Mass casualty incidents (MCIs) in high-risk environments pose major challenges for coordinated emergency response. Training is often infrequent, resource-intensive, and lacks interagency consistency. This study explores the use of Virtual Reality (VR) simulation to train responders in the RAMP triage model across emergency services.

An observational qualitative design was used. Sixteen participants from various emergency services engaged in a VR-based MCI scenario involving 26 patients and hazardous conditions. The scenario required rapid RAMP triage based on essential cues (radial pulse and the ability to follow commands). Structured interviews followed, and data were analyzed thematically.

Three themes emerged: (1) Deficiencies in current training, including inconsistent MCI protocols, lack of guideline familiarity, and limited interagency practice; (2) VR as an effective, low-resource training method enabling repeatable and safe practice—RAMP triage was found intuitive and efficient, even for non-medical personnel; and (3) prerequisites for VR implementation, such as realistic design, technical infrastructure, and stakeholder involvement to support shared understanding.

VR-based MCI training is a feasible and effective supplement to traditional drills. It enables scalable and flexible skill-building, though it should complement and not replace live exercises.

Budget impact analyses are essential for decision-making processes regarding the incorporation of technologies into healthcare systems. Despite advancements driven by the National Committee for Health Technology Incorporation (CONITEC), the Brazilian Network for Health Technology Assessment (REBRATS), and the National Supplementary Health Agency (ANS), challenges persist in evaluating the economic impact of new technologies, including methodological inconsistencies in submitted dossiers.

Reports on the critical analysis of dossiers submitted to the ANS prepared by a health technology assessment center between December 2022 and December 2024 were analyzed. Criticisms of the submitted dossiers were categorized into eleven topics: (i) analytical model; (ii) reference scenario; (iii) alternative scenario; (iv) time horizon; (v) target population; (vi) direct costs; (vii) market behavior; (viii) sensitivity analysis; (ix) input data; (x) output data; and (xi) final decision. Descriptive and quantitative data were used to identify patterns and methodological gaps.

The critical analysis identified recurring issues, including underestimated target populations in 50 percent of cases and inadequate direct cost evaluations in 40 percent. These inadequacies were often linked to outdated data sources, such as the Painel dos Dados do TISS database. Analytical models were deemed adequate in 70 percent of cases, whereas sensitivity analyses were insufficient in 30 percent of cases. Market behavior projections were conservatively estimated in 60 percent of dossiers, affecting the accuracy of budget impact projections. These methodological inconsistencies hindered the reproducibility of analyses and compromised evidence-based decision-making.

The findings highlighted the need for methodological standardization in dossiers submitted to the ANS, emphasizing the need to update data sources and ensure transparency in budget impact calculations. These improvements could enhance evaluation validity and support better decision-making for technology incorporation. Future research should address strategies to reduce uncertainties and inconsistencies in analytical models and cost estimations.

Da Vinci robot-assisted surgery (dV-RAS) has been around for more than 20 years and has become the standard of care for select procedures in specific countries. Recently, health technology assessment agencies have used local evidence to inform their evaluation of dV-RAS. We aimed to compare global, Pan-European (Pan-EU), and Pan-Asian systematic literature reviews on dV-RAS for seven malignant procedures.

The PubMed, Scopus, and Embase databases were systematically searched through to 31 December 2022 following PRISMA and PROSPERO guidelines (CRD42023466759). Studies that compared dV-RAS with laparoscopic or video-assisted thoracoscopic surgery (lap/VATS) or open oncologic surgery and reported on relevant clinical outcomes were included. Studies were checked for the source of clinical data and categorized as being either Pan-EU or Pan-Asian. The global analysis included all eligible studies. Pooled odds ratios or mean differences were calculated for randomized, prospective, and database studies using R software and a fixed-effect or random-effects (when heterogeneity was significant) model. The revised Cochrane risk-of-bias and ROBINS-I tools were used to assess bias.

