Table 1 Examples of LLM applications and how they can potentially support or deter the United Nations’ SDG 3-specific goals
SDG 3: ensure healthy lives and promote well-being for all at all ages | |||
|---|---|---|---|
Specific targets | Supporting examples as enablers | Supporting examples as deterrents | |
3.1. By 2030, reduce the global maternal mortality ratio to less than 70 per 100,000 live births | LLM-based agents developed for maternal health intervention decision support107; conversational agents for women in Ethiopia that support patient education from prenatal to postnatal care108; AI–LLMs to identify abnormal cardiotocography interpretations during labour109 | No relevant studies found. | |
3.2. By 2030, end preventable deaths of newborns and children under 5 years of age, with all countries aiming to reduce neonatal mortality to at least as low as 12 per 1,000 live births and under-5 mortality to at least as low as 25 per 1,000 live births | LLMs to identify factors associated with abusive head trauma in young children presenting to the emergency department110; LLMs to support neonatal care in hospital and community settings111,112; LLM-integrated systems to diagnose potential rare diseases113 | LLMs have been demonstrated to perform poorly on the Neonatal Board Examination, suggesting low reliability in this specialty at the time of testing114. | |
3.3. By 2030, end the epidemics of AIDS, tuberculosis, malaria and neglected tropical diseases and combat hepatitis, waterborne diseases and other communicable diseases | LLMs in answering questions related to hepatitis B or as a medical assistant during medical consultations115,116; vision language models to characterize tuberculosis pathogenesis117; language model-based telehealth services to mitigate HIV spread118 | LLMs have been noted to lack the sociopolitical context and inclusivity essential for effective health communication119, potentially propagating biased and harmful content particularly in marginalized communities. | |
3.4. By 2030, reduce by one third premature mortality from noncommunicable diseases through prevention and treatment and promote mental health and well-being | LLM-based conversational chatbots for mental health coaching120, digital interventions for anxiety and depression121, and suicide risk prediction from clinical notes122; LLM-based decision support for personalized intervention for patients with diabetes123 | In the absence of robust and standardized evaluation criteria and a regulatory and monitoring framework, LLM-based models for mental health coaching are not ready for safe, standalone use120. | |
3.5. Strengthen the prevention and treatment of substance abuse, including narcotic drug abuse and harmful use of alcohol | LLMs to predict persistent postoperative opioid use and opioid use disorder from electronic health records124; LLMs in the management of alcohol use disorder125 | Findings suggest a risky mix of seemingly high-quality, accurate responses on initial inspection that contain inaccurate and potentially fatal medical advice126. | |
3.6. By 2030, halve the number of global deaths and injuries from road traffic accidents | Findings provide insufficient evidence to conclude whether LLMs positively or negatively affect the SDG. A relevant example is: • Integrating a visual language model and reasoning chain for driver behaviour analysis and risk assessment127. | ||
3.7. By 2030, ensure universal access to sexual and reproductive healthcare services, including for family planning, information and education, and the integration of reproductive health into national strategies and programmes | Generative AI to enhance health literacy, including reproductive health128,129 | No relevant studies found. | |
3.8. Achieve universal health coverage, including financial risk protection, access to quality essential healthcare services, and access to safe, effective, quality and affordable essential medicines and vaccines for all | Findings provide insufficient evidence to conclude whether LLMs positively or negatively affect the SDG. Relevant examples include • Foundation models or molecular language models in drug discovery, drug repurposing130, or vaccine design and development131. AI tools may potentially reduce the time required for new drug or vaccine development, thereby lowering the cost of development. • Genetic language models have demonstrated impressive zero-shot performance, potentially advancing our understanding of human diseases and precision care across all domains of medicine132. Advancements in AI and smarter clinical trials are projected to considerably reduce drug development costs, as well as the development costs of products for emerging infectious diseases and maternal health, lowering the average cost per launch by 26% to 39% (https://centerforpolicyimpact.org/our-work/research-and-development). However, there is a need for more representative models, especially in terms of data sources, if current LLMs are to be effectively applied to drug discovery on a global scale. | ||
3.9. By 2030, substantially reduce the number of deaths and illnesses from hazardous chemicals and air, water and soil pollution and contamination | No relevant studies found. | No relevant studies on LLM applications found. However, high computational needs may potentially lead to environmental pollution. Adverse impacts may extend beyond this goal. | |
3.a. Strengthen the implementation of the WHO Framework Convention on Tobacco Control in all countries, as appropriate | No relevant studies found. | ||
3.b. Support the research and development of vaccines and medicines for the communicable and noncommunicable diseases that primarily affect developing countries, provide access to affordable essential medicines and vaccines in accordance with the Doha Declaration on the TRIPS Agreement and Public Health, which affirms the right of developing countries to use to the full the provisions in the Agreement on Trade-Related Aspects of Intellectual Property Rights regarding flexibilities to protect public health and, in particular, provide access to medicines for all | Similar to 3.8, the findings provide insufficient evidence to conclude whether LLMs positively or negatively affect the SDG. Advancements in AI and smarter clinical trials are projected to reduce drug development costs, as well as the development costs of products for emerging infectious diseases and maternal health. However, we are unable to conclude that LLM will improve the affordability of medications and vaccines. | ||
3.c. Substantially increase health financing and the recruitment, development, training and retention of the health workforce in developing countries, especially in least developed countries and small island developing states | LLMs in enhancing medical education133, supporting regions with a shortage of healthcare professionals such as medical specialists or primary care physicians24 | Over-reliance on LLMs could stifle critical thinking and decision-making134, adversely affecting the education and training outcomes of healthcare professionals. | |
3.d. Strengthen the capacity of all countries, in particular developing countries, for early warning, risk reduction and management of national and global health risks | LLMs incorporated into disease outbreak surveillance pipelines135; LLMs in infectious disease transmission modelling; LLMs to disseminate critical health information during a pandemic or crisis136 | Risk of misinformation may lead to worsening health outcomes136. | |
