Comparing Machine Learning to Regression Methods for Mortality Prediction Using Veterans Affairs Electronic Health Record Clinical Data.

BACKGROUND: It is unclear whether machine learning methods yield more accurate electronic health record (EHR) prediction models compared with traditional regression methods. OBJECTIVE: The objective of this study was to compare machine learning and traditional regression models for 10-year mortality prediction using EHR data. DESIGN: This was a cohort study. SETTING: Veterans Affairs (VA) EHR data. PARTICIPANTS: Veterans age above 50 with a primary care visit in 2005, divided into separate training and testing cohorts (n= 124,360 each). MEASUREMENTS AND ANALYTIC METHODS: The primary outcome was 10-year all-cause mortality. We considered 924 potential predictors across a wide range of EHR data elements including demographics (3), vital signs (9), medication classes (399), disease diagnoses (293), laboratory results (71), and health care utilization (149). We compared discrimination (c-statistics), calibration metrics, and diagnostic test characteristics (sensitivity, specificity, and positive and negative predictive values) of machine learning and regression models. RESULTS: Our cohort mean age (SD) was 68.2 (10.5), 93.9% were male; 39.4% died within 10 years. Models yielded testing cohort c-statistics between 0.827 and 0.837. Utilizing all 924 predictors, the Gradient Boosting model yielded the highest c-statistic [0.837, 95% confidence interval (CI): 0.835-0.839]. The full (unselected) logistic regression model had the highest c-statistic of regression models (0.833, 95% CI: 0.830-0.835) but showed evidence of overfitting. The discrimination of the stepwise selection logistic model (101 predictors) was similar (0.832, 95% CI: 0.830-0.834) with minimal overfitting. All models were well-calibrated and had similar diagnostic test characteristics. LIMITATION: Our results should be confirmed in non-VA EHRs. CONCLUSION: The differences in c-statistic between the best machine learning model (924-predictor Gradient Boosting) and 101-predictor stepwise logistic models for 10-year mortality prediction were modest, suggesting stepwise regression methods continue to be a reasonable method for VA EHR mortality prediction model development.

Predicting Life Expectancy to Target Cancer Screening Using Electronic Health Record Clinical Data.

BACKGROUND: Guidelines recommend breast and colorectal cancer screening for older adults with a life expectancy >10 years. Most mortality indexes require clinician data entry, presenting a barrier for routine use in care. Electronic health records (EHR) are a rich clinical data source that could be used to create individualized life expectancy predictions to identify patients for cancer screening without data entry. OBJECTIVE: To develop and internally validate a life expectancy calculator from structured EHR data. DESIGN: Retrospective cohort study using national Veteran's Affairs (VA) EHR databases. PATIENTS: Veterans aged 50+ with a primary care visit during 2005. MAIN MEASURES: We assessed demographics, diseases, medications, laboratory results, healthcare utilization, and vital signs 1 year prior to the index visit. Mortality follow-up was complete through 2017. Using the development cohort (80% sample), we used LASSO Cox regression to select ~100 predictors from 913 EHR data elements. In the validation cohort (remaining 20% sample), we calculated the integrated area under the curve (iAUC) and evaluated calibration. KEY RESULTS: In 3,705,122 patients, the mean age was 68 years and the majority were male (97%) and white (85%); nearly half (49%) died. The life expectancy calculator included 93 predictors; age and gender most strongly contributed to discrimination; diseases also contributed significantly while vital signs were negligible. The iAUC was 0.816 (95% confidence interval, 0.815, 0.817) with good calibration. CONCLUSIONS: We developed a life expectancy calculator using VA EHR data with excellent discrimination and calibration. Automated life expectancy prediction using EHR data may improve guideline-concordant breast and colorectal cancer screening by identifying patients with a life expectancy >10 years.

Do functional status and Medicare claims data improve the predictive accuracy of an electronic health record mortality index? Findings from a national Veterans Affairs cohort.

