Heart failure (HF) hospitalization length of stay (LOS) has been associated with the risk of subsequent readmission and mortality. We identified 19,927 hospitalized patients with HF who were discharged alive from 2008 to 2011 from 3 Kaiser Permanente regions. In adjusted Cox models using LOS 3 to 4 days as the reference category, shorter LOS was not significantly associated with hospital readmissions. LOS of 5 to 10 days was associated with 17% greater risk of readmission within 30 days (hazard ratio [HR] 1.17, 95% confidence interval [CI] 1.07 to 1.28) and 9% greater risk within 1 year (HR 1.09, 95% CI 1.03 to 1.15). LOS of 11 to 29 days was associated with increased readmission risk of 52% at 30 days (HR 1.52, 95% CI 1.30 to 1.76) and 25% at 1 year (HR 1.25, 95% CI 1.16 to 1.35). Mortality risk within 30 days among those with LOS of 1 day was 47% lower (HR 0.53, 95% CI 0.43 to 0.65) and 32% lower at 1 year (HR 0.68, 95% CI 0.62 to 0.74). LOS of 2 days was associated with lower mortality risk of 17% (HR 0.83, 95% CI 0.76 to 0.90) at 1 year. At LOS 5 to 10 days, 30-day and 1-year risk of mortality was increased by 52% (HR 1.52, 95% CI 1.30 to 1.76) and 25% (HR 1.25, 95% CI 1.16 to 1.35), respectively. LOS of 11 to 29 days was associated with 171% higher mortality risk at 30 days (HR 2.71, 95% CI 2.19 to 3.35) and 73% at 1 year (HR 1.73, 95% CI 1.53 to 1.97). Longer LOS during the index HF hospitalization was associated with readmission and mortality within 30 days and 1 year independent of co-morbidities and cardiovascular risk factors. These results suggest that LOS may be a proxy for the severity of HF during the index hospitalization.
Heart failure (HF) is a complex, progressive condition with significant and growing public health and economic burdens. Approximately 20% of US adults will develop HF during their lifetime. By 2030, the total cost of HF is projected to increase markedly by 127% to $70 billion. Much of the cost of HF care is because of hospitalizations and readmissions. Nearly 20% of patients discharged with a principle diagnosis of HF are readmitted within 30 days, whereas 44% are readmitted for any cause within 6 months. A longer length of stay (LOS) during an initial HF hospitalization has been associated with risk of poor outcomes, and it has also been suggested as a predictor of readmission. However, reducing LOS in patients with more severe HF may not reduce risk of readmission or mortality. Therefore, it is important to understand whether LOS is independently associated with hospital readmissions or mortality or is a marker for clinical characteristics that have not been included in previous analyses. We evaluated the importance of HF hospitalization LOS as a predictor of hospital readmission and all-cause mortality within 30 days and 1 year after controlling for variables extracted from patient electronic medical records (EMRs) in a contemporary cohort of patients from an integrated health system operating in 3 distinct regions of the United States.
Methods
The retrospective cohort used in this study comprised members from Kaiser Permanente Southern California, serving approximately 3.6 million subjects of Southern California; Kaiser Permanente Northwest (KPNW), serving approximately 480,000 subjects in the Portland, Oregon, service area; and Kaiser Permanente Georgia, serving approximately 250,000 subjects in the Atlanta, Georgia, metropolitan area. These geographically diverse regions provide care to an ethnically and socioeconomically diverse population. Data on the medical care patients receive are captured through structured administrative and clinical databases and EMRs at each region. A Virtual Data Warehouse (VDW) at each site served as a distributed standardized data resource. The VDW comprised electronic data sets, populated with linked information on demographics, administrative, pharmacy, laboratory, and health care utilization data (including diagnoses and procedures from ambulatory visits and network and non-network hospitalizations). This study was approved by the KPNW Institutional Review Board, and the Kaiser Permanente Southern California and Kaiser Permanente Georgia sites ceded oversight to the KPNW Institutional Review Board. A waiver of informed consent was obtained because of the observational nature of the study.
We identified all members aged ≥18 years of these health systems who had a hospitalization with a primary diagnosis of HF ( International Classification of Diseases, Ninth Edition, Clinical Modification [ ICD-9-CM ], 428.xx) and were discharged alive from January 1, 2008, to December 31, 2011, with no previous hospitalization for HF in the preceding 12 months. Patients with <12 months of health plan membership before the index hospitalization were excluded. We excluded 53 patients with LOS >30 days to avoid undue influence of these uncommon cases on the overall results.
