Pediatric sepsis phenotypes for enhanced therapeutics: An application of clustering to electronic health records.

OBJECTIVE: The heterogeneity of pediatric sepsis patients suggests the potential benefits of clustering analytics to derive phenotypes with distinct host response patterns that may help guide personalized therapeutics. We evaluate the relative performance of latent class analysis (LCA) and K-means, 2 commonly used clustering methods toward the derivation of clinically useful pediatric sepsis phenotypes. METHODS: Data were extracted from anonymized medical records of 6446 pediatric patients that presented to 1 of 6 emergency departments (EDs) between 2013 and 2018 and were thereafter admitted. Using International Classification of Diseases (ICD)-9 and ICD-10 discharge codes, 151 patients were identified with a sepsis continuum diagnosis that included septicemia, sepsis, severe sepsis, and septic shock. Using feature sets used in related clustering studies, LCA and K-means algorithms were used to derive 4 distinct phenotypic pediatric sepsis segmentations. Each segmentation was evaluated for phenotypic homogeneity, separation, and clinical use. RESULTS: Using the 2 feature sets, LCA clustering resulted in 2 similar segmentations of 4 clinically distinct phenotypes, while K-means clustering resulted in segmentations of 3 and 4 phenotypes. All 4 segmentations identified at least 1 high severity phenotype, but LCA-identified phenotypes reflected superior stratification, high entropy approaching 1 (eg, 0.994) indicating excellent separation between estimated phenotypes, and differential treatment/treatment response, and outcomes that were non-randomly distributed across phenotypes (P < 0.001). CONCLUSION: Compared to K-means, which is commonly used in clustering studies, LCA appears to be a more robust, clinically useful statistical tool in analyzing a heterogeneous pediatric sepsis cohort toward informing targeted therapies. Additional prospective studies are needed to validate clinical utility of predictive models that target derived pediatric sepsis phenotypes in emergency department settings.

Effects of Maryland's global budget revenue model on emergency department utilization and revisits.

BACKGROUND: In 2014, Maryland (MD) implemented a "global budget revenue" (GBR) program that prospectively sets hospital budgets. This program introduced incentives for hospitals to tightly control volume and meet budget targets. We examine GBR's effects on emergency department (ED) visits, admissions, and returns. METHODS: We performed an interrupted time-series analysis with difference-in-differences comparisons using 2012 to 2015 Healthcare Cost Utilization and Project data from MD, New York (NY), and New Jersey (NJ). We examined GBR's effects on ED visits/1,000 population, admissions from the ED, and ED returns at 72 h and 9 days. We also examined rates of admission, intensive care unit (ICU) stay, and in-hospital mortality among returns. To evaluate racial/ethnic and payer outcome disparities among ED returns, we performed a triple differences analysis. RESULTS: ED visits decreased with GBR adoption in MD relative to NY and NJ, by five and six visits/1,000 population, respectively. ED admissions declined relative to NY and NJ, by 0.6% and 1.8%, respectively. There was also a post-GBR decline in ED returns by 0.7%. Admissions among returns declined by 2%, while ICU and in-hospital mortality among returns remained relatively stable. ED return outcomes varied by racial/ethnic and payer group. Non-Hispanic Whites and non-Hispanic Blacks experienced a similar decline in returns, while returns remained unchanged among Hispanics/Latinos, widening the disparity gap. Payer group disparities between privately insured and Medicare, Medicaid, and uninsured individuals improved, with the disparity reduction most pronounced among the uninsured. CONCLUSIONS: GBR adoption was associated with lower ED utilization and admissions. ED returns and admissions among returns also decreased, while mortality and ICU stays among returns remained stable, suggesting that GBR has not led to adverse patient outcomes from fewer admissions. However, changes in ED return disparities varied by subgroup, indicating that improvements in care transitions may be uneven across patient populations.

Emergency department care coordination strategies and perceived impact under Maryland's hospital payment reforms.