BACKGROUND: Electronic health record (EHR) prediction models may be easier to use in busy clinical settings since EHR data can be auto-populated into models. This study assessed whether adding functional status and/or Medicare claims data (which are often not available in EHRs) improves the accuracy of a previously developed Veterans Affairs (VA) EHR-based mortality index. METHODS: This was a retrospective cohort study of veterans aged 75 years and older enrolled in VA primary care clinics followed from January 2014 to April 2020 (n = 62,014). We randomly split participants into development (n = 49,612) and validation (n = 12,402) cohorts. The primary outcome was all-cause mortality. We performed logistic regression with backward stepwise selection to develop a 100-predictor base model using 854 EHR candidate variables, including demographics, laboratory values, medications, healthcare utilization, diagnosis codes, and vitals. We incorporated functional measures in a base + function model by adding activities of daily living (range 0-5) and instrumental activities of daily living (range 0-7) scores. Medicare data, including healthcare utilization (e.g., emergency department visits, hospitalizations) and diagnosis codes, were incorporated in a base + Medicare model. A base + function + Medicare model included all data elements. We assessed model performance with the c-statistic, reclassification metrics, fraction of new information provided, and calibration plots. RESULTS: In the overall cohort, mean age was 82.6 years and 98.6% were male. At the end of follow-up, 30,263 participants (48.8%) had died. The base model c-statistic was 0.809 (95% CI 0.805-0.812) in the development cohort and 0.804 (95% CI 0.796-0.812) in the validation cohort. Validation cohort c-statistics for the base + function, base + Medicare, and base + function + Medicare models were 0.809 (95% CI 0.801-0.816), 0.811 (95% CI 0.803-0.818), and 0.814 (95% CI 0.807-0.822), respectively. Adding functional status and Medicare data resulted in similarly small improvements among other model performance measures. All models showed excellent calibration. CONCLUSIONS: Incorporation of functional status and Medicare data into a VA EHR-based mortality index led to small but likely clinically insignificant improvements in model performance.

Comparison of Pharmacy Database Methods for Determining Prevalent Chronic Medication Use.

BACKGROUND: Pharmacy dispensing data are frequently used to identify prevalent medication use as a predictor or covariate in observational research studies. Although several methods have been proposed for using pharmacy dispensing data to identify prevalent medication use, little is known about their comparative performance. OBJECTIVES: The authors sought to compare the performance of different methods for identifying prevalent outpatient medication use. RESEARCH DESIGN: Outpatient pharmacy fill data were compared with medication reconciliation notes denoting prevalent outpatient medication use at the time of hospital admission for a random sample of 207 patients drawn from a national cohort of patients admitted to Veterans Affairs hospitals. Using reconciliation notes as the criterion standard, we determined the test characteristics of 12 pharmacy database algorithms for determining prevalent use of 11 classes of cardiovascular and diabetes medications. RESULTS: The best-performing algorithms included a 180-day fixed look-back period approach (sensitivity, 93%; specificity, 97%; and positive predictive value, 89%) and a medication-on-hand approach with a grace period of 60 days (sensitivity, 91%; specificity, 97%; and positive predictive value, 91%). Algorithms that have been commonly used in previous studies, such as defining prevalent medications to include any medications filled in the prior year or only medications filled in the prior 30 days, performed less well. Algorithm performance was less accurate among patients recently receiving hospital or nursing facility care. CONCLUSION: Pharmacy database algorithms that balance recentness of medication fills with grace periods performed better than more simplistic approaches and should be considered for future studies which examine prevalent chronic medication use.

Frequency of Screening for Colorectal Cancer by Predicted Life Expectancy Among Adults 76-85 Years.

This study assesses whether colorectal cancer screening varied by predicted life expectancy in a national sample of Veterans Affairs patients aged 76 to 85 years.

Estimated GFR Before and After Bariatric Surgery in CKD.

