Patients with heart failure (HF) are at high risk of hospitalization or death. The objective of this study was to develop prediction models to identify patients with HF at highest risk for hospitalization or death. Using clinical and administrative databases, we identified 198,460 patients who received care from the Veterans Health Administration and had ≥1 primary or secondary diagnosis of HF that occurred within 1 year before June 1, 2009. We then tracked their outcomes of hospitalization and death during the subsequent 30 days and 1 year. Predictor variables chosen from 6 clinically relevant categories of sociodemographics, medical conditions, vital signs, use of health services, laboratory tests, and medications were used in multinomial regression models to predict outcomes of hospitalization and death. In patients who were in the ≥95th predicted risk percentile, observed event rates of hospitalization or death within 30 days and 1 year were 27% and 80% respectively, compared to population averages of 5% and 31%, respectively. The c-statistics for the 30-day outcomes were 0.82, 0.80, and 0.80 for hospitalization, death, and hospitalization or death, respectively, and 0.82, 0.76, and 0.77, respectively, for 1-year outcomes. In conclusion, prediction models using electronic health records can accurately identify patients who are at highest risk for hospitalization or death. This information can be used to assist care managers in selecting patients for interventions to decrease their risk of hospitalization or death.
The burden of heart failure (HF) is heavy for the 6 million Americans who are afflicted by this pervasive and costly condition. In the Veterans Health Administration (VHA), the 30-day rehospitalization rate for HF is about 16%. Interventions to decrease hospitalizations and deaths related to HF are desirable but often costly. To increase efficiency, it would be useful to identify patients living in the community who are at highest risk of hospitalization or death, so that interventions could be directed toward patients who might benefit most. Accurately stratifying patients according to their risk of hospitalization and death has been challenging. Previous work in this area has focused primarily on identifying patients likely to be rehospitalized compared to identifying those from an outpatient population at risk for hospitalization or death from HF. Models that can accurately identify highest-risk patients for hospitalization or death are needed but unavailable. This article describes statistical models that predict risk for hospitalization or death in veterans who have HF and receive care from the VHA.
Methods
Using inpatient and outpatient files from the VHA’s Corporate Data Warehouse and National Patient Care database, we identified all patients who had ≥1 primary or secondary diagnosis of HF that occurred 1 year before the index date of June 1, 2009. This study included 198,640 patients with HF as defined by the following International Classification of Diseases, Ninth Revision diagnosis codes: 402.01, 402.11, 402.91, 404.01, 404.03, 404.11, 404.13, 404.91, 404.93, and 428.xx.
Outcomes were the first occurrence of hospitalization from any cause or death from any cause within 30 days and 1 year from the index date through May 31, 2010. If a patient died after he or she was hospitalized, we counted only the first event of hospitalization. Deaths that occurred during hospitalizations were not considered an end point. Thus, we did not calculate overall mortality. The VHA vital status file was used to identify vital status and date of death. We identified hospitalizations that occurred in VHA facilities and those that occurred in non-VHA hospitals that were paid for by the VHA. In this study, there were 6 total end points because there were 3 outcomes of hospitalization, imminent death, and hospitalization or death assessed at 30 days and 1 year.
Using the Corporate Data Warehouse and the National Patient Care database, we extracted predictor variables recorded from June 1, 2008 through May 31, 2009 from 6 clinically relevant categories: demographics, medical conditions, vital signs, use of VHA services, laboratory tests, and medications ( Table 1 ). For vital signs and laboratory data we extracted values most proximal to the index date. Medical conditions were classified using hierarchical condition categories, and the Deyo adaptation of the Charlson score was used to measure co-morbidity.
