Social determinants of health and COVID-19 infection in North Carolina: a geospatial analysis

Study Objectives: The COVID-19 pandemic has demonstrated that social determinants of health (SDOH) are profoundly linked to the spread and outcomes of COVID-19. However, the relationships between these SDOH and COVID-19 spatial outbreaks have yet to be determined. We conducted spatial analyses with geographic information systems (GIS) mapping of county-level SDOH and regional COVID-19 infection outbreaks to demonstrate the most impactful SDOH and to provide a pragmatic visual guide to prevent future outbreaks.

COVID-19 Infection Experiences and Social Determinants of Health in North Carolina: A Qualitative Analysis

Study Objectives: Social determinants of health (SDOH) influence the health outcomes of COVID-19 patients;yet, little is known about how patients at risk of significant disease burden view this relationship. Our study sought to explore patient perceptions of the influence of SDOH on their COVID-19 infection experience and COVID-19 transmission within their communities. Methods: We conducted a qualitative study of patients in a North Carolina health care system’s registry who tested positive for COVID-19 from March 2020 through February 2021. All patients’ addresses across six counties served were geo-referenced and analyzed by Kernel Density Estimation (KDE) to identify population-dense outbreaks of COVID-19 (hotspots). Spatial autocorrelation analysis was performed to identify census area clusters of white, Black and Hispanic populations, based on the 2019 American Community Survey dataset. Patients were identified by a randomized computer-generated sampling method. After informed consent, patients participated in semi-structured phone interviews in English or Spanish based on patient preference by trained bilingual researchers. Each interview was evaluated using a combination of deductive and inductive content analysis to determine prevalent themes related to COVID-19 knowledge and diagnosis, disease experience, and the impact of SDOH. Results: The 10 patients interviewed from our COVID-19 hotspots were of equal distribution by sex, and predominantly Black (70%), ages 22-70 years (IQR 45-62 years), and presented to the ED for evaluation (70%). The respondents were more frequently publicly insured (50% medicaid/medicare;vs 30% uninsured;vs 20% private). The interviews demonstrated themes surrounding the experience and impact of COVID-19. The perceived risk of contracting COVID-19 and knowledge of how to prevent infection varied greatly among our sample, and could be in part explained by SDOH such as their occupation, living conditions and mode of transportation. The experiences of COVID-19 testing, diagnosis, isolation and medical treatment were most influenced by the timing of infection in relation to the study period. For example, in the early months of the pandemic, the knowledge of isolation requirements and available support systems seemed to have negatively impacted the ability to isolate and follow public health guidance, as well as the support mechanisms provided by employers during this period. Communication of infection status once diagnosed varied greatly, with some voicing feelings of shame, and others advocating for sharing of infection experiences to change community behaviors. Suggestions for how to improve the COVID-19 response included improving communication and enforcing public health guidelines, including raising awareness for vulnerable populations on topics like expected symptoms, financial support, increasing testing, and vaccination delivery. Conclusion: Further exploration of important themes and related SDOH that influenced how the participants experienced the COVID-19 pandemic will be necessary to decrease the negative impacts of SDOH in communities that are high-risk for COVID-19 spread.

Prevalence and Predictors of Post-Traumatic Stress Disorder Symptoms Among Emergency Physicians in the United States During the COVID-19 Pandemic

