A modified real AdaBoost algorithm to discover intensive care unit subgroups with a poor outcome.

The Intensive Care Unit (ICU) population is heterogeneous. At individual ICUs, the quality of care may vary within subgroups. We investigate whether poor outcomes of an ICU can be traced back to excess deaths in specific patient subgroups, by discovering candidate subgroups, with a modified adaptive decision tree boosting algorithm applied to 80 Dutch ICUs.

Do intensive care data on respiratory infections reflect influenza epidemics?

Objectives: Severe influenza can lead to Intensive Care Unit (ICU) admission. We explored whether ICU data reflect influenza like illness (ILI) activity in the general population, and whether ICU respiratory infections can predict influenza epidemics.

Methods: We calculated the time lag and correlation between ILI incidence (from ILI sentinel surveillance, based on general practitioners (GP) consultations) and percentages of ICU admissions with a respiratory infection (from the Dutch National Intensive Care Registry) over the years 2003-2011.

Identifying types and causes of errors in mortality data in a clinical registry using multiple information systems.

Errors may occur in the registration of in-hospital mortality, making it less reliable as a quality indicator. We assessed the types of errors made in in-hospital mortality registration in the clinical quality registry National Intensive Care Evaluation (NICE) by comparing its mortality data to data from a national insurance claims database. Subsequently, we performed site visits at eleven Intensive Care Units (ICUs) to investigate the number, types and causes of errors made in in-hospital mortality registration.

Performance of risk-adjusted control charts to monitor in-hospital mortality of intensive care unit patients: a simulation study.

Objectives: Increases in case-mix adjusted mortality may be indications of decreasing quality of care. Risk-adjusted control charts can be used for in-hospital mortality monitoring in intensive care units by issuing a warning signal when there are more deaths than expected. The aim of this study was to systematically assess and compare, by computer simulation, expected delay before a warning signal was given for an upward shift in mortality rate in intensive care mortality data by different risk-adjusted control charts.

A comparison of internal versus external risk-adjustment for monitoring clinical outcomes.

Internal and external prognostic models can be used to calculate severity of illness adjusted mortality risks. However, it is unclear what the consequences are of using an external model instead of an internal model when monitoring an institution's clinical performance. Theoretically, using an internal prognostic model is preferred while external models are often more widely available.