Adaptation of an NLP system to a new healthcare environment to identify social determinants of health.

Social determinants of health (SDoH) are increasingly important factors for population health, healthcare outcomes, and care delivery. However, many of these factors are not reliably captured within structured electronic health record (EHR) data. In this work, we evaluated and adapted a previously published NLP tool to include additional social risk factors for deployment at Vanderbilt University Medical Center in an Acute Myocardial Infarction cohort. We developed a transformation of the SDoH outputs of the tool into the OMOP common data model (CDM) for re-use across many potential use cases, yielding performance measures across 8 SDoH classes of precision 0.83 recall 0.74 and F-measure of 0.78.

The Association between Cytokines and 365-Day Readmission or Mortality in Adult Cardiac Surgery.

Cardiac surgery results in a multifactorial systemic inflammatory response with inflammatory cytokines, such as interleukin-10 and 6 (IL-10 and IL-6), shown to have potential in the prediction of adverse outcomes including readmission or mortality. This study sought to measure the association between IL-6 and IL-10 levels and 1-year hospital readmission or mortality following cardiac surgery. Plasma biomarkers IL-6 and IL-10 were measured in 1,047 patients discharged alive after isolated coronary artery bypass graft surgery from eight medical centers participating in the Northern New England Cardiovascular Disease Study Group between 2004 and 2007. Readmission status and mortality were ascertained using Medicare, state all-payer claims, and the National Death Index. We evaluated the association between preoperative and postoperative cytokines and 1-year readmission or mortality using Kaplan-Meier estimates and Cox's proportional hazards modeling, adjusting for covariates used in the Society of Thoracic Surgeons 30-day readmission model. The median follow-up time was 1 year. After adjustment, patients in the highest tertile of postoperative IL-6 values had a significantly increased risk of readmission or death within 1 year (HR: 1.38; 95% CI: 1.03-1.85), and an increased risk of death within 1 year of discharge (HR: 4.88; 95% CI: 1.26-18.85) compared with patients in the lowest tertile. However, postoperative IL-10 levels, although increasing through tertiles, were not found to be significantly associated independently with 1-year readmission or mortality (HR: 1.25; 95% CI: .93-1.69). Pro-inflammatory cytokine IL-6 and anti-inflammatory cytokine IL-10 may be postoperative markers of cardiac injury, and IL-6, specifically, shows promise in predicting readmission and mortality following cardiac surgery.

Aluminum toxicity risk reduction as a result of reduced acid deposition in Adirondack lakes and ponds.

In 1990, the US Congress amended the Clean Air Act (CAA) to reduce regional-scale ecosystem degradation from SO x and NO x emissions which have been responsible for acid deposition in regions such as the Adirondack Mountains of New York State. An ecosystem assessment project was conducted from 1994 to 2012 by the Darrin Fresh Water Institute to determine the effect of these emission reduction policies on aquatic systems. The project investigated water chemistry and biota in 30 Adirondack lakes and ponded waters. Although regulatory changes made in response to the 1990 CAA amendments resulted in a reduction of acid deposition within the Adirondacks, the ecosystem response to these reductions is complicated. A statistical analysis of SO4, pH, Al, and DOC data collected during this project demonstrates positive change in response to decreased deposition. The changes in water chemistry also have lowered the risk of Al toxicity to brook trout (Salvelinus fontinalis [Mitchill]), which allowed the re-introduction of this species to Brooktrout Lake from which it had been extirpated. However, pH and labile aluminum (Alim) fluctuate and are not strongly correlated to changes in acid deposition. As such, toxicity to S. fontinalis also is cyclic and provides rationale for the difficulties inherent in re-establishing resident populations in impacted aquatic environments. Overall, aquatic ecosystems of the Adirondacks show a positive response to reduced deposition driven by changes in environmental policy, but the response is more complex and indicates an ecosystem-wide interaction between aquatic and watershed components of the ecosystem.

Information Extraction From Electronic Health Records to Predict Readmission Following Acute Myocardial Infarction: Does Natural Language Processing Using Clinical Notes Improve Prediction of Readmission?

