Measuring care trajectories using health administrative databases: a population-based investigation of transitions from emergency to acute care

Utilizing Syndromic Surveillance for Hurricane Irma-Related CO Poisonings in Florida

BackgroundPeople who inject drugs (PWID) have been identified as frequent users of emergency department (ED) and hospital inpatient services. The specific challenges of record linkage in cohorts with numerous administrative health records occurring in close proximity are not well understood. Here, we present a method for patient-specific record linkage of ED and hospital admission data for a cohort of PWID.MethodsData from 688 PWID were linked to two state-wide administrative health databases identifying all ED visits and hospital admissions for the cohort between January 2008 and June 2013. We linked patient-specific ED and hospital admissions data, using administrative date-time timestamps and pre-specified linkage criteria, to identify hospital admissions stemming from ED presentations for a given individual. The ability of standalone databases to identify linked ED visits or hospital admissions was examined.ResultsThere were 3459 ED visits and 1877 hospital admissions identified during the study period. Thirty-four percent of ED visits were linked to hospital admissions. Most links had hospital admission timestamps in-between or identical to their ED visit timestamps (n = 1035, 87%). Allowing 24-h between ED visits and hospital admissions captured more linked records, but increased manual inspection requirements. In linked records (n = 1190), the ED ‘departure status’ variable correctly reflected subsequent hospital admission in only 68% of cases. The hospital ‘admission type’ variable was non-specific in identifying if a preceding ED visit had occurred.ConclusionsLinking ED visits with subsequent hospital admissions in PWID requires access to date and time variables for accurate temporal sorting, especially for same-day presentations. Selecting time-windows to capture linked records requires discretion. Researchers risk under-ascertainment of hospital admissions if using ED data alone.

247EMF Cluster Analysis of Acute Care Utilization Yields Insights for Tailored Pediatric Asthma Interventions

Introduction: Prevalence and incidence of delirium in older patients admitted to acute and long-term care facilities ranges between 9.6% and 89% but little is known in the context of emergency department (ED) incident delirium. Literature regarding the incidence of delirium in the ED and its potential impacts on hospital length of stay (LOS), functional status and unplanned ED readmissions is scant, its consequences have yet to be clearly identified in order to orient modern acute medical care. Methods: This study is part of the multicenter prospective cohort INDEED study. Three Canadian EDs completed the two years prospective study (March-July 2015 and Feb-May 2016). Patients aged 65 years old, initially free of delirium with an ED stay 8hours were followed up to 24h after ward admission. Patients were assessed 2x/day during their entire ED stay and up to 24 hours on hospital ward by research assistants (RA). The primary outcome of this study was incident delirium in the ED or within 24 h of ward admission. Functional and cognitive status were assessed using validated Older Americans’ Resources and Services and the Telephone Interview for Cognitive Status- modified tools. The Confusion Assessment Method (CAM) was used to detect incident delirium. ED and hospital administrative data were collected. Inter-observer agreement was realized among RA. Results: Incident delirium was not different between sites, nor between phases, nor between times from one site to another. All phases confounded, there is between 7 to 11% of ED related incident delirious episodes. Differences were seen in ED LOS between sites in non-delirious patients, but also between some sites for delirious participants (p<0.05). Only one site had a difference in ED LOS between their delirious and non-delirious patients, respectively of 52.1 and 40.1 hours (p<0.05). There is also a difference between sites in the time between arrival to the ED and the incidence of delirium (p=0.003). Kappa statistics were computed to measure inter-rater reliability of the CAM. Based on an alpha of 5%, 138 patients would allow 80% power for an estimated overall incidence proportion of 15 % with 5% precision.. Other predictive delirium variables, such as cognitive status, environmental factors, functional status, comorbidities, physiological status, and ED and hospital length of stay were similar between sites and phases. Conclusion: The fact that incidence of delirium was the same for all sites, despite the differences of ED LOS and different time periods suggest that many other modifiable and non-modifiable factors along LOS influenced the incidence of ED induced delirium. Emergency physician should concentrate on improving senior-friendly environment for the ED.

