Objective: In certain physiologic systems, disease is associated with a loss of complexity in system's output. We test the hypothesis that, in critically ill patients, there is an inverse relation between the complexity of the temperature curve and the clinical status. We also consider whether complexity analysis of the temperature curve may have prognostic value.
Design: Prospective, observational study.
Setting: Intensive care unit of a general hospital in Madrid, Spain.
Patients: Twenty-four successive patients admitted in the intensive care unit with multiple organ failure.
Interventions: Skin temperature was measured every 10 mins from inclusion in the study until discharge or death (median length of stay 18.8 days, interquartile range 86).
Measurements: From the temperature time series, hourly approximate entropy measurements were obtained. Clinical status was evaluated using the Sequential Organ Failure Assessment (SOFA) score.
Main results: A significant inverse relationship between approximate entropy and the attributed SOFA score was observed in 89% of the patients considered. Both mean and minimum approximate entropy were significantly lower in patients who died than in patients who survived (mean approximate entropy, 0.47 vs. 0.61; minimum approximate entropy, 0.24 vs. 0.40; in both cases p < .001). To evaluate the prognostic value of both mean and minimum approximate entropy, we fitted logistic regression models against survival. An increase in 0.1 units in minimum or mean approximate entropy increased 15.4- and 18.5-fold the odds of surviving, respectively.
Conclusions: The clinical status of patients suffering multiple organ failure is inversely correlated to the complexity of the temperature curve expressed as approximate entropy. Reduced complexity has dismal prognostic implications. Its assessment is noninvasive and inexpensive and allows for real-time continuous monitoring of clinical status.