Assessment of retrospective collection of EQ-5D-5L in a US COVID-19 population.

BACKGROUND: It is imperative to evaluate health related quality of life (HRQoL) pre-COVID-19, but there is currently no evidence of the retrospective application of the EuroQol 5-Dimension, 5 level version (EQ-5D-5L) for COVID-19 studies. METHODS: Symptomatic patients with SARS-CoV-2 at CVS Health US test sites were recruited between 01/31/2022-04/30/2022. Consented participants completed the EQ-5D-5L questionnaire twice: a modified version where all the questions were past tense to retrospectively assess pre-COVID-19 baseline QoL, and the standard version in present tense to assess current HRQoL. Duncan's new multiple range test was adopted for post analysis of variance pairwise comparisons of EQ visual analog scale (EQ VAS) means between problem levels for each of 5 domains. A linear mixed model was applied to check whether the relationship between EQ VAS and utility index (UI) was consistent pre-COVID-19 and during COVID-19. Matching-adjusted indirect comparison was used to compare pre-COVID-19 UI and EQ VAS scores with those of the US population. Lastly, Cohen's d was used to quantify the magnitude of difference in means between two groups. RESULTS: Of 676 participants, 10.2% were age 65 or more years old, 73.2% female and 71.9% white. Diabetes was reported by 4.7% participants and hypertension by 11.2%. The estimated coefficient for the interaction of UI-by-retrospective collection indicator (0 = standard prospective collection, 1 = retrospective for pre-COVID-19), -4.2 (SE: 3.2), P = 0.197, indicates that retrospective collection does not significantly alter the relationship between EQ VAS and UI. After adjusting for age, gender, diabetes, hypertension, and percent of mobility problems, the predicted means of pre-COVID-19 baseline EQ VAS and UI were 84.6 and 0.866, respectively. Both means were close to published US population norms (80.4 and 0.851) compared to those observed (87.4 and 0.924). After adjusting for age, gender, diabetes, and hypertension, the calculated ES between pre-COVID-19 and COVID-19 for UI and EQ VAS were 0.15 and 0.39, respectively. Without retrospectively collected EQ-5D-5L, using US population norms tended to underestimate the impact of COVID-19 on HRQoL. CONCLUSION: At a group level the retrospectively collected pre-COVID-19 EQ-5D-5L is adequate and makes it possible to directly evaluate the impact of COVID-19 on HRQoL. ( ClinicalTrials.gov NCT05160636).

Blood Culture Turnaround Time in U.S. Acute Care Hospitals and Implications for Laboratory Process Optimization.

The rapid identification of blood culture isolates and antimicrobial susceptibility test (AST) results play critical roles for the optimal treatment of patients with bloodstream infections. Whereas others have looked at the time to detection in automated culture systems, we examined the overall time from specimen collection to actionable test results. We examined four points of time, namely, blood specimen collection, Gram stain, organism identification (ID), and AST reports, from electronic data from 13 U.S. hospitals for the 11 most common, clinically significant organisms in septic patients. We compared the differences in turnaround times and the times from when specimens were collected and the results were reported in the 24-h spectrum. From January 2015 to June 2016, 165,593 blood specimens were collected, of which, 9.5% gave positive cultures. No matrix-assisted laser desorption ionization-time of flight (MALDI-TOF) mass spectrometry was used during the study period. Across the 10 common bacterial isolates (n = 6,412), the overall median (interquartile range) turnaround times were 0.80 (0.64 to 1.08), 1.81 (1.34 to 2.46), and 2.71 (2.46 to 2.99) days for Gram stain, organism ID, and AST, respectively. For all positive cultures, approximately 25% of the specimens were collected between 6:00 a.m. and 11:59 a.m. In contrast, more of the laboratory reporting times were concentrated between 6:00 a.m. and 11:59 a.m. for Gram stain (43%), organism ID (78%), and AST (82%), respectively (P < 0.001). The overall average turnaround times from specimen collection for Gram stain, organism ID, and AST were approximately 1, 2, and 3 days, respectively. The laboratory results were reported predominantly in the morning hours. Laboratory automation and work flow optimization may play important roles in reducing the microbiology result turnaround time.

The Effect of a Piperacillin/Tazobactam Shortage on Antimicrobial Prescribing and Clostridium difficile Risk in 88 US Medical Centers.