BACKGROUND: Emergency department (ED) care coordination plays an important role in facilitating care transitions across settings. We studied ED care coordination processes and their perceived effectiveness in Maryland (MD) hospitals, which face strong incentives to reduce hospital-based care through global budgets. METHODS: We conducted a qualitative study using semi-structured interviews to examine ED care coordination processes and perceptions of effectiveness. Interviews were conducted from January through October 2019 across MD hospital-based EDs. Results were reviewed to assign analytic domains and identify emerging themes. Descriptive statistics of ED care coordination staffing and processes were also calculated. RESULTS: A total of 25 in-depth interviews across 18 different EDs were conducted with ED physician leadership (n = 14) and care coordination staff (CCS) (n = 11). Across all EDs, there was significant variation in the hours and types of CCS coverage and the number of initiatives implemented to improve care coordination. Participants perceived ED care coordination as effective in facilitating safer discharges and addressing social determinants of health; however, adequate access to outpatient providers was a significant barrier. The majority of ED physician leaders perceived MD's policy reform as having a mixed impact, with improved care transitions and overall patient care as benefits, but increased physician workloads and worsened ED throughput as negative effects. CONCLUSIONS: EDs have responded to the value-based care incentives of MD's global budgeting program with investments to enhance care coordination staffing and a variety of initiatives targeting specific patient populations. Although the observed care coordination initiatives were broadly perceived to produce positive results, MD's global budgeting policies were also perceived to produce barriers to optimizing ED care. Further research is needed to determine the association of the various strategies to improve ED care coordination with patient outcomes to inform practice leaders and policymakers on the efficacy of the various approaches.

The Effects of Global Budgeting on Emergency Department Admission Rates in Maryland.

STUDY OBJECTIVE: In 2014, Maryland launched a population-based payment model that replaced fee-for-service payments with global budgets for all hospital-based services. This global budget revenue program gives hospitals strong incentives to tightly control patient volume and meet budget targets. We examine the effects of the global budget revenue model on rates of admission to the hospital from emergency departments (EDs). METHODS: We used medical record and billing data to examine adult ED encounters from January 1, 2012, to December 31, 2015, in 25 hospital-based EDs, including 10 Maryland global budget revenue hospitals, 10 matched non-Maryland hospitals (primary control), and 5 Maryland Total Patient Revenue hospitals (secondary control). Total Patient Revenue hospitals adopted global budgeting in 2010 under a pilot Maryland program targeting rural hospitals. We conducted difference-in-differences analyses for overall ED admission rates, ED admission rates for ambulatory-care-sensitive conditions and non-ambulatory-care-sensitive conditions, and for clinical conditions that commonly lead to admission. RESULTS: In 3,175,210 ED encounters, the ED admission rate for Maryland global budget revenue hospitals decreased by 0.6% (95% confidence interval -0.8% to -0.4%) compared with that for non-Maryland controls after global budget revenue implementation, a 3.0% relative decline, and decreased by 1.9% (95% confidence interval -2.2% to -1.7%) compared with that for Total Patient Revenue hospitals, a 9.5% relative decline. Relative declines in ED admission rates were similar for ambulatory-care-sensitive-condition and non-ambulatory-care-sensitive-condition encounters. Admission rate declines varied across clinical conditions. CONCLUSION: Implementation of the global budget revenue model led to statistically significant although modest declines in ED admission rates within its first 2 years, with declines in ED admissions most pronounced among certain clinical conditions.

The Impact of Maryland's Global Budget Payment Reform on Emergency Department Admission Rates in a Single Health System.

