4 results
266 Inpatient Quality Indicators Risk-Adjustment Using Interactions Selected by Machine Learning Methods
- Monika Ray, Patrick S. Romano
-
- Journal:
- Journal of Clinical and Translational Science / Volume 6 / Issue s1 / April 2022
- Published online by Cambridge University Press:
- 19 April 2022, pp. 44-45
-
- Article
-
- You have access Access
- Open access
- Export citation
-
OBJECTIVES/GOALS: Predictive models for health outcomes often have poor calibration potentially due to interactions that are ignored by standard methods. Using AHRQ models for Inpatient Quality Indicator (IQI) 11 Abdominal Aortic Aneurysm Repair and IQI 09 Pancreatic Resection mortality, we hypothesize that identifying interactions may improve model calibration. METHODS/STUDY POPULATION: We used adult discharge data from 16 states obtained from AHRQ Healthcare Cost and Utilization Project (State Inpatient Database), California Department of Health Care Access and Information, and New York State Department of Health. We used AHRQ’s v2021-1 Clinical Classifications Software Refined (CCSR) with present on admission flags to create features for risk-adjustment. We compared the performance of Least absolute shrinkage and selection operator (LASSO) model and first-order interaction models estimated using Hierarchical Group Lasso Regression (HGLR), after splitting the data into training and test sets. C-statistics, area under the precision-recall curve and Hosmer-Lemeshow calibration plots are reported. Finally, logistic regression models with selected CCSRs were evaluated on the test set. RESULTS/ANTICIPATED RESULTS: IQI 11 has four strata: open and endovascular repair of ruptured aneurysms (39% and 21% mortality, respectively); open and endovascular repair of unruptured aneurysms (6% and 0.8% mortality, respectively). IQI 09 has two strata: with and without pancreatic cancer (2% and 2.5% mortality, respectively). Comparing the HGLR model (with interaction effects) with Lasso models (without interactions), we noticed meaningful improvements in discrimination and calibration. However, for IQI 09, the extremely low mortality rate did not result in good HGLR or LASSO models. Interactions involving CCSRs could be identified using the novel HGLR method, which improved model performance given a heterogeneous population in IQI 11 with a mix of high and low event rates, unlike the more homogeneous patient population in IQI 09. DISCUSSION/SIGNIFICANCE: Standard implementations of regression models fail to address critical issues that arise in healthcare data – (a) quadratic explosion of potential interactions that cannot be manually identified, and (b) categorical variables with multiple levels or values (e.g., age categories). We propose innovative use of HGLR to robustly address these issues.
Prevalence and Epidemiology of Healthcare-Associated Infections (HAI) in US Nursing Homes (NH), 2017
- Nicola Thompson, Nimalie Stone, Cedric Brown, Taniece Eure, Austin Penna, Grant Barney, Devra Barter, Paula Clogher, Ghinwa Dumyati, Erin Epson, Christina B. Felsen, Linda Frank, Deborah Godine, Lourdes Irizarry, Helen Johnston, Marion Kainer, Linda Li, Ruth Lynfield, J.P. Mahoehney, Joelle Nadle, Valerie Ocampo, Susan Ray, Monika Samper, Sarah Shrum, Marla Sievers, Srinivasan Krithika, Lucy E. Wilson, Alexia Zhang, Shelley Magill
-
- Journal:
- Infection Control & Hospital Epidemiology / Volume 41 / Issue S1 / October 2020
- Published online by Cambridge University Press:
- 02 November 2020, pp. s45-s46
- Print publication:
- October 2020
-
- Article
-
- You have access Access
- Export citation
-
Background: With an aging population, increasingly complex care, and frequent re-admissions, prevention of healthcare-associated infections (HAIs) in nursing homes (NHs) is a federal priority. However, few contemporary sources of HAI data exist to inform surveillance, prevention, and policy. Prevalence surveys (PSs) are an efficient approach to generating data to measure the burden and describe the types of HAI. In 2017, the Centers for Disease Control and Prevention (CDC) performed its first large-scale HAI PS through the Emerging Infections Program (EIP) to measure the prevalence and describe the epidemiology of HAI in NH residents. Methods: NHs from several states (CA, CO, CT, GA, MD, MN, NM, NY, OR, & TN) were randomly selected and asked to participate in a 1-day HAI PS between April and October 2017; participation was voluntary. EIP staff reviewed available medical records for NH residents present on the survey date to collect demographic and basic clinical information and infection signs and symptoms. HAIs with onset on or after NH day 3 were identified using revised McGeer infection definitions applied to data collected by EIP staff and were reported to the CDC through a web-based system. Data were reviewed by CDC staff for potential errors and to validate HAI classifications prior to analysis. HAI prevalence, number of residents with >1 HAI per number of surveyed residents ×100, and 95% CIs were calculated overall (pooled mean) and for selected resident characteristics. Data were analyzed using SAS v9.4 software. Results: Among 15,296 residents in 161 NHs, 358 residents with 375 HAIs were identified. The most common HAI sites were skin (32%), respiratory tract (29%), and urinary tract (20%). Cellulitis, soft-tissue or wound infection, symptomatic UTI, and cold or pharyngitis were the most common individual HAIs (Fig. 1). Overall HAI prevalence was 2.3 per 100 residents (95% CI, 2.1–2.6); at the NH level, the median HAI prevalence was 1.8 and ranged from 0 to 14.3 (interquartile range, 0–3.1). At the resident level (Fig. 2), HAI prevalence was significantly higher in persons admitted for postacute care with diabetes, with a pressure ulcer, receiving wound care, or with a device. Conclusions: In this large-scale survey, 1 in 43 NH residents had an HAI on a given day. Three HAI types comprised >80% of infections. In addition to identifying characteristics that place residents at higher risk for HAIs, these findings provide important data on HAI epidemiology in NHs that can be used to expand HAI surveillance and inform prevention policies and practices.
