ABSTRACT
Background
Clinical guidelines advocate use of validated risk models in patients experiencing heart failure with reduced ejection fraction (HFrEF) to inform prognosis and assist with management. We developed models for worsening HF (WHF) hospitalizations and death within 1 year of incident HFrEF using data available within electronic health records (EHR).
Methods
Adults with incident HFrEF were identified from 2013 to 2022 within an integrated healthcare delivery system. We developed decision tree-based models to estimate risks of WHF hospitalization and death within 1 year of the incident HFrEF date. WHF hospitalizations were ascertained using validated natural language processing algorithms. We evaluated the models using cross-validation and measured final performance (i.e., model discrimination using area under the curve [AUC] and model calibration using the Brier score and calibration plots) on a contemporary hold-out test set of patients from 2021 to 2022.
Results
Among 28,292 adults with incident HFrEF, 17.3% experienced WHF hospitalization and 15.1% all-cause death at 1 year of follow-up. We observed an AUC of 0.698 (95% CI: 0.682-0.714) for WHF hospitalization and 0.849 (95% CI: 0.836-0.861) for death and calibrated with a wide range of predicted risks. In comparison, a claims-based risk score displayed an AUC of 0.577 (95% CI: 0.570-0.606) for WHF hospitalization and a smaller dynamic range. Of patients classified as high risk for WHF hospitalization, only 12.0% were receiving full guideline-directed medical therapy at 6 months after HFrEF diagnosis.
Conclusion
Risk models derived using EHR-based data elements can predict both 1-year WHF hospitalization and all-cause mortality in adults with incident HFrEF more accurately than claims-based approaches. These models can be used to improve population management and better target personalized strategies of care.
Background
Heart failure (HF) with reduced ejection fraction (HFrEF) is 1 of the most rapidly growing chronic conditions and affects more than 3 million adults nationally with a high associated economic burden. HFrEF ranks among the most frequent causes of hospitalization and rehospitalizations in adults, with an annual mortality that ranges from 22 to 30%. ,,, Guideline-directed medical therapies (GDMT) significantly improve survival for many patients with HFrEF. However, gaps persist in the optimal use and dosing of these proven therapies. Patients living with HFrEF require a systematic approach to optimize their management and, if they progress to end-stage HF, appropriately timed referral to an advanced HF specialist to improve outcomes.
Given the heterogeneity and complexity of the HFrEF population, it is critical for physicians to accurately assess risk and provide timely evaluation and treatment across care settings. Clinical practice guidelines recommend integrating risk models into routine practice to provide information on prognosis, enhance outcomes, and guide specialist referrals. However, existing models have faced challenges in internal validity and generalizability across diverse U.S. populations. ,,, With the rising prevalence of HFrEF and the increasing availability of high dimensionality data from electronic health records (EHR), leveraging “big data” techniques hold significant promise for advancing risk prediction models. Natural language processing (NLP) algorithms applied to semi-structured and unstructured EHR data can improve detection of worsening HF (WHF) events with greater diagnostic accuracy than administrative coding alone. In parallel, machine learning (ML) algorithms offers the potential to improve prognostic accuracy at both the individual and population levels. Integrating NLP and ML techniques in large EHR data sets is a critical next step toward improving HF risk stratification informing disease management programs and enhancing patient outcomes.
We aimed to develop an updated risk prediction model using NLP and ML tools, leveraging data routinely captured in the electronic health record (EHR), to identify the subset of patients with incident HFrEF who are at highest risk for worsening heart failure (WHF) hospitalization and all-cause mortality within 1 year, regardless of care setting.
Methods
Source population
Kaiser Permanente Northern CA (KPNC) is a large, integrated health care delivery system with 21 hospitals and >260 outpatient clinics where >4.5 million members currently receive comprehensive (i.e., inpatient, emergency department, and ambulatory) care. The KPNC membership is highly representative of the local surrounding and statewide population with respect to age, sex, race/ethnicity, and socioeconomic status. This study was approved by the KPNC institutional review board, and a waiver of informed consent was obtained, as this was a retrospective, observational data-only study.
Study sample
The initial study sample included all KPNC members aged ≥18 years with their first available left ventricular ejection fraction (LVEF) measurement with a value ≤40% between January 1, 2013 through December 31, 2022. To ensure the population had the clinical syndrome of systolic heart failure, we excluded patients without a qualifying HF encounter at any time before or within 90 days after the initial LVEF ≤40%. Qualifying HF encounters were defined as either a hospitalization with a primary discharge diagnosis code for HF, a hospitalization with a secondary discharge diagnosis code for HF and meeting diagnostic criteria for WHF using NLP algorithms (described below), an emergency department or ambulatory clinic visit with a diagnosis code for HF and meeting criteria for WHF, or the first of ≥3 ambulatory visits with a diagnosis code for HF ( International Classification of Diseases, 9th Edition [ICD-9]: 398.91, 402.x1, 404.01, 404.03, 404.11, 404.13, 404.91, 404.93 and 428.x; and 10th Edition [ICD-10]: I09.81, I11.0, I11.9, I13.0, I13.1, I13.10, I13.11, I13.2, I50, I50.1, I50.2, I50.20, I50.21, I50.22, I50.23, I50.3, I50.30, I50.31, I50.32, I50.33, I50.4, I50.40, I50.41, I50.42, I50.43, I50.9, I97.13). We excluded patients with <12 months of prior continuous KPNC membership and pharmacy benefit to ensure complete capture of baseline covariates, patients with <1 day of follow-up due to in-hospital death or loss of KP membership, and patients with prior end-stage kidney disease (ESKD). ESKD was defined as receipt of chronic dialysis or kidney transplant and was ascertained through a comprehensive health plan ESKD treatment registry. Because qualifying LVEF measurements could be identified within either ambulatory, emergency department, or inpatient encounters, the index date was set as the discharge date from the index encounter (i.e., the same date as LVEF measurement for ambulatory and ED visits, and potentially after the LVEF measurement for hospitalizations).
