Hurdle Poisson Regression Model for Identifying Factors Related to Noncompliance and Waiting Time for Confirmatory Diagnosis in Colorectal Cancer Screening

Int J Technol Assess Health Care. 2019 Jan;35(2):85-91. doi: 10.1017/S0266462319000047. Epub 2019 Mar 1.

Abstract

Objectives: Population-based colorectal cancer (CRC) screening programs that use a fecal immunochemical test (FIT) are often faced with a noncompliance issue and its subsequent waiting time (WT) for those FIT positives complying with confirmatory diagnosis. We aimed to identify factors associated with both of the correlated problems in the same model.

Methods: A total of 294,469 subjects, either with positive FIT test results or having a family history, collected from 2004 to 2013 were enrolled for analysis. We applied a hurdle Poisson regression model to accommodate the hurdle of compliance and also its related WT for undergoing colonoscopy while assessing factors responsible for the mixture of the two outcomes.

Results: The effect on compliance and WT varied with contextual factors, such as geographic areas, type of screening units, and level of urbanization. The hurdle score, representing the risk score in association with noncompliance, and the WT score, reflecting the rate of taking colonoscopy, were used to classify subjects into each of three groups representing the degree of compliance and the level of health awareness.

Conclusion: Our model was not only successfully applied to evaluating factors associated with the compliance and the WT distribution, but also developed into a useful assessment model for stratifying the risk and predicting whether and when screenees comply with the procedure of receiving confirmatory diagnosis given contextual factors and individual characteristics.

Keywords: Colonoscopy; Colorectal cancer screening; Hurdle Poisson model; Noncompliance; Waiting time.

MeSH terms

  • Aged
  • Colonoscopy
  • Colorectal Neoplasms / diagnosis*
  • Early Detection of Cancer / methods*
  • Female
  • Humans
  • Male
  • Middle Aged
  • Models, Statistical*
  • Occult Blood
  • Patient Compliance / statistics & numerical data*
  • Regression Analysis
  • Residence Characteristics
  • Taiwan
  • Time Factors
  • Waiting Lists*