National unintentional carbon monoxide poisoning estimates using hospitalization and emergency department data

ObjeciveIdentify surveillance priorities for emergency department (ED) and Oregon Poison Center (OPC) data ahead of the 2017 Great American Solar Eclipse gatherings in Oregon and create a suite of queries for use in the Health Intelligence Section of the Oregon Public Health Division (OPHD) Incident Management Team (IMT).IntroductionOregon’s statewide syndromic surveillance system (Oregon ESSENCE) has been operational since 2012. Non-federal emergency department data (and several of their associated urgent care centers) are the primary source for the system, although other data sources have been added, including de-identified call data from OPC in 2016 (1).OPHD epidemiologists have experience monitoring mass gatherings (2) and have a strong relationship with OPC, collaborating on a regular basis for routine and heightened public health surveillance. Nevertheless, surveillance for the Great American Solar Eclipse (August 2017) presented a challenge due to the 107 reported simultaneous statewide eclipse-watching events planned for the day of the eclipse (some with estimated attendance of greater than 30,000 people and most in rural or frontier regions of the state).Scientific literature is limited on mass gathering surveillance in the developed world (3), particularly in rural settings (4), so OPC and OPHD worked together to develop a list of health conditions of interest, including some that would warrant both an ED visit and a call to OPC (e.g., snake bites). Monitoring visits in both data sources in would allow for assessment of total burden on the healthcare system, especially in the case of snake bites where only specific bites require administration of anti-venom.MethodsAhead of the planned mass gatherings, OPHD Health Intelligence and OPC compiled a list of expected risks from the literature (4,5) and input from members of the IMT including the Public Information Officer, who monitored media for stories about health. Priority health conditions presented a clear risk to public health (e.g., limited supply of snake anti-venom warranted surveillance of snake bites) or were the subject of substantial media coverage. Query development focused on risks that had specific, well-defined health effects and that would be captured by syndromic ED and OPC data.During an enhanced surveillance period (8/18-8/24), OPHD Health Intelligence reviewed and interpreted trends in common queries with OPC and disseminated a daily statewide surveillance report.ResultsOPHD and OPC created four new queries for both ED and OPC data streams: snake bites, psychedelic mushrooms, 2nd and 3rd degree body burns and eye-related calls and visits. ED queries used chief complaint, discharge diagnosis, or triage note. OPC queries used generic code, therapy and clinical effect.From 8/18-8/22, OPHD Health Intelligence distributed daily surveillance reports to the OPHD IMT and external partners. An increased in eye-related injuries was identified on the day after the eclipse, prompting OPHD Health Intelligence to consult with OPC. ED surveillance data indicated that the increase in eye-related visits was likely a seasonal trend. OPC staff reviewed the charts of patient calls captured by the query and concluded the calls were not related to retinal issues from looking at the sun. No other trends were noted in the joint OPHD/OPC queries.ConclusionsOPHD Health Intelligence piloted four new queries for surveillance during this mass gathering event and exercised the process for disseminating trend information from OPC and ED data. The eclipse event was fairly quiet and very few trends of note were captured by surveillance. Prior to this event, OPC data had not been a part of the Health Intelligence surveillance plan. However, assessing trends in OPC data provides an opportunity to better understand trends seen in ED data (e.g., whether or not a surge in ED visits for snake bites is accompanied by a surge in OPC calls for anti-venom is meaningful). By building a process to review disparate data in tandem, OPHD and OPC strengthened regional surveillance for this event. Applicable queries will continue to be used for planned event surveillance and several additional queries are currently under development.