Follow-up and outcomes
Patients were followed for up to 1 year from the index date of incident HFrEF through December 31, 2023 or until death or health plan disenrollment, if either occurred before the end of the first year of follow-up. Primary outcomes were WHF hospitalization and all-cause death. Death was ascertained using comprehensive information from health plan administrative and clinical databases, member proxy reporting, Social Security Administration vital status files, and state death certificate information.
Identification of WHF hospitalizations
WHF hospitalizations were identified using a combination of symptoms, signs, objective criteria, and changes in HF therapy ascertained from structured and unstructured EHR data using rule-based NLP algorithms. We have previously described and validated this approach in detail. ,, In brief, hospitalizations for WHF were defined as including ≥1 symptom (dyspnea, orthopnea, paroxysmal nocturnal dyspnea, fatigue, weight gain, and/or tachypnea), ≥2 objective findings (tachycardia [heart rate >100 bpm], elevated B-type natriuretic peptide [≥100 ng/L], and/or chest x-ray findings [pulmonary edema, pleural effusion, and/or cardiomegaly]), including ≥1 sign (lower extremity edema, pulmonary rales/wheezing, jugular venous distension, third heart sound [S 3 gallop], hepatomegaly, and/or abdominal swelling), and new administration of intravenous loop diuretics (≥2 doses) and/or new hemodialysis/continuous renal replacement therapy. These diagnostic criteria are based on a standardized definition of inpatient WHF previously developed and validated by a consensus panel of trialists with expertise in clinical endpoint classification in collaboration with the U.S. Food and Drug Administration. , We have previously shown that NLP algorithms based on this definition were highly accurate for identifying WHF compared to diagnosis codes alone, with a sensitivity of 98%, specificity of 95%, positive predictive value of 95%, negative predictive value of 98%, and overall accuracy of 97% for WHF hospitalizations compared to manual adjudication.
Data sources and covariates
We obtained demographic characteristics (i.e., age, sex, self-reported race and Hispanic ethnicity) from health plan databases. We defined relevant comorbidities and cardiac procedures within 5 years before the index date by diagnosis or procedure codes supplemented with laboratory test results, outpatient vital signs, and/or prescribed medications using EHR-based data that was cleaned using standardized procedures and linked at the individual-patient level in the Kaiser Permanente Virtual Data Warehouse as previously described and validated. , Baseline medication use was ascertained using filled outpatient prescriptions within 120 days before the index date (or admission date, if HFrEF was identified while hospitalized). Laboratory values and vital signs were the most recent values within 1 year before the index date. Recent signs and symptoms of HF were identified using the NLP algorithms for WHF in any encounter on or within 30 days before the index date (including the index encounter). Specific echocardiographic parameters were collected from the index LVEF echocardiogram using a separate set of previously described NLP algorithms. If a specific parameter was not recorded within the index echocardiogram, we used the most recent value from any echocardiograms conducted within the prior 2 years.
Statistical analyses
We conducted all analyses using SAS version 9.4 (SAS Institute, Cary, NC) and R version 4.3.1 (R Core Team, 2023). We first compared all characteristics at baseline between patients who were or were not hospitalized with WHF within 1 year using standardized differences derived from Cohen’s D for continuous and binary variables and Cramer’s V for categorical variables. We plotted trends in outpatient GDMT medication use prior to the HFrEF diagnosis (pre-admission or as of 1 day prior to outpatient index echocardiogram), at the time of HFrEF diagnosis (dispensed at discharge or after index echocardiogram), and 6 months after the HFrEF diagnosis across the study period. GDMT medications included angiotensin-converting enzyme inhibitors (ACEi), angiotensin II receptor blockers (ARB), angiotensin/neprilysin inhibitors (ARNi), mineralocorticoid receptor antagonists (MRA), evidence-based beta blockers (bisoprolol, carvedilol, or metoprolol succinate), and sodium-glucose cotransporter-2 inhibitors (SGLT2i).
