Harnessing EHR data for health research.

With the increasing availability of rich, longitudinal, real-world clinical data recorded in electronic health records (EHRs) for millions of patients, there is a growing interest in leveraging these records to improve the understanding of human health and disease and translate these insights into clinical applications. However, there is also a need to consider the limitations of these data due to various biases and to understand the impact of missing information. Recognizing and addressing these limitations can inform the design and interpretation of EHR-based informatics studies that avoid confusing or incorrect conclusions, particularly when applied to population or precision medicine. Here we discuss key considerations in the design, implementation and interpretation of EHR-based informatics studies, drawing from examples in the literature across hypothesis generation, hypothesis testing and machine learning applications. We outline the growing opportunities for EHR-based informatics studies, including association studies and predictive modeling, enabled by evolving AI capabilities-while addressing limitations and potential pitfalls to avoid.

Associations of Everyday and Lifetime Experiences of Discrimination With Willingness to Undergo Alzheimer Disease Predictive Testing.

BACKGROUND AND OBJECTIVES: Rapid developments in Alzheimer disease (AD) biomarker research suggest that predictive testing may become widely available. To ensure equal access to AD predictive testing, it is important to understand factors that affect testing interest. Discrimination may influence attitudes toward AD testing, particularly among racially and ethnically minoritized populations, because of structural racism in health care systems. This study examined whether everyday or lifetime discrimination experiences shape interest in AD predictive testing. METHODS: In the 2010 and 2012 biennial Health and Retirement Study waves, respondents were randomly selected to complete questions on interest in receiving free testing that could determine whether they would develop AD in the future. The exposures were everyday discrimination (6 items) and lifetime discrimination (7 items); both were transformed into a binary variable. Logistic regression models predicting interest in AD testing were controlled for deciles of propensity scores for each discrimination measure. Odds ratios were re-expressed as risk differences (RDs). RESULTS: Our analytic sample included 1,499 respondents. The mean age was 67 (SD = 10.2) years, 57.4% were women, 65.7% were White, and 80% endorsed interest in AD predictive testing. Most of the participants (54.7%) experienced everyday discrimination in at least one domain; 24.1% experienced major lifetime discrimination in at least one domain. Those interested in predictive testing were younger (66 vs 70 years) and more likely to be Black (20% vs 15%) or Latinx (14% vs 8%) than participants uninterested in testing. The probability of wanting an AD test was not associated with discrimination for Black (RD everyday discrimination = -0.026; 95% CI [-0.081 to 0.029]; RD lifetime discrimination = -0.012; 95% CI [-0.085 to 0.063]) or Latinx (RD everyday discrimination = -0.023, 95% CI [-0.082 to 0.039]; RD lifetime discrimination = -0.011; 95% CI [-0.087 to 0.064]) participants. DISCUSSION: Despite historical and contemporary experiences of discrimination, Black and Latinx individuals express interest in AD testing. However, Black and Latinx individuals remain underrepresented in AD research, including research on AD testing. Interest in personalized information about dementia risk may be a pathway to enhance their inclusion in research and clinical trials.

Leveraging electronic health records and knowledge networks for Alzheimer's disease prediction and sex-specific biological insights.

Identification of Alzheimer's disease (AD) onset risk can facilitate interventions before irreversible disease progression. We demonstrate that electronic health records from the University of California, San Francisco, followed by knowledge networks (for example, SPOKE) allow for (1) prediction of AD onset and (2) prioritization of biological hypotheses, and (3) contextualization of sex dimorphism. We trained random forest models and predicted AD onset on a cohort of 749 individuals with AD and 250,545 controls with a mean area under the receiver operating characteristic of 0.72 (7 years prior) to 0.81 (1 day prior). We further harnessed matched cohort models to identify conditions with predictive power before AD onset. Knowledge networks highlight shared genes between multiple top predictors and AD (for example, APOE, ACTB, IL6 and INS). Genetic colocalization analysis supports AD association with hyperlipidemia at the APOE locus, as well as a stronger female AD association with osteoporosis at a locus near MS4A6A. We therefore show how clinical data can be utilized for early AD prediction and identification of personalized biological hypotheses.