BACKGROUND: In 2014, the state of Maryland (MD) moved away from fee-for-service payments and into a global budget revenue (GBR) structure where hospitals have a fixed revenue target, independent of patient volume or services provided. We assess the effects of GBR adoption on emergency department (ED) admission decisions among adult encounters. METHODS: We used hospital medical record and billing data from adult ED encounters from January 1, 2011, through December 31, 2015, with four MD hospitals and two District of Columbia (DC) hospitals within the same health system. We performed difference-in-differences analysis and calculated the effects of the GBR model on ED admission rates (inpatient and observation) using hospital fixed-effect regression adjusted for patient, hospital, and community factors. We also examined changes in the distribution of acuity among ED admissions with GBR adoption. RESULTS: The study sample included 1,492,953 ED encounters with a mean ED admission rate of 20.5%. The ED admission rate difference pre- and post-GBR was -1.14% (95% confidence interval [CI] = -0.89 to -1.40) for MD hospitals and -0.04% (95% CI = -0.24 to 0.32) for DC hospitals with a difference-in-differences result of -1.10% (95% CI = -1.34 to -0.86). This change was attributable to a -3.3% (95% CI = -3.54 to -3.08) decline in inpatient admissions and 2.7% (95% CI = 2.53 to 2.79) increase in observation admissions. Declines in admissions were observed primarily among mild-to-moderate severity of illness encounters with a low risk of mortality. CONCLUSIONS: Within the same health system, implementation of global budgeting in MD hospitals was associated with a decline in ED admissions-particularly lower-acuity admissions-compared to DC hospitals that remained under fee-for-service payments.

Towards Reliable ARDS Clinical Decision Support: ARDS Patient Analytics with Free-text and Structured EMR Data.

In this work, we utilize a combination of free-text and structured data to build Acute Respiratory Distress Syndrome(ARDS) prediction models and ARDS phenotype clusters. We derived 'Patient Context Vectors' representing patientspecific contextual ARDS risk factors, utilizing deep-learning techniques on ICD and free-text clinical notes data. The Patient Context Vectors were combined with structured data from the first 24 hours of admission, such as vital signs and lab results, to build an ARDS patient prediction model and an ARDS patient mortality prediction model achieving AUC of 90.16 and 81.01 respectively. The ability of Patient Context Vectors to summarize patients' medical history and current conditions is also demonstrated by the automatic clustering of ARDS patients into clinically meaningful phenotypes based on comorbidities, patient history, and presenting conditions. To our knowledge, this is the first study to successfully combine free-text and structured data, without any manual patient risk factor curation, to build real-time ARDS prediction models.

The Impact of Hospital and Patient Factors on the Emergency Department Decision to Admit.

BACKGROUND: Substantial variation exists in rates of emergency department (ED) admission. We examine this variation after accounting for local and community characteristics. OBJECTIVES: Elucidate the factors that contribute to admission variation that are amenable to intervention with the goal of reducing variation and health care costs. METHODS: We conducted a retrospective cross-sectional study of 1,412,340 patient encounters across 18 sites from 2012-2013. We calculated the adjusted hospital-level admission rates using multivariate logistic regression. We adjusted for patient, provider, hospital, and community factors to compare admission rate variation and determine the influence of these characteristics on admission rates. RESULTS: The average adjusted admission rate was 22.9%, ranging from 16.1% (95% confidence interval [CI] 11.5-22%) to 32% (95% CI 26.0-38.8). There were higher odds of hospital admission with advancing age, male sex (odds ratio [OR] 1.20, 95% CI 1.91-1.21), and patients seen by a physician vs. mid-level provider (OR 2.26, 95% CI 2.23-2.30). There were increased odds of admission with rising ED volume, at academic institutions (OR 2.23, 95% CI 2.20-2.26) and at for-profit hospitals (OR 1.15, 95% CI 1.12-1.18). Admission rates were lower in communities with a higher per capita income, a higher rate of uninsured patients, and in more urban hospitals. In communities with the most primary providers, there were lower odds of admission (OR 0.60, 95% CI 0.57-0.68). CONCLUSION: Variation in hospital-level admission rates is associated with a number of local and community characteristics. However, the presence of persistent variation after adjustment suggests there are other unmeasured variables that also affect admission rates that deserve further study, particularly in an era of cost containment.

The relative contribution of provider and ED-level factors to variation among the top 15 reasons for ED admission.

