There are limited data on acute decompensated heart failure (ADHF) that develops after hospital admission. This study sought to compare patient characteristics, co-morbidities, mortality, and length of stay by timing of ADHF onset. The surveillance component of the Atherosclerosis Risk in Communities study (2005 to 2011) sampled, abstracted, and adjudicated hospitalizations with select International Classification of Disease, Ninth Revision, Clinical Modification discharge codes from 4 United States communities among those aged ≥55 years. We included 5,602 validated ADHF hospitalizations further classified as preadmission or postadmission onset. Vital status was assessed up to 1 year since admission. We estimated multivariate-adjusted associations of in-hospital mortality and 28- and 365-day case fatalities with timing of ADHF onset (postadmission vs preadmission). All analyses were weighted to account for the stratified sampling design. Of 25,862 weighted ADHF hospitalizations, 7% had postadmission onset of ADHF. Patients with postadmission ADHF were more likely to be older, white, and women. The most common primary discharge diagnosis codes for those with postadmission ADHF included diseases of the circulatory or digestive systems or infectious diseases. Short-term mortality among postadmission ADHF was almost 3 times that of preadmission ADHF (in-hospital mortality: odds ratio 2.7, 95% confidence interval 1.9 to 3.9; 28-day case fatality: odds ratio 2.6, 95% confidence interval 1.8 to 3.7). The average hospital stay was almost twice as long among postadmission as preadmission ADHF (9.6 vs 5.0 days). In conclusion, postadmission onset of ADHF is characterized by differences in co-morbidities and worse short-term prognosis, and opportunities for reducing postadmission ADHF occurrence and associated risks need to be studied.
Heart failure (HF) is the most common reason for hospitalization in the United States (US) among those ≥65 years and has a relatively poor prognosis. Studies of hospitalized HF focus mainly on patients who present to the hospital with symptoms of acute decompensated heart failure (ADHF); however, ADHF may develop after admission to the hospital. Postadmission onset of ADHF may be iatrogenic such as precipitated by procedures or surgery, intravenous fluid administration, changes in medications, or it may be precipitated by a co-morbid medical problem such as myocardial infarction (MI), pulmonary embolus, or atrial fibrillation. Although empirical evidence is limited, it is likely that such in-hospital occurrences of ADHF are associated with increased length of hospital stay, morbidity, and mortality. Precipitating factors and clinical outcomes are important in informing HF management, and additional investigation into the patient’s characteristics and outcomes associated with postadmission onset of ADHF may improve our understanding of its causes and opportunities for prevention. In this study, we describe the prevalence, patient characteristics and co-morbidities, and prognosis associated with postadmission onset of ADHF compared with preadmission onset in the community surveillance component of the Atherosclerosis Risk in Communities (ARIC) study. This is the first such population-based study of hospitalized patients with ADHF in the US.
Methods
Beginning in 2005 and ongoing, the ARIC study samples hospital discharges for HF events occurring in all hospitals serving 4 geographically defined US communities (Forsyth County, North Carolina; Jackson, Mississippi; suburbs of Minneapolis, Minnesota; and Washington County, Maryland). Surveillance procedures for event identification, investigation, and classification have been previously described. Briefly, eligible hospital discharges include those among patients ≥55 years residing in 1 of the 4 ARIC communities with an HF-related International Classification of Disease, Ninth Revision, Clinical Modification (ICD-9-CM) code in any position (ICD-9-CM code[s] of 398.91, 402.01, 402.11, 402.91, 404.01, 404.03, 404.11, 404.13, 404.91, 404.93, 415.0, 416.9, 425.4, 428.x, 518.4, or 786.0x). Hospitalizations are randomly sampled within strata defined by targeted HF ICD-9-CM codes, age, gender, race, and community of residence. Sampling probabilities varied by strata and were selected to achieve similar standard errors for HF event rates across strata. Trained personnel abstract medical records for HF symptoms, diagnostic imaging results, levels of cardiac biomarkers, medical history, and other patient characteristics. Abstractors are trained to identify evidence of new or progressive signs and symptoms of HF and determine whether signs and/or symptoms were present at admission or developed after admission. Fully abstracted cases are then classified by physician review or computer algorithm as definite decompensated HF, probable decompensated HF, chronic stable HF, unlikely HF, and unclassifiable. Confirmed HF diagnoses are sent to the National Death Index to obtain vital status information up to 1 year after admission.
