JAMIA has been a premier venue for publication of scholarly work on clinical decision support systems since its inception. With an initial emphasis on knowledge-based systems, i.e., encoding of clinicians’ knowledge into rules that would trigger alerts and reminders, JAMIA evolved into a phase of an increasing number of articles reporting on the use of EHR “big data” to build and validate predictive models that recognize patterns in large amounts of data to derive actionable recommendations for clinicians. Interestingly, this is happening at the same time that health information technology (HIT) is still evolving, data quality in EHRs is being improved, and health information exchange (HIE) continues to be evaluated for cost effectiveness.
Key recommendations for HIT optimization are proposed by Chresswell (p. 186). Additionally, Wright (p. 192) advocates for more testing of EHR systems, Alexander (p. 69) focuses on IT in nursing homes, and Kharrazi (p. 2) promotes an agenda for population health informatics. HIT use by office-based physicians and healthcare reform programs is described in a study by Heisey-Grove (p. 133). Improving the quality of EHRs is addressed by Van der Bij (p. 84) and Jamieson (p. 126). The latter describes a randomized trial on the quality of admission notes from EHRs. A systematic review of EHR usability is reported by Ellsworth (p. 222), Yadav (p. 143) compares EHRs with paper records in the documentation of physical exams, Denny (p. 165) describes hypertension “phenotyping” from EHRs, and Das (p. 24) proposes how to generate discharge recommendations.
HIT is now pervasive in healthcare and this issue of JAMIA features several articles on the use of HIT for predictive modeling: Goldstein reviews risk prediction (p. 202) and describes the challenges in predicting mortality over time horizons (p. 180), Gillame-Bert (p. 48) learns temporal rules that predict instability in patients undergoing continuous monitoring, and Lennon (p. 148) shows predictive value in particular combinations of pathology markers for pancreatic cysts. Manaktala (p. 91) proposes a clinical decision support system for sepsis mortality, and Haslam (p. 13) discovers disease relationships from clinical trial data.
HIE systems help clinicians share information from a particular patient with each other. However, the cost effectiveness of various types of HIEs remains hard to measure. Downing (p. 116) reports on the policies of 11 health systems, Dixon (p. 99) describes the characteristics of veterans who enroll in HIE, Vest (p. 39) describes the organizational capacity and utility of clinical event notification, and Slovis (p. 30) studies the rate of duplicative CT exams.
Pharmacy informatics is an important sub-field of specialization. Nelson (p. 197) describes the interaction between pharmacists and the EHR, and White (p. 175) reports on outcomes of a computer-based system for Vitamin D prescriptions. Medication reconciliation is systematically reviewed by Marien (p. 231). Also related to pharmacy are reports from Eschmann (p. 62), who describes a system to predict hyperkalemia from drug interactions, Heringa (p. 55), who shows how clustering related drug interaction alerts helps reduce the number of alerts, and Manzi (p. 77), who describes a clinical pharmacogenomics service.
Other types of computer-based interaction with providers and patients are reported in this issue of the journal: McGrath (p. 213) systematically reviews computer-aided instruction in oral health, Ratanawongsa (p. 109) studies computer use and literacy in safety net outpatient communications, and Kelly (p. 156) reports how families stay engaged in pediatric care through a patient portal.
HIT evolved rapidly in the past decade and is likely to continue to evolve at this pace until its many challenges are overcome. Our responsibility as professionals is to help drive HIT, clinical decision support, and predictive analytics to the next level, and to educate the next generation so that they can fill the many knowledge gaps that still exist.