Predicting serious complications in patients with cancer and pulmonary embolism using decision tree modelling: the EPIPHANY Index

Abstract

Background:Our objective was to develop a prognostic stratification tool that enables patients with cancer and pulmonary embolism (PE), whether incidental or symptomatic, to be classified according to the risk of serious complications within 15 days.Methods:The sample comprised cases from a national registry of pulmonary thromboembolism in patients with cancer (1075 patients from 14 Spanish centres). Diagnosis was incidental in 53.5% of the events in this registry. The Exhaustive CHAID analysis was applied with 10-fold cross-validation to predict development of serious complications following PE diagnosis.Results:About 208 patients (19.3%, 95% confidence interval (CI), 17.1–21.8%) developed a serious complication after PE diagnosis. The 15-day mortality rate was 10.1%, (95% CI, 8.4–12.1%). The decision tree detected six explanatory covariates: Hestia-like clinical decision rule (any risk criterion present vs none), Eastern Cooperative Group performance scale (ECOG-PS; <2 vs ⩾2), O2 saturation (<90 vs ⩾90%), presence of PE-specific symptoms, tumour response (progression, unknown, or not evaluated vs others), and primary tumour resection. Three risk classes were created (low, intermediate, and high risk). The risk of serious complications within 15 days increases according to the group: 1.6, 9.4, 30.6% P<0.0001. Fifteen-day mortality rates also rise progressively in low-, intermediate-, and high-risk patients: 0.3, 6.1, and 17.1% P<0.0001. The cross-validated risk estimate is 0.191 (s.e.=0.012). The optimism-corrected area under the receiver operating characteristic curve is 0.779 (95% CI, 0.717–0.840).Conclusions:We have developed and internally validated a prognostic index to predict serious complications with the potential to impact decision-making in patients with cancer and PE.

Clinical prediction rules for mortality in patients with pulmonary embolism and cancer to guide outpatient management: a meta‐analysis

“…). Eight studies evaluating 10 CPRs in 3974 (range 124 to 1075) unique patients reported mortality data in low‐ and higher‐risk PE patients with cancer and were included in our meta‐analysis (Table ) [; E.R. Weeda, unpublished data].…”

“…Extensive rules may not be easily applicable in a general busy practice. The EPIPHANY index is a multi‐step decision tree , which may make it more difficult to use, as clinicians have displayed low recall for specific elements of CPRs . Nonetheless, its creators note that five of its six variables (Eastern Cooperative Oncology Group‐Performance Status [ECOG‐PS], tumor response assessment, previous tumor resection, oxygen saturation and the presence of PE‐specific symptoms) can be assessed at the patient's bedside.…”

“…Nonetheless, its creators note that five of its six variables (Eastern Cooperative Oncology Group‐Performance Status [ECOG‐PS], tumor response assessment, previous tumor resection, oxygen saturation and the presence of PE‐specific symptoms) can be assessed at the patient's bedside. The authors chose to pursue a decision tree model to imitate real‐world decision‐making . CPR pragmatism must also be balanced with validity.…”

Decision tree model for predicting in‐hospital cardiac arrest among patients admitted with acute coronary syndrome

“…Decision tree model was given to the interest in generating a classification that would reasonably imitate authentic decision‐making. This means that, compared to a binary logistic regression, which postulates the existence of additive effects that contribute to explaining an outcome, decision trees factor in the existence of strong interactions between variables, and are better suited to elaborating decision‐making algorithms that follow the same structure …”

C‐reactive protein to albumin ratio is a key indicator in a predictive model for anastomosis leakage after esophagectomy: Application of classification and regression tree analysis

“…logistic regression, Cox proportional hazards regression) and some serological and biochemical indicators However, shortcomings in traditional regression analysis exist. Particularly, in this study, CAR as a composite indicator has a multicollinearity relationship with CRP and albumin, which may lead to the wrong identification of relevant predictors in regression analysis Classification and regression tree (CART), as a data mining technique, is ideally suitable for the generation of clinical rules, which often enables the detection of complex interactions between predictors (including predictors with multicollinearity), which may be difficult or impossible to uncover using traditional statistical techniques CART has been shown to perform as well or better than other traditional statistical techniques, such as logistic regression analysis, and is increasingly being applied to diagnose disease and predict outcomes or complications in patients, including diabetes, trauma, and cancer . To the best of our knowledge, CART analysis has not previously been used to assess the predictors of AL in patients after esophagectomy.…”