Faculty Bibliography

INTRODUCTION/BACKGROUND:In early March 2020, coronavirus 2019 (COVID-19) spread rapidly in New York City. Shortly thereafter, in response to the shelter-in-place orders and concern for infection, emergency department (ED) volumes decreased. While a connection between severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) infection and hyperglycemia/insulin deficiency is well described, its direct relation to diabetic ketoacidosis (DKA) is not. In this study we describe trends in ED volume and admitted patient diagnoses of DKA among five of our health system's EDs, as they relate to peak SARS-CoV-2 activity in New York City. METHODS:For the five EDs in our hospital system, deidentified visit data extracted for routine quality review was made available for analysis. We looked at total visits and select visit diagnoses related to DKA, across the months of March, April and May 2019, and compared those counts to the same period in 2020. RESULTS:A total of 93,218 visits were recorded across our five EDs from March 1-May 31, 2019. During that period there were 106 diagnoses of DKA made in the EDs (0.114% of visits). Across the same period in 2020 there were 59,009 visits, and 214 diagnoses of DKA (0.363% of visits) CONCLUSION: Despite a decrease in ED volume of 26.9% across our system during this time period, net cases of DKA diagnoses rose drastically by 70.1% compared to the prior year.

DESCRIPTION OF PROBLEM/UNASSIGNED:Streamlining communication between radiology and referring services is vital to ensure appropriate care with minimal delays. Increased subspecialization has led to compartmentalization of the radiology department with many physicians working in disparate areas. At our hospital, we anecdotally noted that a significant portion of incoming phone calls were misdirected to the wrong workstations. This resulted in wasted time, unnecessary interruptions, and delays in care because the referring clinicians could not efficiently navigate the radiology department staffing structure. Our quality improvement project involved developing a web-based tool allowing the emergency department (ED) to more efficiently contact the appropriate radiology desk and reduce misdirected phone calls. INSTITUTIONAL APPROACH EMPLOYED TO ADDRESS THE PROBLEM/UNASSIGNED:Surveys were sent to radiology residents and ED providers (attendings, residents, physician assistants) to assess how often phone calls were misdirected to the wrong radiology station. Radiology residents were asked which stations received the most misdirected phone calls, and what station the caller was often looking for. ED providers were asked which stations they intended when they were told they called the wrong station, and a series of questions in the survey assessed their knowledge of commonly called radiology station (Plain Film, CT Body, Ultrasound, Neuoradiology, Pediatrics, and Overnight Desk). ED and radiology physicians worked together to design a simple, easily accessed web-based tool that allowed the ED clinicians to determine which station should be called during for each hour of the day, which integrated differences in staffing by radiology throughout the day. After the tool had been implemented for 8 months, surveys were again sent to radiology residents and ED clinicians asking the same questions as before to assess for any significant change in response. Additional questions were added to the ED survey to assess awareness of the new tool. DESCRIPTION OF OUTCOMES IN CHANGE OF PRACTICE/UNASSIGNED:An interactive, easily updated schedule with optimal contact numbers was made available through the ED intranet. The design allowed for easy modification of contact numbers over time to accommodate changes in coverage location or staffing models. Prior to implementation contact information was presented on a static screen, which was unable to be changed and included multiple incorrect and defunct numbers. Additionally, contact defaulted to a general radiology pager, which was carried by a resident only responsible for plain films for most of the day. Numbers included in the new intranet tool were all pertinent reading room stations, all scheduling desks, and all technologist workspaces. Different schedules were provided for weekdays and weekends. Initial survey results showed that prior to the intervention, 74% of radiology residents said they received misdirected phone calls at least twice a day, and 57.9% of ED respondents reached the wrong recipient at least once per day. Frequencies of misdirected calls dropped to 58.4% of radiology residents (P = 0.37) and 17.9% of ED respondents (P < 0.01) on follow-up surveys 8 months after the tool was established. After establishing the new tool, 82.1% of ED respondents were aware of the new intranet contact tool and were using it to contact radiology. On the series of questions assessing ED respondents' knowledge of radiology numbers, over 50% of respondents knew the correct answer or answered using the call sheet after implementation; this resulted in statistically significant increases in accuracy for Body, Neuroradiology, and Pediatric radiology stations. Furthermore, with the exception of ED plain films, there was a statistically significant reduction in number of responses who said the general radiology pager should be called for reads. Fifty percent of radiology residents believed there was a reduction in the number of misdirected phone calls from the ED with this tool. CONCLUSION, LIMITATIONS, AND DESCRIPTIONS OF FUTURE DIRECTIONS/UNASSIGNED:Our tool was successful in accomplishing multiple goals. First, over 80% of ED respondents adopted the new tool. Second, the number of misdirected phone calls based on the subjective perception of ED respondents and radiology residents was reduced. Third, we objectively improved the ED respondents' behavior pattern in contacting the radiology department by either calling the correct number using the call tool, and by reducing the number of respondents who use the pager. Going forward, we hope to be able to expand use of this tool throughout the hospital in order to provide more timely and efficient care with other services by streamlining access between referring services and the appropriate radiology recipients.

