R.D. Harrington’s research while affiliated with University of Washington and other places

Objective Identifying study-eligible patients within clinical databases is a critical step in clinical research. However, accurate query design typically requires extensive technical and biomedical expertise. We sought to create a system capable of generating data model-agnostic queries while also providing novel logical reasoning capabilities for complex clinical trial eligibility criteria. Materials and Methods The task of query creation from eligibility criteria requires solving several text-processing problems, including named entity recognition and relation extraction, sequence-to-sequence transformation, normalization, and reasoning. We incorporated hybrid deep learning and rule-based modules for these, as well as a knowledge base of the Unified Medical Language System (UMLS) and linked ontologies. To enable data-model agnostic query creation, we introduce a novel method for tagging database schema elements using UMLS concepts. To evaluate our system, called LeafAI, we compared the capability of LeafAI to a human database programmer to identify patients who had been enrolled in 8 clinical trials conducted at our institution. We measured performance by the number of actual enrolled patients matched by generated queries. Results LeafAI matched a mean 43% of enrolled patients with 27 225 eligible across 8 clinical trials, compared to 27% matched and 14 587 eligible in queries by a human database programmer. The human programmer spent 26 total hours crafting queries compared to several minutes by LeafAI. Conclusions Our work contributes a state-of-the-art data model-agnostic query generation system capable of conditional reasoning using a knowledge base. We demonstrate that LeafAI can rival an experienced human programmer in finding patients eligible for clinical trials.

Objective: Identifying study-eligible patients within clinical databases is a critical step in clinical research. However, accurate query design typically requires extensive technical and biomedical expertise. We sought to create a system capable of generating data model-agnostic queries while also providing novel logical reasoning capabilities for complex clinical trial eligibility criteria. Materials and Methods: The task of query creation from eligibility criteria requires solving several text-processing problems, including named entity recognition and relation extraction, sequence-to-sequence transformation, normalization, and reasoning. We incorporated hybrid deep learning and rule-based modules for these, as well as a knowledge base of the Unified Medical Language System (UMLS) and linked ontologies. To enable data-model agnostic query creation, we introduce a novel method for tagging database schema elements using UMLS concepts. To evaluate our system, called LeafAI, we compared the capability of LeafAI to a human database programmer to identify patients who had been enrolled in 8 clinical trials conducted at our institution. We measured performance by the number of actual enrolled patients matched by generated queries. Results: LeafAI matched a mean 43% of enrolled patients with 27,225 eligible across 8 clinical trials, compared to 27% matched and 14,587 eligible in queries by a human database programmer. The human programmer spent 26 total hours crafting queries compared to several minutes by LeafAI. Conclusions: Our work contributes a state-of-the-art data model-agnostic query generation system capable of conditional reasoning using a knowledge base. We demonstrate that LeafAI can rival a human programmer in finding patients eligible for clinical trials.

Background The HIV reservoir of latently infected CD4+ T cells represents the barrier to cure. CD4+ T cell proliferation is a mechanism that sustains the reservoir even during prolonged antiretroviral therapy (ART). Blocking proliferation may therefore deplete the reservoir. Methods We conducted an unblinded, uncontrolled, clinical trial of mycophenolate, a T cell anti-proliferative compound, in people with HIV on chronic suppressive ART. Study drug dose selection was based on calibration to an observed ex vivo anti-proliferative effect. The primary outcome was clinically significant reduction (>0.25 log10) in the HIV reservoir measured by total and intact HIV DNA per million T cells in blood over 48 weeks. Results Five participants enrolled in the trial. Four participants took mycophenolate mofetil (MMF). One had a per protocol switch to enteric-coated mycophenolate sodium (Myfortic®) due to nausea but left the study for personal reasons. One participant developed finger cellulitis, but there were no opportunistic infections. In the four participants who completed the protocol, there was no clinically significant reduction in total or intact HIV DNA. There was no change in blood CD4+ T cell subset composition within the HIV reservoir or the entire CD4+ T cell population, although total CD4+ T cells decreased slightly in all four participants. An ex vivo anti-proliferative effect was observed using participant serum obtained one hour after dosing, but this effect was severely diminished at drug trough. Conclusions Mycophenolate given over 48 weeks did not reduce the volume or composition of the HIV reservoir.