The data set was split based on the year of HFrEF diagnosis, with 2012-2020 comprising the training set and 2021-2022 the hold-out test set. In the training set, we trained a series of models using 5-fold cross-validation: a regularized logistic regression model, a random forest model, a gradient boosted machine (GBM), and an extreme gradient boosted machine (XGBoost). Hyperparameters for each model type were tuned using a grid search method that tested values for each hyperparameter across a prespecified range to find the optimally performing model. All models were derived using the training set, and model performance was measured using both cross-validation on the training set and on the hold-out test set, to evaluate potential data drift from 2012-2020 to 2021-2022. We calculated variable importance for the best performing model as the relative gains in squared error for each split during decision-tree construction for each variable. All models were trained using the H2O package in R (H2O.ai [2020], version 3.36.1.1). We addressed missing data using approaches tailored to the model type. For the regularized logistic regression, missing values were imputed using the median value. Tree-based models are natively able to handle missing data by placing all observations with missing data in the child node that best minimizes prediction error.
To compare the performance of our models with the performance of an existing model for WHF hospitalization, we used the Readmission Risk Score for Heart Failure ( https://www.readmissionscore.org/heart_failure.php ), a model developed from chart-abstracted data and approved by the National Quality Form for public reporting of hospital-level readmission rates by the Centers for Medicare and Medicaid Services. This model was selected because all data elements can be derived from electronic administrative claims data, which forms the basis of many risk stratification tools currently used in U.S. health systems. The claims-based model was not intended as a direct benchmark for our HF-specific outcomes. Instead, it served as a point of reference to assess potential performance differences between widely used claims-based approaches and more granular EHR-derived models. We obtained coefficients for the logistic regression models from the original paper’s tables and applied the model to the test set to generate predicted risks of 1-year WHF hospitalization and to calculate the AUC and Brier score. We visually examined and compared model calibration of the best performing models from each approach by plotting the mean predicted risk of each outcome with the observed proportion of patients experiencing each outcome within the deciles of the predicted risk.
Finally, we evaluated potential care gaps among those with a high predicted risk of WHF. In the contemporary test set (2021-2022), we categorized patients into low (0-9%), moderate (10-19%), and high (≥20%) risk using the predicted risks of 1-year WHF events from the GBM model. Within each group, we calculated the proportions of GDMT adherence at 6 months after the HFrEF diagnosis as quadruple therapy (ACEi/ARB/ARNi, MRA, evidence-based beta blocker, and SGLT2i), triple therapy (ACEi/ARB/ARNi, MRA, evidence-based beta blocker), any 2 therapies, single therapy, or no therapy. We excluded patients who died prior to 6 months in this analysis.
Results
Patient characteristics