The global systematic literature review identified 230 studies (34 randomized, 74 prospective, 122 database), including 78 Pan-EU and 35 Pan-Asian studies. Results from the global and regional studies agreed on rates of conversions and blood transfusions and lengths of hospital stay. The global, Pan-EU, and Pan-Asian studies did not align on: operative time (Global: dV-RAS longer; Pan-EU: dV-RAS longer versus open; Pan-Asian: dV-RAS longer versus lap/VATS); 30-day complications (Global and Pan-Asian: dV-RAS lower; Pan-EU: dV-RAS lower versus open); 30-day readmissions (Global and Pan-EU: dV-RAS lower; Pan-Asian: dV-RAS lower versus open); 30-day reoperations (Global: dV-RAS lower versus open; Pan-EU and Pan-Asian: dV-RAS similar); 30-day mortality (Global: dV-RAS lower; Pan-EU: dV-RAS lower versus open; Pan-Asian: dV-RAS longer versus lap/VATS).

Our analysis highlighted that global and regional evidence aligned on select outcomes. The differences we observed might be associated with surgeon experience, limited evidence, or other unknown regional biases. These findings should be considered by health technology assessment agencies when looking at regional data versus global evidence.

Generative artificial intelligence (AI) is revolutionizing real-world evidence generation in health care. This study compared chemotherapy recommendations for women with breast cancer across different clinical risk profiles and Oncotype DX Breast Recurrence Score® test results, as obtained from a Delphi panel of experts, with recommendations generated by ChatGPT. The objective was to analyze concordances and differences between AI-generated and expert-driven insights.

An online survey of 10 independent breast cancer experts, blinded to each other’s responses, assessed chemotherapy recommendations for patients with early stage, node-negative breast cancer. Responses were analyzed by clinical risk (high versus low), age group (≤50 versus >50 years), and recurrence score (RS) (<11, 11 to 25, >25) from Oncotype DX. ChatGPT, using automated prompt engineering, addressed the same scenarios as agent-based oncologists. Expert recommendations were summarized as arithmetic means, while ChatGPT responses were analyzed for concordance. A one-sample t-test compared mean estimates between the Delphi panel and ChatGPT results, highlighting differences in recommendations across groups.

The Delphi panel and ChatGPT provided clinically similar chemotherapy recommendations for patients evaluated with Oncotype DX, with no statistically significant differences in eight out of 12 scenarios. Both agreed on zero percent chemotherapy for patients with low clinical risk (RS<11) and showed comparable results for high clinical risk (RS>25), including patients under 50 years (8% versus 10%) and over 50 years (6% versus 5%). The largest divergence, which was statistically significant, was observed for patients with low clinical risk (RS 11 to 25) who were over 50 years (1% versus 20%). High-risk patients consistently received strong recommendations, with near perfect agreement for those with high clinical risk (RS>25) who were under 50 years (98% versus 95%).

From a decision-making perspective, the responses from ChatGPT and the Delphi panel were very similar, suggesting that AI can effectively support the health technology assessment process. This alignment highlights AI’s potential to accelerate decision-making, offering a faster alternative to the traditional, time-consuming Delphi model while maintaining reliable chemotherapy recommendations.

Despite the significant impact of the media on individuals with mental illness, newspaper articles related to antidepressants have not been systematically studied. The present study aimed to analyze Brazilian journalistic coverage on the use of antidepressants to understand how the news presents and shapes the topic of antidepressants for its readers.

This qualitative study evaluated journalistic content on antidepressant use from Folha de São Paulo, a newspaper available in digital format. Articles published between 1 January 2019 and 31 December 2023 were collected and stored in a database for analysis. Natural language processing (NLP) techniques, combined with machine learning, were applied. The R software (version 4.3.3) with tidytext, tm, tidyverse, and stringr packages was used for the analyses. The study identified patterns and trends in language usage, focusing on frequently occurring terms to understand how antidepressants are portrayed in media content.

The initial research was conducted using the keyword “antidepressants” in the search engine of the Folha de São Paulo. Of the articles assessed for eligibility, 182 were included in the study. Across all the articles, the analysis aimed to identify the most frequently used words, resulting in a word cloud. The most used words in the text body were “treatment,” followed by “years,” “anxiety,” “women,” and “pandemic.” Regarding the most frequent words in the titles, they were “depression,” “health,” and “study.” On average, there was no significant change in the total number of n-grams used.

The findings illustrated how Brazilian media frames antidepressant use, revealing potential misinformation or stigma. Understanding these representations can guide strategies for improving public awareness and reducing stigma around mental health. By highlighting trends in journalistic narratives, this study contributes to public health policies that promote accurate, responsible communication about antidepressant use.

