Admission platelet count as a prognostic indicator in intensive care

Introduction

Abnormal platelet counts are common findings in ICU patients. Thrombocytopenia is associated with a poor outcome [1]. Conversely, thrombocytosis may be associated with an improved outcome [2]. We therefore conducted a retrospective observational study in our own unit to investigate this further.

Methods

All patients admitted to the ICU of a large district general hospital (Royal Cornwall Hospital) from January 2002 to April 2005 were included in this retrospective study. We collected data on age, sex, admission category, platelet count, APACHE II score, APACHE II predicted mortality, and hospital mortality. The platelet value was taken as the lowest platelet count obtained within the first 24 hours of ICU admission. The primary outcome was hospital mortality. Statistical analysis was conducted with SPSS version 15.0 using logistical regression models.

Results

A total of 1,767 patients were admitted during the study period. We excluded 119 patients with no recorded platelet data. We found a strong negative correlation between the admission platelet count and mortality, which was significant (P = 0.001, logistic regression). To test this relationship with actual hospital mortality we divided the cohort into deciles of platelet count and plotted the data against mortality. Those with platelet counts below 67 had a mortality rate of 57.2%. This was substantially higher then the remaining deciles (P = 0.0001, Fisher's exact test). We did not demonstrate any significant reduction in mortality in patients with thrombocytosis (P = 0.523, Fisher's exact test). We compared medical versus surgical patients and found that, for any given platelet value, the predicted outcome for surgical patients was better (P = 0.008, t test). We analysed a model that included platelets as an additional indicator for outcome. In binary logistic regression analysis there was a significant association between platelet count and mortality (coefficient = 0.998, CI = 0.996–0.999). This association remained significant in a multiple logistic regression model, which included APACHE II (P < 0.001). A model including both APACHE II and platelet count improved the proportion of deaths correctly predicted from 69.5% with APACHE II alone to 71.3% with platelets included.

Conclusion

We confirmed previous findings that there is a correlation between low platelet counts and adverse outcome, and we have further demonstrated that the correlation between platelet count and predicted mortality exists across the spectrum of platelet values. In addition we have demonstrated a difference in mortality between medical and surgical patients for any given admission platelet values. Finally, we have demonstrated that platelet values provide additional prognostic information above the APACHE II score.