We identified 28,292 eligible patients with new-onset HFrEF (LVEF ≤40%) from January 1, 2013 to December 31, 2022, with 42.2% identified during a HF hospitalization and 57.8% identified following an emergency department or outpatient visit. Baseline characteristics are presented in Table 1 . Mean age was 69.4 ± 14 years, 34.8% were women, and 13.9% identified as Asian or Pacific Islander, 13.9% as Hispanic, and 11.1% as Non-Hispanic Black. The mean LVEF of the cohort was 31 ± 7%. Overall, 4894 (17.3%) patients experienced a WHF hospitalization within 1 year within a median (q1, q3) of 57 (19, 146) days, and 4282 (15.1%) patients died within 1 year within a median (q1, q3) of 97 (34, 205) days. A total of 1514 patients (5.4%) experienced both outcomes within 1 year.
Table 1
Baseline characteristics stratified by worsening heart failure.
| Characteristics | Overall | WHF Hospitalization | No WHF Hospitalization | P-Value | Effect Size |
|---|---|---|---|---|---|
| ( n = 28,292) | ( n = 4894) | ( n = 23,398) | |||
| Demographics | |||||
| Age, years, mean (SD) | 69.4 (14.6) | 72.9 (13.6) | 68.6 (14.7) | <.001 | 0.3 |
| Median (interquartile range) | 70.4 (59.6, 80.6) | 74.8 (64.4, 83.4) | 69.4 (58.9, 79.9) | <.001 | |
| Women, n (%) | 9849 (34.8) | 1758 (35.9) | 8091 (34.6) | .07 | 0.03 |
| Self-reported race, n (%) | <.001 | 0.04 | |||
| American Indian/AK Native | 126 (0.4) | 19 (0.4) | 107 (0.5) | ||
| Asian | 3645 (12.9) | 573 (11.7) | 3072 (13.1) | ||
| Black | 3141 (11.1) | 657 (13.4) | 2484 (10.6) | ||
| HIan/Pacific Islander | 277 (1.0) | 37 (0.8) | 240 (1.0) | ||
| Multi-racial | 1212 (4.3) | 246 (5.0) | 966 (4.1) | ||
| White | 17,056 (60.3) | 2933 (59.9) | 14,123 (60.4) | ||
| Unknown | 2835 (10.0) | 429 (8.8) | 2406 (10.3) | ||
| Hispanic, n (%) | 3922 (13.9) | 694 (14.2) | 3228 (13.8) | .48 | 0.01 |
| Tobacco use, n (%) | 3444 (12.2) | 590 (12.1) | 2854 (12.2) | .78 | 0 |
| Alcohol use, n (%) | 11,500 (40.6) | 1752 (35.8) | 9748 (41.7) | <.001 | 0.12 |
| Illicit drug use, n (%) | 1574 (5.6) | 274 (5.6) | 1300 (5.6) | .91 | 0 |
| Incident HFrEF characteristics | |||||
| Initial LVEF <40% measured while hospitalized | 15,828 (55.9) | 2864 (58.5) | 12,964 (55.4) | <.001 | 0.06 |
| Prior duration of diagnosed heart failure (days from first HF diagnosis to first LVEF <40% measurement) | |||||
| Mean (SD) | 137.9 (416.6) | 191.4 (473.2) | 126.7 (402.9) | <.001 | 0.15 |
| Median (interquartile range) | 1.0 (−1.0, 12.0) | 1.0 (0.0, 62.0) | 1.0 (−1.0, 8.0) | <.001 | |
| Type of initial qualifying HF encounter | <0.001 | 0.11 | |||
| Worsening heart failure hospitalization defined by symptoms, signs, objective criteria, and changes in HF-related therapy | 10,945 (38.7) | 2410 (49.2) | 8535 (36.5) | ||
| Worsening heart failure emergency department visit | 2523 (8.9) | 487 (10.0) | 2036 (8.7) | ||
| Worsening heart failure ambulatory visit | 1841 (6.5) | 324 (6.6) | 1517 (6.5) | ||
| HF hospitalization not meeting specific criteria for WHF but with a primary discharge diagnosis of HF | 986 (3.5) | 155 (3.2) | 831 (3.6) | ||
| ≥3 ambulatory visits with HF diagnoses | 11,997 (42.4) | 1518 (31.0) | 10,479 (44.8) | ||
| Timing of qualifying HF | <.001 | 0.08 | |||
| EF ≥40 prior to first <40 | 3506 (12.4) | 891 (18.2) | 2615 (11.2) | ||
| First HF code ≥1 year prior to first EF | 1006 (3.6) | 190 (3.9) | 816 (3.5) | ||
| HF code ≤1 year prior to first EF | 16,410 (58.0) | 2629 (53.7) | 13,781 (58.9) | ||
| HF code after EF | 7370 (26.0) | 1184 (24.2) | 6186 (26.4) | ||
| Recent Signs and Symptoms of Heart Failure, n (%) | |||||