BACKGROUND: Anti-infective shortages are a pervasive problem in the United States. The objective of this study was to identify any associations between changes in prescribing of antibiotics that have a high risk for CDI during a piperacillin/tazobactam (PIP/TAZO) shortage and hospital-onset Clostridium difficile infection (HO-CDI) risk in 88 US medical centers. METHODS: We analyzed electronically captured microbiology and antibiotic use data from a network of US hospitals from July 2014 through June 2016. The primary outcome was HO-CDI rate and the secondary outcome was changes in antibiotic usage. We fit a Poisson model to estimate the risk of HO-CDI associated with PIP/TAZO shortage that were associated with increased high-risk antibiotic use while controlling for hospital characteristics. RESULTS: A total of 88 hospitals experienced PIP/TAZO shortage and 72 of them experienced a shift toward increased use of high-risk antibiotics during the shortage period. The adjusted relative risk (RR) of HO-CDI for hospitals experiencing a PIP/TAZO shortage was 1.03 (95% confidence interval [CI], .85-1.26; P = .73). The adjusted RR of HO-CDI for hospitals that both experienced a shortage and also showed a shift toward increased use of high-risk antibiotics was 1.30 (95% CI, 1.03-1.64; P < .05). CONCLUSIONS: Hospitals that experienced a PIP/TAZO shortage and responded to that shortage by shifting antibiotic usage toward antibiotics traditionally known to place patients at greater risk for CDI experienced greater HO-CDI rates; this highlights an important adverse effect of the PIP/TAZO shortage and the importance of antibiotic stewardship when mitigating drug shortages.

Predicting Readmission at Early Hospitalization Using Electronic Clinical Data: An Early Readmission Risk Score.

BACKGROUND: Identifying patients at high risk for readmission early during hospitalization may aid efforts in reducing readmissions. We sought to develop an early readmission risk predictive model using automated clinical data available at hospital admission. METHODS: We developed an early readmission risk model using a derivation cohort and validated the model with a validation cohort. We used a published Acute Laboratory Risk of Mortality Score as an aggregated measure of clinical severity at admission and the number of hospital discharges in the previous 90 days as a measure of disease progression. We then evaluated the administrative data-enhanced model by adding principal and secondary diagnoses and other variables. We examined the c-statistic change when additional variables were added to the model. RESULTS: There were 1,195,640 adult discharges from 70 hospitals with 39.8% male and the median age of 63 years (first and third quartile: 43, 78). The 30-day readmission rate was 11.9% (n=142,211). The early readmission model yielded a graded relationship of readmission and the Acute Laboratory Risk of Mortality Score and the number of previous discharges within 90 days. The model c-statistic was 0.697 with good calibration. When administrative variables were added to the model, the c-statistic increased to 0.722. CONCLUSIONS: Automated clinical data can generate a readmission risk score early at hospitalization with fair discrimination. It may have applied value to aid early care transition. Adding administrative data increases predictive accuracy. The administrative data-enhanced model may be used for hospital comparison and outcome research.

Global Health Education in Gastroenterology Fellowship: A National Survey.

BACKGROUND: Interest in global health (GH) education is increasing across disciplines. AIMS: To assess exposure to and perception of GH training among gastroenterology fellows and program directors across the USA. METHODS: Design: Electronic survey study. SETTING: The questionnaire was circulated to accredited US gastroenterology fellowship programs, with the assistance of the American Gastroenterological Association. PARTICIPANTS: Gastroenterology program directors and fellows. RESULTS: The questionnaire was returned by 127 respondents (47 program directors, 78 fellows) from 55 training programs (36 % of all training programs). 61 % of respondents had prior experience in GH. 17 % of programs offered GH curriculum with international elective (13 %), didactic (9 %), and research activity (7 %) being the most common. Fellows had adequate experience managing hepatitis B (93 %), cholangiocarcinoma (84 %), and intrahepatic duct stones (84 %). 74, 69 and 68 % reported having little to no experience managing hepatitis E, tuberculosis mesenteritis, or epidemic infectious enteritis, respectively. Most fellows would participate in an elective in an underserved area locally (81 %) or a 4-week elective abroad (71 %), if available. 44 % of fellows planned on working or volunteering abroad after fellowship. Barriers to establishing GH curriculum included funding (94 %), scheduling (88 %), and a lack of standardized objectives (78 %). Lack of interest, however, was not a concern. Fellows (49 %), more than faculty (29 %) (χ 2 = 21.9; p = 0.03), believed that GH education should be included in fellowship curriculum. CONCLUSIONS: Program directors and trainees recognize the importance of GH education. However, only 17 % of ACGME-approved fellowship programs offer the opportunity. Global health curriculum may enhance gastroenterology training.

Development and validation of a mortality risk-adjustment model for patients hospitalized for exacerbations of chronic obstructive pulmonary disease.