Background Infective endocarditis (IE) remains highly morbid, but few studies have evaluated factors associated with IE mortality. We examined correlates of 90-day mortality among people who inject drugs (PWID) and do not inject drugs (non-PWID). Methods We queried the electronic medical record for cases of IE among adults ≥18 years of age at two academic medical centers in Seattle, Washington from January 1, 2014, to July 31, 2019. Cases were reviewed to confirm a diagnosis of IE and drug use status. Deaths were confirmed through the Washington State death index. Descriptive statistics were used to characterize IE in PWID and non-PWID. Kaplan-Meier log rank tests and Cox proportional hazard models were used to assess correlates of 90-day mortality. Results We identified 507 patients with IE, 213 (42%) of whom were PWID. Sixteen percent of patients died within 90 days of admission, including 14% of PWID and 17% of non-PWID (p=0.50). In a multivariable Cox proportional hazard model, injection drug use was associated with a higher mortality within the first 14 days of admission (adjusted HR [aHR] 2.33 [95% CI: 1.16 – 4.65], p=0.02); however, there was no association between injection drug use and mortality between 15 and 90 days of admission (aHR 0.63 [95% CI: 0.31 – 1.30], p=0.21). Conclusions Overall 90-day mortality did not differ between PWID and non-PWID with IE, although PWID experienced a higher risk of death within 14 days of admission. These findings suggest early IE diagnosis and treatment among PWID is critical to improving outcomes.

Nor98-like atypical scrapie is a sporadic disease that affects the central nervous system of sheep and goats that, in contrast to classical scrapie, is not generally regarded as naturally transmissible. However, infectivity has been demonstrated via bioassay not only of brain tissue but also of certain peripheral nerves, lymphoid tissues, and muscle. This study examines placental tissue, a well characterized route of natural transmission for classical scrapie. Further, this study was conducted in sheep homozygous for the classical scrapie resistant ARR genotype and is the first to characterize the transmission of Nor98-like scrapie between homozygous-ARR sheep. Nor98-like scrapie isolated from a United States ARR/ARR sheep was transmitted to four ARR/ARR ewes via intracerebral inoculation of brain homogenate. These ewes were followed and observed to 8 years of age, remained non-clinical but exhibited progression of infection that was consistent with Nor98-like scrapie, including characteristic patterns of PrPSc accumulation in the brain and a lack of accumulation in peripheral lymphoid tissues as detected by conventional methods. Immunoblots of placental tissues from the infected ewes revealed accumulation of a distinct conformation of PrPres, particularly as the animals aged; however, the placenta showed no infectivity when analyzed via ovinized mouse bioassay. Taken together, these results support a low risk for natural transmission of Nor98-like scrapie in ARR/ARR sheep.

To examine the relationship between African birth and HIV outcomes and comorbidities among individuals accessing care at the University of Washington. Patients who received a diagnosis of HIV at the University of Washington from 1995 to 2018 were identified. African-born patients were defined as those with recorded birthplace or primary language belonging to an African country. This cohort was compared to all non-African-born patients for initial CD4 count < 200 cells/mL, time from diagnosis to viral suppression, and prevalence of comorbid conditions. We identified 357 African-born and 3710 non-African-born patients. Over the time period, African-born patients were more likely to present with initial CD4 counts < 200 cells/mL (31% vs 19%, p < 0.01), but had shorter time to viral suppression (HR 1.31, [95% CI: 1.14–1.56]). African-born patients had higher rates of hepatitis B and tuberculosis (12% vs. 7% p < 0.01 and 13% vs. 3% p < 0.01). African-born patients living in the Seattle area have better HIV outcomes, but low initial CD4 counts suggest that they are presenting to care late. Increased efforts to engage this population in HIV, hepatitis B, and tuberculosis screening are warranted.