BACKGROUND: Several reviews have recently detailed the beneficial effects of weight loss surgery for kidney function. However, these studies have a number of limitations, including small sample size, few done in chronic kidney disease (CKD) stages 3 and 4, and many not including the main bariatric surgery procedures used in the United States today. STUDY DESIGN: This was an observational retrospective cohort study comparing propensity score-matched bariatric surgery patients and nonsurgery control patients who were referred for, but did not have, surgery. Roux-en-Y gastric bypass (RYGB) and sleeve gastrectomy were also compared using propensity matching. SETTING & PARTICIPANTS: Patients (714 surgery patients; 714 controls) were from a large integrated health care system, a mean of 58±8 (SD) years old, and mostly women (77%) and non-Hispanic whites (56%) and had diabetes mellitus (66%) and/or hypertension (91%). PREDICTOR: Predictors at the time of surgery or referral to surgery were age, sex, race/ethnicity, weight, and presence of diabetes and/or hypertension. OUTCOMES: The primary outcome for this study was change in estimated glomerular filtration rate (eGFR) from serum creatinine level over a median 3-year follow-up period. MEASUREMENTS: Serum creatinine was used to calculate eGFR using the CKD-EPI (CKD Epidemiology Collaboration) creatinine equation. RESULTS: Surgery patients had 9.84 (95% CI, 8.05-11.62) mL/min/1.73m2 greater eGFRs than controls at a median 3 years' follow-up and RYGB patients had 6.60 (95% CI, 3.42-9.78) mL/min/1.73m2 greater eGFRs than sleeve gastrectomy patients during the same period. LIMITATIONS: This study is limited by its nonrandomized observational study design, estimation of GFR, and large changes in muscle mass, which may affect serum creatinine level independent of changes in kidney function. CONCLUSIONS: Bariatric surgery, especially the RYGB procedure, results in significant improvements for up to 3 years in eGFRs for patients with CKD stages 3 and 4.

Which older adults are at highest risk of prescribing cascades? A national study of the gabapentinoid-loop diuretic cascade.

BACKGROUND: Prescribing cascades are important contributors to polypharmacy. Little is known about which older adults are at highest risk of experiencing prescribing cascades. We explored which older veterans are at highest risk of the gabapentinoid (including gabapentin and pregabalin)-loop diuretic (LD) cascade, given the dramatic increase in gabapentinoid prescribing in recent years. METHODS: Using Veterans Affairs and Medicare claims data (2010-2019), we performed a prescription sequence symmetry analysis (PSSA) to assess loop diuretic initiation before and after gabapentinoid initiation among older veterans (≥66 years). To identify the cascade, we calculated the adjusted sequence ratio (aSR), which assesses the temporality of LD relative to gabapentinoid initiation. To explore high-risk groups, we used multivariable logistic regression with prescribing order modeled as a binary dependent variable. We calculated adjusted odds ratios (aORs), measuring the extent to which factors are associated with one prescribing order versus another. RESULTS: Of 151,442 veterans who initiated a gabapentinoid, there were 1,981 patients who initiated a LD within 6 months after initiating a gabapentinoid compared to 1,599 patients who initiated a LD within 6 months before initiating a gabapentinoid. In the gabapentinoid-LD group, the mean age was 73 years, 98% were male, 13% were Black, 5% were Hispanic, and 80% were White. Patients in each group were similar across patient and health utilization factors (standardized mean difference <0.10 for all comparisons). The aSR was 1.23 (95% CI: 1.13, 1.34), strongly suggesting the cascade's presence. People age ≥85 years were less likely to have the cascade (compared to 66-74 years; aOR 0.74, 95% CI: 0.56-0.96), and people taking ≥10 medications were more likely to have the cascade (compared to 0-4 drugs; aOR 1.39, 95% CI: 1.07-1.82). CONCLUSIONS: Among older adults, those who are younger and taking many medications may be at higher risk of the gabapentinoid-LD cascade, contributing to worsening polypharmacy and potential drug-related harms. We did not identify strong predictors of this cascade, suggesting that prescribing cascade prevention efforts should be widespread rather than focused on specific subgroups.