BACKGROUND: Acute exacerbation of chronic obstructive pulmonary disease (AECOPD) is a leading cause of hospitalization and death. We sought to develop and validate a mortality risk-adjustment model to enhance hospital performance measurement and to support comparative effectiveness research. METHODS: Using a derivation cohort of 69,299 AECOPD admissions in 2005-2006 across 172 hospitals, we developed a logistic regression model with age, sex, laboratory results, vital signs, and secondary diagnosis-based comorbidities as covariates. We converted the model coefficients into a score system and validated it using 33,327 admissions from 2007. We used the c-statistic to assess model fit. RESULTS: In the derivation and validation cohorts, the median (interquartile range) age was 72 (range, 63-79) versus 71 (range, 62-79) years; 45.6% versus 45.9% were male; and in-hospital mortality rates were 3.2% versus 2.9%, respectively. The predicted probability of deaths for individuals ranged from 0.004 to 0.942 versus 0.001 to 0.933, respectively. The relative contribution of variables to the predictive ability of the derivation model was age (18.3%), admission laboratory results (39.9%), vital signs (14.7%), altered mental status (7.1%), and comorbidities (19.9%). The model c-statistic was 0.83 (95% CI: 0.82, 0.84) versus 0.84 (95% CI: 0.83, 0.85), respectively, with good calibration for both cohorts. CONCLUSIONS: A mortality prediction model combining clinical and administrative data that can be obtained from electronic health records demonstrated good discrimination among patients hospitalized for AECOPD. The addition of admission vital signs and laboratory results enhanced clinical validity and could be applied to future comparative effectiveness research and hospital profiling efforts.

Using enriched observational data to develop and validate age-specific mortality risk adjustment models for hospitalized pediatric patients.

BACKGROUND: Growth and development in early childhood are associated with rapid physiological changes. We sought to develop and validate age-specific mortality risk adjustment models for hospitalized pediatric patients using objective physiological variables on admission in addition to administrative variables. METHODS: Age-specific laboratory and vital sign variables were crafted for neonates (up to 30 d old), infants/toddlers (1-23 mo), and children (2-17 y). We fit 3 logistic regression models, 1 for each age group, using a derivation cohort comprising admissions from 2000-2001 in 215 hospitals. We validated the models with a separate validation cohort comprising admissions from 2002-2007 in 62 hospitals. We used the c statistic to assess model fit. RESULTS: The derivation cohort comprised 93,011 neonates (0.55% mortality), 46,152 infants/toddlers (0.37% mortality), and 104,010 children (0.40% mortality). The corresponding numbers of admissions (mortality rates) for the validation cohort were 162,131 (0.50%), 33,818 (0.09%), and 73,362 (0.20%), respectively. The c statistics for the 3 models were 0.94, 0.91, and 0.92, respectively, for the derivation cohort and 0.91, 0.86, and 0.93, respectively, for the validation cohort. The relative contributions of physiological versus administrative variables to the model fit were 52% versus 48% (neonates), 93% versus 7% (infants/toddlers), and 82% versus 18% (children). CONCLUSIONS: The thresholds for physiological determinants varied by age. Common physiological variables assessed on admission contributed significantly to predicting mortality for hospitalized pediatric patients. These models may have practical utility in risk adjustment for pediatric outcomes and comparative effectiveness research when physiological data are captured through the electronic medical record.

Impact of COVID-19 and effects of booster vaccination with BNT162b2 on six-month long COVID symptoms, quality of life, work productivity and activity impairment during Omicron.