Baseline patient demographic characteristics, medical history, and clinical data were extracted from the VDW during the index hospitalization or within 12 months before the index hospitalization if not available during the hospitalization. Demographic variables included age, gender, and race/ethnicity. We ascertained risk factors for HF including ambulatory systolic and diastolic blood pressures, body mass index, smoking status, LDL cholesterol, and heart rate. History of HF ( ICD-9-CM 428.xx) and co-morbidities including coronary heart disease ( ICD-9-CM 410.x to 414.x), arrhythmia ( ICD-9-CM 427.x), stroke ( ICD-9-CM code 430.x to 432.x and 434.x to 436.x), diabetes mellitus ( ICD-9-CM 250.xx), hypertension ( ICD-9 code 401.x to 405.x), chronic kidney disease ( ICD-9-CM 585.x), and depression ( ICD-9-CM code 296.2 to 296.8 and 311) were extracted from inpatient EMR, outpatient EMR, and claims diagnoses. We characterized baseline kidney function using outpatient serum creatinine concentration values and estimated glomerular filtration rate using the Modification Diet in Renal Disease equation. We also collected other laboratory measures particularly relevant to HF, including sodium, hemoglobin, and B-type natriuretic peptide and HbA1c, fasting glucose, troponin, and C-reactive protein (CRP). Baseline exposure to relevant cardiovascular prescription medications including angiotensin-converting enzyme inhibitors, angiotensin receptor blockers, diuretics, β blockers, calcium channel blockers, digoxin, and diabetic medications were extracted from ambulatory pharmacy records.
We initially evaluated 3 study outcomes: (1) hospital readmission with a primary diagnosis of HF, (2) hospital readmission for any reason, and (3) all-cause mortality, each over 30 days, 6 months, and 1 year after the index hospitalization. However, all associations between LOS and hospital readmission with a primary diagnosis of HF were nonsignificant. Furthermore, results of 6-month and 1-year time periods for all outcomes were nearly identical. Therefore, we focused on the following 4 outcomes: (1) hospital readmission for any reason within 30 days; (2) hospital readmission for any reason within 1 year; (3) all-cause mortality within 30 days; and (4) all-cause mortality within 1 year. Deaths were identified from health plan databases, state death registries, and Social Security Administration Death Master files.
Baseline patient characteristics are described in terms of means and SDs for continuous variables and percentages for categorical variables. We plotted cumulative incidence functions (CIF = 1 − Kaplan-Meier) to calculate unadjusted hospital readmission and all-cause mortality rates. For hospital readmission, the estimated CIF was created using the SAS macro %CIF that implements appropriate nonparametric methods for estimating CIF and accounts for the competing risk of mortality.
The primary variable of interest was LOS (defined as the difference between the discharge and admission dates) during the index hospitalization for HF. Preliminary analyses indicated that the relation between LOS and all outcomes was nonlinear, and inclusion of a quadratic term was not statistically significant. Therefore, we created 5 categories of LOS: 1, 2, 3 to 4, 5 to 10, and 11 to 30 days. To isolate the impact of LOS on the outcomes, we constructed a series of Cox proportional hazards models of the association between LOS categories and risk of readmission and all-cause mortality within 30 days and 1 year, using 3 to 4 days as the reference category. We evaluated the univariate associations between patient characteristics and outcomes. In multivariable analyses, the initial model (model A) included adjustment for variables that are easily obtainable at any given clinic encounter including age, gender, race, Hispanic ethnicity, body mass index, current cigarette smoking, blood pressure, and heart rate. The subsequent model (model B) added use of pharmaceutical agents for the treatment of cardiovascular diseases and diabetes including ACE inhibitors, angiotensin receptor blockers, β blockers, diuretics, digoxin, calcium channel blockers, statins, metformin, sulfonylureas, and insulin and the presence of co-morbidities including existing HF, coronary artery disease, arrhythmia, stroke diabetes, hypertension, chronic kidney disease, and depression. The final model (model C) added routinely measured laboratory data including estimated glomerular filtration rate, LDL cholesterol, sodium, and hemoglobin. Laboratory variables with >25% of the data missing (hemoglobin A1c, glucose, B-type natriuretic peptide, troponin, and CRP) were not considered routinely measured and were not included in the final model. All analyses were conducted using SAS software, version 9.2 (SAS Institute, Cary, North Carolina).