Incorporating patient preferences into health technology assessment represents a promising avenue for its enhancement. However, the complexity and diversity of methods for evaluating preferences, coupled with uncertainties regarding their impact on decision-making processes, present significant challenges to their effective integration. Consequently, the objective of this study was to evaluate the application of the MaxDiff analysis methodology for determining patient preferences, with a focus on pharmaceutical services as a case study.

The experimental design for scaling MaxDiff was developed using Sawtooth Lighthouse software. The design parameters included 12 pharmaceutical services. Descriptive statistics, hierarchical Bayesian analysis, latent class analysis, and logistic regression methods were employed to generate preference results.

The most preferred services were identified as follows: the creation of a personalized list of safe medications to address patient needs, with a quantitative contribution of 0.5; and the provision of rapid tests (for influenza and Helicobacter) and stroke risk assessment (quantitative contribution of 0.31). In contrast, patients did not prioritize services related to contraception selection or weight control and weight loss program development. Descriptive statistics, hierarchical Bayesian analysis, and logistic regression methods sequentially identified the most preferred services. Latent class analysis revealed two distinct consumer segments: Segment 1 (39%) and Segment 2 (60%), which differed in their sociodemographic characteristics and preferred services.

The obtained results can be utilized as patient-based evidence to understand patients’ unmet needs in the assessment of technologies such as pharmaceutical services, taking into account the local context. Despite the simplicity of the MaxDiff analysis methodology, the resource intensity and the level of knowledge among specialists and patients regarding preference collection methods remain significant barriers to its implementation.

Over the years many researchers have sought to understand the decision-making process of CONITEC, the Brazilian health technology assessment (HTA) agency. The emergence of high-cost technologies, such as monoclonal antibodies (mAbs), has heightened the importance of delineating the decision-making criteria. This study aimed to elucidate the criteria employed by CONITEC for the incorporation of mAbs in Brazil.

All CONITEC reports published between 2019 and 2024 on mAbs were included. Descriptive statistics were utilized to summarize the data. The statistical significance between outcomes and covariates was assessed using the Kruskal-Wallis test for continuous variables and the chi-square test for categorical variables. Logistic regressions were produced to evaluate the impact of each covariable on the recommendations. Results with a p-value of less than 0.10 were considered statistically significant due to the small sample size. All analyses were conducted using R software.

After cleaning the database, 53 reports were included, encompassing 37 mAbs evaluated for 45 indications. Sixteen submissions (30.1%) received a positive recommendation. Efficacy (p=0.009) and the preliminary decisions (p<0.001) were significantly associated with the final recommendation. Further analysis showed that an incremental cost-effectiveness ratio (ICER) below BRL150,000 (USD65,020) was associated with a higher chance of a positive recommendation (p=0.050). Logistic regression revealed that the logarithm of the ICER was significantly associated with recommendations (odds ratio 0.618; p=0.031), indicating that increasing the ICER by 2.7 times was associated with a 38.2 percent lower chance of listing.

Efficacy is important for decision-makers in Brazil. However, the results of economic analyses also influence the recommendation of mAbs. An ICER above BRL150,000 (USD65,020) per quality-adjusted life year significantly reduced the chances of a mAb being listed.

The healthcare and rehabilitation processes for individuals with low back pain impose significant costs on healthcare systems worldwide. Understanding the economic burden of this condition can help policymakers implement preventive strategies and promote a more rational allocation of health resources. The aim of this study was to investigate the costs of managing low back pain in Brazil between 2010 and 2019.

This study evaluated the costs of low back pain from the perspective of the Brazilian Unified Health System (SUS) in the outpatient setting. Data were collected from the Ambulatory Information System and analyzed by sex, age group, and health condition (low back pain). Ambulatory care data were presented as number of visits, total annual cost, and number of procedures performed. In addition, linear regression analysis using generalized linear models (gamma distribution with identity link) was performed to examine the association between total outpatient care costs and predictors such as age, gender, and race.

Between 2010 and 2019, the SUS spent more than USD189 million to treat low back pain in adults. During this period, more than 16 million physical therapy sessions, 7,000 surgeries, and one million imaging studies were performed. These procedures accounted for approximately 85 percent of total spending on low back pain in Brazil. Despite the high number of physical therapy sessions, significant expenditures were still observed for surgical procedures. Regression analysis showed that men had higher costs per procedure than women.

