Anna Aronsson Dannewitz: Optimized prediction of mortality by use of register-based information in an intensive care unit population

Abstract

The primary objective of this thesis was to explore the importance of comorbidity for long-term survival following admission to the intensive care unit (ICU). By use of routinely collected laboratory biomarkers and by utilising a more detailed analysis of the patient’s hospital discharge history, we aim to challenge traditional risk prediction models and measures of comorbidity. We also apply the more comprehensive prediction model to understand how preexisting comorbidities may interact with a critical illness in older ICU patients and compare long-term mortality with the general population. Finally, we aim to understand the importance of comorbidity in relation to socioeconomic status (SES) when the measurement of comorbidity is optimised.

The study was performed by linking data from the Swedish Intensive Care Register (SIR), a hospital clinical laboratory database and several national public authority registers. Routinely collected laboratory biomarkers and quantitative comorbidity measures were compared in Cox regression models adjusting for age, sex and baseline comorbidities with Charlson Comorbidity Index (CCI) and the Simplified Acute Physiology Score (SAPS) version 2 or 3. The mortality in older ICU patients was then compared with individuals from the general population with a landmark of one year. Also, associations between socioeconomic factors and mortality were estimated by using Cox regression models adjusting for age, sex, SAPS3 and baseline comorbidity.

Routinely collected biomarkers discriminate both short- and long-term mortality in general ICU patients, almost as well as the SAPS II. The more comprehensive comorbidity prediction model provides a separation of risk categories within strata of age, the CCI, and intermediate SAPS3 strata. Older patients admitted to the ICU, and who survive the first year after an ICU admission, return to close to the mortality rate of the general population having similar comorbidity. Low educational level was associated with an increased long-term mortality rate after ICU admission. In a Landmark analysis, the association was weaker during the first year after ICU admission than after the first year, suggesting that risk factors other than those specifically related to ICU admission may be important.

By utilising extensive population-based data, the project contributes to the development of methods within registry-based epidemiological research. The importance of specific comorbidities in defined subgroups of intensive care patients should be of interest not only to intensive care specialists but also in a broader healthcare perspective.