conclusion: Based on the Bradford Hill criteria and modelling, HAdv alone is unlikely to be the cause of the recent increase in hepatitis in children. The causal link between SARS-CoV-2, and to a lesser degree AAV-2, appears substantially stronger but remains unproven. Hepatitis is a known complication of multisystem inflammatory syndrome in children following COVID-19, and SARS-CoV-2 has been linked to increased susceptibility to infection post-COVID-19, which may suggest complex causal pathways including a possible interaction with AAV-2 infection/reactivation in hosts that are genetically susceptible or sensitised to infection.

Background: As the COVID-19 pandemic began, cases of respiratory illness of unknown cause could have presented early disease signals. The detection of such signals can provide useful clues regarding disease emergence and opportunities for early outbreak control. This study assessed trends in the global incidence of reports of respiratory illness of unknown cause as early indicators of COVID-19 in late 2019, using the early warning system, EPIWATCH, to investigate the disease’s origins and early epidemiology. Methods: Global news articles published between 1-Oct and 31-Dec in 2019 and collected by EPIWATCH, an open-source Artificial Intelligence (AI)-driven epidemic surveillance system, were searched for keywords relating to unknown respiratory illness. Control periods representing the pre-COVID period (1-Oct to 31-Dec 2016-2018) were used to compare report trends over time. Data on report location, case demographics, outbreak setting, and type of illness, were extracted and descriptively analysed. Fisher’s exact tests were conducted to assess proportional differences between years. Results: Overall, 113 EPIWATCH reports of unknown respiratory illness were identified from 12 countries in 2019. Most reports (n=111 [98.2%]) were published in Northern hemisphere countries, in particular Russia (n=A72 [63.7%]) and the United States (n=17 [15.0%]). China experienced a month-on-month increase in reports. Schools were the most reported outbreak setting (n=49 [43.4%]). The proportion of included reports in 2019 was significantly higher compared to 2016-2018 (p<.001). Conclusion: Potential early COVID-19 signals were detected from various countries before the end of 2019. AI-driven open-source disease surveillance systems, such as EPIWATCH, can provide early warnings to enhance disease surveillance and improve outbreak prevention and response.

We estimated the effectiveness of influenza vaccines in preventing laboratory-confirmed influenza among older adults in aged care. Electronic database searches were conducted using search terms, and studies were selected as per the selection criteria. Fourteen studies were included for final review. The studies exhibited considerable variation in reported vaccine effectiveness (VE) across different seasons. Among the observational studies, VE ranged from 7.2% to 89.8% against laboratory-confirmed influenza across different vaccines. Randomized clinical trials demonstrated a 17% reduction in infection rates with the adjuvanted trivalent vaccine. The limitations include the small number of included studies conducted in different countries or regions, varied seasons, variations in diagnostic testing methods, a focus on the A/H3N2 strain, and few studies available on the effectiveness of enhanced influenza vaccines in aged care settings. Despite challenges associated with achieving optimal protection, the studies showed the benefits of influenza vaccination in the elderly residents.

Background. Healthcare workers (HCWs) are at risk of SARS-CoV-2 infections due to occupational exposure. The use of airborne personal protective equipment (PPE) significantly reduces this risk. In June 2021, an epidemic of the Delta variant began in New South Wales (NSW), Australia. Concurrent PPE guidelines, set by the Clinical Excellence Commission (CEC), restricted the use of respirators. Objective. To understand the relationship of PPE guidelines with workplace-acquired HCW SARS-CoV-2 infections in different clinical settings and to examine the relationship between rates of community transmission and workplace-acquired HCW infections during the Delta outbreak in NSW. Methods. Total SARS-CoV-2 HCW infections between 13 June and 30 October 2021 (first four months of the Delta wave) were estimated from the government COVID-19 surveillance reports and compared with the surveillance reports of community transmission. In the absence of a detailed reporting of HCW infections, open-source data including news articles, media releases, and epidemiological surveillance reports were also collected. Data were extracted on HCW cases of SARS-CoV-2 from four hospitals, including the number of HCW cases (per NSW Health definition), clinical setting, PPE guidelines, and evidence of increasing local transmission. Results. SARS-CoV-2 infections in HCW identified as workplace-acquired infections (n=177) and those without a known transmission source (n=532) increased during the period of increasing community transmission (n=75,014) in NSW. Four hospital COVID-19 clusters affecting 20 HCWs were identified between June and October 2021. HCW clusters occurred in general wards where staff were recommended to wear surgical masks. No workplace-acquired HCW infections were reported in these hospitals from critical care wards, where respirators were recommended during the same outbreak weeks. Conclusions. Differences in PPE policy across different wards may leave healthcare staff at risk of SARS-CoV-2 infection. During periods of high community transmission, respirators should be provided to protect hospital staff. Formal reporting of HCW infections should occur.