Mastorakos, George

Patient portals have been widely used by patients to enable timely communications with their providers via secure messaging for various issues including transportation barriers. The large volume of portal messages offers an invaluable opportunity for studying transportation barriers reported by patients. In this work, we explored the feasibility of cutting-edge deep learning techniques for identifying transportation issues mentioned in patient portal messages with deep semantic embeddings. The successful creation of annotated corpus and identification of 7 transportation issues showed the feasibility of this strategy. The developed annotated corpus could aid in developing an artificial intelligence tool to automatically identify transportation issues from millions of patient portal messages. The identified specific transportation issues and the analysis of patient demographics could shed light on how to reduce transportation gaps for patients.

Background . Patients increasingly use asynchronous communication platforms to converse with care teams. Natural language processing (NLP) to classify content and automate triage of these messages has great potential to enhance clinical efficiency. We characterize the contents of a corpus of portal messages generated by patients using NLP methods. We aim to demonstrate descriptive analyses of patient text that can contribute to the development of future sophisticated NLP applications. Methods . We collected approximately 3,000 portal messages from the cardiology, dermatology, and gastroenterology departments at Mayo Clinic. After labeling these messages as either Active Symptom, Logistical, Prescription, or Update, we used NER (named entity recognition) to identify medical concepts based on the UMLS library. We hierarchically analyzed the distribution of these messages in terms of departments, message types, medical concepts, and keywords therewithin. Results . Active Symptom and Logistical content types comprised approximately 67% of the message cohort. The “Findings” medical concept had the largest number of keywords across all groupings of content types and departments. “Anatomical Sites” and “Disorders” keywords were more prevalent in Active Symptom messages, while “Drugs” keywords were most prevalent in Prescription messages. Logistical messages tended to have the lower proportions of “Anatomical Sites,”, “Disorders,”, “Drugs,”, and “Findings” keywords when compared to other message content types. Conclusions . This descriptive corpus analysis sheds light on the content and foci of portal messages. The insight into the content and differences among message themes can inform the development of more robust NLP models.

OBJECTIVEThe level of radiation awareness by surgeons and residents in spinal surgery does not match the ubiquity of fluoroscopy in operating rooms in the United States. The present method of monitoring radiation exposure may contribute to the current deficiency in radiation awareness. Current dosimeters involve a considerable lag from the time that the surgical team is exposed to radiation to the time that they are provided with that exposure data. The objective of the current study was to assess the feasibility of monitoring radiation exposure in operating room personnel during lateral transpsoas lumbar interbody fusion (LLIF) and minimally invasive transforaminal lumbar interbody fusion (MI-TLIF) procedures by using a wearable personal device with real-time feedback.METHODSOperating room staff participating in minimally invasive surgical procedures under a single surgeon during a 6-month period were prospectively enrolled in this study. All radiation dose exposures were recorded for each member of the surgical team (surgeon, assistant surgeon, scrub nurse, and circulating nurse) using a personal dosimeter (DoseAware). Radiation doses were recorded in microsieverts (μSv). Comparisons between groups were made using ANOVA with the Tukey post hoc test and Student t-test.RESULTSThirty-nine patients underwent interbody fusions: 25 underwent LLIF procedures (14 LLIF alone, 11 LLIF with percutaneous screw placement [PSP]) and 14 underwent MI-TLIF. For each operative scenario per spinal level, the surgeon experienced significantly higher (p < 0.035) average radiation exposure (LLIF: 167.9 μSv, LLIF+PSP: 424.2 μSv, MI-TLIF: 397.9 μSv) than other members of the team, followed by the assistant surgeon (LLIF: 149.7 μSv, LLIF+PSP: 242.3 μSv, MI-TLIF: 274.9 μSv). The scrub nurse (LLIF: 15.4 μSv, LLIF+PSP: 125.7 μSv, MI-TLIF: 183.0 μSv) and circulating nurse (LLIF: 1.2 μSv, LLIF+PSP: 9.2 μSv, MI-TLIF: 102.3 μSv) experienced significantly lower exposures. Radiation exposure was not correlated with the patient’s body mass index (p ≥ 0.233); however, it was positively correlated with increasing patient age (p ≤ 0.004).CONCLUSIONSReal-time monitoring of radiation exposure is currently feasible and shortens the time between exposure and the availability of information regarding that exposure. A shortened feedback loop that offers more reliable and immediate data would conceivably raise the level of concern for radiation exposure in spinal surgeries and could alter patterns of behavior, leading to decreased exposures. Further studies are ongoing to determine the effect of real-time dosimetry in spinal surgery.

