In this study, I aim to create a better understanding of the information behavior of users engaging with historical artifacts in both digital and non-digital forms. By information behavior, I refer to how people interact with, seek and utilize information. Studying people’s interactions with historical artifacts is important to demonstrate varied ways of using the information in different contexts and areas of knowledge. Considering the versatility of historical corpora, they can be used in multiple areas of study. The research is performed by conducting interviews with past users of content from the Svoboda Diaries Project, which aims to digitize personal diaries from 19th century Ottoman Iraq. We employ an iterative interview process using the case study method to study examples of how scholars, professors, and students from a variety of disciplines engage with historical content. These users of information hold different perspectives and have taken separate paths towards using historical documents that are part of the collection. This research is based on several phases: (1) pilot interviews, (2) interviews, (3) data analysis, and (4) formulating conclusions about the potential uses of the historical corpus. The broader goal is to examine the diverse uses of historical artifacts in different academic fields. These references could inform the future use of this and other collections. In the conference poster, I will describe the first phase of the study and explain how the current and past interns of the Svoboda Diaries Project have interacted with the historical information from the diaries to develop projects in diverse disciplines of academia.

Randomized controlled trials form the basis of translating research data into clinical practice. Adequately powered trials are essential to draw a precise and accurate conclusion. Our study aims to determine the proportion of randomized controlled trials published in the field of emergency medicine that were sufficiently powered to detect a true 25% difference in outcomes between study groups. We conducted a PubMed search to identify randomized trials related to emergency care published in 5 top-ranked general medical and emergency medicine journals in the last 10 years. Standard statistical techniques were used to calculate the sample size required to have at least a 90% probability of detecting a 25% difference in the primary outcome between study groups. Adequate power was defined as a planned sample size larger than the sample size required to detect this difference. We found that approximately half of the studies that met inclusion criteria reported no significant difference between study groups. 36.3% of these “negative” studies had adequate power to detect a 25% difference between study groups. When grouped by study setting, 26.4% of Emergency Medical Services (EMS) based studies as compared to 22.5% of Emergency Department (ED) studies had adequate power to have at least a 90% chance of detecting a 25% difference between study groups (p=0.11). Therefore, we concluded that a large proportion of randomized trials in the medical literature had inadequate power to detect a clinically significant difference between study groups. Our study would help to strengthen research practice in the field of emergency medicine and to advance knowledge in this field. 

Hidden bias, also known as implicit or unconscious bias, affects attitudes, thinking, and behaviors in everyday interactions. It contributes to poor continuity and quality of care, and mistrusting relationships between health care providers and patients. Patients may not be treated equitably due to different identities (race, ethnicity, gender, etc) or different diseases (obesity, diabetes, hypertension, etc). These negative outcomes lead to health disparities and inequities. Despite this evidence, training strategies to detect and address hidden bias in patient-provider interactions are not well characterized and do not fully utilize innovative informatics and technology approaches. Can we leverage innovative technology to identify implicit bias from nonverbal cues in interpersonal interactions? Can we then provide feedback that raises awareness of those biases? The UnBIASED project will develop computational sensing tools to assess nonverbal communication signals associated with implicit bias and provide feedback to patients and providers. This approach could shape the next generation of training strategies for hidden healthcare bias. Documenting the range and utility of strategies in prior work upon which this innovative approach expands is important. To characterize existing training strategies for hidden bias and ways that technology can help, I report on a literature review of existing interventions and recommendations to combat implicit bias in clinical settings. Using dimensions, such as format of intervention- paper, technology, interactional (e.g., standardized patients), I characterize training strategies and their utility from prior work. Through this literature review, I aim to identify the gaps in existing work that illustrate opportunities for informatics and technology innovations for addressing implicit bias in healthcare. This review will provide practical insights for academic medical systems and programs on ways that technology can extend medical education curriculum to address implicit healthcare bias.

Despite systemic focus on avoiding potentially preventable hospitalizations, little is known about how decisions towards admitting patients are made. This project seeks to identify factors associated with admission or discharge from the Emergency Department (ED), of patients. Within the population of patients who present to the ED but are not medically ill, a subset is admitted anyways. We hypothesized that there would be identifiable demographic, socioeconomic or access differences between those who were admitted vs. those who were not admitted. This was a retrospective cross-sectional study of referrals for admission to the Acute Care Medicine center specifically from the ED. Using an existing database, we identified 452 cases fitting the aforementioned referral inclusion criteria. We performed in-depth chart review to identify relevant factors from demographics, measures of illness and other systems level factors. A chi-squared and t-test analysis was used to compare categorical and continuous variables for statistically significant differences, while a logistic regression model was used to measure admission odds ratios. Our results showed that 349 of the 452 referrals (77.9%) were admitted. Versus discharged patients, admitted patients were typically older (56.6 vs. 53.0 years) and experienced homelessness prior to ED presentation (36.7% vs. 20.4%). Predictors of admission included age over 65 (OR 3.5 [95%CI 1.1-1.6], p = 0.039) and night/weekend presentation (OR 2.0 [95%CI 1.1- 3.5], p = 0.020). The most common contributing factors that physicians identified as key to admission among referrals included lack of outpatient social support (35.8%), homelessness (33.0%) and substance abuse (23.5%). These results suggest that physician admission is predicated upon more than medical acuity, and incorporates patient characteristics and social determinants of health. With respect to the limitations of our setting and study design, this implies a need for closer integration and knowledge of social services into clinical settings.