Study Objectives: Our study aims to identify the prevalence of post-traumatic stress disorder (PTSD) symptoms among emergency physicians in the United States following the COVID-19 pandemic, and explore related factors and predictors of PTSD symptoms. Methods: Study participants included board-certified & board-eligible emergency physicians’ residents, and non-emergency physicians working in an EM setting, who were practicing in the US. Convenience sampling recruitment via multiple national EM listservs was used to complete an anonymous, online self-report survey from September 2020 to April 2021. Research data was stored on Qualtrics, a secure, password-protected multi-user database with access granted to the study team only. Surveys included demographics, a binary PTSD experience question and a PSS-I-5 scale pre-piloted for ease of use and comprehension. Based on Diagnostic and Statistical Manual of Mental Disorders-Fifth Edition (DSM-5) criteria, we asked participants to dichotomously self-report trauma. We used the American Psychological Association-approved PTSD Symptom Scale (PSS-I-5), a validated, reliable tool, to assess the severity of the PTSD symptoms in frontline health care workers during the COVID-19 pandemic categorized into minimal 0-8, mild 9-18, moderate 19-30, severe 31-45, and very severe 46-80. Descriptive analyses were reported with percentages, means, and medians using RStudio. Figures were used to visualize variations in reported PSS-I-5 scores through the course of the pandemic. Results: Our sample included 315 total complete surveys of the 362 initiated surveys. PSS-I-5 scores ranged from 0-67 (IQR 4-27, median=13, mean=17.2). The majority of participants are age 35-50 (45.7%), EM board-certified (69.5%), white (86.4%), practice at urban level 1 trauma centers (44.8%), and do not have previous PTSD (91.8%) or other mental health diagnoses (73.3%). More than half (55.9%) of the respondents self-identified as having experienced trauma based on the DSM-5 criteria. PSS-I-5 scores were higher from those completing the survey in March-April 2021 (median=13, mean=17.3) compared to those sent in September-October 2020 (median=11, mean=16.5). Most participants had PTSD symptoms (92.1%);40.7% with minimal (129), 22.1% mild (70), followed by moderate (57, 18.0%), severe (39, 12.3%), and very severe (23, 7.3%). A higher proportion of those reporting severe and very severe PTSD symptoms are female and practice at level 3/4 trauma centers. Of non-emergency physicians who participated in the study, a majority reported severe symptoms (median=31, mean=25.4). Major perceived causes of stress included shift acuity, crowding, fear of self/family/friends getting sick, lack of personal protective equipment, and dissatisfaction with hospital administration. Conclusion: The prevalence of PTSD symptoms among our sample following the COVID-19 pandemic is 92.1%, with higher PSS-I-5 scores generally reported later in the pandemic. Race, age, and practice setting all appear to be associated with more severe PTSD symptoms. More research is needed to describe and reduce the burden of PTSD among those on the COVID front lines in the ED. [Formula presented] Figure 1. Frequency histogram of PSS scores [Formula presented] Figure 2. PSS mean (circle) and median (star) scores per month of survey end date [Formula presented]

COVID-19 Infection Experiences and Social Determinants of Health in North Carolina: A Qualitative Analysis

Study Objectives: Social determinants of health (SDOH) influence the health outcomes of COVID-19 patients;yet, little is known about how patients at risk of significant disease burden view this relationship. Our study sought to explore patient perceptions of the influence of SDOH on their COVID-19 infection experience and COVID-19 transmission within their communities. Methods: We conducted a qualitative study of patients in a North Carolina health care system’s registry who tested positive for COVID-19 from March 2020 through February 2021. All patients’ addresses across six counties served were geo-referenced and analyzed by Kernel Density Estimation (KDE) to identify population-dense outbreaks of COVID-19 (hotspots). Spatial autocorrelation analysis was performed to identify census area clusters of white, Black and Hispanic populations, based on the 2019 American Community Survey dataset. Patients were identified by a randomized computer-generated sampling method. Patients participated in semi-structured phone interviews in English or Spanish based on patient preference by trained bilingual researchers. Each interview was evaluated using a combination of deductive and inductive content analysis to determine prevalent themes related to COVID-19 knowledge and diagnosis, disease experience, and the impact of SDOH. Results: The 10 patients interviewed from our COVID-19 hotspots were of equal distribution by sex, and predominantly Black (70%), ages 22-70 years (IQR 45-62 years), and presented to the ED for evaluation (70%). The respondents were more frequently publicly insured (50% medicaid/medicare;vs 30% uninsured;vs 20% private). The interviews demonstrated themes surrounding the experience and impact of COVID-19. The perceived risk of contracting COVID-19 and knowledge of how to prevent infection varied greatly and could be in part explained by SDOH such as their occupation and living conditions. The experiences of COVID-19 testing, diagnosis, isolation and treatment were most influenced by the timing of infection in relation to the study period. Earlier in the pandemic, the knowledge of isolation requirements and available support systems seemed to have negatively impacted the ability to isolate and follow public health guidance, as well as the support mechanisms provided by employers during this period. Communication of infection status once diagnosed varied greatly, with some voicing feelings of shame, and others advocating for sharing of infection experiences to change community behaviors. Suggestions for how to improve the COVID-19 response included improving communication and enforcing public health guidelines, including raising awareness for vulnerable populations. Conclusion: Further exploration of important themes and related SDOH that influenced how the participants experienced the COVID-19 pandemic will be necessary to decrease the negative impacts of SDOH in communities that are high-risk for COVID-19 spread.