ObjectiveEpidemiologists will understand the differences between syndromic and discharge emergency department data sources, the strengths and limitations of each data source, and how each of these different emergency department data sources can be best applied to inform a public health response to the opioid overdose epidemic.IntroductionTimely and accurate measurement of overdose morbidity using emergency department (ED) data is necessary to inform an effective public health response given the dynamic nature of opioid overdose epidemic in the United States. However, from jurisdiction to jurisdiction, differing sources and types of ED data vary in their quality and comprehensiveness. Many jurisdictions collect timely emergency department data through syndromic surveillance (SyS) systems, while others may have access to more complete, but slower emergency department discharge datasets. State and local epidemiologists must make decisions regarding which datasets to use and how to best operationalize, interpret, and present overdose morbidity using ED data. These choices may affect the number, timeliness, and accuracy of the cases identified.MethodsCDC partnered with 45 states and the District of Columbia to combat the worsening opioid overdose epidemic through three cooperative agreements: Prevention for States (PFS), Data Driven Prevention Initiative (DDPI), and Enhanced State Opioid Overdose Surveillance (ESOOS). To support funded jurisdictions in monitoring non-fatal opioid overdoses, CDC developed two different sets of indicator guidance for measuring non-fatal opioid overdoses using ED data, with each focusing on different ED data sources (SyS and discharge). We report on the following attributes for each type of ED data source1,2: 1) timeliness; 2) data quality (e.g., percent completeness by field); 3) validity; and 4) representativeness (e.g., percent of facilities included).ResultsWhen comparing timeliness across data sources, SyS data has clear advantages, with many jurisdictions receiving data within 24 hours of an event. For discharge data, timeliness is more variable with some jurisdictions receiving data within weeks while others wait over 1.5 years before receiving a complete discharge dataset. Data quality and completeness tends to be stronger in discharge datasets as facilities are required to submit complete discharge records with valid ICD-10-CM codes in order to be reimbursed by payers. By contrast, for SyS data systems, participating facilities may not consistently submit data for all possible fields, including diagnosis. Validity is dependent on the data source as well as the case definition or syndrome definition used; with this in mind, SyS data overdose indicators are designed to have high sensitivity, with less attention to specificity. Discharge data overdose indicators are designed to have a high positive predictive value, while sensitivity and specificity are both important considerations. Discharge datasets often include records for 100% of ED visits from all nonfederal, acute care-affiliated facilities in a state included. By contrast, representativeness of facilities in SyS data systems varies widely across states with some states having less than 50% of facilities reporting.ConclusionsCDC funded partners share overdose morbidity data with CDC using either ED SyS data, ED discharge data, or both. CDC indicator guidance for ED discharge data is designed for states to track changes in health outcomes over time for descriptive, performance monitoring, and evaluation purposes and to create rates that are more comparable across injury category, time, and place. Considering these objectives, CDC placed a higher priority on data quality, validity (i.e., positive predictive value), and representativeness, all of which are stronger attributes of discharge data. CDC’s indicator guidance for ED SyS data is designed for states to rapidly identify changes in nonfatal overdoses and to identify areas within a particular state that are experiencing rapid change in the frequency or types of overdose events. When considering these needs, CDC prioritized timeliness and validity in terms of sensitivity, both of which are stronger attributes of SyS data. SyS and discharge ED data each lend themselves to different informational applications and interpretations based on the strengths and limitations of each dataset. An effective, informed public health response to the opioid overdose epidemic requires continued investment in public health surveillance infrastructure, careful consideration of the needs of the data user, and transparency regarding the unique strengths and limitations of each dataset.