Certain medical schools have begun teaching courses in health systems science (HSS) to train medical students in skills aimed to improve health care in the United States. Although substantial research has been done on the potential benefit of HSS courses, reactions from students have not been reported. In this Invited Commentary, five medical students who have completed the first year of a longitudinal HSS course at the Arizona campus of the Mayo Clinic School of Medicine offer their observations of how early exposure to HSS affected their reactions to subsequent course work and current events in health care. The authors describe the HSS course and outline three benefits they have observed from their experience so far: (1) thinking more critically about health care delivery during all educational experiences, (2) gaining a better understanding of the complexity of the health care system, and (3) having a greater consideration for the many facets of health care delivery. The HSS course helped the authors identify health systems problems, develop solutions that incorporated diverse domains of health care delivery, and recognize the role and responsibility of the physician as an agent of change in a health care system.

ABSTRACT Defective interfering RNAs (DI-RNAs) of the viral genome can form during infections of negative-strand RNA viruses and outgrow full-length viral genomes, thereby modulating the severity and duration of infection. Here we document the frequent de novo generation of copy-back DI-RNAs from independent rescue events both for a vaccine measles virus (vac2) and for a wild-type measles virus (IC323) as early as passage 1 after virus rescue. Moreover, vaccine and wild-type C-protein-deficient (C-protein-knockout [C KO ]) measles viruses generated about 10 times more DI-RNAs than parental virus, suggesting that C enhances the processivity of the viral polymerase. We obtained the nucleotide sequences of 65 individual DI-RNAs, identified breakpoints and reinitiation sites, and predicted their structural features. Several DI-RNAs possessed clusters of A-to-G or U-to-C transitions. Sequences flanking these mutation sites were characteristic of those favored by adenosine deaminase acting on RNA-1 (ADAR1), which catalyzes in double-stranded RNA the C-6 deamination of adenosine to produce inosine, which is recognized as guanosine, a process known as A-to-I RNA editing. In individual DI-RNAs the transitions were of the same type and occurred on both sides of the breakpoint. These patterns of mutations suggest that ADAR1 edits unencapsidated DI-RNAs that form double-strand RNA structures. Encapsidated DI-RNAs were incorporated into virus particles, which reduced the infectivity of virus stocks. The C KO phenotype was dominant: DI-RNAs derived from vac2 with a C KO suppressed the replication of vac2, as shown by coinfections of interferon-incompetent lymphatic cells with viruses expressing different fluorescent reporter proteins. In contrast, coinfection with a C-protein-expressing virus did not counteract the suppressive phenotype of DI-RNAs. IMPORTANCE Recombinant measles viruses (MVs) are in clinical trials as cancer therapeutics and as vectored vaccines for HIV-AIDS and other infectious diseases. The efficacy of MV-based vectors depends on their replication proficiency and immune activation capacity. Here we document that copy-back defective interfering RNAs (DI-RNAs) are generated by recombinant vaccine and wild-type MVs immediately after rescue. The MV C protein interferes with DI-RNA generation and may enhance the processivity of the viral polymerase. We frequently detected clusters of A-to-G or U-to-C transitions and noted that sequences flanking individual mutations contain motifs favoring recognition by the adenosine deaminase acting on RNA-1 (ADAR1). The consistent type of transitions on the DI-RNAs indicates that these are direct substrates for editing by ADAR1. The ADAR1-mediated biased hypermutation events are consistent with the protein kinase R (PKR)-ADAR1 balancing model of innate immunity activation. We show by coinfection that the C-defective phenotype is dominant.