STUDY OBJECTIVE: We examine adult emergency department (ED) admission rates for the top 15 most frequently admitted conditions, and assess the relative contribution in admission rate variation attributable to the provider and hospital. METHODS: This was a retrospective, cross-sectional study of ED encounters (≥18years) from 19 EDs and 603 providers (January 2012-December 2013), linked to the Area Health Resources File for county-level information on healthcare resources. "Hospital admission" was the outcome, a composite of inpatient, observation, or intra-hospital transfer. We studied the 15 most commonly admitted conditions, and calculated condition-specific risk-standardized hospital admission rates (RSARs) using multi-level hierarchical generalized linear models. We then decomposed the relative contribution of provider-level and hospital-level variation for each condition. RESULTS: The top 15 conditions made up 34% of encounters and 49% of admissions. After adjustment, the eight conditions with the highest hospital-level variation were: 1) injuries, 2) extremity fracture (except hip fracture), 3) skin infection, 4) lower respiratory disease, 5) asthma/chronic obstructive pulmonary disease (A&C), 6) abdominal pain, 7) fluid/electrolyte disorders, and 8) chest pain. Hospital-level intra-class correlation coefficients (ICC) ranged from 0.042 for A&C to 0.167 for extremity fractures. Provider-level ICCs ranged from 0.026 for abdominal pain to 0.104 for chest pain. Several patient, hospital, and community factors were associated with admission rates, but these varied across conditions. CONCLUSION: For different conditions, there were different contributions to variation at the hospital- and provider-level. These findings deserve consideration when designing interventions to optimize admission decisions and in value-based payment programs.

How (ED) admission decisions differ when the same physician works in two different emergency department.

STUDY OBJECTIVE: Emergency physicians often work in multiple hospital emergency departments (EDs). We study how emergency physician admission decisions vary in different settings. METHODS: We conducted a retrospective, cross-sectional study over two years (2012-3) in six EDs in three states. Included physicians had ≥200 encounters per site in two different EDs. "Admissions" were ED encounters resulting in admission to the hospital or transfer to another hospital. The primary outcome was the adjusted admission rate difference between the two sites. Hierarchical logistic regression analysis was used to calculate adjusted admission rates for each physician, which were then tabulated for each physician and compared across sites. RESULTS: In 51,807 ED encounters seen by 16 physicians the average admission rate was 20.0%, and unadjusted admission rates differed between sites by 2.9% (range 0-8.4%) for the same physician. The adjusted admission rate was 19.3% and differed between sites by 2.1% (range 0.4%-6.2%). CONCLUSION: In this sample, some ED physicians made similar admission decisions in different settings while others increased or decreased their admission rates up to 25% when practicing in a different ED.

Costs of ED episodes of care in the United States.

BACKGROUND: Emergency department (ED) care is a focus of cost reduction efforts. Costs for acute care originating in the ED, including outpatient and inpatient encounters (i.e. ED episodes), have not been estimated. OBJECTIVE: We estimate total US costs of ED episodes, potentially avoidable costs, and proportional costs of national health expenditures (NHEs). METHODS: We conducted a secondary analysis of 2010 data from the Medical Expenditure Panel Survey, National Hospital Ambulatory Medical Care Survey, and the Healthcare Cost and Utilization Project's Nationwide Inpatient Sample. Outpatient ED encounters were categorized based on the New York University algorithm and admissions by ambulatory care-sensitive condition (ACSC) vs non-ACSC. Potentially avoidable encounters were nonemergent ED visits and ACSC hospital admissions. Using the Medical Expenditure Panel Survey, we determined mean per-visit payments for each visit type. Using the National Hospital Ambulatory Medical Care Survey and Nationwide Inpatient Sample, we estimated aggregate expenditures and proportional costs of NHE by visit category. RESULTS: Emergency department episodes of care accounted for $328.1 billion in payments in 2010. This represented 12.5% of NHE; ED admissions were 8.3% and outpatient ED care was 4.2%. Nonemergent outpatient visits were the most common, comprising 30.4% of ED episodes, and non-ACSC admissions were the most costly at $188.3 billion. Potentially avoidable encounters accounted for $64.4 billion, 19.6% of ED episodes, and 2.4% of NHE. CONCLUSIONS: More than 1 in 10 health care dollars is spent on ED episodes of care. Of this, less than 1 in 5 dollars is potentially avoidable; therefore, efforts to reduce ED visits through improved primary care may have little impact on overall costs.