Social Determinants of Health and COVID-19 Infection in North Carolina: A Geospatial Analysis

Study Objectives: The COVID-19 pandemic has demonstrated that social determinants of health (SDOH) are profoundly linked to the spread and outcomes of COVID-19. However, the relationships between these SDOH and COVID-19 spatial outbreaks have yet to be determined. We conducted spatial analyses with geographic information systems (GIS) mapping of county-level SDOH and regional COVID-19 infection outbreaks to demonstrate the most impactful SDOH and to provide a pragmatic visual guide to prevent future outbreaks. Methods: We analyzed the geospatial associations of COVID-19 infections and SDOH to identify areas of overlap. Our sample comprised all patients in a North Carolina health care system’s registry who tested positive for COVID-19 from March 2020-February 2021. Patients’ addresses were geo-referenced and analyzed by Kernel Density Estimation (KDE) to identify population-dense outbreaks of COVID-19 (hotspots). A set of 12 SDOH variables for each county were collected from the American Community Survey (ACS-5) and the Economic Research Service. Principal Component Analysis was applied to SDOH variables in order to reduce dimensions down to 3 geographical SDOH categories: Protective SDOH, High-Risk SDOH and Increased Vulnerability for Infection (Table 1). Using Multivariate Clustering Analysis (MCA), three clusters of census tracts were categorized according to SDOH indicators: decreased social disparities (DSD), equivocal social disparities (ESD) and increased social disparities (ISD) (Image A). Kruskal-Wallis and Dunn's Post-Hoc adjusted with Bonferroni were utilized to verify any difference in the proportion of patients residing in the different clusters (significance p<0.05). Results: A total of 13,733 patients were included in the study. The patients predominantly reside in Durham County (55.4%), are women (56.96%), and between 40 and 69 years old (41.9%). Further, patients are predominantly white (38.7%), non-Hispanic (79.63%), and single (49.6%). The concomitant analysis of KDE and MCA showed an overlap of COVID-19 hotspots with areas of ISD (Image B). The MCA revealed that there are 308 census tracts constituted by six counties, in which 40 form ISD clusters (vs. 109 ESD;vs. 159 DSD). In addition, ISD clusters have the highest rates of infection, with 179.8 patients per 10,000 (vs. 81.7 ESD;vs. 60.1 DSD). The ISD cluster was the most densely populated and was significantly more densely populated from the ESD and DSD clusters (p=0.01). Conclusion: In this sampling of COVID-19 patients, a disproportionate amount of patients come from areas with increased social disparities, suggesting further research and health policy will need to be directed towards addressing negative and vulnerability SDOH to curtail pandemic-related outbreaks. [Formula presented] [Formula presented]

181EMF Prospective Evaluation of Novice-Acquired 3D Ultrasound for Identification of Upper Extremity Fractures