Importance:As coronavirus disease 2019 (COVID-19) spread throughout the US in the early months of 2020, acute care delivery changed to accommodate an influx of patients with a highly contagious infection about which little was known. Objective:To examine trends in emergency department (ED) visits and visits that led to hospitalizations covering a 4-month period leading up to and during the COVID-19 outbreak in the US. Design, Setting, and Participants:This retrospective, observational, cross-sectional study of 24 EDs in 5 large health care systems in Colorado (n = 4), Connecticut (n = 5), Massachusetts (n = 5), New York (n = 5), and North Carolina (n = 5) examined daily ED visit and hospital admission rates from January 1 to April 30, 2020, in relation to national and the 5 states' COVID-19 case counts. Exposures:Time (day) as a continuous variable. Main Outcomes and Measures:Daily counts of ED visits, hospital admissions, and COVID-19 cases. Results:A total of 24 EDs were studied. The annual ED volume before the COVID-19 pandemic ranged from 13 000 to 115 000 visits per year; the decrease in ED visits ranged from 41.5% in Colorado to 63.5% in New York. The weeks with the most rapid rates of decrease in visits were in March 2020, which corresponded with national public health messaging about COVID-19. Hospital admission rates from the ED were stable until new COVID-19 case rates began to increase locally; the largest relative increase in admission rates was 149.0% in New York, followed by 51.7% in Massachusetts, 36.2% in Connecticut, 29.4% in Colorado, and 22.0% in North Carolina. Conclusions and Relevance:From January through April 2020, as the COVID-19 pandemic intensified in the US, temporal associations were observed with a decrease in ED visits and an increase in hospital admission rates in 5 health care systems in 5 states. These findings suggest that practitioners and public health officials should emphasize the importance of visiting the ED during the COVID-19 pandemic for serious symptoms, illnesses, and injuries that cannot be managed in other settings.

OBJECTIVE:Clinical decision support (CDS) tools have been shown to reduce inappropriate imaging orders. We hypothesized that CDS may be especially effective for house staff physicians who are prone to overuse of resources. MATERIALS AND METHODS:Our hospital implemented CDS for CT and MRI orders in the emergency department with scores based on the American College of Radiology's Appropriateness Criteria (range, 1-9; higher scores represent more-appropriate orders). Data on CT and MRI orders from April 2013 through June 2016 were categorized as pre-CDS or baseline, post-CDS period 1 (i.e., intervention with active feedback for scores of ≤ 4), and post-CDS period 2 (i.e., intervention with active feedback for scores of ≤ 6). Segmented regression analysis with interrupted time series data estimated changes in scores stratified by house staff and non-house staff. Generalized linear models further estimated the modifying effect of the house staff variable. RESULTS:Mean scores were 6.2, 6.2, and 6.7 in the pre-CDS, post-CDS 1, and post-CDS 2 periods, respectively (p < 0.05). In the segmented regression analysis, mean scores significantly (p < 0.05) increased when comparing pre-CDS versus post-CDS 2 periods for both house staff (baseline increase, 0.41; 95% CI, 0.17-0.64) and non-house staff (baseline increase, 0.58; 95% CI, 0.34-0.81), showing no differences in effect between the cohorts. The generalized linear model showed significantly higher scores, particularly in the post-CDS 2 period compared with the pre-CDS period (0.44 increase in scores; p < 0.05). The house staff variable did not significantly change estimates in the post-CDS 2 period. CONCLUSION:Implementation of active CDS increased overall scores of CT and MRI orders. However, there was no significant difference in effect on scores between house staff and non-house staff.

Lyme disease (LD) is the most common tick-borne illness in the United States. Although appropriate antibiotic treatment is effective for most cases, up to 20% of patients develop post-treatment Lyme disease syndrome (PTLDS). There is an urgent need to improve clinical management of LD using precise understanding of disease and patient stratification. We applied machine-learning to electronic medical records to better characterize the heterogeneity of LD and developed predictive models for identifying medications that are associated with risks of subsequent comorbidities. For broad disease categories, we identified 3, 16, and 17 comorbidities within 2, 5, and 10 years of diagnosis, respectively. At a higher resolution of ICD-9 codes, we identified known associations with LD including chronic pain and cognitive disorders, as well as particular comorbidities on a timescale that matched PTLDS symptomology. We identified 7, 30, and 35 medications associated with risks of these comorbidities within 2, 5, and 10 years, respectively. For instance, the first-line antibiotic doxycycline exhibited a consistently protective association for typical symptoms of LD, including backache. Our approach and findings may suggest new hypotheses for more personalized treatments regimens for LD patients.

