This survey was a nation-wide population-based register study of patients with IE. Our cohort consisted of patients who have been hospitalized and treated for IE during 1997 to 2007 in Sweden. The Swedish Hospital Discharge Register collects individual data from all hospitals and more than 99% of the discharges in somatic care are covered by the register [9] (link). In this study exposure was defined as a diagnosis of infective endocarditis (IE), and outcome was death or non-death by the end of the follow-up period. Exposure data (IE diagnosis) was collected from the nationwide inpatient register and outcome was then cross-referenced from the Swedish population register. Cases of IE were identified using ICD-10 codes: I33, I38, and I39. In order to refine the resulting mortality rates the cohort has been divided in several categories: native-valve IE, prosthetic-valve IE, intravenous drug users (IVDUs); as well as two age-groups; <65 years and ≥65 years. It can readily be assumed that different patient categories have different mortality risks, and therefore we wanted to isolate the rates specific to a particular patient profile. In addition to exposure (IE diagnosis) and outcome data, we also collected information on whether valve surgery was performed, or not performed, at any time point during follow-up up after initial hospital admission. The surgical operations of interest were identified using ICD-10 codes: FG, FJ, FK, and FM.
For the patients in our cohort we have established the date, but not specific cause, of death (if it occurred before the end of our follow-up period). Therefore it was the all-cause attributable IE mortality rates which were determined. Crude rates were determined by counting the number of deaths in our patient cohort within a specific time period after IE was diagnosed. The crude mortality rates represented the absolute mortality risk in the cohort. Another way to illustrate the absolute mortality risk associated with IE is to look at the Kaplan-Meier survival estimator, where the crude survival rate is 1- crude mortality rate.
In order to explore possible increases in long-term relative mortality risks, the crude mortality rates were then directly standardized using age- and sex- stratified mortality rates from the general population of Sweden as the reference population. This data was available from the Statistics Sweden. The standardized mortality ratio (SMR), was then the ratio between the observed number of deaths and the expected number of deaths, where the expected number of deaths was obtained by multiplying the person-years in the cohort with the age- and sex specific mortality rates in the Swedish general population. Ninety-five percent confidence intervals for the SMR were calculated assuming that the observed number of deaths was Poisson distributed. Comparisons of mortality between early surgery and medical therapy were done within each patient category (native and prosthetic valve IE) using age- and sex- stratified Mantel-Haenszel estimates of the odds ratio. The time trend for the annual incidence and mortality rate of IE was explored in a linear regression model using a quasipoisson distribution and t-test for significance.
For the patients in our cohort we have established the date, but not specific cause, of death (if it occurred before the end of our follow-up period). Therefore it was the all-cause attributable IE mortality rates which were determined. Crude rates were determined by counting the number of deaths in our patient cohort within a specific time period after IE was diagnosed. The crude mortality rates represented the absolute mortality risk in the cohort. Another way to illustrate the absolute mortality risk associated with IE is to look at the Kaplan-Meier survival estimator, where the crude survival rate is 1- crude mortality rate.
In order to explore possible increases in long-term relative mortality risks, the crude mortality rates were then directly standardized using age- and sex- stratified mortality rates from the general population of Sweden as the reference population. This data was available from the Statistics Sweden. The standardized mortality ratio (SMR), was then the ratio between the observed number of deaths and the expected number of deaths, where the expected number of deaths was obtained by multiplying the person-years in the cohort with the age- and sex specific mortality rates in the Swedish general population. Ninety-five percent confidence intervals for the SMR were calculated assuming that the observed number of deaths was Poisson distributed. Comparisons of mortality between early surgery and medical therapy were done within each patient category (native and prosthetic valve IE) using age- and sex- stratified Mantel-Haenszel estimates of the odds ratio. The time trend for the annual incidence and mortality rate of IE was explored in a linear regression model using a quasipoisson distribution and t-test for significance.
Free full text: Click here