Study Objectives: The initial imaging modality recommended by the American College of Radiology for suspected hand and wrist fractures is radiography. As there are radiation exposure risks associated with radiography, 2D ultrasound (2DUS) has also been investigated for diagnosis of these injuries. While sensitive and specific, 2DUS is operator dependent, requiring expertise to acquire and interpret images. 3DUS by novices is little studied in orthopedic evaluation. We aimed to determine whether novice-acquired 3DUS with expert or novice readers can identify hand and wrist fractures. We hypothesized that expert and novice interpretations of novice-acquired 3DUS of orthopedic injuries would show high agreement with each other and with the reference standard. Methods: The STARD criteria for studies of diagnostic tests were applied. Following IRB approval and informed consent, we prospectively enrolled subjects at a tertiary care academic medical center and an associated orthopedic clinic. We estimated a sample size of 70 subjects for an intraclass correlation coefficient (ICC) 0.7 (with alpha of 0.5 and power 0.8) and to detect kappa of 0.8. A single novice operator third-year medical student (MS3) performed all image acquisitions without any specific effort to identify anatomy or injuries during acquisition. 2D B mode US images were acquired using a Philips Lumify L12-4 transducer connected to a smartphone, and paired to an inertial measurement unit. All scans were reconstructed in volume rendering mode and displayed in 3DSlicer, an open-source visualization tool. Scans were interpreted by three groups of readers: 2 MS3s (novice), 3 emergency physicians with US fellowship training, and 2 board certified radiologists with musculoskeletal fellowship training (expert). The reference standard was board-certified radiologist interpretation of x-rays obtained during routine clinical care. Readers were blinded to all clinical data and x-ray diagnosis and rated 3DUS volumes for the presence or absence of fracture, fracture characteristics when present, and additional findings. Agreement between novices and experts in 3DUS interpretation and between 3DUS and x-ray findings are reported (kappa/ICC). Sensitivity/specificity/LR+/LR- with 95% CI were calculated. Time to perform and interpret 3DUS were reported. Results: 22 subjects were enrolled before the study was suspended due to the COVID-19 pandemic, with 90 3DUS volumes available for interpretation. Analysis is ongoing as results continue to be submitted, precluding calculation of kappa/ICC at this time. Expert 1 had sensitivity 0.8 (0.28, 0.99), specificity 0.69 (0.39, 0.91), LR+ 2.58 (1.03, 6.49), and LR- 0.29 (0.05, 1.73). Novice 1 had sensitivity 0.4 (0.05, 0.85), specificity 0.31 (0.09, 0.61), LR+ 0.58 (0.19, 1.79), and LR- 1.94 (0.66, 5.70). Interpretation times declined by over 50% for both novice and expert readers with an increasing number of scans interpreted. Mean acquisition time was 97 seconds per volume (median 97, IQR 57.75) with a mean of 2.5 volumes acquired per subject (median 2, IQR 1.25). Conclusion: Novice-acquired 3DUS by augmentation of 2DUS was rapid, and interpretation times decreased rapidly with experience. Preliminary results show a promising LR+ when scans are interpreted by an expert reader.

166EMF Prospective Comparison of 3D Point of Care Ultrasound and CT Angiography for Carotid Stenosis