Background People who inject drugs (PWID) are at high risk for IE and account for a growing proportion of IE cases in the United States. We describe key characteristics of IE and predictors of 90-day mortality among people who do and do not inject drug at two large academic medical centers. Methods We used a string-searching and pattern-matching algorithm within all discharge (DC) summaries to query the electronic medical record (EMR) for cases of IE among adults ≥18 years of age at two academic medical centers in Seattle, Washington from December 1, 2013 to July 31, 2019. All cases were chart reviewed by a member of the study team to confirm a clinical diagnosis of IE and verify housing and PWID status, the latter defined as any injection drug use in the 3 months prior to admission. Microbiology and valve involvement were extracted from DC summaries and chart-reviewed where needed. Deaths were obtained from Washington state death index, which links to our EMR. Descriptive statistics were used to compare PWID and non-PWID with IE, and Kaplan-Meier log rank tests and Cox proportional hazard models were used to assess for predictors of 90-day mortality. Results We identified 387 patients with IE, 44% (n=166) of whom were PWID. When compared to non-PWID, PWID were younger (median age 33 vs. 55 years, p< 0.001) and more likely to be female (48% vs. 31%, p=0.001), homeless (41% vs. 9%, p< 0.001), have coagulase-positive Staphylococcal IE (69% vs. 32%, p< 0.001), and have right sided IE (66% vs. 26%, p< 0.001). Seventeen percent (n=64) of patients died within 90 days of admission, including 14% (n=23) of PWID and 19% (n=41) of non-PWID, with no difference in 90-day mortality between these groups (log-rank p=0.3). In univariate analyses, having left sided IE was the only predictor of 90-day mortality (HR 4.79, 95% CI 2.18 – 10.5). Conclusion Despite PWID being significantly younger and having a much higher frequency of right sided IE, they had similar 90-day mortality to non-PWID in this contemporary, urban cohort of hospitalized IE patients. Table 1. Demographic Characteristics of People Who Do and Do Not Inject Drugs with Infective Endocarditis at Two Seattle Hospitals, 2014 – 2019 Disclosures All Authors: No reported disclosures

Quality improvement and patient safety (QIPS) are core components of graduate medical education (GME). Training programs and affiliated medical centers must partner to create an environment in which trainees can learn while meaningfully contributing to QIPS efforts, to further the shared goal of improving patient care. Numerous challenges have been identified in the literature, including lack of resources, lack of faculty expertise, and siloed QIPS programs. In this article the authors describe a framework for integrated QIPS training for residents in the University of Washington Internal Medicine Residency Program, beginning in 2014 with the creation of a dedicated QIPS chief resident position and assistant program director for health systems position, the building of a formal curriculum, and integration with medical center QIPS efforts. The post-graduate year (PGY) 1 curriculum focused on the culture of patient safety and entering traditional patient safety event (PSE) reports. The PGY2 curriculum highlighted QIPS methodology and how to conduct mentored PSE reviews of cases that were of educational value to trainees and a clinical priority to the medical center. Additional PGY2/PGY3 training focused on the active report, presentation, and evaluation of cases during morbidity and mortality conferences while on clinical services, as well as how to lead longitudinal QIPS work. Select residents led mentored QI projects as part of an additional elective. The hallmark feature of this framework was the depth of integration with medical center priorities, which maximized educational and operational value. Evaluation of the program demonstrated improved attitudes, knowledge, and behavior changes in trainees, and significant contributions to medical center QIPS work. This specialty-agnostic framework allowed for training program and medical center integration, as well as horizontal integration across GME specialties, and can be a model for other institutions.

Clinical pathways can be useful when disparate clinical -pathologic groups converge on a common diagnostic and therapeutic trajectory. The progressive increase in the incidence of endocarditis in the United States has included higher risk subjects whose candidacy for aggressive cardiac surgical intervention may be highly resource-intensive, prohibitively high risk, or delayed and possibly deferred by comorbidities. We sought to define the sequence, application and resolution of multidisciplinary endocarditis team decision making in four distinct clinical groups.

We launched Infectious Disease electronic consultations (eConsults) in 2018. Over 15.5 months, PCPs submitted 328 eConsults; most frequent reasons were a positive culture or PCR result, syphilis, and latent tuberculosis. Practitioners commonly requested advice on antimicrobial choice, clinical evaluation, and indications for treatment. Internal phone consultations decreased after eConsult implementation.