Abstract: Objective The aim of this study was to identify the direct and indirect factors influencing adverse outcomes in patients with colorectal cancer after treatment,and to explore the causal effects of these factors on adverse outcomes,providing a basis for improving patient prognosis. Methods Medical record information of patients admitted to Harbin Medical University Cancer Hospital and diagnosed with colorectal cancer from 2013 to 2015 was collected.Adverse outcomes were defined as death,metastasis,or recurrence within two years after treatment.The fast greedy equivalence search algorithm was used to construct a causal graphical model and analyze the direct and indirect factors influencing adverse outcomes.Based on this,the intervention calculus when the directed acyclic graph is absent(IDA)algorithm was employed to evaluate the causal effects of these factors on adverse outcomes. Results A total of 2,332 patients were included,with an average age of(68.0±10.9)years and an adverse outcome incidence of 6.22%.The causal graph contained 20 nodes and 36 edges.The direct factors influencing adverse outcomes included chemotherapy,pathological type,surgical treatment,and length of hospital stay(｜IDA｜ values were 0.039,0.059,0.255,and 0.054,respectively).The indirect influencing factors included age,alcohol consumption,body mass index,differentiation degree,radiotherapy,and surgery nature(｜IDA｜ values were 0.011,0.021,0.012,0.042,0.021,and 0.030,respectively). Conclusions Based on identifying key factors influencing adverse outcomes in colorectal cancer through the causal graph,the IDA algorithm can quantify the causal effects of influencing factors on adverse outcomes.The study suggests that in the clinical treatment of colorectal cancer,increasing the acceptance rate of surgery and chemotherapy in patients without surgical or chemotherapy contraindications can reduce the incidence of adverse outcomes and thereby improve prognosis.

Key words: colorectal cancer, causal diagram model, fast greedy equivalence search algorithm, intervention calculus algorithm without causal graph, causal effect