Degree Granting Department
Industrial and Management Systems Engineering
Jose L. Zayas-Castro
Data mining, Hemodynamic parameters, ICU, Patient monitoring, Platelet count
Laboratory tests are a primary resource for diagnosing patient diseases. However, physicians often make decisions based on a single laboratory result and have a limited perspective of the role of commonly-measured parameters in enhancing the diagnostic process. By providing a dynamic patient profile, the diagnosis could be more accurate and timely, allowing physicians to anticipate changes in the recovery trajectory and intervene more effectively.
The assessment and monitoring of the circulatory system is essential for patients in intensive care units (ICU). One component of this system is the platelet count, which is used in assessing blood clotting. However, platelet counts represent a dynamic equilibrium of many simultaneous processes, including altered capillary permeability, inflammatory cascades (sepsis), and the coagulation process. To characterize the value of dynamic changes in platelet count, analytical methods are applied to datasets of critically-ill patients in (1) a homogeneous population of ICU cardiac surgery patients and (2) a heterogeneous group of ICU patients with different conditions and several hospital admissions.
The objective of this study was to develop a methodology to anticipate adverse events using metrics that capture dynamic changes of platelet counts in a homogeneous population, then redefine the methodology for a more heterogeneous and complex dataset. The methodology was extended to analyze other important physiological parameters of the circulatory system (i.e., calcium, albumin, anion gap, and total carbon dioxide). Finally, the methodology was applied to simultaneously analyze some parameters enhancing the predictive power of various models.
This methodology assesses dynamic changes of clinical parameters for a heterogeneous population of ICU patients, defining rates of change determined by multiple point regression and by the simpler fixed time parameter value ratios at specific time intervals. Both metrics provide prognostic information, differentiating survivors from non-survivors and have demonstrated being more predictive than complex metrics and risk assessment scores with greater dimensionality.
The goal was to determine a minimal set of biomarkers that would better assist care providers in assessing the risk of complications, allowing them alterations in the management of patients. These metrics should be simple and their implementation would be feasible in any environment and under uncertain conditions of the specific diagnosis and the onset of an acute event that causes a patient's admission to the ICU.
The results provide evidence of the different behaviors of physiologic parameters during the recovery processes for survivors and non-survivors. These differences were observed during the first 8 to 10 days after a patient's admission to the ICU. The application of the presented methodology could enhance physicians' ability to diagnose more accurately, anticipate changes in recovery trajectories, and prescribe effective treatment, leading to more personalized care and reduced mortality rates.
Scholar Commons Citation
Puertas, Monica A., "Statistical and Prognostic Modeling of Clinical Outcomes with Complex Physiologic Data" (2014). Graduate Theses and Dissertations.