The study revealed significant costs for low back pain in Brazil, with higher expenditures for men and those aged 34 to 63 years. Major expenditures were associated with surgery and physical therapy, whereas the prescription of imaging studies decreased, which is in line with recommendations. These findings highlight the need for targeted strategies to effectively manage resources for low back pain in the ambulatory care setting.

Diabetes is a condition that affects public health based on its increased incidence, prevalence, morbidity, and mortality. According to the latest Atlas of the International Diabetes Federation, it is estimated that there are 537 million adults with the condition; for Brazil, the estimate is 16 million. Investment in diabetes and its complications exceeds USD42 billion in Brazil.

This national study was conducted from 1 July to 22 August 2024. Interviews were conducted online with 1,843 Brazilians over 18 years of age with diabetes. The study’s objective was to identify barriers to diabetes treatment in Brazil after the incorporation of health technologies in the Unified Health System (SUS) aimed at diabetes. Data were processed according to the profile of reported diabetes diagnosis in the National Health Survey 2019, conducted by the Brazilian Institute of Geography and Statistics.

The preliminary study indicated that 56 percent of those diagnosed with diabetes are over 60 years of age, 58 percent have elementary education, 54 percent have a monthly family income of up to two minimum wages, 75 percent have type 2 diabetes, 67 percent say they have hypertension, and 63 percent have high cholesterol. Obesity was reported by 39 percent, 82 percent use oral medication, 67 percent undergo tests through the SUS, and 84 percent obtain free medication through the SUS. Of those interviewed, 60 percent cite the availability of medication as the main factor in monitoring the condition, followed by attention from doctors (53%).

The aging population poses challenges for diabetes control in Brazil and will further strain the public health system. Current barriers, such as bottlenecks in care and lack of medicines and doctors, are likely to worsen. Vulnerable groups with lower levels of education and income are currently the most affected. With the increasing age of the population, the system will spend more resources and may have difficulty sustaining itself in a few years.

Skin-related neglected tropical diseases (skin NTDs) are very prevalent in endemic areas. Resources to manage them are very scarce. The World Health Organization’s Skin NTDs app is designed to help frontline health workers in identifying skin NTDs (n=13) and common skin conditions (n=24). A beta version including artificial intelligence (AI) was developed, and its accuracy and usability was assessed in real life conditions.

The Skin NTDs app usability and user experience was assessed in frontline healthcare workers (n=38) in Kenya. Participants answered the user Mobile App Rating Scale (uMARS) questionnaire. Focus group discussions (n=4) and semi-structured interviews (n=15) were used to get an in-depth understanding of the user experience. To assess accuracy of the AI algorithm, 40 participants from five counties in Kenya used the app for five months, uploading photographs of the skin lesion (n=605) to an external platform. AI algorithm accuracy was calculated based on the gold standard of consensus diagnosis reached by three independent dermatologists.

The Skin NTDs app received high scores on the uMARS questionnaire (mean app quality 3.82/5 and perceived impact 4.1/5; n=38). Focus group discussions and interview responses aligned with the uMARS findings, reinforcing the positive assessment of the app. It helped to empower professionals, increased their knowledge about skin diseases, and improved their communications skills with patients. The app was time saving and reduced referral of patients to specialists. Some features to be improved were identified. Overall accuracy of the app was found to be 80 percent in diseases where AI has been trained with a higher number of photos of endemic skin conditions.

The Skin NTDs app showed commendable quality and holds potential to be scaled up and implemented at the national level in Kenya and globally. It performs well as a clinical decision support system to help frontline healthcare workers identify potential skin diseases that patients suffer from and to reduce the number of referrals to dermatologists in contexts where there is a lack of specialized professionals.

Chile has been conducting various efforts toward reforming the healthcare system, yet significant discrepancies persist in the equitable access to and utilization of healthcare resources. The aim of this scoping review was to map the existing literature on hearing health care and treatments for hearing loss (HL) in Chile to identify local determinants that may contribute to stratification of access to hearing health care.