Funding: None
Disclosures: None
Appropriateness of Initiating Antibiotics for Urinary Tract Infection Among Nursing Home Residents
- Taniece R. Eure, Nicola D. Thompson, Austin Penna, Wendy M. Bamberg, Grant Barney, Devra Barter, Paula Clogher, Malini DeSilva, Ghinwa Dumyati, Erin Epson, Christina B. Felsen, Linda Frank, Deborah Godine, Lourdes Irizarry, Helen Johnston, Marion A. Kainer, Linda Li, Ruth Lynfield, JP Mahoehney, Joelle Nadle, Valerie L. S. Ocampo, Susan M. Ray, Monika E. Samper, Sarah Shrum Davis, Marla Sievers, Krithika Srinivasan, Lucy E. Wilson, Alexia Y. Zhang, Shelley S. Magill, Nimalie D. Stone
-
- Journal:
- Infection Control & Hospital Epidemiology / Volume 41 / Issue S1 / October 2020
- Published online by Cambridge University Press:
- 02 November 2020, pp. s127-s128
- Print publication:
- October 2020
-
- Article
-
- You have access Access
- Export citation
-
Background: Antibiotics are among the most commonly prescribed drugs in nursing homes; urinary tract infections (UTIs) are a frequent indication. Although there is no gold standard for the diagnosis of UTIs, various criteria have been developed to inform and standardize nursing home prescribing decisions, with the goal of reducing unnecessary antibiotic prescribing. Using different published criteria designed to guide decisions on initiating treatment of UTIs (ie, symptomatic, catheter-associated, and uncomplicated cystitis), our objective was to assess the appropriateness of antibiotic prescribing among NH residents. Methods: In 2017, the CDC Emerging Infections Program (EIP) performed a prevalence survey of healthcare-associated infections and antibiotic use in 161 nursing homes from 10 states: California, Colorado, Connecticut, Georgia, Maryland, Minnesota, New Mexico, New York, Oregon, and Tennessee. EIP staff reviewed resident medical records to collect demographic and clinical information, infection signs, symptoms, and diagnostic testing documented on the day an antibiotic was initiated and 6 days prior. We applied 4 criteria to determine whether initiation of treatment for UTI was supported: (1) the Loeb minimum clinical criteria (Loeb); (2) the Suspected UTI Situation, Background, Assessment, and Recommendation tool (UTI SBAR tool); (3) adaptation of Infectious Diseases Society of America UTI treatment guidelines for nursing home residents (Crnich & Drinka); and (4) diagnostic criteria for uncomplicated cystitis (cystitis consensus) (Fig. 1). We calculated the percentage of residents for whom initiating UTI treatment was appropriate by these criteria. Results: Of 248 residents for whom UTI treatment was initiated in the nursing home, the median age was 79 years [IQR, 19], 63% were female, and 35% were admitted for postacute care. There was substantial variability in the percentage of residents with antibiotic initiation classified as appropriate by each of the criteria, ranging from 8% for the cystitis consensus, to 27% for Loeb, to 33% for the UTI SBAR tool, to 51% for Crnich and Drinka (Fig. 2). Conclusions: Appropriate initiation of UTI treatment among nursing home residents remained low regardless of criteria used. At best only half of antibiotic treatment met published prescribing criteria. Although insufficient documentation of infection signs, symptoms and testing may have contributed to the low percentages observed, adequate documentation in the medical record to support prescribing should be standard practice, as outlined in the CDC Core Elements of Antibiotic Stewardship for nursing homes. Standardized UTI prescribing criteria should be incorporated into nursing home stewardship activities to improve the assessment and documentation of symptomatic UTI and to reduce inappropriate antibiotic use.
Funding: None
Disclosures: None
Protoplanetary Disk Evolution: Singles vs. Binaries
- Sebastian Daemgen, Ray Jayawardhana, Monika G. Petr-Gotzens, Elliot Meyer
-
- Journal:
- Proceedings of the International Astronomical Union / Volume 10 / Issue S314 / November 2015
- Published online by Cambridge University Press:
- 27 January 2016, pp. 135-138
- Print publication:
- November 2015
-
- Article
-
- You have access Access
- Export citation
-
Based on a large number of observations carried out in the last decade it appears that the fraction of stars with protoplanetary disks declines steadily between ~1 Myr and ~10 Myr. We do, however, know that the multiplicity fraction of star-forming regions can be as high as >50% and that multiples have reduced disk lifetimes on average. As a consequence, the observed roughly exponential disk decay can be fully attributed neither to single nor binary stars and its functional form may need revision. Observational evidence for a non-exponential decay has been provided by Kraus et al. (2012), who statistically correct previous disk frequency measurements for the presence of binaries and find agreement with models that feature a constantly high disk fraction up to ~3 Myr, followed by a rapid (≲2 Myr) decline.
We present results from our high angular resolution observational program to study the fraction of protoplanetary disks of single and binary stars separately. We find that disk evolution timescales of stars bound in close binaries (<100 AU) are significantly reduced compared to wider binaries. The frequencies of accretors among single stars and wide binaries appear indistinguishable, and are found to be lower than predicted from planet forming disk models governed by viscous evolution and photoevaporation.