Antihypertensive Medication and Fracture Risk in Older Veterans Health Administration Nursing Home Residents.

Importance: Limited evidence exists on the association between initiation of antihypertensive medication and risk of fractures in older long-term nursing home residents. Objective: To assess the association between antihypertensive medication initiation and risk of fracture. Design, Setting, and Participants: This was a retrospective cohort study using target trial emulation for data derived from 29 648 older long-term care nursing home residents in the Veterans Health Administration (VA) from January 1, 2006, to October 31, 2019. Data were analyzed from December 1, 2021, to November 11, 2023. Exposure: Episodes of antihypertensive medication initiation were identified, and eligible initiation episodes were matched with comparable controls who did not initiate therapy. Main Outcome and Measures: The primary outcome was nontraumatic fracture of the humerus, hip, pelvis, radius, or ulna within 30 days of antihypertensive medication initiation. Results were computed among subgroups of residents with dementia, across systolic and diastolic blood pressure thresholds of 140 and 80 mm Hg, respectively, and with use of prior antihypertensive therapies. Analyses were adjusted for more than 50 baseline covariates using 1:4 propensity score matching. Results: Data from 29 648 individuals were included in this study (mean [SD] age, 78.0 [8.4] years; 28 952 [97.7%] male). In the propensity score-matched cohort of 64 710 residents (mean [SD] age, 77.9 [8.5] years), the incidence rate of fractures per 100 person-years in residents initiating antihypertensive medication was 5.4 compared with 2.2 in the control arm. This finding corresponded to an adjusted hazard ratio (HR) of 2.42 (95% CI, 1.43-4.08) and an adjusted excess risk per 100 person-years of 3.12 (95% CI, 0.95-6.78). Antihypertensive medication initiation was also associated with higher risk of severe falls requiring hospitalizations or emergency department visits (HR, 1.80 [95% CI, 1.53-2.13]) and syncope (HR, 1.69 [95% CI, 1.30-2.19]). The magnitude of fracture risk was numerically higher among subgroups of residents with dementia (HR, 3.28 [95% CI, 1.76-6.10]), systolic blood pressure of 140 mm Hg or higher (HR, 3.12 [95% CI, 1.71-5.69]), diastolic blood pressure of 80 mm Hg or higher (HR, 4.41 [95% CI, 1.67-11.68]), and no recent antihypertensive medication use (HR, 4.77 [95% CI, 1.49-15.32]). Conclusions and Relevance: Findings indicated that initiation of antihypertensive medication was associated with elevated risks of fractures and falls. These risks were numerically higher among residents with dementia, higher baseline blood pressures values, and no recent antihypertensive medication use. Caution and additional monitoring are advised when initiating antihypertensive medication in this vulnerable population.

Effects of residential socioeconomic polarization on high blood pressure among nursing home residents.

OBJECTIVE: Neighborhood concentration of racial, income, education, and housing deprivation is known to be associated with higher rates of hypertension. The objective of this study is to examine the association between tract-level spatial social polarization and hypertension in a cohort with relatively equal access to health care, a Veterans Affairs nursing home. METHODS: 41,973 long-term care residents aged ≥65 years were matched with tract-level Indices of Concentration at the Extremes across four socioeconomic domains. We modeled high blood pressure against these indices controlling for individual-level cardiovascular confounders. RESULTS: We found participants who had resided in the most disadvantaged quintile had a 1.10 (95% 1.01, 1.19) relative risk of high blood pressure compared to those in the other quintiles for the joint measuring race/ethnicity and income domain. CONCLUSIONS: We achieved our objective by demonstrating that concentrated deprivation is associated with worse cardiovascular outcomes even in a population with equal access to care. Measures that jointly consider economic and racial/ethnic polarization elucidate larger disparities than single domain measures.

Complex Patterns of Antihypertensive Treatment Changes in Long-Term Care Residents.