Study Objectives: Carotid ultrasound using dedicated 3D systems is more reproducible and better quantifies disease compared to 2D Doppler ultrasound, but 3D system costs limit access. Low-cost point-of-care 3D ultrasound (POC 3DUS) can augment any 2D ultrasound. This system previously had near-perfect agreement for fetal measurements between novice and expert operators. We hypothesized that carotid assessment would not differ between novice-acquired 3DUS interpreted by novices and experts and CT angiography (CTA) interpreted by radiologists. Methods: We adhered to STARD criteria. Enrollment was by prospective convenience sample at a single medical center;any patient with recent/upcoming head and neck CTA was eligible. 2D B mode US acquisitions used a linear probe coupled to a screen capture device or smartphone, plus an orientation sensor and 3D reconstruction software. Scans were displayed as 2D stacks and intersecting cardinal planes (Figure). 3DUS were interpreted by medical students (novice), US fellowship trained emergency physicians, and radiologists (expert). CTAs were interpreted by neuroradiologists. Readers described NASCET stenosis, plaque, intimal-medial thickness, and minimum luminal cross-sectional area. Inter-reader reliability was measured by intraclass correlation coefficient (ICC)/kappa. We determined a sample size of 50 subjects for ICC 0.7 (alpha 0.05, power 0.8) and kappa 0.8. 3DUS sensitivity/specificity/LRs were estimated with CTA as the reference standard. Anonymous patient satisfaction surveys were administered. Results: Due to COVID-19, enrollment ended after 30 subjects (144 3DUS, 33 CTAs). Of the 60 arteries imaged, 21 had plaque on clinical CTA interpretation. Analysis is still in process. Mean 3DUS acquisition and reconstruction times were 13.1 sec (median 12.7, IQR 9.1-17.3) and 7.9 sec (med 8.0, IQR 5.0-10.3). Mean 3DUS interpretation time was 3m, 52s (med 3:06, IQR 2:14-4:49) for the first 497 3DUS reads. 13 patient surveys were completed. Mean subject willingness to repeat 3DUS was 8.1/10 (med 10, IQR 6.1-10). 2 subjects reported increased discomfort during the exam (mean change 0, med 0, IQR 0-0). 9 of 11 (81.8%) perceived a shorter scan time for 3DUS than for CTA, MRA, and/or 2DUS (2 declined to answer). CTA inter-reader agreement on plaque presence is 11/14 (0.79, 95% CI 0.52-0.92). Expert interpretations of the first 120 3DUS agreed on 55 (0.45, 95% CI 0.37-0.55), disagreed on 35 (0.29, 95% CI 0.22-0.38), and one or both readers were “unsure” on 30 (0.25, 95% CI 0.18-0.33). Of 90 3DUS where both readers answered with certainty, there was 61% raw agreement (95% CI 0.51-0.71). For the first 264 expert 3DUS interpretations, sensitivity is 0.77 (95% CI 0.66-0.87), specificity 0.59 (95% CI 0.50-0.67), +LR 0.47, -LR 0.84, using the original CTA read as reference standard (excluding 42 “unsure”). Conclusion: POC 3DUS is time-efficient with good patient satisfaction and promising sensitivity. Potential applications include initial diagnostic evaluation for neurologic symptoms or carotid bruit in low-resource settings. [Formula presented]

National COVID-19 vaccination plan: using artificial spatial intelligence to overcome challenges in Brazil

This article explores the use of spatial artificial intelligence to estimate the resources needed to implement Brazil's COVID-19 immu nization campaign. Using secondary data, we conducted a cross-sectional ecological study adop ting a time-series design. The unit of analysis was Brazil's primary care centers (PCCs). A four-step analysis was performed to estimate the popula tion in PCC catchment areas using artificial in telligence algorithms and satellite imagery. We also assessed internet access in each PCC and con ducted a space-time cluster analysis of trends in cases of SARS linked to COVID-19 at municipal level. Around 18% of Brazil's elderly population live more than 4 kilometer from a vaccination point. A total of 4,790 municipalities showed an upward trend in SARS cases. The number of PCCs located more than 5 kilometer from cell towers was largest in the North and Northeast regions. Innovative stra tegies are needed to address the challenges posed by the implementation of the country's National COVID-19 Vaccination Plan. The use of spatial artificial intelligence-based methodologies can help improve the country's COVID-19 response.

Machine learning analysis to predict health outcomes among emergency department users in southern brazil: A protocol study

Objective: Emergency services are essential to the organization of the health care system. Nevertheless, they face different operational difficulties, including overcrowded services, largely explained by their inappropriate use and the repeated visits from users. Although a known situation, information on the theme is scarce in Brazil, particularly regarding longitudinal user monitoring. Thus, this project aims to evaluate the predictive performance of different machine learning algorithms to estimate the inappropriate and repeated use of emergency services and mortality. Methods: To that end, a study will be conducted in the municipality of Pelotas, Rio Grande do Sul, with around five thousand users of the municipal emergency department. Results: If the study is successful, we will provide an algorithm that could be used in clinical practice to assist health professionals in decision-making within hospitals. Different knowledge dissemination strategies will be used to increase the capacity of the study to produce innovations for the organization of the health system and services. Conclusion: A high performance predictive model may be able to help decision-making in the emergency services, improving quality of care.