To explore variation in rates of acute care utilization for mental health conditions, including hospitalizations and emergency department (ED) visits, across high-income countries before and during the COVID-19 pandemic. Administrative patient-level data between 2017 and 2020 of eight high-income countries: Canada, England, Finland, France, New Zealand, Spain, Switzerland, and the United States (US). Multi-country retrospective observational study using a federated data approach that evaluated age-sex standardized rates of hospitalizations and ED visits for mental health conditions. There was significant variation in rates of acute mental health care utilization across countries. Among the subset of four countries with both hospitalization and ED data, the US had the highest pre-COVID-19 combined average annual acute care rate of 1613 episodes/100,000 people (95% CI: 1428, 1797). Finland had the lowest rate of 776 (686, 866). When examining hospitalization rates only, France had the highest rate of inpatient hospitalizations of 988/100,000 (95% CI 858, 1118) while Spain had the lowest at 87/100,000 (95% CI 76, 99). For ED rates for mental health conditions, the US had the highest rate of 958/100,000 (95% CI 861, 1055) while France had the lowest rate with 241/100,000 (95% CI 216, 265). Notable shifts coinciding with the onset of the COVID-19 pandemic were observed including a substitution of care setting in the US from ED to inpatient care, and overall declines in acute care utilization in Canada and France. The study underscores the importance of understanding and addressing variation in acute care utilization for mental health conditions, including the differential effect of COVID-19, across different health care systems. Further research is needed to elucidate the extent to which factors such as workforce capacity, access barriers, financial incentives, COVID-19 preparedness, and community-based care may contribute to these variations. Approximately one billion people globally live with a mental health condition, with significant consequences for individuals and societies. Rates of mental health diagnoses vary across high-income countries, with substantial differences in access to effective care. The COVID-19 pandemic has exacerbated mental health challenges globally, with varying impacts across countries. This study provides a comprehensive international comparison of hospitalization and emergency department visit rates for mental health conditions across eight high-income countries. It highlights significant variations in acute care utilization patterns, particularly in countries that are more likely to care for people with mental health conditions in emergency departments rather than inpatient facilities The study identifies temporal and cross-country differences in acute care management of mental health conditions coinciding with the onset of the COVID-19 pandemic.

Optimization of Linkage between North Carolina EMS and ED Data: EMS Naloxone Cases

Introduction: For cancer patients undergoing active treatment, emergency department (ED) visits may be an indicator of a breakdown in continuity and quality of care. Palliative care (PC) may be an important resource for patients in need of symptom management even during treatment with curative intent. This study aims to describe ED utilization by cancer patients and determine if PC consults impact ED use. Methods: Patient data from the Tom Baker Cancer Center (TBCC) was linked to PC and ED data as a retrospective cohort study. ED data was obtained from two administrative databases and PC data was obtained from four administrative databases and restricted to the first four hundred days following diagnosis. Univariate and Multivariate analyses were used. Results: Three actively treated cancer patient cohorts were identified based on first presentation following intake at the TBCC: 1) Used ED first (n=1637), 2) Used PC first (n=539), and 3) Only used services at the TBCC (n=2153). Using Multivariate analysis, patients living alone or who had a diagnosis of prostate or breast cancer were more likely to access the ED first or to only use services at the TBCC rather than access PC first. Patients who were divorced, on income support, or diagnosed with a lung or GI cancer, were more likely to access PC first rather than access the ED or only use services at the TBCC. A subgroup analysis was performed on those who accessed the ED at some point during their care, consisting of three groups: 1) ED Only Users (n=1091), 2) ED First Users, who also accessed PC (n=546), and 3) PC First Users, who also accessed the ED. There was a significant difference in rates of ED visits between the three groups: ED Only Users went to the ED at a rate of 3.8 per 1000 patient days; ED First Users, who also accessed PC, went to the ED at a rate of 7.7 per 1000 patient days; and PC First Users, who also accessed the ED, went to the ED at a rate of 9.2 per 1000 patient days (p< 0.001). Conclusion: In a tertiary cancer centre, patients who were divorced, on income support, or diagnosed with lung or GI cancer were more likely to access PC. Amongst those patients who presented to the ED, those who accessed PC first had higher rates of ED use. Further explorations of presenting complaints, utilization patterns, and symptom burdens will be analyzed to determine if early PC consults can influence or decrease ED utilization.

