Registries and linkage procedure
Details of the registries and linkage procedures used to construct nationwide cohorts of patients hospitalized for the first time for heart failure have been previously described [10, 11]. Briefly, the data of the Dutch Hospital Discharge Register (HDR), the Dutch Population Register (PR), and the National Cause of Death Register were linked using a unique record identification number based on a combination of birth, sex and postal code (unique for 84% of the population). The PR was used to obtain data on demographic characteristics, HDR was used to identify patients with a hospital admission for heart failure, and cause of death statistics were used to obtain data on causes of death following admission for heart failure [11]. The PR became electronically available from 1995 onwards. Linkage of the registries is therefore possible from 1995 and onwards. For this study data was available from 1995 to 2015. All linkages and analyses were performed in agreement with the privacy legislation in the Netherlands and conforms with the principles outlined in the Declaration of Helsinki [12].
Study population
A prospective cohort of patients with heart failure was built by selecting patients from the HDR with a primary admission for the following International Classification of Diseases (ICD) 9th revision codes for heart failure: 428.0, 428.1, 428.9, 402.01, 402.11 and 402.91. Those with a hospital admission for heart failure in the previous 5 years were excluded to ensure that the admissions for heart failure were, with a high probability, first new onset admissions. To investigate differences in mortality risk over time, two cohorts were created: one cohort containing information about patients admitted for heart failure between 1 Jan 2000 and 31 Dec 2002 (in short: the 2000–02 cohort) and one cohort containing information from patients admitted between 1 Jan 2008 and 31 Dec 2010 (in short: the 2008–10 cohort). For both cohorts, patients were additionally divided in isolated left-sided heart failure (ICD-9: 428.1) and other heart failure (ICD-9: 428.0, 428.9, 402.01, 402.11 and 402.91) to allow evaluation of the value of this ICD subdivision.
Outcomes
The main outcomes were 30-day, 1-year and 5-year overall mortality. Follow-up was defined as time from hospital admission for heart failure to the day the patients died or the end of study period. Cause specific mortality is reported for cardiovascular mortality (separately for heart failure, ischemic heart disease, cerebrovascular disease, and other cardiovascular disease), cancer mortality (separately lung cancer) and respiratory mortality (separately chronic obstructive pulmonary disease (COPD)), and chronic kidney disease/renal failure mortality. All ICD codes used are mentioned in Appendix 1.
Other characteristics
Demographic information comprises age, sex, and marital status. We determined the presence of comorbidity by the Charlson comorbidity index based on previous hospital admissions [13], which is considered a valid measure to estimate comorbidity in clinical research [14]. The mean Charlson comorbidity index was calculated as well as the proportion of patients that had an index score of 1 or more. Data on the duration of the hospital admission was available. No information was available on severity of heart failure at the time of admission, nor was data available to allow for differentiation between heart failure with preserved ejection fraction and reduced ejection fraction.
Validation of heart failure discharge codes
The accuracy of the heart failure discharge codes were assessed in a dedicated validation study. For each precision digit of code ICD-9 code 428, 50 patients of the University Medical Center Utrecht were randomly selected and the medical records of these patients were manually checked for correct discharge ICD-9 code and discharge date. These codes were 428.0 (congestive heart failure, unspecified), 428.1 (left heart failure), and 428.2 (heart failure, unspecified).
Data analysis
Baseline characteristics are presented as absolute numbers and percentages for both the 2000–02 and the 2008–10 cohorts. Secondly, we provided absolute numbers and percentages of all-cause mortality, cardiovascular mortality, cancer mortality, respiratory mortality and renal mortality of patients who died within 30-days, 1-year and 5-years after admission for heart failure in the recent cohort, and presented that by sex. Next, we estimated the 30-days, 1-year and 5-year mortality risk after first admission for heart failure in the 2000–02 and the 2008–10 cohorts and stratified these results by age and sex. Potential differences in mortality in sex and age groups were tested with logistic regression analyses. All analyses were adjusted for the Charlson Comorbidity Index (Table 3). To explore whether change in mortality over time was statistically different between men and women, we added an interaction term between sex and time and compared this model with the model without the interaction term using the likelihood ratio test. Then, we investigated whether a previous hospital admission for overall cardiovascular disease, acute myocardial infarction, or chronic pulmonary disease was associated with increased 30-day, 1-year and 5-year mortality in men and women using Cox proportional hazard models adjusted for age. Finally, we estimated the mortality risks of isolated left-sided heart failure (ICD-9: 428.1) and other heart failure (ICD-9: 428.0, 428.9, 402.01, 402.11 and 402.91), by age and sex. All analyses were performed using SPSS 22.0 (SPSS Inc., Chicago, IL, USA) and a p-value < 0.05 was considered statistically significant.