OBJECTIVES: Antihypertensive treatment changes are common in long-term care residents, yet data on the frequency and predictors of changes are lacking. We described the patterns of antihypertensive changes and examined the triggering factors. DESIGN: Retrospective cohort study. SETTING AND PARTICIPANTS: A total of 24,870 Department of Veterans Affairs (VA) nursing home residents aged ≥65 years with long-term stays (≥180 days) from 2006 to 2019. METHODS: We obtained data from the VA Corporate Data Warehouse. Based on Bar Code Medication Administration medication data, we defined 2 types of change events in 180 days of admission: deprescribing (reduced number of antihypertensives or dose reduction of ≥30% compared with the previous week and maintained for at least 2 weeks) and intensification (opposite of deprescribing). Mortality was identified within 2 years after admission. RESULTS: More than 85% of residents were prescribed antihypertensives and 68% of them experienced ≥1 change event during the first 6 months of the nursing home stay. We categorized residents into 10 distinct patterns: no change (27%), 1 deprescribing (11%), multiple deprescribing (5%), 1 intensification (10%), multiple intensification (7%), 1 deprescribing followed by 1 intensification (3%), 1 intensification followed by 1 deprescribing (4%), 3 changes with mixed events (7%), >3 changes with mixed events (10%), and no antihypertensive use (15%). Treatment changes were more frequent in residents with better physical function and/or cognitive function. Potentially triggering factors differed by the type of antihypertensive change: incident high blood pressure and cardiovascular events were associated with intensification, and low blood pressure, weight loss, and falls were associated with deprescribing. Death occurred in 7881 (32%) residents over 2 years. The highest mortality was for those without antihypertensive medication (incidence = 344/1000 person-years). CONCLUSIONS AND IMPLICATIONS: Patterns of medication changes existing in long-term care residents are complex. Future studies should explore the benefits and harms of these antihypertensive treatment changes.

Tutorial: Assessing the impact of nonignorable missingness on regression analysis using Index of Local Sensitivity to Nonignorability.

Data sets with missing observations are common in psychology research. One typically analyzes such data by applying statistical methods that rely on the assumption that the missing observations are missing at random (MAR). This assumption greatly simplifies analysis but is unverifiable from the data at hand, and assuming it incorrectly may lead to bias. Thus we often wish to conduct sensitivity analyses to judge whether conclusions are robust to departures from MAR-that is, whether key findings would hold up even if MAR does not in fact hold. This article describes a class of sensitivity analyses derived from a measure of robustness called the Index of Local Sensitivity to Nonignorability (ISNI). ISNI is straightforward to compute and avoids the estimation of complicated non-MAR missing-data models. The accompanying R package isni implements the method for a range of commonly used regression models; the syntax is simple and similar to that for the regular analysis that assumes MAR. We illustrate the application of the method and software to address the credibility of MAR analyses in a series of analyses of real-world data sets from psychology research. (PsycInfo Database Record (c) 2023 APA, all rights reserved).

Can markers of disease severity improve the predictive power of claims-based multimorbidity indices?

BACKGROUND: Claims-based measures of multimorbidity, which evaluate the presence of a defined list of diseases, are limited in their ability to predict future outcomes. We evaluated whether claims-based markers of disease severity could improve assessments of multimorbid burden. METHODS: We developed 7 dichotomous markers of disease severity which could be applied to a range of diseases using claims data. These markers were based on the number of disease-associated outpatient visits, emergency department visits, and hospitalizations made by an individual over a defined interval; whether an individual with a given disease had outpatient visits to a specialist who typically treats that disease; and ICD-9 codes which connote more versus less advanced or symptomatic manifestations of a disease. Using Medicare claims linked with Health and Retirement Study data, we tested whether including these markers improved ability to predict ADL decline, IADL decline, hospitalization, and death compared to equivalent models which only included the presence or absence of diseases. RESULTS: Of 5012 subjects, median age was 76 years and 58% were female. For a majority of diseases tested individually, adding each of the 7 severity markers yielded minimal increase in c-statistic (≤0.002) for outcomes of ADL decline and mortality compared to models considering only the presence versus absence of disease. Gains in predictive power were more substantial for a small number of individual diseases. Inclusion of the most promising marker in multi-disease multimorbidity indices yielded minimal gains in c-statistics (<0.001-0.007) for predicting ADL decline, IADL decline, hospitalization, and death compared to indices without these markers. CONCLUSIONS: Claims-based markers of disease severity did not contribute meaningfully to the ability of multimorbidity indices to predict ADL decline, mortality, and other important outcomes.