BACKGROUND: Longitudinal estimates of long COVID burden during Omicron remain limited. This study characterized long-term impacts of COVID-19 and booster vaccination on symptoms, Health-Related Quality of Life (HRQoL), and Work Productivity Activity Impairment (WPAI). METHODS: Outpatients with ≥ 1 self-reported symptom and positive SARS-CoV-2 test at CVS Health United States test sites were recruited between 01/31 and 04/30/2022. Symptoms, EQ-5D and WPAI were collected via online surveys until 6 months following infection. Both observed and model-based estimates were analyzed. Effect sizes based on Cohen's d quantified the magnitude of outcome changes over time, within and between vaccination groups. Mixed models for repeated measures were conducted for multivariable analyses, adjusting for covariates. Logistic regression assessed odds ratio (OR) of long COVID between vaccination groups. RESULTS: At long COVID start (Week 4), 328 participants included 87 (27%) Boosted with BNT162b2, 86 (26%) with a BNT162b2 primary series (Primed), and 155 (47%) Unvaccinated. Mean age was 42.0 years, 73.8% were female, 26.5% had ≥ 1 comorbidity, 36.9% prior infection, and 39.6% reported ≥ 3 symptoms (mean: 3.1 symptoms). At Month 6, among 260 participants, Boosted reported a mean of 1.1 symptoms versus 3.4 and 2.8 in Unvaccinated and Primed, respectively (p < 0.001). Boosted had reduced risks of ≥ 3 symptoms versus Unvaccinated (observed: OR 0.22, 95% CI 0.10-0.47, p < 0.001; model-based: OR 0.36, 95% CI 0.15-0.87, p = 0.019) and Primed (observed: OR 0.29, 95% CI 0.13-0.67, p = 0.003; model-based: OR 0.59, 95% CI 0.21-1.65, p = 0.459). Results were consistent using ≥ 2 symptoms. Regarding HRQoL, among those with long COVID, Boosted had higher EQ-5D Utility Index (UI) than Unvaccinated (observed: 0.922 vs. 0.731, p = 0.014; model-based: 0.910 vs. 0.758, p-value = 0.038) and Primed (0.922 vs. 0.648, p = 0.014; model-based: 0.910 vs. 0.708, p-value = 0.008). Observed and model-based estimates for EQ-VAS and UI among Boosted were comparable with pre-COVID since Month 3. Subjects vaccinated generally reported better WPAI scores. CONCLUSIONS: Long COVID negatively impacted HRQoL and WPAI. The BNT162b2 booster could have a beneficial effect in reducing the risk and burden of long COVID. Boosted participants reported fewer and less durable symptoms, which contributed to improve HRQoL and maintain WPAI levels. Limitations included self-reported data and small sample size for WPAI.

Impact of COVID-19 and effects of BNT162b2 on patient-reported outcomes: quality of life, symptoms, and work productivity among US adult outpatients.

BACKGROUND: Although there is extensive literature on the clinical benefits of COVID-19 vaccination, data on humanistic effects are limited. This study evaluated the impact of SARS-CoV-2 infection on symptoms, Health-Related Quality of Life (HRQoL) and Work Productivity and Impairment (WPAI) prior to and one month following infection between individuals vaccinated with BNT162b2 and those unvaccinated. METHODS: Subjects with ≥ 1 self-reported symptom and positive RT-PCR for SARS-CoV-2 at CVS Health US test sites were recruited between 01/31/2022 and 04/30/2022. Socio-demographics, clinical characteristics and vaccination status were evaluated. Self-reported symptoms, HRQoL, and WPAI outcomes were assessed using questionnaires and validated instruments (EQ-5D-5L, WPAI-GH) across acute COVID time points from pre-COVID to Week 4, and between vaccination groups. Mixed models for repeated measures were conducted for multivariable analyses, adjusting for several covariates. Effect size (ES) of Cohen's d was calculated to quantify the magnitude of outcome changes within and between vaccination groups. RESULTS: The study population included 430 subjects: 197 unvaccinated and 233 vaccinated with BNT162b2. Mean (SD) age was 42.4 years (14.3), 76.0% were female, 38.8% reported prior infection and 24.2% at least one comorbidity. Statistically significant differences in outcomes were observed compared with baseline and between groups. The EQ-Visual analogue scale scores and Utility Index dropped in both cohorts at Day 3 and increased by Week 4 but did not return to pre-COVID levels. The mean changes were statistically lower in the BNT162b2 cohort at Day 3 and Week 4. The BNT162b2 cohort reported lower prevalence and fewer symptoms at index date and Week 4. At Week 1, COVID-19 had a large impact on all WPAI-GH domains: the work productivity time loss among unvaccinated and vaccinated was 65.0% and 53.8%, and the mean activity impairment was 50.2% and 43.9%, respectively. Except for absenteeism at Week 4, the BNT162b2 cohort was associated with statistically significant less worsening in all WPAI-GH scores at both Week 1 and 4. CONCLUSIONS: COVID-19 negatively impacted HRQoL and work productivity among mildly symptomatic outpatients. Compared with unvaccinated, those vaccinated with BNT162b2 were less impacted by COVID-19 infection and recovered faster.

A simple risk score accurately predicts in-hospital mortality, length of stay, and cost in acute upper GI bleeding.