| Shortness of breath | 18,789 (66.4) | 3363 (68.7) | 15,426 (65.9) | <.001 | 0.06 |
| Paroxysmal nocturnal dyspnea | 3993 (14.1) | 689 (14.1) | 3304 (14.1) | .94 | 0 |
| Orthopnea | 7755 (27.4) | 1371 (28.0) | 6384 (27.3) | .3 | 0.02 |
| Rales | 12,468 (44.1) | 2408 (49.2) | 10,060 (43.0) | <.001 | 0.12 |
| S3 gallop | 574 (2.0) | 94 (1.9) | 480 (2.1) | .56 | 0.01 |
| Lower extremity edema | 14,255 (50.4) | 2775 (56.7) | 11,480 (49.1) | <.001 | 0.15 |
| Jugular vein distension | 6149 (21.7) | 1144 (23.4) | 5005 (21.4) | <.01 | 0.05 |
| Hepatomegaly | 186 (0.7) | 27 (0.6) | 159 (0.7) | .31 | 0.02 |
| Recent weight gain | 2525 (8.9) | 453 (9.3) | 2072 (8.9) | .37 | 0.01 |
| Abdominal swelling | 2288 (8.1) | 455 (9.3) | 1833 (7.8) | <.001 | 0.05 |
| Tachypnea | 5933 (21.0) | 1169 (23.9) | 4764 (20.4) | <.001 | 0.09 |
| Fatigue | 6088 (21.5) | 1102 (22.5) | 4986 (21.3) | .06 | 0.03 |
| Pulmonary edema | 10,452 (36.9) | 1971 (40.3) | 8481 (36.2) | <.001 | 0.08 |
| Cardiomegaly | 9217 (32.6) | 1746 (35.7) | 7471 (31.9) | <.001 | 0.08 |
| Pleural effusion | 11,820 (41.8) | 2186 (44.7) | 9634 (41.2) | <.001 | 0.07 |
| Hypoxia | 8221 (29.1) | 1603 (32.8) | 6618 (28.3) | <.001 | 0.1 |
| Medical History, n (%) | |||||
| Atrial fibrillation or flutter | 9459 (33.4) | 1728 (35.3) | 7731 (33.0) | <.01 | 0.05 |
| Ventricular fibrillation or tachycardia | 620 (2.2) | 86 (1.8) | 534 (2.3) | <.05 | 0.04 |
| Ischemic stroke or transient ischemic attack | 1582 (5.6) | 334 (6.8) | 1248 (5.3) | <.001 | 0.06 |
| Acute myocardial infarction | 4826 (17.1) | 892 (18.2) | 3934 (16.8) | <.05 | 0.04 |
| Coronary artery disease | 8763 (31.0) | 1730 (35.3) | 7033 (30.1) | <.001 | 0.11 |
| Cardiac arrest | 613 (2.2) | 94 (1.9) | 519 (2.2) | .19 | 0.02 |
| Mitral or aortic valvular disease | 3949 (14.0) | 991 (20.2) | 2958 (12.6) | <.001 | 0.21 |
| Peripheral artery disease | 1836 (6.5) | 465 (9.5) | 1371 (5.9) | <.001 | 0.14 |
| Venous thromboembolism | 1920 (6.8) | 405 (8.3) | 1515 (6.5) | <.001 | 0.07 |
| Other thromboembolic events | 164 (0.6) | 42 (0.9) | 122 (0.5) | <.01 | 0.04 |
| Diabetes mellitus | 10,812 (38.2) | 2313 (47.3) | 8499 (36.3) | <.001 | 0.22 |
| Hypertension | 20,581 (72.7) | 3983 (81.4) | 16,598 (70.9) | <.001 | 0.25 |
| Dyslipidemia | 22,114 (78.2) | 4070 (83.2) | 18,044 (77.1) | <.001 | 0.15 |
| Hyperthyroidism | 826 (2.9) | 142 (2.9) | 684 (2.9) | .93 | 0 |
| Hypothyroidism | 3981 (14.1) | 803 (16.4) | 3178 (13.6) | <.001 | 0.08 |
| Chronic liver disease | 1504 (5.3) | 313 (6.4) | 1191 (5.1) | <.001 | 0.06 |
| Chronic lung disease | 8394 (29.7) | 1752 (35.8) | 6642 (28.4) | <.001 | 0.16 |
| COPD | 4008 (14.2) | 991 (20.2) | 3017 (12.9) | <.001 | 0.2 |
| Diagnosed dementia | 1729 (6.1) | 340 (6.9) | 1389 (5.9) | <.01 | 0.04 |
| Diagnosed depression | 4316 (15.3) | 872 (17.8) | 3444 (14.7) | <.001 | 0.08 |
| Hospitalized bleed | 1194 (4.2) | 284 (5.8) | 910 (3.9) | <.001 | 0.09 |
| Arthritis | 7421 (26.2) | 1542 (31.5) | 5879 (25.1) | <.001 | 0.14 |
| Frailty | 1881 (6.6) | 464 (9.5) | 1417 (6.1) | <.001 | 0.13 |
| Hearing impairment | 6284 (22.2) | 1250 (25.5) | 5034 (21.5) | <.001 | 0.1 |
| Visual impairment | 18,818 (66.5) | 3586 (73.3) | 15,232 (65.1) | <.001 | 0.18 |
| Osteoporosis | 2898 (10.2) | 577 (11.8) | 2321 (9.9) | <.001 | 0.06 |
| Cardiac Procedure History, n (%) | |||||
| Coronary artery bypass graft | 1209 (4.3) | 153 (3.1) | 1056 (4.5) | <.001 | 0.07 |
| Percutaneous coronary intervention | 3690 (13.0) | 658 (13.4) | 3032 (13.0) | .36 | 0.01 |