From 2005 to 2011, 5,944 events were identified as definite or probable ADHF from 15,655 eligible discharges sampled for HF ascertainment. In this sample, 50% of hospitalizations were reviewed by a physician and 64% of final ADHF diagnoses were made by physician review. We excluded patients transferred to or from another hospital (n = 116), missing information on timing of decompensation onset (n = 89), and race other than black or white (n = 137) for a final unweighted sample of 5,602 events. These hospitalizations were further classified as preadmission or postadmission onset of ADHF. The primary outcomes of this study included in-hospital mortality and 28- and 365-day case fatalities. At the time of this analysis, events through 2010 had been sent for vital status search. Therefore, events occurring in 2011 were not included in the case fatality analyses. Hospital length of stay (LOS) was defined as the time, in days, from hospital admission to discharge. All covariates were abstracted from the medical record by trained personnel and included patient demographics (age, gender, and race), insurance status, current smoking, clinical measurements (body mass index and serum creatinine), previous diagnosis or hospitalization for HF, systolic HF (ejection fraction <50%), medical history (hypertension, diabetes, chronic obstructive pulmonary disease, MI, coronary artery disease, atrial fibrillation, valvular heart disease, and dialysis), medications (before hospitalization or progression of in-hospital HF), and discharge diagnoses (up to 26 listed). Primary discharge diagnoses (in first position) were grouped according to chapters of the ICD-9-CM. Furthermore, for each hospitalization, the entire list of discharge diagnoses was scanned for relevant codes and grouped into co-morbid conditions (e.g., atrial fibrillation, chronic kidney disease) and cardiac procedures (i.e., coronary artery bypass graft and percutaneous coronary intervention) according to the Centers for Medicare and Medicaid Services Chronic Conditions Data Warehouse definitions and Agency for Healthcare Research and Quality’s inpatient quality indicators.
Statistical analyses accounted for the stratified sampling design of ARIC HF surveillance and weighted by the inverse of the sampling probability. The Rao-Scott chi-square test statistics were used to compute p values for comparing covariates with timing of ADHF onset (postadmission vs preadmission). Weighted logistic regression was used to estimate odds ratios with 95% confidence intervals of in-hospital mortality and case fatality with timing of ADHF onset; controlling for potential confounders of age, gender, race; HF type and history; and medical history of smoking, hypertension, diabetes, chronic obstructive pulmonary disease, MI, coronary artery disease, atrial fibrillation, valvular heart disease, and dialysis. Age-adjusted survival curves stratified by ADHF onset were constructed based on a weighted Cox proportional hazards model. Extreme values of hospital LOS were truncated at 300 days (n = 5). Due to positive skewness, LOS was log (base 10) transformed, and geometric means and 95% confidence interval were estimated by timing of ADHF onset. Statistical analyses were conducted using survey procedures in SAS version 9.3 (SAS Institute Inc., Cary, North Carolina).
Results
The weighted sample represented 25,862 ADHF hospitalizations. Of these, 1,902 (7%, 377 unweighted) had postadmission onset of ADHF. The median days from hospital admission to postadmission onset were 2 (interquartile range 1 to 5 days). Patients with postadmission ADHF were more likely to be older and white women ( Table 1 ). Also, among patients with postadmission ADHF, a lower proportion had a documented previous diagnosis of HF (57%) compared with preadmission ADHF (73%). In contrast, patients with preadmission ADHF were more likely to have a documented history of smoking, chronic obstructive pulmonary disease, and valvular heart disease. Use of certain medications before hospitalization or progression of HF was mostly similar regardless of the timing of ADHF onset; however, administration of intravenous inotropes was more common in patients with postadmission ADHF.