Background Social risk factors influence rehospitalization rates yet are challenging to incorporate into prediction models. Integration of social risk factors using natural language processing (NLP) and machine learning could improve risk prediction of 30-day readmission following an acute myocardial infarction. Methods and Results Patients were enrolled into derivation and validation cohorts. The derivation cohort included inpatient discharges from Vanderbilt University Medical Center between January 1, 2007, and December 31, 2016, with a primary diagnosis of acute myocardial infarction, who were discharged alive, and not transferred from another facility. The validation cohort included patients from Dartmouth-Hitchcock Health Center between April 2, 2011, and December 31, 2016, meeting the same eligibility criteria described above. Data from both sites were linked to Centers for Medicare & Medicaid Services administrative data to supplement 30-day hospital readmissions. Clinical notes from each cohort were extracted, and an NLP model was deployed, counting mentions of 7 social risk factors. Five machine learning models were run using clinical and NLP-derived variables. Model discrimination and calibration were assessed, and receiver operating characteristic comparison analyses were performed. The 30-day rehospitalization rates among the derivation (n=6165) and validation (n=4024) cohorts were 15.1% (n=934) and 10.2% (n=412), respectively. The derivation models demonstrated no statistical improvement in model performance with the addition of the selected NLP-derived social risk factors. Conclusions Social risk factors extracted using NLP did not significantly improve 30-day readmission prediction among hospitalized patients with acute myocardial infarction. Alternative methods are needed to capture social risk factors.

Development of Electronic Health Record-Based Prediction Models for 30-Day Readmission Risk Among Patients Hospitalized for Acute Myocardial Infarction.

Importance: In the US, more than 600 000 adults will experience an acute myocardial infarction (AMI) each year, and up to 20% of the patients will be rehospitalized within 30 days. This study highlights the need for consideration of calibration in these risk models. Objective: To compare multiple machine learning risk prediction models using an electronic health record (EHR)-derived data set standardized to a common data model. Design, Setting, and Participants: This was a retrospective cohort study that developed risk prediction models for 30-day readmission among all inpatients discharged from Vanderbilt University Medical Center between January 1, 2007, and December 31, 2016, with a primary diagnosis of AMI who were not transferred from another facility. The model was externally validated at Dartmouth-Hitchcock Medical Center from April 2, 2011, to December 31, 2016. Data analysis occurred between January 4, 2019, and November 15, 2020. Exposures: Acute myocardial infarction that required hospital admission. Main Outcomes and Measures: The main outcome was thirty-day hospital readmission. A total of 141 candidate variables were considered from administrative codes, medication orders, and laboratory tests. Multiple risk prediction models were developed using parametric models (elastic net, least absolute shrinkage and selection operator, and ridge regression) and nonparametric models (random forest and gradient boosting). The models were assessed using holdout data with area under the receiver operating characteristic curve (AUROC), percentage of calibration, and calibration curve belts. Results: The final Vanderbilt University Medical Center cohort included 6163 unique patients, among whom the mean (SD) age was 67 (13) years, 4137 were male (67.1%), 1019 (16.5%) were Black or other race, and 933 (15.1%) were rehospitalized within 30 days. The final Dartmouth-Hitchcock Medical Center cohort included 4024 unique patients, with mean (SD) age of 68 (12) years; 2584 (64.2%) were male, 412 (10.2%) were rehospitalized within 30 days, and most of the cohort were non-Hispanic and White. The final test set AUROC performance was between 0.686 to 0.695 for the parametric models and 0.686 to 0.704 for the nonparametric models. In the validation cohort, AUROC performance was between 0.558 to 0.655 for parametric models and 0.606 to 0.608 for nonparametric models. Conclusions and Relevance: In this study, 5 machine learning models were developed and externally validated to predict 30-day readmission AMI hospitalization. These models can be deployed within an EHR using routinely collected data.

End-of-Life Spending and Healthcare Utilization Among Older Adults with Chronic Obstructive Pulmonary Disease.