Costs associated with ambulatory care sensitive conditions across hospital-based settings.

OBJECTIVES: Ambulatory care sensitive conditions (ACSCs) are acute care diagnoses that could potentially be prevented through improved primary care. This study investigated how payments and charges for these ACSC visits differ by three hospital-based settings (outpatient, emergency department [ED], and inpatient) and examined differences in payments and charges by their physician and facility components. METHODS: This was a secondary analysis of data (2005 through 2010) from the Medical Expenditure Panel Survey. Multiple linear regression models were used to assess differences in the mean-adjusted payments and charges for ACSC visits by clinical setting and further divided payments and charges into physician and facility components. RESULTS: Of all ACSC visits from 2005 through 2010, 41% were outpatient visits, 36% were ED visits, and 23% were hospital admissions. After adjusting for patient demographics and comorbid conditions, charges for an inpatient ACSC visit were four times higher ($11,414 vs. $2,563) and payments were five times higher ($4,325 vs. $859) when compared to an ED visit. By comparison, charges for an ACSC ED visit were two times higher ($2,563 vs. $1,084) and payments 2.5 times higher ($859 vs. $341) relative to an ACSC visit managed in an outpatient hospital-based clinic. Across all clinical settings, hospital facility fees account for 77% to 94% of the charge differences and 81% to 93% of the payment differences. CONCLUSIONS: For hospital-based ACSC visits, inpatient hospitalizations are by far the most expensive. Finding ways to expand outpatient resources and improve the health management of the chronically ill may avoid conditions that lead to more expensive hospital-based encounters. Across all hospital-based settings, facility fees are the major contributor of expense.

Anticipated changes in reimbursements for US outpatient emergency department encounters after health reform.

STUDY OBJECTIVE: We study how reimbursements to emergency departments (EDs) for outpatient visits may be affected by the insurance coverage expansion of the Patient Protection and Affordable Care Act as previously uninsured patients gain coverage either through the Medicaid expansion or through health insurance exchanges. METHODS: We conducted a secondary analysis of data (2005 to 2010) from the Medical Expenditure Panel Survey. We specified multiple linear regression models to examine differences in the payments, charges, and reimbursement ratios by insurance category. Comparisons were made between 2 groups to reflect likely movements in insurance status after the Patient Protection and Affordable Care Act implementation: (1) the uninsured who will be Medicaid eligible afterward versus Medicaid insured, and (2) the uninsured who will be Medicaid ineligible afterward versus the privately insured. RESULTS: From 2005 to 2010, as a percentage of total ED charges, outpatient ED encounters for Medicaid beneficiaries reimbursed 17% more than for uninsured individuals who will become Medicaid eligible after Patient Protection and Affordable Care Act implementation: 40.0% versus 34.0%, mean absolute difference=5.9%, 95% confidence interval 5.7% to 6.2%. During the same period, the privately insured reimbursed 39% more than for uninsured individuals who will not be Medicaid eligible after Patient Protection and Affordable Care Act implementation: 54.0% versus 38.8%, mean absolute difference=15.2%, 95% confidence interval 12.8% to 17.6%. CONCLUSION: Assuming historical reimbursement patterns remain after Patient Protection and Affordable Care Act implementation, outpatient ED encounters could reimburse considerably more for both the previously uninsured patients who will obtain Medicaid insurance and for those who move into private insurance products through health insurance exchanges. Although our study does provide insight into the future, multiple factors will ultimately influence reimbursements after implementation of the act.