Systolic blood pressure, antihypertensive treatment, and cardiovascular and mortality risk in VA nursing home residents.

BACKGROUND: Optimal systolic BP (SBP) control in nursing home residents is uncertain, largely because this population has been excluded from clinical trials. We examined the association of SBP levels with the risk of cardiovascular (CV) events and mortality in Veterans Affairs (VA) nursing home residents on different numbers of antihypertensive medications. METHODS: Our study included 36,634 residents aged ≥65 years with a VA nursing home stay of ≥90 days from October 2006-June 2019. SBP was averaged over the first week after admission and divided into categories. Cause-specific hazard ratios (HRs) of SBP categories with CV events (primary outcome) and all-cause mortality (secondary outcome) were examined using Cox regression and multistate modeling stratified by the number of antihypertensive medications used at admission (0, 1 or 2, and ≥3 medications). RESULTS: More than 76% of residents were on antihypertensive therapy and 20% received ≥3 medications. In residents on antihypertensive therapy, a low SBP < 110 mmHg (compared with SBP 130 ~ 149 mmHg) was associated with a greater CV risk (adjusted HR [95% confidence interval]: 1.47 [1.28-1.68] in 1 or 2 medications group, and 1.41 [1.19-1.67] in ≥3 medications group). In residents on no antihypertensives, both low SBP < 110 mmHg and high SBP ≥ 150 mmHg were associated with higher mortality; while in residents receiving any antihypertensives, a low SBP was associated with higher mortality and the highest point estimates were for SBP < 110 mmHg (1.36 [1.28-1.45] in 1 or 2 medications group, and 1.47 [1.31-1.64] in ≥3 medications group). CONCLUSIONS: The associations of SBP with CV and mortality risk varied by the intensity of antihypertensive treatment among VA nursing home residents. A low SBP among those receiving antihypertensives was associated with increased CV and mortality risk, and untreated high SBP was associated with higher mortality. More research is needed on the benefits and harms of SBP lowering in long-term care populations.

Development and validation of novel multimorbidity indices for older adults.

BACKGROUND: Measuring multimorbidity in claims data is used for risk adjustment and identifying populations at high risk for adverse events. Multimorbidity indices such as Charlson and Elixhauser scores have important limitations. We sought to create a better method of measuring multimorbidity using claims data by incorporating geriatric conditions, markers of disease severity, and disease-disease interactions, and by tailoring measures to different outcomes. METHODS: Health conditions were assessed using Medicare inpatient and outpatient claims from subjects age 67 and older in the Health and Retirement Study. Separate indices were developed for ADL decline, IADL decline, hospitalization, and death, each over 2 years of follow-up. We validated these indices using data from Medicare claims linked to the National Health and Aging Trends Study. RESULTS: The development cohort included 5012 subjects with median age 76 years; 58% were female. Claims-based markers of disease severity and disease-disease interactions yielded minimal gains in predictive power and were not included in the final indices. In the validation cohort, after adjusting for age and sex, c-statistics for the new multimorbidity indices were 0.72 for ADL decline, 0.69 for IADL decline, 0.72 for hospitalization, and 0.77 for death. These c-statistics were 0.02-0.03 higher than c-statistics from Charlson and Elixhauser indices for predicting ADL decline, IADL decline, and hospitalization, and <0.01 higher for death (p < 0.05 for each outcome except death), and were similar to those from the CMS-HCC model. On decision curve analysis, the new indices provided minimal benefit compared with legacy approaches. C-statistics for both new and legacy indices varied substantially across derivation and validation cohorts. CONCLUSIONS: A new series of claims-based multimorbidity measures were modestly better at predicting hospitalization and functional decline than several legacy indices, and no better at predicting death. There may be limited opportunity in claims data to measure multimorbidity better than older methods.

