Significance and background. Electronic health records (EHRs) are becoming increasingly common for storing massive amounts of longitudinal patient data. Efficient standardization, processing, interpretation, and visualization of EHRs is critical for generating new medical knowledge and facilitating predictive, personalized, and participatory healthcare. However, the heterogeneous nature of EHRs makes it a challenge to capture disease activity and progression in a reliable and valid way.

Clinical problem: Otitis media, inflammation of the the middle ear, is second only to acute respiratory infections as the most frequently diagnosed illness in children¹. Otoscopy is the key to diagnosing these pathologies. Most common ear pathologies seen include acute otitis media (AOM), an infection of the middle ear, and serious otitis media (SOM), fluid in the middle ear space. In both of these cases, diagnosis requires that the practitioner visualize the tympanic membrane to ascertain the status of the middle ear space.

Clinical problem: Foreign body aspiration (FBA) is a potentially life-threatening condition seen primarily in children less than three years of age.1 Prompt diagnosis is critical in reducing morbidity, as delays can lead to sequelae such as pneumonia, respiratory tract injury, obstructive emphysema, and respiratory failure.2 Unfortunately, identification of a FBA remains a challenge due to limited history taking in this age group, unreliable physical exams, and supporting radiology

The Clinical Problem

Sepsis is the single most expensive disease process in U.S. hospitalizations. It is also one of the deadliest.  Sepsis that develops after a patient has been admitted to the hospital is called hospital acquired sepsis (HAS). While it only accounts for 11-15% of all cases of sepsis, it is associated with significantly higher mortality, costs, and length of stay compared to community acquired sepsis.

In the U.S., approximately 47% of the population has hypertension and hypertension is a leading preventable cause of death contributing to over a half million deaths a year (1).  Effective medications are readily available to treat hypertension and guidelines for managing hypertension have been published (2, 3).  Nonetheless, finding an effective treatment regimen for an individual often requires multiple rounds of medication optimization including varying both drugs and dosages (4,5).

Removed given end of review

Background:
Maternal morbidity and mortality are increasing disproportionately in the United States as compared to developing countries. Recent studies have found that most cases of maternal mortality occur in the postpartum period with cardiovascular and infectious complications identified as primary causes. As such, professional organizations, such as the American College of Obstetricians and Gynecologists (ACOG) and the Society for Maternal-Fetal Medicine, have renewed focus on the postpartum period as time where interventions can improve outcomes.

The clinical problem: At UCSF, Hospitalists often work in 5-10 day blocks, providing complex inpatient care to as many as 15 patients at a time. These patients are all acutely ill, and most have a long list of comorbidities, medications, and abnormal lab values that must be remembered and accounted for. Caring for these patients requires knowing these details intimately, as well as many others. However, because Hospitalists work a predefined number of days, they often provide care to patients for only a portion of their admission.

Each year, the United States spends more than $20 billion for patients hospitalized with cirrhosis. Despite year-on-year increases in spending, in-hospital mortality for inpatients with cirrhosis remains at 7-10%.  Practice variations between hospitals account for 85% of differences in in-hospital mortality.  Herein lies the opportunity to leverage novel clinical informatics methodologies to improve quality of care for inpatients with cirrhosis.  In this proposal, our objective is to construct and develop an AI-driven clinical decision support (CDS) application, called “CirrhosisTx,” built on the SMART-on-FHIR application programming interface to improve the quality of inpatient cirrhosis care by “nudging” providers to align practices with AGA and AASLD practice guidelines. 

Background and Significance: Pain is ubiquitous, with over 50% of hospitalized patients reporting significant pain (1).