| Implantable cardioverter defibrillator | 538 (1.9) | 71 (1.5) | 467 (2.0) | <.05 | 0.04 |
| Coronary angiography | 8795 (31.1) | 1419 (29.0) | 7376 (31.5) | <.001 | 0.06 |
| Cardiac resynchronization therapy pacemaker | 8 (0.0) | 1 (0.0) | 7 (0.0) | .72 | 0.01 |
| Cardiac resynchronization therapy defibrillator | 68 (0.2) | 10 (0.2) | 58 (0.2) | .57 | 0.01 |
| Pacemaker | 1632 (5.8) | 316 (6.5) | 1316 (5.6) | <.05 | 0.03 |
| Vital Signs, mean (SD) | |||||
| Body mass index, kg/m2 | 29.5 (7.3) | 29.0 (7.2) | 29.6 (7.3) | <.001 | 0.09 |
| Missing, n (%) | 3011 (10.6) | 435 (8.9) | 2576 (11.0) | ||
| Systolic blood pressure, mmHg | 132.5 (21.7) | 132.2 (22.0) | 132.6 (21.6) | .32 | 0.02 |
| Missing, n (%) | 3319 (11.7) | 479 (9.8) | 2840 (12.1) | ||
| Diastolic blood pressure, mmHg | 77.0 (16.0) | 74.6 (15.7) | 77.5 (16.0) | <.001 | 0.18 |
| Missing, n (%) | 3328 (11.8) | 480 (9.8) | 2848 (12.2) | ||
| Heart rate, bpm | 90.3 (21.8) | 89.9 (20.7) | 90.4 (22.1) | .1 | 0.03 |
| Missing, n (%) | 2177 (7.7) | 314 (6.4) | 1863 (8.0) | ||
| Respiratory rate, breaths/min | 20.5 (5.3) | 20.6 (5.3) | 20.5 (5.3) | .3 | 0.02 |
| Missing, n (%) | 10,550 (37.3) | 1563 (31.9) | 8987 (38.4) | ||
| Medications, n (%) | |||||
| Angiotensin-converting enzyme inhibitor | 14,027 (49.6) | 2150 (43.9) | 11,877 (50.8) | <.001 | 0.14 |
| Angiotensin II receptor blocker | 5971 (21.1) | 1046 (21.4) | 4925 (21.0) | .61 | 0.01 |
| Angiotensin-neprilysin inhibitors | 759 (2.7) | 73 (1.5) | 686 (2.9) | <.001 | 0.1 |
| Aldosterone receptor antagonist | 4015 (14.2) | 522 (10.7) | 3493 (14.9) | <.001 | 0.13 |
| Diuretic | 20,611 (72.9) | 3848 (78.6) | 16,763 (71.6) | <.001 | 0.16 |
| Any beta blocker | 21,830 (77.2) | 3605 (73.7) | 18,225 (77.9) | <.001 | 0.1 |
| Evidence-based beta blocker (Bisoprolol, Carvedilol, Metoprolol Succinate) | 13,455 (47.6) | 2151 (44.0) | 11,304 (48.3) | <.001 | 0.09 |
| Other beta blocker | 11,519 (40.7) | 2077 (42.4) | 9442 (40.4) | <.01 | 0.04 |
| Calcium channel blocker | 6524 (23.1) | 1487 (30.4) | 5037 (21.5) | <.001 | 0.2 |
| Alpha blocker | 1580 (5.6) | 364 (7.4) | 1216 (5.2) | <.001 | 0.09 |
| Any anti-hypertensive drugs | 26,748 (94.5) | 4625 (94.5) | 22,123 (94.6) | .89 | 0 |
| Central alpha-2 adrenergic agonists | 513 (1.8) | 127 (2.6) | 386 (1.6) | <.001 | 0.07 |
| Antiarrhythmic drug | 3052 (10.8) | 474 (9.7) | 2578 (11.0) | <.01 | 0.04 |
| Anticoagulant | 8931 (31.6) | 1535 (31.4) | 7396 (31.6) | .74 | 0.01 |
| Antiplatelet drug | 4830 (17.1) | 879 (18.0) | 3951 (16.9) | .07 | 0.03 |
| Non-steroidal anti-inflammatory drugs | 2578 (9.1) | 360 (7.4) | 2218 (9.5) | <.001 | 0.08 |
| Acetaminophen | 5543 (19.6) | 1075 (22.0) | 4468 (19.1) | <.001 | 0.07 |
| Aspirin | 7633 (27.0) | 1190 (24.3) | 6443 (27.5) | <.001 | 0.07 |
| Statins | 18,577 (65.7) | 3340 (68.2) | 15,237 (65.1) | <.001 | 0.07 |
| Other lipid-lowering drugs | 573 (2.0) | 135 (2.8) | 438 (1.9) | <.001 | 0.06 |
| Diabetic therapy | 7970 (28.2) | 1729 (35.3) | 6241 (26.7) | <.001 | 0.19 |
| SGLT-2 inhibitors | 892 (3.2) | 121 (2.5) | 771 (3.3) | <.01 | 0.05 |
| Hydralazine | 2924 (10.3) | 777 (15.9) | 2147 (9.2) | <.001 | 0.2 |
| Vasodilators | 6749 (23.9) | 1416 (28.9) | 5333 (22.8) | <.001 | 0.14 |
| Nitrates | 5866 (20.7) | 1200 (24.5) | 4666 (19.9) | <.001 | 0.11 |
| Laboratory Values, mean (SD) | |||||
| B-type natriuretic peptide (ng/L) | 945.6 (905.8) | 1124.7 (977.3) | 905.5 (884.1) | <.001 | 0.24 |
