Abstract:
The purpose of this correlational research was to develop a predictive model for chemotherapy induced neutropenia in leukemic and lymphoma patients. The study was carried out retrospectively and all patients' data were collected from Phrapokklao Hospital during January 1, 2002 to December 31, 2006. The investigator reviewed the medical records of 175 leukemic and lymphoma patients receiving intravenous chemotherapy and evaluated patients risk factors that associated with neutropenia event. A neutropenia event was operationally defined as an ANC nadir 500 cells/mm[superscript3]. The risk factors that related to neutropenia event were first analyzed and identified by using univariate logistic regression and then significant risk factors were evaluated by using multivariate logistic regression model. In addition, the predictive model was validated by using the receiver operating characteristic (ROC) curves. Consequently, the final model was as follows : predictive score = (2 x BM.involvement) + (7 x Regimen) + (12 x ANC) + (2 x RBC) + (2 x platelet) + (Hgb) + (2 x LDH) + (RDI) The predictive model demonstrated that the risk factors had satisfactory fit to the data (model chi-square = 88.986, p < 0.001). Area under ROC curves for the validated model was 0.683. When the cut-off point of 7 was chosen and used to predict the presence of neutropenia event, the model had sensitivity, specificity, and accuracy values of 94.74%, 38.46% and 62.22%, respectively. In conclusion, this study has created a model with preliminary evidences of validity. Further refinement and testing of the model would provide more comprehensive evidence for its validity and generalizability when identifying patients who have the chance of chemotherapy-induced neutropenia.