| Variable | Preadmission ADHF (N ∗ = 23,959) | Postadmission ADHF (N ∗ = 1,902) | p-Value |
|---|---|---|---|
| Age 55–69 (years) | 7,415 (31%) | 392 (21%) | <.0001 |
| 70–79 | 6,774 (28%) | 496 (26%) | |
| 80+ | 9,770 (41%) | 1,015 (53%) | |
| Black female | 3,805 (16%) | 265 (14%) | <.0001 |
| Black male | 3,511 (15%) | 117 (6%) | |
| White female | 8,675 (36%) | 847 (45%) | |
| White male | 7,968 (33%) | 673 (35%) | |
| Private insurance or Medicare | 21,177 (89%) | 1,718 (92%) | 0.10 |
| Systolic HF (ejection fraction <50%) | 12,200 (56%) | 918 (55%) | 0.63 |
| Body mass index † (kg/m 2 , mean [SE]) | 29.3 (0.1) | 28.4 (0.5) | <.0001 |
| Serum creatinine (mg/dL, mean [SE]) | 2.1 (0.03) | 2.2 (0.08) | 0.70 |
| Prior diagnosis of HF | 17,518 (73%) | 1,082 (57%) | <.0001 |
| Prior hospitalization for HF | 9,071 (38%) | 391 (21%) | <.0001 |
| Current smoker | 3,360 (14%) | 141 (7%) | 0.001 |
| Hypertension | 20,328 (85%) | 1,643 (86%) | 0.52 |
| Diabetes mellitus | 11,484 (48%) | 857 (45%) | 0.35 |
| Chronic obstructive pulmonary disease | 8,655 (36%) | 423 (22%) | <.0001 |
| Myocardial infarction | 6,514 (27%) | 548 (29%) | 0.57 |
| Coronary artery disease | 11,142 (47%) | 853 (45%) | 0.58 |
| Atrial fibrillation | 8,947 (37%) | 660 (35%) | 0.38 |
| Valvular heart disease | 6,194 (26%) | 391 (21%) | 0.04 |
| Dialysis | 1,692 (7%) | 96 (5%) | 0.16 |
| Medications prior to hospitalization or progression of HF | |||
| Angiotensin-converting enzyme inhibitor or angiotensin receptor blocker | 12,001 (50%) | 836 (44%) | 0.04 |
| Beta blockers | 15,948 (67%) | 1,243 (65%) | 0.67 |
| Diuretics, oral or intravenous | 21,925 (92%) | 1,764 (93%) | 0.45 |
| Intravenous diuretics | 19,336 (81%) | 1,561 (82%) | 0.56 |
| Intravenous inotropes | 1,468 (6%) | 203 (11%) | 0.001 |
∗ n is weighted as discussed in the Methods section.
† Extreme values of body mass index (<15.7 kg/m 2 [first percentile] or >58.9 kg/m 2 [ninety-ninth percentile]) coded to missing (overall 18% missing).
Patients with both preadmission and postadmission ADHF had diseases of the circulatory system listed as the leading primary discharge diagnosis (63% and 22%, respectively; Table 2 ). Diseases of the respiratory system (13%) and infectious diseases (13%) were also common among patients with preadmission ADHF. As expected, we observed a wider distribution of primary discharge diagnoses among those with postadmission ADHF, such as infectious diseases (16%), digestive diseases (15%), and injuries and poisonings (10%). Furthermore, patients with postadmission ADHF were more likely to have an acute MI, renal failure, gastrointestinal hemorrhage, and a cardiac procedure (coronary artery bypass graft or percutaneous coronary intervention) during the hospitalization.