Results
Baseline characteristics of the 19,927 patients with a hospital discharge for HF are presented in Table 1 . Risk factors for HF (blood pressure, heart rate, LDL cholesterol, and cigarette smoking) were well controlled. Nearly 50% of the cohort was prescribed medications for cardiovascular diseases. Co-morbidity burden was high and 57% had a preexisting diagnosis of HF. Use of potentially predictive laboratory results (HbA1c, fasting glucose, B-type natriuretic peptide, troponin, and CRP) was infrequent in this population.
| Variable | Mean (SD) or % | % missing |
|---|---|---|
| Number of patients | 19,927 | — |
| Mean LOS of Index Hospitalization | 3.8 (4.8) | 0 |
| LOS > 7 Days | 12.7% | 0 |
| Age (years) | 73.9 (13.1) | 0 |
| Female | 47.1% | 0 |
| Hispanic | 17.0% | 0 |
| African-American | 15.3% | 0 |
| Body Mass Index (kg/m 2 ) | 30 (8) | 6.4% |
| Current Smoker | 8.0% | 0 |
| Systolic Blood Pressure (mm Hg) | 130 (16) | 3.6% |
| Diastolic Blood Pressure (mm Hg) | 69 (10) | 3.6% |
| Heart Rate (beats per minute) | 75 (13) | 3.7% |
| ACEIs or ARBs | 46.2% | 0 |
| β-Blockers | 50.2% | 0 |
| Diuretics | 47.1% | 0 |
| Digoxin | 10.1% | 0 |
| Insulin | 15.7% | 0 |
| Statins | 47.0% | 0 |
| Metformin | 9.7% | 0 |
| Sulfonylureas | 15.3% | 0 |
| Calcium Channel Blockers | 26.6% | 0 |
| Existing Heart Failure | 57.1% | 0 |
| Coronary Artery Disease | 45.4% | 0 |
| Arrhythymia | 47.7% | 0 |
| Stroke | 8.7% | 0 |
| Diabetes | 46.8% | 0 |
| Hypertension | 85.7% | 0 |
| Chronic Kidney Disease | 62.3% | 0 |
| Depression | 23.3% | 0 |
| eGFR (mL/min/1.73m 2 ) | 59 (25) | 14.2% |
| LDL Cholesterol (mg/dL) | 92 (33) | 22.4% |
| Sodium (mmol/L) | 139 (3) | 20.1% |
| Hemoglobin | 12.6 (1.7) | 24.9% |
| Hemoglobin A1c (%) | 7.2 (1.5) | 49.6% |
| Fasting Glucose (mg/dL) | 119 (45) | 64.0% |
| B-Type Natriuretic Peptide (pg/mL) | 758 (816) | 84.5% |
| Troponin (ng/mL) | 0.10 (.28) | 95.8% |
| C-Reactive Protein (mg/dL) | 2.1 (3.0) | 97.9% |
Within 30 days, 21% of patients had a hospital readmission and 60% had a readmission within 1 year ( Table 2 , Figure 1 ). Within 30 days, 6% of patients had died and by 1 year, 25% of patients had died ( Table 2 , Figure 2 ).
| All-Cause Readmission | All-Cause Mortality | |||
|---|---|---|---|---|
| 30 Days | 1 Year | 30 Days | 1 Year | |
| All Patients (n=19,927) | 20.8% (20.2 – 21.4) | 60.0% (59.3 – 60.6) | 6.3% (6.0 – 6.7) | 25.7% (25.1 – 26.3) |
| Index Length of Stay (days): | ||||
| 1 (n=4,457) | 19.4% (18.2 – 20.5) | 60.9% (59.5 – 62.4) | 3.1% (2.7 – 3.7) | 19.0% (17.9 – 20.2) |
| 2 (n=4,872) | 19.1% (18.0 – 20.2) | 59.9% (58.5 – 61.3) | 5.2% (4.6 – 5.9) | 22.9% (21.7 – 24.1) |
| 3-4 (n=5,748) | 20.5% (19.5 – 21.6) | 59.0% (57.7 – 60.3) | 6.2% (5.6 – 6.9) | 27.2% (26.0 – 28.4) |
| 5-10 (n=4,064) | 23.3% (22.0 – 24.6) | 59.7% (58.2 – 61.3) | 9.4% (8.5 – 10.3) | 31.6% (30.1 – 33.0) |
| 11-30 (n=786) | 28.9% (25.7 – 32.1) | 62.5% (59.0 – 65.8) | 15.6% (13.1 – 18.2) | 39.4% (35.9 – 42.9) |
Stay updated, free articles. Join our Telegram channel
Full access? Get Clinical Tree