The Joanna Briggs Institute guidance for scoping reviews was followed. The PCC criteria (Population, Concept, Context) was used to guide development of the search strategy. Searches were conducted in MEDLINE (PubMed), the Cochrane Library, and Science Direct databases and supplemented by a manual search. Searches were limited to publications from 2000 to June 2023, with no restrictions on language or publication type. Two independent reviewers screened all retrieved references, assessed the eligibility following the PCC criteria, and charted data of the eligible publications. Disagreements were solved through discussion with a third reviewer. A structured narrative synthesis of findings was conducted.

The search yielded 506 unique records for screening, of which 104 full-text publications were assessed and 39 were included in the review. The evidence revealed that the treatment of HL is publicly financed for children younger than four years (any type of HL, degree moderate or higher), the elderly aged 65 years or older (bilateral HL requiring a hearing aid) and people at least four years of age diagnosed with total deafness (cochlear implant candidates). However, for individuals aged between four and 64 years who have HL but are not totally deaf, there is not a clear rehabilitation pathway or dedicated public reimbursement.

Age and degree of HL might be the most prominent determinants of access to hearing health care in Chile. Diagnosis and treatment of people aged four to 64 years with moderate to severe HL might incur out-of-pocket expenses or even have no access to care for HL. These findings highlight the need for public health policies to promote equitable access to hearing health care in Chile.

Rare diseases (RDs) present unique challenges in health technology assessment (HTA) due to small, heterogeneous populations and limited clinical data, which complicate economic modeling and delay access to innovative interventions. This survey, conducted by the HTAi Rare Diseases Interest Group, explored key challenges in economic modeling for RDs, the application of solutions, and the use of evidence assessment frameworks in HTA practice.

A mixed-method survey was developed with multiple-choice, Likert-scale, and open-text questions using the SurveyMonkey platform. Responses were collected from members of HTAi and the International Network of Agencies for Health Technology Assessment at conference sessions. Additional respondents were recruited via professional networks and direct emails. Options exceeding 50 percent of responses and rate weighted averages guided the interpretation and identification of areas of agreement.

In total, 36 individuals across all continents, mostly from consultancy and HTA agencies, responded to the survey. Key challenges in economic modeling included insufficient quality of clinical trials (duration and sample size), scarcity of robust data, and lack of health-related quality of life metrics. The most common solutions included using proxy diseases, pooled data, and qualitative research. Broader impacts were integrated through societal perspectives, decision modifiers, and deliberative processes. GRADE was the most commonly cited explicit evidence assessment framework. In some cases, implicit approaches dominated HTA practice where challenges such as single-arm trials and reliance on real-world evidence prevailed, though patient-reported outcomes had gained some acceptance.

Scarcity and quality of clinical data remain significant barriers to economic modeling for RDs. Solutions such as qualitative methods and pooling data show promise. The broader impacts of RDs are increasingly being considered, but further research is essential to refine methods, embrace non-traditional data, and develop appropriate frameworks to assess evidence for RDs in HTA.

Chronic lymphocytic leukemia (CLL) is a malignant hematologic disorder that affects older adults. CLL is the most common leukemia in adults in the Western world. The genomic landscape of CLL is heterogeneous. The aim of this report was to identify the clinically relevant molecular alterations in CLL and define the role of targeted next-generation sequencing (NGS) in routine care.

The evaluation included an analysis of systematic reviews, meta-analyses, clinical guidelines, and relevant molecular alterations using OncoKB™ and TOPOGRAPH classifications. Approvals from the French National Authority for Health Transparency Committee or compassionate use decisions from the French National Agency for Medicines and Health Products Safety were also considered.

Targeted NGS in CLL detected molecular alterations to determine prognosis and guide treatment decisions with a higher sensitivity than Sanger sequencing. The recommended analyses include:

• • TP53 and IGHV mutations before first-line treatment;

• • TP53 mutation before treatment modification in relapse cases;

• • TP53, BTK, and PLCG2 mutations after treatment with BTK inhibitors;

• • TP53 and BCL2 mutations after treatment with BCL2 inhibitors; or

• • TP53, BTK, PLCG2, and BCL2 mutations after treatment with both BTK and BCL2 inhibitors.

Targeted NGS is essential for managing CLL. The gene panel will be updated dynamically based on new scientific evidence and regulatory approvals.