Identifying older emergency department (ED) patients with clinical features associated with adverse postdischarge outcomes may lead to improved clinical reasoning and better targeting for preventative interventions. Previous studies have used single-country samples to identify limited sets of determinants for a limited number of proxy outcomes. The objective of this study was to identify and compare geriatric syndromes that influence the probability of postdischarge outcomes among older ED patients from a multinational context. A multinational prospective cohort study of ED patients aged 75 years or older was conducted. A total of 13 ED sites from Australia, Belgium, Canada, Germany, Iceland, India, and Sweden participated. Patients who were expected to die within 24 hours or did not speak the native language were excluded. Of the 2,475 patients approached for inclusion, 2,282 (92.2%) were enrolled. Patients were assessed at ED admission with the interRAI ED Contact Assessment, a geriatric ED assessment. Outcomes were examined for patients admitted to a hospital ward (62.9%, n=1,436) or discharged to a community setting (34.0%, n=775) after an ED visit. Overall, 3% of patients were lost to follow-up. Hospital length of stay (LOS) and discharge to higher level of care was recorded for patients admitted to a hospital ward. Any ED or hospital use within 28 days of discharge was recorded for patients discharged to a community setting. Unadjusted and adjusted odds ratios (ORs) were used to describe determinants using standard and multilevel logistic regression. A multi-country model including living alone (OR=1.78, p≤0.01), informal caregiver distress (OR=1.69, p=0.02), deficits in ambulation (OR=1.94, p≤0.01), poor self-report (OR = 1.84, p≤0.01), and traumatic injury (OR=2.18, p≤0.01) best described older patients at risk of longer hospital lengths of stay. A model including recent ED visits (OR=2.10, p≤0.01), baseline functional impairment (OR=1.68, p≤0.01), and anhedonia (OR=1.73, p≤0.01) best described older patients at risk of proximate repeat hospital use. A sufficiently accurate and generalizable model to describe the risk of discharge to higher levels of care among admitted patients was not achieved. Despite markedly different health care systems, the probability of long hospital lengths of stay and repeat hospital use among older ED patients is detectable at the multinational level with moderate accuracy. This study demonstrates the potential utility of incorporating common geriatric clinical features in routine clinical examination and disposition planning for older patients in EDs.

Statement of purpose The NC Division of Public Health, in collaboration with UNC Chapel Hill, is improving injury surveillance data as part of the NC Surveillance Quality Improvement (SQI) Project. The project has focused on improving emergency department (ED) data in the statewide public health surveillance system NC DETECT. Unlike statewide mortality and hospital discharge data, NC DETECT ED data are available in near real time with over 75% of ED visits assigned at least one billing code within two weeks of the visit. One task/goal of the NC SQI project was the development of 12 poisoning and drug overdose surveillance case definitions. Methods/Approach The case definitions drew from existing definitions developed by state and national organisations; content experts in injury epidemiology, surveillance methods, and public health informatics; and end user feedback. Nine of the definitions incorporate diagnosis and/or E-codes (poisoning, unintentional poisoning, acute alcohol poisoning, drug overdose, unintentional drug overdose, opioid overdose, prescription analgesic opioid overdose, methadone overdose, and heroin overdose). Two definitions use a combination of diagnosis codes, E-codes, and keyword searches (drug overdose and heroin-related ED visits). One definition consists of only a keyword search (Narcan/naloxone). Results The new case definitions were added to the NC DETECT web portal in summer 2014. Authorised users can access both current and historical ED data. Authorised users from local health departments can access line-listing data for their county and compare aggregate data to other counties and the state. These case definitions may be revised based on user feedback. In addition, custom-reports can be developed to address specific poisoning topics (e.g. fentanyl overdoses). Conclusions NC DETECT ED data are vital to NC for public health surveillance. The development of 12 poisoning and drug overdose case definitions has streamlined poisoning surveillance activities. Significance and contribution to the field Given the variation among poisoning case definitions available nationwide, the NC SQI project has developed definitions for use in NC, tested the efficacy of these definitions using ED data, and revised these definitions based on expert and user feedback. It is hoped that these definitions may inform surveillance activities in other states.