There is evidence that the decision to admit patients to the hospital is not always correlated with their medical acuity. Emerging evidence suggests that a patient’s environment or psychosocial support may affect their likelihood of admission. However, there has been no comprehensive description of this body of literature. Given the breadth of this field, completion of a scoping review to map the current literature is warranted. The research question which this scoping review explores is: what factors other than medical acuity have been studied in relation to the physician’s decision to admit adult patients into the hospital? The scoping review method used in this review involves the following steps: identifying the research question, identifying and selecting studies that were relevant, charting and summarising the results extracted from these studies, and lastly consultation with stakeholders in order to add valuable insight. We collaborated with a research librarian to perform a comprehensive search of seven databases (PubMed, CINAHL, PsycINFO, EMBASE, Web of Science, Sociological Abstracts, and Social Science Abstracts) that resulted in 15,571 unique abstracts. A two-step screening process is currently underway. First, each abstract was reviewed independently by two research team members. From this, 1,368 abstracts were selected and are now being reviewed in full-text to confirm appropriateness for inclusion in the final analysis. We will next begin creating a narrative summary to describe the study designs used, factors studied, the terminology used, and outcomes identified. This research will provide a comprehensive view of trends that affect hospitalization and can be used to build new policies for treatment practices to improve healthcare in all communities.

Graphical dashboards have become a common way for patients to manage their quality of life following cancer treatment. Line charts have proven to help patients track trends in the severity of symptoms following cancer treatment. However, little is known about how to design such resources for people who have limited experience with traditional graphics (i.e., limited numeracy and graph literacy). The purpose of this project is to design and develop a web prototype that communicates prostate cancer quality of life over time for patients with limited numeracy and graph literacy. We define “quality of life” as a self-reported function after prostate cancer treatment. The quality of life data conveyed in prototypes are patient-reported function scores from the Expanded Prostate Cancer Index Composite (EPIC) survey. We are designing prototypes that do not require a high numeracy or graph literacy skills for comprehension through the use of audio, animation, pictographs, isotypes, and other formats. Our ultimate goal is to inform the design of innovative tools that accurately convey EPIC scores over time in ways that facilitate understanding and healthcare engagement among patients with limited numeracy and graph literacy. In this project, we focus specifically on combining (1) design activities and (2) web prototyping activities to create a set of functional prototypes for usability testing and feedback from underserved prostate cancer patients. We will use findings to enhance the design of a revised set of prototypes that better meet the needs of this important but often overlooked group.

Though there has been significant attention to preventing unnecessary hospital admissions, emerging evidence indicates that some patients who lack acuity of illness warranting admission continue to be hospitalized. In this study, we use a qualitative approach to further explore the hospitalist perspective and determine what factors influence the disposition decision for patients thought not to be appropriate for inpatient hospitalization. All admission calls to the acute care Medicine service at our hospital are logged into a Triage Database on an ongoing basis. This database allows the assessing physician to record a brief written narrative about the case and issues with making admission decisions. Entries were included in the study if admission calls originated from the Emergency Department and if the hospitalist assessed that the degree of patient’s medical acuity did not warrant hospitalization. Three hundred entries met inclusion criteria and were exported to Dedoose for qualitative analysis. A comprehensive codebook was iteratively created through independent coding accompanied by weekly meetings, with a final inter-rater reliability kappa score of 0.79. Using a grounded theory approach, a conceptual model of the admission decision-making process was constructed. We found that many factors are considered during admission assessment, both within and beyond physician control. The decision appears to occur in two distinct steps. First, an initial assessment considers patient factors, unmet medical and social needs, assessment of risk for discharge, and/or system-level barriers. A final admission decision is made after considering the influence of outside pressures such as: emergency department activity, coordination of care, resource constraint, and/or physician disagreements. Some of these factors may strongly outweigh the lack of medical acuity. Future work should investigate how these factors may vary at institutions with differing patient demographics and the effectiveness of possible interventions to alleviate the pressures of non-acute admission assessment factors.

 Many hospitalized patients with advanced cancer report feeling unprepared for discussions with their clinicians about goals of care. They also struggle with identifying what questions to ask their clinicians in these discussions, to get the information they need to make decisions. Employing a user-centered design approach, we developed an interactive guide that will be accessed through an iPad to help prepare patients for these discussions and the topics that may come up about goals of care without overwhelming or distressing them. The target population will be hospitalized patients with advanced cancer who are having a planned goals-of-care discussion (usually in the form of a scheduled meeting involving patients, family members, and clinicians). We are currently developing a prototype on a WordPress platform. We obtained feedback on the content of the prototype through focus groups conducted with both clinicians and family members of cancer patients. A key finding that emerged from these data was that patients often struggled with being burdened with information overload while at the same time feeling uninformed about their cancer. As we develop a high fidelity prototype, we intend to leverage the interactivity of our medium as a way to combat these issues.