Internationally, there is a growing drive to integrate patient preference information (PPI) into health technology assessment (HTA) decision-making. However, it is unclear how PPI is used in real-world contexts and the extent to which it can reshape and influence HTA decisions. Understanding how PPI can be meaningfully used in HTA is fundamental for supporting its practical application and increasing trust in its use.

The aim was to enhance understanding of how HTA committees interpret and incorporate PPI into decision-making. Mock deliberation workshops were conducted by members of the HTAi Patient Preference Project Subcommittee with experts from four HTA agencies. For the workshops, a mock evidence package was developed that included PPI evidence alongside key clinical and economic evidence to simulate decision-making processes. The mock deliberations were analyzed using thematic analysis.

The mock deliberation workshops (25 participants) highlighted both opportunities and challenges in integrating PPI into HTA processes. While PPI was acknowledged as valuable evidence, its influence on decision-making in the scenarios varied from being cited by a few participants as informing the decision to most participants indicating the evidence was not persuasive. Participants pointed out the need for more structured guidance on how to incorporate PPI alongside clinical and economic evidence. The findings highlighted the importance of standardizing the path to integration, as well as the content and format of PPI, to support consistent interpretation and application in HTA.

By examining how PPI is weighed alongside clinical and economic evidence, the project offered valuable insights for stakeholders aiming to generate more pertinent PPI studies for incorporation into the decision-making process. The key takeaways for HTA agencies included strategies for effectively balancing PPI with clinical and economic evidence, highlighting a shift toward incorporating more diverse and inclusive forms of evidence.

Oncology economic modeling often struggles with estimating clinical benefits like overall survival (OS) and progression-free survival (PFS), especially with short-duration trials involving extrapolating survival curves. This study analyzed methods used by the Brazilian National Committee for Health Technology Incorporation (CONITEC) in health technology assessment (HTA) to adapt economic models to evolving healthcare needs.

A descriptive analysis was conducted on the economic evaluations from CONITEC’s recommendation reports on oncology drugs published between January 2019 and December 2023. Three independent reviewers extracted data on outcomes, survival curve distributions, validation methods, and time horizons. The study focused on identifying patterns in modeling practices and explored whether these methods integrated with the expanding scope of HTA, including the use of non-traditional data sources and adaptive techniques.

Thirty CONITEC reports were analyzed, evaluating 32 oncology treatments across 45 indications. The most common economic models were partitioned survival analysis (PartSA) (58%) and Markov models (33%). The Weibull distribution was the most frequent in OS extrapolations (50% for PartSA; 84% for Markov). PFS extrapolations showed a preference for exponential distribution in PartSA (35%) and Weibull in Markov models (54%). Methods for model validation were reported in 69 percent of Markov models and 100 percent of PartSA models. Extrapolation horizons ranged from 10 years (75% of models) to 30 years (31%).

The analysis highlighted the reliance on traditional models, such as PartSA and Markov, for modeling long-term cancer progression. However, these models may not fully capture treatment complexity. Emerging methods and the integration of diverse data sources, such as real-world data, are vital for improving HTA relevance and addressing the evolving demands of healthcare systems and next-generation evidence.

Hereditary breast and ovarian cancer, linked to BRCA gene mutations, poses significant health risks. Australian breast and ovarian cancer mortality rates emphasize the public health importance of effective screening. This study evaluated the cost effectiveness of population-wide BRCA screening and its equity implications, focusing on reducing health disparities across socioeconomic groups through early detection and prevention strategies.

A cost-effectiveness analysis was conducted comparing population-wide BRCA screening to family-history-based testing for Australian women aged 30 to 50 years. A decision-analytic model included a decision tree and two Markov models simulating outcomes over a lifetime horizon. Transition probabilities captured risk-reducing interventions. Outcomes included quality-adjusted life years (QALYs) and costs in 2024 AUD, discounted at five percent. Socioeconomic quintiles were defined using Socio-Economic Indexes for Areas, and distributional impacts were assessed with net health benefits and equally distributed equivalent health (EDEH) using the Atkinson index. Model inputs were derived from census data and empirical studies. Opportunity costs used UK data as a proxy.