In Maine, rabies postexposure prophylaxis (PEP) administration is reportable to public health. We sought to determine the objectives of the Maine Center for Disease Control and Prevention's (Maine CDC's) PEP administration surveillance system and whether the method of conducting surveillance through a manual health care provider (hereinafter, provider) reporting system meets these objectives. We also compared provider-reported PEP administrations with administrations identified in emergency department (ED) data. During September 2022, we interviewed 8 Maine CDC epidemiologists to determine system objectives. We obtained and compared PEP administration data from provider reporting system and ED data and summarized each dataset by year, exposing animal, and facility. We assessed the ability of each source to address surveillance system objectives by comparing data elements with each objective. Maine CDC epidemiologists described the following objectives of the surveillance system: (1) track potential human exposures to rabid or potentially rabid animals, (2) document PEP administration trends, and (3) ensure PEP is correctly administered. They determined the third objective is not being achieved by the current system. During January 2018-June 2022, we identified 538 provider-reported PEP administrations and 1191 PEP administrations through ED data. ED data were more timely than provider reports and identified more PEP administrations, but 28% of ED records did not contain information on the exposing animal. Maine CDC can use ED data to document PEP administration trends in near-real time. ED data obtained from syndromic surveillance might be used in tandem with or in place of Maine CDC's traditional PEP surveillance system. We are building more complex queries that more fully capture PEP administrations to have a thorough understanding of PEP administered in Maine.

Understanding the level of recording of acute serious events in general practice electronic health records (EHRs) is critical for making decisions about the suitability of general practice datasets to address research questions and requirements for linking general practice EHRs with other datasets. To examine data source agreement of five serious acute events (myocardial infarction, stroke, venous thromboembolism (VTE), pancreatitis and suicide) recorded in general practice EHRs compared with hospital, emergency department (ED) and mortality data. Data from 61 general practices routinely contributing data to the MedicineInsight database was linked with New South Wales administrative hospital, ED and mortality data. The study population comprised patients with at least three clinical encounters at participating general practices between 2019 and 2020 and at least one record in hospital, ED or mortality data between 2010 and 2020. Agreement was assessed between MedicineInsight diagnostic algorithms for the five events of interest and coded diagnoses in the administrative data. Dates of concordant events were compared. The study included 274,420 general practice patients with at least one record in the administrative data between 2010 and 2020. Across the five acute events, specificity and NPV were excellent (>98%) but sensitivity (13%-51%) and PPV (30%-75%) were low. Sensitivity and PPV were highest for VTE (50.9%) and acute pancreatitis (75.2%), respectively. The majority (roughly 70-80%) of true positive cases were recorded in the EHR within 30 days of administrative records. Large proportions of events identified from administrative data were not detected by diagnostic algorithms applied to general practice EHRs within the specific time period. EHR data extraction and study design only partly explain the low sensitivities/PPVs. Our findings support the use of Australian general practice EHRs linked to hospital, ED and mortality data for robust research on the selected serious acute conditions.

Introduction: Data collected from telephone helplines that provide health advice can improve the timeliness and accuracy of disease surveillance, contributing to an appropriate and rapid public health response. We show how these data can forewarn health professionals of increased rates of influenza-like illness (ILI) in the community and discuss implications for COVID-19 syndromic surveillance. Methods: The healthdirect helpline (HH) captures demographic details and characteristics of symptoms from users in 6 Australian states and territories. We compare ILI activity in the HH with ILI activity in emergency department (ED), laboratory, Flutracking and general practice (GP) data using cross-correlation functions. Results: Helpline data correlated strongly with ED data (range in yearly correlations from 0.82-0.99), GP data (0.66 – 0.95) and Flutracking data (0.62-0.89), but yearly correlations with laboratory data varied (0.49-0.95). The highest correlation with laboratory and GP data occurred when HH activity was 1-2 weeks in advance of these data, while correlations with ED and Flutracking data were strongest with no time lag. Discussion: Our analysis demonstrates that the number of ILI-related calls to the HH is a reliable indicator of ILI incidence in Australia. An increase in calls is likely to occur simultaneously with an increase in visits to EDs and prior to an increase in positive laboratory influenza tests and visits to GPs. A surveillance system including these data would assist health practitioners to receive timely and accurate estimates of the level of ILI in the community to better respond to and prepare for seasonal and epidemic influenza.