| Median (interquartile range) | 660.0 (341.0, 1219.0) | 815.0 (450.0, 1458.0) | 622.0 (324.0, 1167.0) | <.001 | |
| Missing, n (%) | 5293 (18.7) | 688 (14.1) | 4605 (19.7) | ||
| Hemoglobin, g/dL | 12.6 (2.2) | 12.0 (2.3) | 12.8 (2.2) | <.001 | 0.35 |
| Missing, n (%) | 831 (2.9) | 100 (2.0) | 731 (3.1) | ||
| Hemoglobin A1C, % | 6.8 (1.7) | 6.9 (1.8) | 6.7 (1.7) | <.001 | 0.11 |
| Missing, n (%) | 9098 (32.2) | 1459 (29.8) | 7639 (32.6) | ||
| Glucose, mg/dL | 122.5 (47.4) | 126.6 (51.2) | 121.6 (46.6) | <.001 | 0.1 |
| Missing, n (%) | 2303 (8.1) | 342 (7.0) | 1961 (8.4) | ||
| High density lipoprotein, mg/dL | 45.6 (15.3) | 45.2 (15.2) | 45.7 (15.3) | .07 | 0.04 |
| Missing, n (%) | 10,164 (35.9) | 1898 (38.8) | 8266 (35.3) | ||
| Low density lipoprotein, mg/dL | 89.9 (38.0) | 86.0 (38.9) | 90.7 (37.8) | <.001 | 0.12 |
| Missing, n (%) | 13,363 (47.2) | 2295 (46.9) | 11,068 (47.3) | ||
| Total cholesterol, mg/dL | 158.0 (47.0) | 154.0 (48.3) | 158.9 (46.6) | <.001 | 0.1 |
| Missing, n (%) | 9970 (35.2) | 1857 (37.9) | 8113 (34.7) | ||
| Triglycerides, mg/dL | 125.0 (91.4) | 122.0 (78.9) | 125.6 (93.7) | <.05 | 0.04 |
| Missing, n (%) | 10,990 (38.8) | 2046 (41.8) | 8944 (38.2) | ||
| Hematocrit, % | 38.7 (6.5) | 37.0 (6.7) | 39.1 (6.4) | <.001 | 0.33 |
| Missing, n (%) | 829 (2.9) | 100 (2.0) | 729 (3.1) | ||
| White blood cells, 1000/uL | 8.3 (4.2) | 8.3 (5.1) | 8.3 (4.0) | .55 | 0.01 |
| Missing, n (%) | 874 (3.1) | 107 (2.2) | 767 (3.3) | ||
| Platelets, 1000/uL | 235.4 (90.2) | 234.1 (94.4) | 235.7 (89.3) | .29 | 0.02 |
| Missing, n (%) | 879 (3.1) | 108 (2.2) | 771 (3.3) | ||
| Prothrombin International Normalized Ratio | 1.4 (0.6) | 1.4 (0.6) | 1.4 (0.6) | .08 | 0.03 |
| Missing, n (%) | 9574 (33.8) | 1418 (29.0) | 8156 (34.9) | ||
| Alanine aminotransferase, units/L | 37.3 (77.9) | 33.8 (63.5) | 38.0 (80.7) | <.001 | 0.06 |
| Missing, n (%) | 5228 (18.5) | 782 (16.0) | 4446 (19.0) | ||
| Serum potassium, mmol/L | 4.2 (0.5) | 4.2 (0.5) | 4.2 (0.5) | .98 | 0 |
| Missing, n (%) | 663 (2.3) | 80 (1.6) | 583 (2.5) | ||
| Serum sodium, mmol/L | 138.3 (3.7) | 138.0 (3.9) | 138.4 (3.6) | <.001 | 0.1 |
| Missing, n (%) | 1865 (6.6) | 226 (4.6) | 1639 (7.0) | ||
| Serum calcium, mmol/L | 8.9 (0.6) | 8.9 (0.7) | 8.9 (0.6) | <.001 | 0.1 |
| Missing, n (%) | 12,055 (42.6) | 1781 (36.4) | 10,274 (43.9) | ||
| Serum phosphate, mmol/L | 3.7 (0.9) | 3.8 (0.9) | 3.7 (0.9) | <.01 | 0.06 |
| Missing, n (%) | 14,737 (52.1) | 2257 (46.1) | 12,480 (53.3) | ||
| Thyroid stimulating hormone, mg/dL | 2.8 (5.7) | 3.0 (7.4) | 2.8 (5.2) | .06 | 0.04 |
| Missing, n (%) | 7790 (27.5) | 1326 (27.1) | 6464 (27.6) | ||
| Serum albumin, mg/dL | 4.1 (45.4) | 3.4 (0.6) | 4.2 (50.9) | .12 | 0.02 |
| Missing, n (%) | 16,565 (58.6) | 2478 (50.6) | 14,087 (60.2) | ||
| Serum creatinine, mg/dL | 1.2 (0.7) | 1.4 (0.9) | 1.2 (0.7) | <.001 | 0.26 |
| Missing, n (%) | 343 (1.2) | 40 (0.8) | 303 (1.3) | ||
| Estimated glomerular filtration rate, mL/min/1.73m2 | 70.1 (24.7) | 61.8 (25.4) | 71.8 (24.2) | <.001 | 0.41 |
| Missing, n (%) | 343 (1.2) | 40 (0.8) | 303 (1.3) | ||
| Blood urea nitrogen, mg/dL | 25.7 (15.4) | 30.3 (17.5) | 24.7 (14.7) | <.001 | 0.35 |
| Missing, n (%) | 3361 (11.9) | 449 (9.2) | 2912 (12.4) | ||
| Urinary dipstick protein excretion, n (%) | <.001 | 0.1 | |||
| Negative | 9735 (34.4) | 1669 (34.1) | 8066 (34.5) | ||
| Trace | 1688 (6.0) | 302 (6.2) | 1386 (5.9) | ||
| 1+ | 3807 (13.5) | 764 (15.6) | 3043 (13.0) | ||
| 2+ | 3256 (11.5) | 742 (15.2) | 2514 (10.7) | ||