Background/UNASSIGNED:The opioid epidemic in the United States is averaging over 100 deaths per day due to overdose. The effectiveness of opioids as pain treatments, and the drug-seeking behavior of opioid addicts, leads physicians in the United States to issue over 200 million opioid prescriptions every year. To better understand the biomedical profile of opioid-dependent patients, we analyzed information from electronic health records (EHR) including lab tests, vital signs, medical procedures, prescriptions, and other data from millions of patients to predict opioid substance dependence. Results/UNASSIGNED:We trained a machine learning model to classify patients by likelihood of having a diagnosis of substance dependence using EHR data from patients diagnosed with substance dependence, along with control patients with no history of substance-related conditions, matched by age, gender, and status of HIV, hepatitis C, and sickle cell disease. The top machine learning classifier using all features achieved a mean area under the receiver operating characteristic (AUROC) curve of ~ 92%, and analysis of the model uncovered associations between basic clinical factors and substance dependence. Additionally, diagnoses, prescriptions, and procedures prior to the diagnoses of substance dependence were analyzed to elucidate the clinical profile of substance-dependent patients, relative to controls. Conclusions/UNASSIGNED:The predictive model may hold utility for identifying patients at risk of developing dependence, risk of overdose, and opioid-seeking patients that report other symptoms in their visits to the emergency room.

BACKGROUND:The misuse of prescription opioids (MUPO) is a leading public health concern. Social media are playing an expanded role in public health research, but there are few methods for estimating established epidemiological metrics from social media. The purpose of this study was to demonstrate that the geographic variation of social media posts mentioning prescription opioid misuse strongly correlates with government estimates of MUPO in the last month. METHODS:We wrote software to acquire publicly available tweets from Twitter from 2012 to 2014 that contained at least one keyword related to prescription opioid use (n = 3,611,528). A medical toxicologist and emergency physician curated the list of keywords. We used the semantic distance (SemD) to automatically quantify the similarity of meaning between tweets and identify tweets that mentioned MUPO. We defined the SemD between two words as the shortest distance between the two corresponding word-centroids. Each word-centroid represented all recognized meanings of a word. We validated this automatic identification with manual curation. We used Twitter metadata to estimate the location of each tweet. We compared our estimated geographic distribution with the 2013-2015 National Surveys on Drug Usage and Health (NSDUH). RESULTS:Tweets that mentioned MUPO formed a distinct cluster far away from semantically unrelated tweets. The state-by-state correlation between Twitter and NSDUH was highly significant across all NSDUH survey years. The correlation was strongest between Twitter and NSDUH data from those aged 18-25 (r = 0.94, p < 0.01 for 2012; r = 0.94, p < 0.01 for 2013; r = 0.71, p = 0.02 for 2014). The correlation was driven by discussions of opioid use, even after controlling for geographic variation in Twitter usage. CONCLUSIONS:Mentions of MUPO on Twitter correlate strongly with state-by-state NSDUH estimates of MUPO. We have also demonstrated that a natural language processing can be used to analyze social media to provide insights for syndromic toxicosurveillance.

The feasibility of using mobile health applications to conduct observational clinical studies requires rigorous validation. Here, we report initial findings from the Asthma Mobile Health Study, a research study, including recruitment, consent, and enrollment, conducted entirely remotely by smartphone. We achieved secure bidirectional data flow between investigators and 7,593 participants from across the United States, including many with severe asthma. Our platform enabled prospective collection of longitudinal, multidimensional data (e.g., surveys, devices, geolocation, and air quality) in a subset of users over the 6-month study period. Consistent trending and correlation of interrelated variables support the quality of data obtained via this method. We detected increased reporting of asthma symptoms in regions affected by heat, pollen, and wildfires. Potential challenges with this technology include selection bias, low retention rates, reporting bias, and data security. These issues require attention to realize the full potential of mobile platforms in research and patient care.

In our recent Asthma Mobile Health Study (AMHS), thousands of asthma patients across the country contributed medical data through the iPhone Asthma Health App on a daily basis for an extended period of time. The collected data included daily self-reported asthma symptoms, symptom triggers, and real time geographic location information. The AMHS is just one of many studies occurring in the context of now many thousands of mobile health apps aimed at improving wellness and better managing chronic disease conditions, leveraging the passive and active collection of data from mobile, handheld smart devices. The ability to identify patient groups or patterns of symptoms that might predict adverse outcomes such as asthma exacerbations or hospitalizations from these types of large, prospectively collected data sets, would be of significant general interest. However, conventional clustering methods cannot be applied to these types of longitudinally collected data, especially survey data actively collected from app users, given heterogeneous patterns of missing values due to: 1) varying survey response rates among different users, 2) varying survey response rates over time of each user, and 3) non-overlapping periods of enrollment among different users. To handle such complicated missing data structure, we proposed a probability imputation model to infer missing data. We also employed a consensus clustering strategy in tandem with the multiple imputation procedure. Through simulation studies under a range of scenarios reflecting real data conditions, we identified favorable performance of the proposed method over other strategies that impute the missing value through low-rank matrix completion. When applying the proposed new method to study asthma triggers and symptoms collected as part of the AMHS, we identified several patient groups with distinct phenotype patterns. Further validation of the methods described in this paper might be used to identify clinically important patterns in large data sets with complicated missing data structure, improving the ability to use such data sets to identify at-risk populations for potential intervention.