Mutational robustness of an RNA virus influences sensitivity to lethal mutagenesis.

The ability to extinguish a viral population of fixed reproductive capacity by causing small changes in the mutation rate is referred to as lethal mutagenesis and is a corollary of population genetics theory. Here we show that coxsackievirus B3 (CVB3) exhibits reduced mutational robustness relative to poliovirus, manifesting in enhanced sensitivity of CVB3 to lethal mutagens that is dependent on the size of the viral population. We suggest that mutational robustness may be a useful measure of the sensitivity of a virus to lethal mutagenesis.

Environmental and socio-economic factors as contributors to racial disparities in diabetes prevalence.

BACKGROUND: We deployed a study design that attempts to account for racial differences in socioeconomic and environmental risk exposures to determine if the diabetes race disparity reported in national data is similar when black and white Americans live under similar social conditions. DESIGN & METHODS: We compared data from the 2003 National Health Interview Survey (NHIS) with the Exploring Health Disparities in Integrated Communities-Southwest Baltimore (EHDIC-SWB) Study, which was conducted in a racially-integrated urban community without race differences in socioeconomic status. RESULTS: In the NHIS, African Americans had greater adjusted odds of having diabetes compared to whites (OR: 1.61, 95% CI: 1.26-2.04); whereas, in EHDIC-SWB white and African Americans had similar odds of having diabetes (OR: 1.07, 95% CI: 0.71-1.58). Diabetes prevalence for African Americans was similar in NHIS and EHDIC-SWB (10.4%, 95%CI: 9.5-11.4 and 10.5%, 95%CI: 8.5-12.5, respectively). Diabetes prevalence among whites differed for NHIS (6.6%, 95%CI: 6.2-6.9%) and EHDIC-SWB (10.1%, 95%CI: 7.6-12.5%). CONCLUSIONS: Race disparities in diabetes may stem from differences in the health risk environments that African Americans and whites live. When African Americans and whites live in similar risk environments, their health outcomes are more similar.

Lethal mutagenesis of picornaviruses with N-6-modified purine nucleoside analogues.

RNA viruses exhibit extraordinarily high mutation rates during genome replication. Nonnatural ribonucleosides that can increase the mutation rate of RNA viruses by acting as ambiguous substrates during replication have been explored as antiviral agents acting through lethal mutagenesis. We have synthesized novel N-6-substituted purine analogues with ambiguous incorporation characteristics due to tautomerization of the nucleobase. The most potent of these analogues reduced the titer of poliovirus (PV) and coxsackievirus (CVB3) over 1,000-fold during a single passage in HeLa cell culture, with an increase in transition mutation frequency up to 65-fold. Kinetic analysis of incorporation by the PV polymerase indicated that these analogues were templated ambiguously with increased efficiency compared to the known mutagenic nucleoside ribavirin. Notably, these nucleosides were not efficient substrates for cellular ribonucleotide reductase in vitro, suggesting that conversion to the deoxyriboucleoside may be hindered, potentially limiting genetic damage to the host cell. Furthermore, a high-fidelity PV variant (G64S) displayed resistance to the antiviral effect and mutagenic potential of these analogues. These purine nucleoside analogues represent promising lead compounds in the development of clinically useful antiviral therapies based on the strategy of lethal mutagenesis.

Synthesis and antiviral activity of 5-substituted cytidine analogues: identification of a potent inhibitor of viral RNA-dependent RNA polymerases.

As part of our studies of lethal viral mutagens, a series of 5-substituted cytidine analogues were synthesized and evaluated for antiviral activity. Among the compounds examined, 5-nitrocytidine was effective against poliovirus (PV) and coxsackievirus B3 (CVB3) and exhibited greater activity than the clinically employed drug ribavirin. Instead of promoting viral mutagenesis, 5-nitrocytidine triphosphate inhibited PV RNA-dependent RNA polymerase (K(d) = 1.1 +/- 0.1 microM), and this inhibition is sufficient to explain the observed antiviral activity.