BACKGROUND: Although the early use of a risk stratification score in upper GI bleeding is recommended, existing risk scores are not widely used in clinical practice. OBJECTIVE: We sought to develop and validate an easily calculated bedside risk score, AIMS65, by using data routinely available at initial evaluation. DESIGN: Data from patients admitted from the emergency department with acute upper GI bleeding were extracted from a database containing information from 187 U.S. hospitals. Recursive partitioning was applied to derive a risk score for in-hospital mortality by using data from 2004 to 2005 in 29,222 patients. The score was validated by using data from 2006 to 2007 in 32,504 patients. Accuracy to predict mortality was assessed by the area under the receiver operating characteristic (AUROC) curve. MAIN OUTCOME MEASUREMENTS: Mortality, length of stay (LOS), and cost of admission. RESULTS: The 5 factors present at admission with the best discrimination were albumin less than 3.0 g/dL, international normalized ratio greater than 1.5, altered mental status, systolic blood pressure 90 mm Hg or lower, and age older than 65 years. For those with no risk factors, the mortality rate was 0.3% compared with 31.8% in patients with all 5 (P < .001). The model had a high predictive accuracy (AUROC = 0.80; 95% CI, 0.78-0.81), which was confirmed in the validation cohort (AUROC = 0.77, 95% CI, 0.75-0.79). Longer LOS and increased costs were seen with higher scores (P < .001). LIMITATIONS: Database data used does not include outcomes such as rebleeding. CONCLUSIONS: AIMS65 is a simple, accurate risk score that predicts in-hospital mortality, LOS, and cost in patients with acute upper GI bleeding.

Relationships between creatinine increase and mortality rates in patients given vancomycin in 76 hospitals: The increasing role of infectious disease pharmacists.

PURPOSE: Vancomycin is a commonly used antimicrobial with the potential for renal toxicity. We evaluated vancomycin duration, changes in renal function after vancomycin initiation ("post-vancomycin" renal function changes), and associated mortality risk among hospitalized patients. METHODS: We analyzed data from 76 hospitals and excluded patients with a baseline serum creatinine concentration (SCr) of >3.35 mg/dL. We estimated mortality risk relative to vancomycin duration and the magnitude of post-vancomycin SCr change, controlling for demographics, baseline SCr, underlying diseases, clinical acuity, and comorbidities. RESULTS: Among 128,993 adult inpatients treated with vancomycin, 49.0% did not experience SCr elevation. Among the remaining patients, 26.0%, 11.4%, 8.8% and 4.8% experienced increases in post-vancomycin SCr of 1% to 20%, 21% to 40%, 41% to 100%, and greater than 100%, respectively. Compared to mortality risk among patients with a vancomycin therapy duration between 4 and 5 days (the lowest-mortality group), longer vancomycin therapy duration was not independently associated with higher mortality risk after adjusting for confounders. In contrast, there was a graded relationship between post-vancomycin SCr elevation and mortality. Multivariable adjusted mortality odds ratios ranged from 1.60 to 13.66, corresponding to SCr increases of 10% and greater than 200%, respectively. CONCLUSION: Half of patients given vancomycin did not experience SCr elevation and had the lowest mortality, suggesting that vancomycin can be used safely if renal function is stabilized. In the large study cohort, vancomycin duration itself was not an independent predictor of mortality. Post-vancomycin SCr elevation appeared to be a driver of in-hospital mortality. Even a 10% post-vancomycin SCr increase was associated with an increased mortality risk. This finding stresses the importance of closely monitoring renal function and may support the value of pharmacokinetic dosing.

An automated model to identify heart failure patients at risk for 30-day readmission or death using electronic medical record data.

BACKGROUND: A real-time electronic predictive model that identifies hospitalized heart failure (HF) patients at high risk for readmission or death may be valuable to clinicians and hospitals who care for these patients. METHODS: An automated predictive model for 30-day readmission and death was derived and validated from clinical and nonclinical risk factors present on admission in 1372 HF hospitalizations to a major urban hospital between January 2007 and August 2008. Data were extracted from an electronic medical record. The performance of the electronic model was compared with mortality and readmission models developed by the Center for Medicaid and Medicare Services (CMS models) and a HF mortality model derived from the Acute Decompensated Heart Failure Registry (ADHERE model). RESULTS: The 30-day mortality and readmission rates were 3.1% and 24.1% respectively. The electronic model demonstrated good discrimination for 30 day mortality (C statistic 0.86) and readmission (C statistic 0.72) and performed as well, or better than, the ADHERE model and CMS models for both outcomes (C statistic ranges: 0.72-0.73 and 0.56-0.66 for mortality and readmissions respectively; P < 0.05 in all comparisons). Markers of social instability and lower socioeconomic status improved readmission prediction in the electronic model (C statistic 0.72 vs. 0.61, P < 0.05). CONCLUSIONS: Clinical and social factors available within hours of hospital presentation and extractable from an EMR predicted mortality and readmission at 30 days. Incorporating complex social factors increased the model's accuracy, suggesting that such factors could enhance risk adjustment models designed to compare hospital readmission rates.