| 3+ | 948 (3.4) | 260 (5.3) | 688 (2.9) | ||
| Unknown | 8858 (31.3) | 1157 (23.6) | 7701 (32.9) | ||
| Urine protein-to-creatinine ratio, mg/mg | 1.7 (2.4) | 2.0 (2.7) | 1.6 (2.3) | <.001 | 0.17 |
| Missing, n (%) | 22,897 (80.9) | 3553 (72.6) | 19,344 (82.7) | ||
| Urine albumin-to-creatinine ratio, mg/g | 105.5 (113.5) | 128.3 (114.5) | 99.3 (112.5) | <.001 | 0.26 |
| Missing, n (%) | 18,759 (66.3) | 2875 (58.7) | 15,884 (67.9) | ||
| Echocardiographic Parameters | |||||
| Left ventricular ejection fraction | 31.0 (7.4) | 31.4 (7.3) | 30.9 (7.4) | <.001 | 0.06 |
| Median (q1, q3) | 32.0 (25.0, 37.0) | 32.5 (25.0, 38.0) | 32.0 (25.0, 37.0) | <.001 | |
| Aortic stenosis, n (%) | 2777 (9.8) | 799 (16.3) | 1978 (8.5) | <.001 | 0.24 |
| Aortic stenosis severity | <.001 | 0.11 | |||
| None | 23,292 (82.3) | 3812 (77.9) | 19,480 (83.3) | ||
| Mild | 946 (3.3) | 227 (4.6) | 719 (3.1) | ||
| Mild to Moderate | 222 (0.8) | 66 (1.3) | 156 (0.7) | ||
| Moderate | 511 (1.8) | 140 (2.9) | 371 (1.6) | ||
| Moderate to Severe | 243 (0.9) | 76 (1.6) | 167 (0.7) | ||
| Severe | 731 (2.6) | 261 (5.3) | 470 (2.0) | ||
| Unknown | 2347 (8.3) | 312 (6.4) | 2035 (8.7) | ||
| Aortic valve area | 2.1 (0.8) | 1.9 (0.9) | 2.1 (0.8) | <.001 | 0.23 |
| Missing, n (%) | 12,568 (44.4) | 1863 (38.1) | 10,705 (45.8) | ||
| Aortic valve maximum velocity | 1.5 (0.7) | 1.7 (0.9) | 1.5 (0.7) | <.001 | 0.26 |
| Missing, n (%) | 6467 (22.9) | 918 (18.8) | 5549 (23.7) | ||
| Aortic valve velocity time Integral | 29.5 (17.8) | 33.8 (21.6) | 28.6 (16.6) | <.001 | 0.27 |
| Missing, n (%) | 9508 (33.6) | 1390 (28.4) | 8118 (34.7) | ||
| Mean aortic valve gradient | 6.7 (8.6) | 8.7 (10.8) | 6.3 (8.0) | <.001 | 0.25 |
| Missing, n (%) | 9333 (33.0) | 1346 (27.5) | 7987 (34.1) | ||
| Peak aortic valve gradient | 10.9 (13.5) | 13.9 (16.9) | 10.3 (12.5) | <.001 | 0.24 |
| Missing, n (%) | 6568 (23.2) | 940 (19.2) | 5628 (24.1) | ||
| Bicuspid aortic valve | 240 (0.8) | 56 (1.1) | 184 (0.8) | <.05 | 0.04 |
| Prosthetic valve | 535 (1.9) | 118 (2.4) | 417 (1.8) | <.01 | 0.04 |
| Left ventricular hypertrophy | <.001 | 0.06 | |||
| None | 8675 (30.7) | 1457 (29.8) | 7218 (30.8) | ||
| Mild | 6866 (24.3) | 1310 (26.8) | 5556 (23.7) | ||
| Mild to Moderate | 687 (2.4) | 145 (3.0) | 542 (2.3) | ||
| Moderate | 1725 (6.1) | 394 (8.1) | 1331 (5.7) | ||
| Moderate to Severe | 199 (0.7) | 35 (0.7) | 164 (0.7) | ||
| Severe | 387 (1.4) | 92 (1.9) | 295 (1.3) | ||
| Unknown | 9753 (34.5) | 1461 (29.9) | 8292 (35.4) | ||
| End diastolic diameter | 5.4 (0.9) | 5.3 (0.9) | 5.4 (0.9) | <.001 | 0.1 |
| Missing, n (%) | 4145 (14.7) | 590 (12.1) | 3555 (15.2) | ||
| End diastolic volume | 136.1 (55.7) | 133.5 (55.7) | 136.6 (55.7) | <.01 | 0.06 |
| Missing, n (%) | 10,334 (36.5) | 1651 (33.7) | 8683 (37.1) | ||
| End systolic volume | 90.0 (45.9) | 87.1 (45.4) | 90.6 (46.0) | <.001 | 0.08 |
| Missing, n (%) | 10,597 (37.5) | 1686 (34.5) | 8911 (38.1) | ||
| End systolic diameter | 4.4 (1.0) | 4.3 (1.0) | 4.4 (1.0) | <.001 | 0.12 |
| Missing, n (%) | 10,853 (38.4) | 1763 (36.0) | 9090 (38.8) | ||
| Left ventricular outflow tract diameter | 2.1 (0.2) | 2.1 (0.2) | 2.1 (0.2) | <.001 | 0.08 |
| Missing, n (%) | 9676 (34.2) | 1386 (28.3) | 8290 (35.4) | ||
| Left ventricular outflow tract velocity time integral | 15.4 (5.2) | 15.8 (5.4) | 15.3 (5.2) | <.001 | 0.1 |
| Missing, n (%) | 9180 (32.4) | 1339 (27.4) | 7841 (33.5) |
Stay updated, free articles. Join our Telegram channel
Full access? Get Clinical Tree
