The predictive ability of the Acute Physiology and Chronic Health Evaluation (APACHE II) score for mortality in the Intensive Care Unit in Kimberley hospital
Introduction: The aim of this study was to assess the Acute Physiology and Chronic Health Evaluation (APACHE II) prognostic index in the Intensive Care Unit of Kimberley Hospital Complex (KHC) on admission. The study was more specifically aimed at patients meeting criteria for the Systemic Inflammatory Response Syndrome (SIRS), as patients admitted to KHC ICU frequently meet the criteria and often progress to sepsis, severe sepsis and septic shock. Design: A cohort study on South African patients meeting SIRS criteria, including all races and gender. Setting: Intensive Care Unit of Kimberley Hospital Complex, provincial hospital in the Northern Cape province, South Africa. Patients and measurements: Consecutive patients meeting the criteria for SIRS on admission to ICU between August 2006 and May 2007 were included. For each patient the diagnosis, physiological and chronic health data necessary for the APACHE score was gathered and recorded by the doctor on duty on time of admission. Predicted and actual mortality rates were calculated. Data was provided to the department of Biostatistics of the UFS for processing. Results were summarised by means, standard variations and percentiles (numerical variables) and frequencies and percentages (categorical variables). Results Of the 160 patients included in the study, 59 died (36.9%). Patients discharged from the unit before 14 days were followed up in the ward until 14 days or discharge from hospital (whichever came first). 77 patients were discharged from ICU within 14 days of which 3 (1.9%) died in the ward within the 14-day period. 74 of the discharged patients (46.3%) were alive after 14 days. 24 patients (14%) participating in the trial were still in ICU after 14 days and mortality not recorded. The counting of patients who survived and those who died, for each level of death risk predicted, allowed the calculation of sensitivity, specificity and the percentage of correct predictions for each level of predicted death risk. The sensitivity of the calculated death risk was higher at scores below 8, gradually decreasing as scores increased, reaching 50.9% at score >21. Conversely the specificity increased from 1% for scores <5, reaching 79.2% for death risk at scores >21. The most accurate combination of sensitivity and specificity was found at scores of 16-18, with the positive prediction value ranging from 51.3-54.4% and the negative prediction value ranging from 76.1-77.5%. There was a meaningful connection between APACHE II scores and the mortality rate, for all patients and each diagnostic group. In each successive APACHE II score interval the mortality rate was higher than that of the preceding interval. Thus, the result has confirmed the capability of this index to stratify such patients according to the degree of severity of their health condition. Conclusion The APACHE II scoring system may be usefully applied in Intensive Care Units for predicting mortality, classifying and assessing severity of disease and evaluating performance. It must however be used with caution for planning department resource allocation and decision making regarding admission of patients to Intensive Care.