| Primary Discharge Diagnosis Grouped by ICD-9-CM Chapter | Preadmission ADHF (N ∗ = 23,582) | Postadmission ADHF (N ∗ = 1,866) | p-Value |
|---|---|---|---|
| Infectious and parasitic diseases (001–139) | 2,973 (13%) | 301 (16%) | <.0001 † |
| Neoplasms (140–239) | 193 (1%) | 88 (5%) | |
| Endocrine, nutritional and metabolic diseases, and immunity disorders (240–279) | 347 (1%) | 96 (5%) | |
| Diseases of the blood and blood-forming organs (280–289) | 77 (0%) | 16 (1%) | |
| Mental disorders (290–319) | 46 (0%) | 2 (0%) | |
| Diseases of the nervous system and sense organs (320–389) | 39 (0%) | 16 (1%) | |
| Diseases of the circulatory system (390–459) | 14,755 (63%) | 415 (22%) | |
| Diseases of the respiratory system (460–519) | 3,069 (13%) | 120 (6%) | |
| Diseases of the digestive system (520–579) | 415 (2%) | 277 (15%) | |
| Diseases of the genitourinary system (580–629) | 718 (3%) | 123 (7%) | |
| Diseases of the skin and subcutaneous tissue (680–709) | 75 (0%) | 51 (3%) | |
| Diseases of the musculoskeletal system and connective tissue (710–739) | 73 (0%) | 124 (7%) | |
| Symptoms, signs, and ill-defined conditions (780–799) | 435 (2%) | 34 (2%) | |
| Injury and poisoning (800–999) | 324 (1%) | 179 (10%) | |
| Supplementary classification of factors influencing health status and contact with health services (V01–V89) | 12 (0%) | 10 (1%) | |
| Supplementary classification of external cause of injury and poisoning (E800–E999) | 4 (0%) | — | |
| Procedures (00–99) | 28 (0%) | 14 (1%) | |
| Comorbid conditions and procedures ‡ (ICD-9-CM code(s)) | |||
| Ischemic heart disease | 12,867 (54%) | 1,064 (56%) | 0.55 |
| Acute myocardial infarction (410.x) | 2,080 (9%) | 1,501 (21%) | <.0001 |
| Atrial fibrillation (427.31) | 9,566 (40%) | 808 (42%) | 0.39 |
| Chronic kidney disease | 8,815 (37%) | 792 (42%) | 0.10 |
| Renal failure (584.x or 586.x) | 3,242 (14%) | 506 (27%) | <.0001 |
| Chronic obstructive pulmonary disease and bronchiectasis | 5,385 (22%) | 293 (15%) | 0.008 |
| Pneumonia | 3,097 (13%) | 244 (13%) | 0.96 |
| Gastrointestinal hemorrhage | 466 (2%) | 123 (6%) | <.0001 |
| Hip or pelvic fracture | 51 (0%) | 29 (2%) | <.0001 |
| Any procedure | 6,456 (27%) | 957 (50%) | <.0001 |
| Operation on cardiovascular system (35.x–39.x) | 4,958 (21%) | 652 (34%) | <.0001 |
| Coronary artery bypass graft or percutaneous coronary intervention | 536 (2%) | 105 (6%) | 0.005 |
∗ n is weighted as discussed in the Methods section.
† Calculated by grouping ICD-9-CM chapters on diseases of the blood and blood-forming organs; mental disorders; diseases of the nervous system and sense organs; diseases of the skin and subcutaneous tissue; symptoms, signs, and ill-defined conditions; supplementary classification of factors influencing health status and contact with health services; supplementary classification of external cause of injury and poisoning; and procedures to eliminate zero and low frequencies.
‡ Presence of discharge diagnosis or procedure ICD-9-CM code in any position; diagnosis codes based on Centers for Medicare and Medicaid Services Chronic Condition Warehouse categories; pneumonia, gastrointestinal hemorrhage, and cardiac procedure codes based on Agency for Healthcare Research and Quality’s inpatient quality indicators.
Overall, in-hospital mortality for patients with ADHF was 7%, 28-day case fatality was 12%, and 365-day case fatality was 37%. Patients with postadmission ADHF had substantially worse short-term prognosis; 16% of these patients did not survive to hospital discharge and 27% died within 28 days (compared with 6% and 11%, respectively, among preadmission ADHF). Furthermore, 45% of postadmission ADHF were deceased by 1 year since admission (compared with 37% among postadmission ADHF). After adjusting for potential confounders, we observed strong relative differences between postadmission and preadmission ADHF in in-hospital mortality and 28-day case fatality ( Table 3 ). The association of 365-day case fatality with postadmission onset was much weaker, although still statistically significant. The age-adjusted survival curves show a sharp, early divergence between preadmission and postadmission onset of ADHF, whereas no further divergence after 2 to 3 months since admission ( Figure 1 ).