| Predictor Variables | Patients | Percentage of Patients With End Points | |||
|---|---|---|---|---|---|
| Hospitalization (days) | Imminent Death (days) | ||||
| 30 | 365 | 30 | 365 | ||
| All | 100% | 4.2% | 23.6% | 0.9% | 7.1% |
| Demographics | |||||
| Men | 98.1% | 4.2% | 23.6% | 0.9% | 7.1% |
| Age (years) | |||||
| <65 | 29.6% | 5.7% | 31.1% | 0.5% | 3.0% |
| ≥65 | 70.4% | 3.6% | 20.5% | 1.1% | 8.8% |
| Currently married | 54.4% | 3.3% | 19.3% | 0.9% | 7.5% |
| Service connection ≥50% | 24.3% | 5.7% | 31.3% | 0.8% | 5.2% |
| Veterans Health Administration care enrollment priority groups | |||||
| 1 | 21.4% | 5.9% | 31.9% | 0.9% | 5.8% |
| 2, 3 | 13.1% | 4.0% | 22.7% | 0.9% | 7.1% |
| 4 | 8.1% | 7.3% | 35.6% | 1.5% | 10.1% |
| 5 | 34.2% | 4.5% | 25.7% | 0.8% | 6.5% |
| 6–8 | 23.2% | 1.4% | 9.4% | 0.8% | 8.2% |
| Medical diagnoses and co-morbidity index | |||||
| Coronary atherosclerosis | 61.2% | 4.7% | 25.7% | 0.9% | 7.1% |
| Acute myocardial infarction | 2.7% | 11.2% | 47.7% | 1.4% | 6.7% |
| Unstable angina pectoris | 5.0% | 10.0% | 45.9% | 1.0% | 5.2% |
| Respiratory failure | 6.3% | 11.3% | 47.1% | 2.1% | 8.8% |
| Valvular heart disease | 11.7% | 7.2% | 35.0% | 1.1% | 7.3% |
| Hypertension | 81.0% | 4.6% | 25.5% | 0.9% | 6.7% |
| Stroke | 7.3% | 6.9% | 34.9% | 1.3% | 8.3% |
| Renal failure | 25.4% | 7.7% | 36.3% | 1.4% | 8.8% |
| Chronic obstructive pulmonary disease | 30.9% | 6.6% | 32.8% | 1.3% | 8.6% |
| Pneumonia | 8.7% | 11.6% | 48.6% | 2.3% | 8.8% |
| Diabetes mellitus | 49.7% | 4.9% | 27.2% | 0.9% | 7.0% |
| Malnutrition | 1.3% | 12.7% | 49.6% | 3.5% | 11.9% |
| Dementia | 8.3% | 6.4% | 32.6% | 2.1% | 12.4% |
| Functional disease ⁎ | 5.6% | 8.8% | 41.8% | 1.6% | 8.3% |
| Peripheral vascular disease | 20.4% | 7.2% | 35.0% | 1.2% | 7.7% |
| Metastatic cancer | 2.7% | 9.7% | 38.7% | 3.8% | 16.7% |
| Trauma | 13.9% | 8.8% | 43.0% | 1.2% | 6.2% |
| Psychiatric disorders | 9.6% | 7.7% | 39.5% | 1.0% | 6.3% |
| Liver disease | 2.7% | 11.3% | 48.2% | 1.6% | 6.9% |
| Coronary artery bypass grafts | 13.9% | 6.0% | 30.7% | 0.9% | 6.8% |
| Atrial fibrillation | 29.1% | 5.4% | 28.0% | 1.1% | 8.2% |
| Depression | 20.3% | 6.3% | 33.0% | 0.9% | 6.6% |
| Mental disorders | 44.0% | 6.0% | 31.8% | 1.0% | 6.8% |
| Deyo-Charlson index ≥3 | 58.2% | 5.8% | 30.5% | 1.1% | 7.9% |
| Vital signs | |||||
| Systolic blood pressure (mm Hg) | |||||
| <110 | 17.2% | 5.3% | 26.0% | 1.4% | 10.2% |
| >140 | 17.2% | 5.1% | 28.2% | 0.7% | 5.4% |
| Diastolic blood pressure (mm Hg) | |||||
| <60 | 18.7% | 4.8% | 24.8% | 1.3% | 10.3% |
| ≥90 | 5.0% | 5.9% | 31.2% | 0.8% | 4.1% |
| Heart rate (beats/min) | |||||
| <60 | 14.3% | 3.4% | 21.9% | 0.6% | 5.9% |
| >85 | 16.2% | 6.1% | 28.8% | 1.5% | 8.3% |
| Respiration rate (breaths/min) | |||||
| <18 | 22.3% | 3.5% | 20.8% | 0.8% | 6.8% |
| >20 | 6.7% | 6.2% | 29.4% | 2.0% | 11.2% |
