Symptom-Based Predictive Model for Colorectal Cancer Diagnosis: Optimization According to Chilean Public Health Policy Guidelines.

Abstract

In Chile, the public health policy known as "Explicit Health Guarantees" (GES) allows the referral of patients with suspected Colorectal Cancer (CRC) to a tertiary center for colonoscopy within 45 days of evaluation. The correlation of these symptoms with the diagnosis has not been analized.

Aim: This study aims to analyze variables linked to CRC diagnosis, those in GES guidelines and other clinically important, and to develop a symptom-based predictive model for CRC diagnosis.

Methods: A retrospective analytical study was conducted from July 2016 to December 2021. Inclusion criteria were patients referred for colonoscopy as per GES guidelines. Colonoscopy variables were evaluated for test quality. Sixteen variables were included in the predictive model, ten from the GES guidelines and six of clinical interest. Statistical univariate analysis with SPSS 26® (p<0.05). Multivariate analysis used binary logistic regression with ROC analysis and 95% confidence intervals for comparison.

Results: The cohort included 1099 patients with a mean age of 63.9±13.3 years; 61.1% were female. 148 patients (13%) were diagnosed with neoplasia with 66.9% stage III-IV. Significant variables in the predictive model included age, gender, diarrhea, lower gastrointestinal bleeding, compromised general condition, anemia, palpable rectal mass, suggestive ultrasound, and CT scan, with an AUC of 0.86 (95%CI 0.83-0.89). A model without imaging variables achieved an AUC of 0.81 (95%CI 0.78-0.85).

Conclusion: The GES policy enabled a CRC detection rate of 13% in this cohort. Predictive models were developed to optimize referrals for colonoscopy and enhance public health policy. The models require validation in an independent cohort to determine their real-world applicability.