Maximizing data use to propel informatics practice and research

In this issue of JAMIA we present several articles focused on PersonGenerated Health and Wellness Data for Healthcare, a topic of increasing relevance for informatics, given several advances in personal monitoring devices, electronic diaries, social networks, and online surveys. The context for these special-focus articles is presented in a guest editorial (see page 438). Related to this topic, some authors show how web-based interviews through patient portals can improve health care proxy documentation (see page 580). This issue also contains several articles on predictive modeling and clinical decision support. Predictive models require a large amount of data, and it is exciting to see how some authors are combining public data and the literature to generate new knowledge (see page 617) and how some government agencies such as the U.S. Food and Drug Administration are making drug and device data available for research (see page 596). Other authors are using data from the web or tweets to study the association of parity and cancer risk (see page 582), predict new cases of back pain (see page 644), or develop large cohorts of patients with a particular health condition (see page 635). Computing with distributed data is often required to comply with institutional requirements, but is also very challenging. A novel algorithm for vertical grid logistic regression is presented (see page 570). Electronic health records (EHRs) are increasingly being utilized in practice, for example in developing countries (see page 544) and in the United States (see page 562), and are changing the clinical encounter (see page 654). Even though EHR data are known to have high variability in terms of quality, particularly for documentation of race and ethnicity (see page 627), they can be mined to recognize patterns used in predicting mortality, readmission, and length of stay (see page 553) and to assess appropriate use of diagnostic imaging studies (see page 649). EHR systems can embed models that assist in prescribing medications for patients with kidney disease (see page 579). However, not all clinical decision support is well accepted: overrides related to drug allergy alerts are increasing over time (see page 601). As the articles in this issue of the journal illustrate, biomedical informatics addresses a wide range of problems using many different approaches. Biomedical informatics is at the intersection of health science and data science, and JAMIA is proud to publish a wide range of highly innovative scholarly work representing the full spectrum of our field.