Medication misuse and overuse in community-dwelling persons with dementia.

BACKGROUND: Persons with dementia (PWD) have high rates of polypharmacy. While previous studies have examined specific types of problematic medication use in PWD, we sought to characterize a broad spectrum of medication misuse and overuse among community-dwelling PWD. METHODS: We included community-dwelling adults aged ≥66 in the Health and Retirement Study from 2008 to 2018 linked to Medicare and classified as having dementia using a validated algorithm. Medication usage was ascertained over the 1-year prior to an HRS interview date. Potentially problematic medications were identified by: (1) medication overuse including over-aggressive treatment of diabetes/hypertension (e.g., insulin/sulfonylurea with hemoglobin A1c < 7.5%) and medications inappropriate near end of life based on STOPPFrail and (2) medication misuse including medications that negatively affect cognition and medications from 2019 Beers and STOPP Version 2 criteria. To contextualize, we compared medication use to people without dementia through a propensity-matched cohort by age, sex, comorbidities, and interview year. We applied survey weights to make our results nationally representative. RESULTS: Among 1441 PWD, median age was 84 (interquartile range = 78-89), 67% female, and 14% Black. Overall, 73% of PWD were prescribed ≥1 potentially problematic medication with a mean of 2.09 per individual in the prior year. This was notable across several domains, including 41% prescribed ≥1 medication that negatively affects cognition. Frequently problematic medications included proton pump inhibitors (PPIs), non-steroidal anti-inflammatory drugs (NSAIDs), opioids, antihypertensives, and antidiabetic agents. Problematic medication use was higher among PWD compared to those without dementia with 73% versus 67% prescribed ≥1 problematic medication (p = 0.002) and mean of 2.09 versus 1.62 (p < 0.001), respectively. CONCLUSION: Community-dwelling PWD frequently receive problematic medications across multiple domains and at higher frequencies compared to those without dementia. Deprescribing efforts for PWD should focus not only on potentially harmful central nervous system-active medications but also on other classes such as PPIs and NSAIDs.

Clinical Outcomes of Intensive Inpatient Blood Pressure Management in Hospitalized Older Adults.

Importance: Asymptomatic blood pressure (BP) elevations are common in hospitalized older adults, and widespread heterogeneity in the clinical management of elevated inpatient BPs exists. Objective: To examine the association of intensive treatment of elevated inpatient BPs with in-hospital clinical outcomes of older adults hospitalized for noncardiac conditions. Design, Setting, and Participants: This retrospective cohort study examined Veterans Health Administration data between October 1, 2015, and December 31, 2017, for patients aged 65 years or older hospitalized for noncardiovascular diagnoses and who experienced elevated BPs in the first 48 hours of hospitalization. Interventions: Intensive BP treatment following the first 48 hours of hospitalization, defined as receipt of intravenous antihypertensives or oral classes not used prior to admission. Main Outcome and Measures: The primary outcome was a composite of inpatient mortality, intensive care unit transfer, stroke, acute kidney injury, B-type natriuretic peptide elevation, and troponin elevation. Data were analyzed between October 1, 2021, and January 10, 2023, with propensity score overlap weighting used to adjust for confounding between those who did and did not receive early intensive treatment. Results: Among 66 140 included patients (mean [SD] age, 74.4 [8.1] years; 97.5% male and 2.6% female; 17.4% Black, 1.7% Hispanic, and 75.9% White), 14 084 (21.3%) received intensive BP treatment in the first 48 hours of hospitalization. Patients who received early intensive treatment vs those who did not continued to receive a greater number of additional antihypertensives during the remainder of their hospitalization (mean additional doses, 6.1 [95% CI, 5.8-6.4] vs 1.6 [95% CI, 1.5-1.8], respectively). Intensive treatment was associated with a greater risk of the primary composite outcome (1220 [8.7%] vs 3570 [6.9%]; weighted odds ratio [OR], 1.28; 95% CI, 1.18-1.39), with the highest risk among patients receiving intravenous antihypertensives (weighted OR, 1.90; 95% CI, 1.65-2.19). Intensively treated patients were more likely to experience each component of the composite outcome except for stroke and mortality. Findings were consistent across subgroups stratified by age, frailty, preadmission BP, early hospitalization BP, and cardiovascular disease history. Conclusions and Relevance: The study's findings indicate that among hospitalized older adults with elevated BPs, intensive pharmacologic antihypertensive treatment was associated with a greater risk of adverse events. These findings do not support the treatment of elevated inpatient BPs without evidence of end organ damage, and they highlight the need for randomized clinical trials of inpatient BP treatment targets.