BACKGROUND: End-of-life spending and healthcare utilization among older adults with COPD have not been previously described. METHODS: We examined data on Medicare beneficiaries aged 65 years or older with chronic obstructive pulmonary disease (COPD) who died during the period of 2013-2014. End-of-life measures were retrospectively reviewed for 2 years prior to death. Hospital referral regions (HRRs) were categorized into quintiles of age-sex-race-adjusted overall spending during the last 2 years of life. Geographic quintile variation in spending and healthcare utilization was examined across the continuum. RESULTS: We investigated data on 146,240 decedents with COPD from 306 HRRs. Age-sex-race-adjusted overall spending per decedent during the last 2 years of life varied significantly nationwide ($61,271±$11,639 per decedent; range: $48,288±$3,665 to $79,453±$9,242). Inpatient care accounted for 40.2% of spending ($24,626±$6,192 per decedent). Overall, 82%±4% of decedents were admitted to the hospital for 13.7±3.1 days, and 55%±11% were admitted to an intensive care unit for 5.4±2.5 days. Compared with HRRs in the lowest spending quintile, HRRs in the highest spending quintile had a 1.5-fold longer hospital length of stay. Skilled nursing facilities accounted for 11.6% of spending ($7101±$2403 per decedent), and these facilities were utilized by 38%±7% of decedents for 18.7±4.9 days. Hospice accounted for 10.3% of spending ($6,307±$2,201 per decedent) and was utilized by 47%±9% of decedents for 39.7±14.8 days. Significant geographic variation in hospice utilization existed nationwide. CONCLUSIONS: End-of-life spending and healthcare utilization among older adults with COPD varied substantially nationwide. Decedents with COPD frequently utilized acute care near the end of life. Hospice utilization was higher than expected, with significant geographic disparities.

Are Urinary Biomarkers Better Than Acute Kidney Injury Duration for Predicting Readmission?

BACKGROUND: Acute kidney injury (AKI) is a common complication of cardiac surgery. Postprocedural AKI is a risk factor for 30-day readmission. We sought to examine the association of AKI and kidney injury biomarkers with readmission after cardiac surgery. METHODS: Patients alive at discharge who underwent cardiac surgery from the Translational Research Investigating Biomarker Endpoints-AKI cohort were enrolled from six medical centers in the United States and Canada. AKI duration was defined as the total number of days AKI was present during index admission (no AKI, 1-2, 3-6, and 7+ days). Preoperative and postoperative urinary levels were collected for interleukin-18, neutrophil gelatinase-associated lipocalin, kidney injury molecule-1, liver-fatty-acid-binding protein, cystatin C, microalbumin, creatinine, and albumin-to-creatinine ratio. Readmission and death events were identified through US (Medicare) and Canadian administrative databases at 30 days and 365 days after discharge. RESULTS: Of 968 patients 15.9% were readmitted or died within 30 days of discharge and 35.9% were readmitted or died within 365 days. AKI duration of 3 to 6 days was significantly associated with 30-day readmission or death (adjusted odds ratio, 1.82%; 95% confidence interval, 1.08-3.05). Patients with AKI duration ≥ 7 days had increased odds of readmission or death at both 30 days (adjusted odds ratio, 2.49%; 95% confidence interval, 1.15-5.43) and 365 days (adjusted odds ratio, 3.67%; 95% confidence interval, 1.73-7.79). Urinary biomarkers had no association with readmission and death. CONCLUSIONS: AKI duration ≥ 3 days, and not kidney biomarkers, was strongly associated with readmission or death. These clinical outcomes are potentially due to cardiovascular or hemodynamic causes rather than intrinsic injury to the kidney parenchyma.

Long-term dataset on aquatic responses to concurrent climate change and recovery from acidification.

Concurrent regional and global environmental changes are affecting freshwater ecosystems. Decadal-scale data on lake ecosystems that can describe processes affected by these changes are important as multiple stressors often interact to alter the trajectory of key ecological phenomena in complex ways. Due to the practical challenges associated with long-term data collections, the majority of existing long-term data sets focus on only a small number of lakes or few response variables. Here we present physical, chemical, and biological data from 28 lakes in the Adirondack Mountains of northern New York State. These data span the period from 1994-2012 and harmonize multiple open and as-yet unpublished data sources. The dataset creation is reproducible and transparent; R code and all original files used to create the dataset are provided in an appendix. This dataset will be useful for examining ecological change in lakes undergoing multiple stressors.