Attributable burden in patients with carbapenem-nonsusceptible gram-negative respiratory infections.

OBJECTIVE: We aimed to describe the clinical and economic burden attributable to carbapenem-nonsusceptible (C-NS) respiratory infections. METHODS: This retrospective matched cohort study assessed clinical and economic outcomes of adult patients (aged ≥18 years) who were admitted to one of 78 acute care hospitals in the United States with nonduplicate C-NS and carbapenem-susceptible (C-S) isolates from a respiratory source. A subset analysis of patients with principal diagnosis codes denoting bacterial pneumonia or other diagnoses was also conducted. Isolates were classified as community- or hospital-onset based on collection time. A generalized linear mixed model method was used to estimate the attributable burden for mortality, 30-day readmission, length of stay (LOS), cost, and net gain/loss (payment minus cost) using propensity score-matched C-NS versus C-S cohorts. RESULTS: For C-NS cases, mortality (25.7%), LOS (29.4 days), and costs ($81,574) were highest in the other principal diagnosis, hospital-onset subgroup; readmissions (19.4%) and net loss (-$9522) were greatest in the bacterial pneumonia, hospital-onset subgroup. Mortality and readmissions were not significantly higher for C-NS cases in any propensity score-matched subgroup. Significant C-NS-attributable burden was found for both other principal diagnosis subgroups for LOS (hospital-onset: 3.7 days, P = 0.006; community-onset: 1.5 days, P<0.001) and cost (hospital-onset: $12,777, P<0.01; community-onset: $2681, P<0.001). CONCLUSIONS: Increased LOS and cost burden were observed in propensity score-matched patients with C-NS compared with C-S respiratory infections; the C-NS-attributable burden was significant only for patients with other principal diagnoses.

Contributing Factors to the Clinical and Economic Burden of Patients with Laboratory-Confirmed Carbapenem-Nonsusceptible Gram-Negative Respiratory Infections.

PURPOSE: This study examined patient- and hospital-level predictor variables that contribute to worse clinical and economic outcomes in patients with carbapenem-nonsusceptible respiratory infections. PATIENTS AND METHODS: Electronic data (January 2013 to September 2015) were from 78 US hospitals. Nonduplicate, gram-negative respiratory isolates were considered carbapenem-nonsusceptible if they tested resistant/intermediate to imipenem, meropenem, doripenem, or ertapenem. Potential predictors of outcomes (in-hospital mortality, 30-day readmission, length of stay [LOS], hospital total cost, and net gain/loss per patient) were examined using univariate analysis and generalized linear mixed models. Statistical significance and model goodness-of-fit criteria were used to identify significant predictors. RESULTS: A total of 1488 carbapenem-nonsusceptible respiratory patients were identified. Overall, the mortality rate was 13.7%, 30-day readmission rate was 20.6%, mean LOS was 20 days, mean total cost was $54,158, and mean net loss was $139 per patient. Our models showed that hospital-onset infection, higher clinical severity, mechanical ventilation/intensive care unit status, polymicrobial infection, and underlying diseases were all significant predictors for mortality, LOS, and total cost. Hospital-onset infections were also associated with a significantly greater net loss (P≤.01), and underlying disease significantly impacted readmissions (P=.03). The number of prior admissions, hospital characteristics, and payer type were also found to significantly impact measured outcomes. CONCLUSION: Carbapenem-nonsusceptible respiratory infections are associated with a considerable clinical and economic burden. The impact of hospital-onset infections on both clinical and economic outcomes highlights the continued need for action on this modifiable risk factor through antimicrobial stewardship and optimal therapy, thereby reducing the burden in this patient population.

Contributing Factors to the Clinical and Economic Burden of Patients with Laboratory-Confirmed Carbapenem-Nonsusceptible Gram-Negative Urinary Tract Infections.