Screening at the age of 30 years was cost saving (incremental cost-effectiveness ratio [ICER] −AUD8,331 [−USD5,482] per QALY), while screening at ages 35 and 40 years was cost effective (ICERs: AUD28,231 [USD18,577] and AUD40,086 [USD26,384] per QALY, respectively). Screening at ages 45 and 50 years exceeded the threshold (ICERs: AUD78,666 [USD51,778] and AUD155,937 [USD102,640] per QALY). Distributional analysis showed screening at the age of 30 years increased average quality-adjusted life expectancy and EDEH, reducing health inequality between socioeconomic groups. In contrast, screening at older ages resulted in diminished benefits and exacerbated inequalities, highlighting the importance of early intervention in achieving both cost-effectiveness and equity goals.

Population-wide BRCA screening was cost effective for Australian women at younger ages, particularly at the age of 30 years, and reduced health inequalities. Policymakers should prioritize early screening to address hereditary breast and ovarian cancer, promote health equity, and alleviate disease burden. Further strategies to enhance access among disadvantaged groups could amplify these benefits, advancing a more equitable healthcare system.

Health technology assessment (HTA) often involves complex cost-effectiveness analyses (CEA) that can be challenging for non-experts. This study demonstrated the use of artificial-intelligence (AI)-powered prompts to simplify CEA processes. By integrating tools for generating efficiency frontiers and net-benefit analyses, this approach enables stakeholders, such as non-modeling HTA specialists or decision-makers, to understand therapeutic scenarios and make informed adjustments to analyses.

The prompts were created using ChatGPT-4o and tested for usability and reproducibility in ChatGPT-4o and the free version 3.5. The prompts were designed to automate key steps in economic analysis, including calculating net monetary and health benefits, performing cost-effectiveness analyses applying dominance and extended dominance concepts, and generating the efficient frontier plot. Ten HTA experts with no modeling experience evaluated the prompts using predefined scenarios with hypothetical datasets. The results of both versions of ChatGPT were compared to the expected results. The usability and accuracy of the prompts were assessed during the evaluation.

ChatGPT-4o achieved 100 percent accuracy in calculating net health and monetary benefits (NHB and NMB) and correctly applied the concepts of dominance and extended dominance in 83 percent of cases. It calculated the incremental cost-effectiveness ratio (ICER) for non-dominated therapies in 50 percent of situations and successfully generated the efficiency frontier graphic in half of cases. In contrast, ChatGPT 3.5 achieved 50 percent accuracy for NHB and NMB calculations, only 17 percent for applying dominance and extended dominance concepts, and failed to calculate ICERs or generate the efficiency frontier plot as expected.

AI-powered prompts simplified cost-effectiveness analyses by enabling non-technical stakeholders to create and adjust efficiency frontiers and benefit analyses. ChatGPT-4o demonstrated improved reliability. Limitations in ChatGPT 3.5, particularly with ICER calculation and graphical outputs, indicated the need to adapt prompts to specific AI tools. Future developments may increase the robustness and usability of these types of prompts across platforms.

Generative artificial intelligence (AI) holds promise in aiding development of health economic models. Our objective was to explore the feasibility of using generative AI to replicate health economic models based on previously published models.

We replicated a Markov model of ulcerative colitis described in the literature using a two-step approach. First, we used Python for large language model interactions and utilized ValueGen.AI, a GPT-4-based platform with multi-agent pipelines (CrewAI, LangChain, and OpenAI libraries), to extract model structure and parameters from the source. These parameters were implemented in R’s heemod package to construct and run the Markov model. Next, we repeated the experiment using a more detailed technical report of the same model. We evaluated generative AI’s performance by comparing its conceptualization and parameterization with the original sources.

Using the publication, the generative AI platform effectively extracted costs and quality-of-life inputs linked to health states. However, for health states and transition probabilities, initial attempts were less successful due to limited descriptions in the text, resulting in misinterpretations and conflicting health states. Performance improved considerably when we used the detailed technical report, which offered clearer and more structured information. Generative AI could not successfully extract all transition probability calculations for the defined model cycle length based on the formulas provided in the documents.

Generative AI holds significant potential for replicating previously published health economic models. However, challenges remain in capturing detailed model parameters, particularly when description of modeling approach lacks clarity and transparency. Improving standardized reporting practices within the health economics and outcomes research field is needed to enable generative AI to better support stakeholders during HTA processes.