| Body mass index (kg/m 2 ) | |||||
| <25 | 16.3% | 6.0% | 29.0% | 1.7% | 10.7% |
| ≥35 | 17.0% | 4.9% | 28.3% | 0.5% | 4.1% |
| Health care usage | |||||
| All hospitalization in previous year | 23.0% | 10.7% | 48.5% | 1.4% | 6.5% |
| Mental health hospitalization in previous year | 10.9% | 11.7% | 51.1% | 1.6% | 6.7% |
| Primary care visits in 1 year | |||||
| 1, 2 | 33.0% | 1.7% | 10.3% | 0.9% | 8.8% |
| ≥3 | 64.8% | 5.5% | 30.2% | 0.9% | 6.2% |
| Emergency visits in 1 year | |||||
| 1 | 15.6% | 5.0% | 31.9% | 0.8% | 5.5% |
| ≥2 | 22.5% | 11.1% | 50.6% | 1.3% | 5.7% |
| Cardiology visits in 1 year | 42.1% | 7.0% | 36.8% | 0.9% | 4.9% |
| Pulmonary visits in 1 year | 11.2% | 8.3% | 41.5% | 1.2% | 6.2% |
| Telephone visits in 1 year | |||||
| 1–3 | 21.6% | 5.4% | 29.3% | 1.0% | 6.9% |
| ≥4 | 12.9% | 6.9% | 36.5% | 1.2% | 7.6% |
| Other nonface-to-face visits in 1 year † | |||||
| 1, 2 | 25.3% | 2.8% | 18.5% | 0.8% | 7.2% |
| ≥3 | 48.4% | 6.6% | 34.1% | 1.0% | 6.5% |
| Mental health visits in 1 year | |||||
| 1–3 | 11.1% | 6.6% | 33.9% | 1.0% | 6.5% |
| ≥4 | 8.6% | 7.5% | 39.6% | 0.5% | 3.9% |
| Other visits in 1 year ‡ | |||||
| 1–3 | 22.8% | 1.8% | 12.7% | 0.8% | 7.6% |
| ≥4 | 56.9% | 6.5% | 34.9% | 1.0% | 6.1% |
| Total outpatient visits in 1 year | |||||
| ≤5 | 24.3% | 0.5% | 3.9% | 0.8% | 9.2% |
| 6–19 | 35.5% | 2.5% | 17.5% | 0.7% | 6.9% |
| ≥20 | 40.2% | 8.0% | 40.9% | 1.1% | 6.1% |
| Emergency visits in 1 month | 7.3% | 15.8% | 56.3% | 2.0% | 6.5% |
| Primary care visits in 1 month | 29.6% | 6.3% | 32.0% | 0.8% | 6.3% |
| Cardiology visits in 1 month | 8.3% | 10.5% | 44.3% | 0.9% | 4.3% |
| Pulmonary visits in 1 month | 1.7% | 10.8% | 47.4% | 1.5% | 6.5% |
| Telephone visits in 1 month | 9.0% | 7.7% | 36.9% | 1.2% | 7.4% |
| Other nonface-to-face visits in 1 month † | |||||
| 1 | 15.5% | 6.4% | 32.5% | 0.9% | 6.2% |
| ≥2 | 8.9% | 11.0% | 44.8% | 1.6% | 6.8% |
| Other visits in 1 month ‡ | |||||
| 1 | 19.6% | 4.9% | 29.8% | 0.6% | 5.6% |
| ≥2 | 18.4% | 10.3% | 45.7% | 1.5% | 6.4% |
| Total outpatient visits in 1 month | |||||
| 1 | 23.6% | 2.7% | 20.2% | 0.6% | 6.5% |
| ≥2 | 38.3% | 8.1% | 39.0% | 1.1% | 6.1% |
| Providers for prescriptions ≥3 | 54.1% | 6.5% | 34.6% | 0.9% | 5.7% |
| Laboratory tests | |||||
| Laboratory test performed in previous year | 89.9% | 4.6% | 25.7% | 0.9% | 6.7% |
| Albumin <2.5 g/dl | 1.3% | 14.3% | 46.0% | 7.0% | 16.8% |
| Serum urea nitrogen >40 mg/dl | 8.0% | 7.0% | 33.6% | 1.9% | 12.9% |
| Creatinine >2 mg/dl | 11.8% | 6.1% | 31.0% | 1.2% | 9.0% |
| Potassium (mEq/L) | |||||
| <3.5 | 2.8% | 8.0% | 34.1% | 1.8% | 8.4% |
| >6 | 0.2% | 8.5% | 36.3% | 2.8% | 9.3% |
| White blood cell count (thousands/mm 3 ) | |||||
| <2 | 1.2% | 5.6% | 30.8% | 0.8% | 5.6% |
| >14 | 2.2% | 8.0% | 35.1% | 2.4% | 10.7% |
Stay updated, free articles. Join our Telegram channel
Full access? Get Clinical Tree