PURPOSE: We explored patient- and hospital-level predictor variables for worse clinical and economic outcomes in carbapenem-nonsusceptible urinary tract infections (UTIs). PATIENTS AND METHODS: We used electronic data (January 2013-September 2015; 78 US hospitals) from a large multicenter clinical database. Nonduplicate gram-negative isolates were considered carbapenem-nonsusceptible if they had resistant/intermediate susceptibility. Potential predictors of outcomes (mortality, 30-day readmissions, length of stay [LOS], hospital total cost, and net gain/loss per case) were examined using generalized linear mixed models. Significant predictors were identified based on statistical significance and model goodness-of-fit criteria. RESULTS: A total of 1439 carbapenem-nonsusceptible urine cases were identified. The mortality rate was 5.5%; the hospital readmission rate was 25.0%. Mean (standard deviation [SD]) LOS, total cost, and loss per case were 12 (14) days, $21,502 ($37,172), and $5828 ($26,540), respectively. Hospital-onset (vs community-onset) infection significantly impacted all outcomes: mortality (odds ratio [OR], 2.21; 95% confidence interval [CI], 1.19-4.11; P=.01), 30-day readmissions (OR, 2.35; 95% CI, 1.49-3.71; P<.001), LOS (25.7 vs 10.2 days; P<.001), hospital total cost ($67,810 vs $22,141; P<.001), and loss per case (-$28,054 vs -$10,809; P<.001). Mechanical ventilation/intensive care unit status, neoplasms, and other underlying diseases were also common predictors for worse outcomes overall; polymicrobial infection was significantly associated with worse economic outcomes. Other key predictors were >1 prior hospitalization for 30-day readmissions, high Acute Laboratory Risk of Mortality Score for mortality, LOS, cost, and hospital teaching status for cost. CONCLUSION: Hospital-onset infections, polymicrobial infections, higher clinical severity, and underlying diseases are key predictors for worsened overall burden of carbapenem-nonsusceptible gram-negative UTIs.

Predicting hospital readmission in patients with mental or substance use disorders: A machine learning approach.

OBJECTIVE: Mental or substance use disorders (M/SUD) are major contributors of disease burden with high risk for hospital readmissions. We sought to develop and evaluate a readmission model using a machine learning (ML) approach. METHODS: We analyzed patients with continuous enrollment for three years and at least one episode of M/SUD as the primary reason for hospital admission. The outcome was readmission within 30-days from discharge. Model performance was evaluated using the Area under the Receiver Operating Characteristic (AUROC). We compared the AUROCs of an extreme gradient boosted tree (XGBoost) model to generalized linear model with elastic net regularization (GLMNet). RESULTS: We analyzed 65,426 unique patients and 97,688 admissions. Patients with mental disorders accounted for 66 % (13.2 % readmission rate) and substance use disorders, 34 % (22.3 % readmission rate). Among all those who had readmissions, 70.7 %, 17.0 %, and 12.4 % had 1, 2, or 3+ readmissions, respectively. Previous hospitalizations, hospital utilization, discharge disposition, diagnosis category, and comorbidity were among the highest important features in the XGBoost model. The XGBoost model AUROC was 0.737 (95 % CI: 0.732 to 0.742) versus the GLMNet 0.697 (95 % CI: 0.690 to 0.703). The AUROC of the final XGBoost model on the testing set was 0.738 (95 % CI: 0.730 to 0.748), higher than published readmission models for mental health patients. CONCLUSIONS: The XGBoost model has a better performance than GLMNet and previously published models in predicting readmissions in mental health patients. Our model may be further tested to aid targeted demographic initiatives to reduce M/SUDs readmissions and benchmarking.

Burden of early-onset candidemia: analysis of culture-positive bloodstream infections from a large U.S. database.

OBJECTIVES: : To characterize the epidemiology and burden of early-onset, nonnosocomial candidemia. DESIGN: : Retrospective review of Cardinal Health Outcomes Research Database, which comprises all acute care admissions at participating hospitals. SETTING: : A total of 176 acute care hospitals. PATIENTS: : All patients admitted from 2000 through 2005 who had early-onset bloodstream infection, defined as presence of both a positive blood culture drawn within 1 day before or within 48 hrs after hospital admission and an appropriate diagnostic code for infection. INTERVENTION: : None. MEASUREMENTS AND MAIN RESULTS: : To evaluate the impact of different pathogens on clinical and economic outcomes, we performed mixed-effect logistic and linear regression analyses and controlled for potential confounding factors. Of 64,307 early-onset bloodstream infections, 738 (1.2%) were positive for Candida. The rate of early-onset candidemia nearly doubled between 2000 and 2003 (p < .001) and then stabilized. Crude in-hospital mortality was higher for candidemia than for bacterial bloodstream infection (28.3% vs. 15.0%; p < .0001). Compared with patients with bacterial bloodstream infections, patients with candidemia were more likely to have been admitted within 30 days and to have been transferred from another healthcare facility. Compared with Gram-negative bacterial bloodstream infection and after controlling for other risk factors, candidemia was associated with increased mortality risk (odds ratio, 2.38; 95% confidence interval, 1.94-2.91; p < .0001), longer attributable hospital stay (4.8 days; 95% confidence interval, 4.1-5.5; p < .0001), and higher attributable hospital costs ($12,617; 95% confidence Interval, $10,755-$14,479; p < .0001). CONCLUSIONS: : Early-onset candidemia seems to be a distinct entity, which is increasing in frequency and is associated with increased mortality risk, longer hospital stay, and higher hospital costs relative to bacterial bloodstream infection.

