Document Type : Methodological papers
Authors
1
Ph.D. candidate in Epidemiology and Biostatistics, Faculty of Public Health, Khon Kean University, Khon Kaen, Thailand.
2
ASEAN Cancer Epidemiology and Prevention Research Group, Khon Kaen University, Khon Kaen, Thailand.
3
Department of Epidemiology and Biostatistics, Faculty of Public Health, Khon Kaen University, Khon Kaen, Thailand.
Abstract
Background: Multilevel analysis, in several forms, has been extensively utilized over the past few decades. While utilizing for colorectal cancer (CRC) screening may be unclear, especially at community level. The study aimed to explain the use of multilevel negative binomial analysis, developed as a practical guide through data obtained from a study of CRC screening in Thailand. Method: We analyzed the data of 2,475 fecal immunochemical test (FIT) cases in treatment arms from a population-based randomized controlled trial for CRC screening in the Khon Kaen province of Thailand. We summarized the statistical methodology, highlighting the advantages and disadvantages of data analysis using a multilevel negative binomial method compared with a standard negative binomial approach based on the data obtained in the randomized controlled trial for CRC screening; where active smoking and fecal hemoglobin (f-Hb) concentration were considered as the main exposure and outcome, respectively. Results: Our findings showed differences of significant value and magnitude in the effects of both methods. Active smoking was statistical significantly with an f-Hb concentration IRRadj = 1.47 (95%, CI: 1.01-2.14) through the use of the standard negative binomial method, whereas the multilevel negative binomial approach produced a non-statistical significance of IRRadj = 1.30 (95%, CI: 0.89-1.90). Conclusion: Utilizing a standard statistical approach in CRC screening, the data analyzed were equal to zero. Hierarchical data, based on contextual factors and using a multilevel modeling approach, must be addressed. The f-Hb concentration, occurred over-dispersion, which implies that further studies utilize over-dispersion for improved appropriate statistical analysis.
Keywords
Main Subjects
)