AI methods and their application to health and prevention using open data.

Abstract

Artificial Intelligence (AI) and machine learning have revolutionized society, impacting everything from search engines and media consumption to loan approval processes. These algorithms are powered by data and can solve problems that have been impossible for humans to solve. In the health sector, the use of AI has been growing exponentially, and investments in data collection and open data are crucial for its progress. The thesis discussed in the article uses ML models on open data sets to address various health-related questions. In chapter 1, the focus is on sudden infant death syndrome (SIDS) and sudden unexpected infant death (SUID) and the effect of maternal smoking during pregnancy. By using a glass-box model known as generalized additive models (GAMs), the authors found that SUID risk doubles with any maternal smoking during pregnancy, and each cigarette increases the odds by 0.07, with 22% of SUID deaths in the United States attributed to maternal smoking. In chapters 2 and 3, the authors differentiate between two distinct subpopulations of SUID deaths, depending on the age of the infant at the time of death. In chapter 3, the authors focus on sudden unexpected postnatal collapse (SUPC), a subcategory of SUID, and found that SUPC deaths differed statistically from SUID deaths occurring from 7-364 days of age, highlighting the need for adequate nurse staffing during the immediate recovery period and postpartum stay. Chapters 4 and 5 focus on COVID-19 and the use of open data to infer causality from observational data using synthetic controls and Bayesian structural time series. Chapter 4 analyzes the relationship between vaccination and reducing COVID-19 deaths in Washington state, finding a 27.1% decrease in deaths compared to the synthetic control. In chapter 5, the authors analyze the association between COVID-19 public health measures and the epidemiology of infectious conjunctivitis, finding a 34% decrease in search activity and a 37% decrease in emergency department encounters for infectious conjunctivitis after the adoption of COVID-19-associated public health measures. In conclusion, the use of AI and machine learning in health has immense potential, but investments in data collection and open data are essential. The thesis discussed in the article shows how machine learning can be used to address critical health-related questions and underscores the need for continued investment in this field. AI in health is still in its early stages, and we can expect further advancements in the future.