Synthetic Health Data Can Augment Community Research Efforts to Better Inform the Public During Emerging Pandemics.

The COVID-19 pandemic had disproportionate effects on the Veteran population due to the increased prevalence of medical and environmental risk factors. Synthetic electronic health record (EHR) data can help meet the acute need for Veteran population-specific predictive modeling efforts by avoiding the strict barriers to access, currently present within Veteran Health Administration (VHA) datasets. The U.S. Food and Drug Administration (FDA) and the VHA launched the precisionFDA COVID-19 Risk Factor Modeling Challenge to develop COVID-19 diagnostic and prognostic models; identify Veteran population-specific risk factors; and test the usefulness of synthetic data as a substitute for real data. The use of synthetic data boosted challenge participation by providing a dataset that was accessible to all competitors. Models trained on synthetic data showed similar but systematically inflated model performance metrics to those trained on real data. The important risk factors identified in the synthetic data largely overlapped with those identified from the real data, and both sets of risk factors were validated in the literature. Tradeoffs exist between synthetic data generation approaches based on whether a real EHR dataset is required as input. Synthetic data generated directly from real EHR input will more closely align with the characteristics of the relevant cohort. This work shows that synthetic EHR data will have practical value to the Veterans' health research community for the foreseeable future.

Sliding scale insulin use in a national cohort study of nursing home residents with type 2 diabetes.

BACKGROUND: Guidelines discourage sliding scale insulin (SSI) use after the first week of a nursing home (NH) admission. We sought to determine the prevalence of SSI and identify factors associated with stopping SSI or transitioning to another short-acting insulin regimen. METHODS: In an observational study from October 1, 2013, to June 30, 2017 of non-hospice Veterans Affairs NH residents with type 2 diabetes and an NH admission over 1 week, we compared the weekly prevalence of SSI versus two other short-acting insulin regimens - fixed dose insulin (FDI) or correction dose insulin (CDI, defined as variable SSI given alongside fixed doses of insulin) - from week 2 to week 12 of admission. Among those on SSI in week 2, we examined factors associated with stopping SSI or transitioning to other regimens by week 5. Factors included demographics (e.g., age, sex, race/ethnicity), frailty-related factors (e.g., comorbidities, cognitive impairment, functional impairment), and diabetes-related factors (e.g., HbA1c, long-acting insulin use, hyperglycemia, and hypoglycemia). RESULTS: In week 2, 21% of our cohort was on SSI, 8% was on FDI, and 7% was on CDI. SSI was the most common regimen in frail subgroups (e.g., 18% of our cohort with moderate-severe cognitive impairment was on SSI vs 5% on FDI and 4% on CDI). SSI prevalence decreased steadily from 21% to 16% at week 12 (p for linear trend <0.001), mostly through stopping SSI. Diabetes-related factors (e.g., hyperglycemia) were more strongly associated with continuing SSI or transitioning to a non-SSI short-acting insulin regimen than frailty-related factors. CONCLUSIONS: SSI is the most common method of administering short-acting insulin in NH residents. More research needs to be done to explore why sliding scale use persists weeks after NH admission and explore how we can replace this practice with safer, more effective, and less burdensome regimens.