Artificial intelligence: opportunities and risks for public health

Artificial Intelligence has been applied in academic research and in inference tasks across the broader economy with demonstrable success, 1Panch T Szolovits P Atun R Artificial intelligence, machine learning and health systems. J Glob Health. 2018; 8: 020303Crossref PubMed Scopus (178) Google Scholar but less so for the core functions of public health, namely protecting and promoting the health of populations. 2Acheson D Committee of Inquiry into the Future Development of the Public Health FunctionPublic health in England: the report of the Committee of Inquiry into the Future Development of the Public Health Function. Her Majesty's Stationery Office, London1988Google Scholar To date, vision statements on the future of public health3Khoury MJ Iademarco MF Riley WT Precision public health for the era of precision medicine. Am J Prev Med. 2016; 50: 398-401Summary Full Text Full Text PDF PubMed Scopus (292) Google Scholar have focused on the technical possibilities of artificial intelligence and less on how social determinants might influence outcomes achieved by it. Artificial intelligence has the potential to improve the efficiency and effectiveness of processes across an expanded public health continuum4Burton H Personalised prevention and public health: an urgent agenda. PHG Foundation, CambridgeMarch 6, 2015www. phgfoundation. org/blog/personalised-prevention-and-public-health-an-urgent-agendaDate accessed: December 14, 2018Google Scholar (table) to make possible personalised predict and prevent approaches, 5Newton J Ekpe M Bradley P Predictive prevention and the drive for precision public health. Public Health England, LondonNov 20, 2018https: //publichealthmatters. blog. gov. uk/2018/11/20/predictive-prevention-and-the-drive-for-precision-public-health/Date accessed: December 14, 2018Google Scholar, 6Department of Health and Social CarePrevention is better than cure. https: //assets. publishing. service. gov. uk/government/uploads/system/uploads/attachmentdata/file/753688/Preventionᵢsbetterₜhancure₅-11. pdfDate: Nov 5, 2018Date accessed: December 14, 2018Google Scholar applied differentially across populations to match preventive services to individual need. This approach is potentially a radical expansion in the scope of public health and many of these activities will be led by organisations beyond the established public health institutions. TablePublic health domains and potential uses of artificial intelligence within themPotential use of artificial intelligenceExampleHealth protectionAnalysing patterns of data for almost real-time surveillance and disease detectionUsing Google search and phone GPS information to predict restaurants that are causing foodborne illness7Sadilek A Caty S DiPrete L et al. Machine-learned epidemiology: real-time detection of foodborne illness at scale. npj Digit Med. 2018; (published online Dec 6; article number 36. ) http: //www. nature. com/articles/s41746-018-0045-1Date accessed: December 14, 2018Crossref Scopus (42) Google ScholarHealth promotionOffering targeted and personalised health advice based on personal risk profile and behavioural patternsUsing machine learning to generate improved cardiovascular disease risk models8Weng SF Reps J Kai J Garibaldi JM Qureshi N Can machine-learning improve cardiovascular risk prediction using routine clinical data? . PLoS One. 2017; 12: e0174944Crossref PubMed Scopus (572) Google ScholarIncreasing efficiency of health services (1) Using machine learning to detect abnormalities in screening tests such as mammography or cervical cytology; (2) machine learning-facilitated automated evidence synthesis (1) Deep learning algorithms for detecting diabetic retinopathy;9De Fauw J Ledsam JR Romera-Paredes B et al. Clinically applicable deep learning for diagnosis and referral in retinal disease. Nat Med. 2018; 24: 1342-1350Crossref PubMed Scopus (1116) Google Scholar, 10Gulshan V Peng L Coram M et al. Development and validation of a deep learning algorithm for detection of diabetic retinopathy in retinal fundus photographs. JAMA. 2016; 316: 2402-2410Crossref PubMed Scopus (3460) Google Scholar (2) the Human Behaviour-Change Project uses machine learning for evidence synthesis and interpretation around behaviour change11Michie S Thomas J Johnston M et al. The Human Behaviour-Change Project: harnessing the power of artificial intelligence and machine learning for evidence synthesis and interpretation. Implement Sci. 2017; 12: 121Crossref PubMed Scopus (111) Google Scholar Open table in a new tab Equity must be central to the implementation of artificial intelligence across health systems. Large datasets are pivotal to the development of these technologies, but must be representative of the population to ensure all can benefit. Typically, minority groups are less represented in datasets used to develop artificial intelligence algorithms and the health challenges for these communities are less obvious to data science teams, which tend not to be representative of these populations. The rhetoric around artificial intelligence involves greater emphasis on personalised recommendations and individual action, however, this should not undermine the importance of continued collective action to address social and structural determinants of health. Achievement of primary prevention benefits depends more on social factors than secondary prevention irrespective of the marginal benefits of artificial intelligence. It is plausible that the effect of artificial intelligence on public health could be principally indirect. Broad automation of manual jobs through artificial intelligence might cause near-term unemployment in low-income communities, with adverse health effects. However, automation will probably augment efficiency of logistics and human resources, with profound benefits through improved productivity and performance of health systems. However, the net effect of these trends is difficult to predict, especially in the context of political and economic uncertainty. Even if the primary effects of artificial intelligence are beyond the operations of public health organisations, as the present trajectory suggests, there will be immediate and enduring consequences for public health. Artificial intelligence has the potential to widen social disparities and divert attention from collective action, ignoring the lessons of decades of public health innovation. The public health community must be actively involved in not just creating the circumstances for the safe and effective development of artificial intelligence that delivers for whole populations but also in the development of artificial intelligence technologies to ensure the narrative around personalisation and the responsibilities of the individual does not distract governments from their continued responsibility for the health of their citizens. The reality of public health organisations is that they face mounting cost constraints and challenges in recruiting the talent and resources necessary for the development of artificial intelligence. Limited resources should not be used to duplicate massive investments that have already been made by the private sector. Hence, collaborations should use these investments while ensuring mutual benefit by aligning the profit motive of private organisations with social responsibility and the advancement of public health. Public health organisations will need to show leadership regarding private sector involvement in the application of artificial intelligence in public health and in the creation of specific contracting instruments that ensure any such partnerships deliver material returns and protections for the health institutions and populations that they serve. This online publication has been corrected. The corrected version first appeared at thelancet. com/digital-health on June 27, 2019 This online publication has been corrected. The corrected version first appeared at thelancet. com/digital-health on June 27, 2019 We declare no competing interests. Correction to Lancet Digital Health 2019; 1: e13–14Panch T, Pearson-Stuttard J, Greaves F, Atun R. Artificial intelligence: opportunities and risks for public health. Lancet Digital Health 2019; 1: e13–14—The second author's name has been corrected to Jonathan Pearson-Stuttard. This correction has been made as of June 27, 2019. Full-Text PDF Open Access