Candidemia on presentation to the hospital: development and validation of a risk score.

INTRODUCTION: Candidemia results in substantial morbidity and mortality, especially if initial antifungal therapy is delayed or is inappropriate; however, candidemia is difficult to diagnose because of its nonspecific presentation. METHODS: To develop a risk score for identifying hospitalized patients with candidemia, we performed a retrospective analysis of a large database of 176 acute-care hospitals in the United States. We studied 64,019 patients with bloodstream infection (BSI) on presentation from 2000 through 2005 (derivation cohort) and 24,685 from 2006 to 2007 (validation cohort). We used recursive partitioning (RPART) to identify the best discriminators for Candida as the cause of BSI. We compared three sets of models (equal-weight, unequal-weight, vs full model with additional variables from logistic regression model) for sensitivity analysis. RESULTS: The RPART identified 6 variables as the best discriminators: age < 65 years, temperature 0.10, indicating predicted and observed candidemia rates did not differ significant across the 7 risk stratus). The full model with 16 risk factors had slightly higher AUROCs (0.74 versus 0.73 for derivation versus validation); however, 7 variables were no longer significant in the recalibrated model for the validation cohort, indicating that the additional items did not materially enhance the model. CONCLUSIONS: A simple equal-weight risk score differentiated patients' risk for candidemia in a graded fashion upon hospital presentation.

Hospital-level high-risk antibiotic use in relation to hospital-associated Clostridioides difficile infections: Retrospective analysis of 2016-2017 data from US hospitals.

OBJECTIVE: Antibiotics are widely used by all specialties in the hospital setting. We evaluated previously defined high-risk antibiotic use in relation to Clostridioides difficile infections (CDIs). METHODS: We analyzed 2016-2017 data from 171 hospitals. High-risk antibiotics included second-, third-, and fourth-generation cephalosporins, fluoroquinolones, carbapenems, and lincosamides. A CDI case was a positive stool C. difficile toxin or molecular assay result from a patient without a positive result in the previous 8 weeks. Hospital-associated (HA) CDI cases included specimens collected >3 calendar days after admission or ≤3 calendar days from a patient with a prior same-hospital discharge within 28 days. We used the multivariable Poisson regression model to estimate the relative risk (RR) of high-risk antibiotic use on HA CDI, controlling for confounders. RESULTS: The median days of therapy for high-risk antibiotic use was 241.2 (interquartile range [IQR], 192.6-295.2) per 1,000 days present; the overall HA CDI rate was 33 (IQR, 24-43) per 10,000 admissions. The overall correlation of high-risk antibiotic use and HA CDI was 0.22 (P = .003), and higher correlation was observed in teaching hospitals (0.38; P = .002). For every 100-day (per 1,000 days present) increase in high-risk antibiotic therapy, there was a 12% increase in HA CDI (RR, 1.12; 95% CI, 1.04-1.21; P = .002) after adjusting for confounders. CONCLUSIONS: High-risk antibiotic use is an independent predictor of HA CDI. This assessment of poststewardship implementation in the United States highlights the importance of tracking trends of antimicrobial use over time as it relates to CDI.

Reductions in mortality associated with intensive public reporting of hospital outcomes.

It is unclear whether public reporting of hospital and physician performance has improved outcomes for the conditions being reported. We studied the effect of intensive public reporting on hospital mortality for 6 high-frequency, high-mortality medical conditions. Patients in Pennsylvania were matched to patients in other states with varying public reporting environments using propensity score methods. The effect of public reporting was estimated using a difference in differences approach. Patients treated at hospitals subjected to intensive public reporting had significantly lower odds of in-hospital mortality when compared with similar patients treated at hospitals in environments with no public reporting or only limited reporting. Overall, the 2000-2003 in-hospital mortality odds ratio for Pennsylvania patients versus non-Pennsylvania patients ranged from 0.59 to 0.79 across 6 clinical conditions (all P < .0001). For the same comparison using the 1997-1999 period, odds ratios ranged from 0.72 to 0.90, suggesting improvement when intensive public reporting occurred.