Examining Bias and Reporting in Oral Health Prediction Modeling Studies.

Area of Science:

• Oral health research
• Biostatistics
• Epidemiology

Background:

• Prediction modeling studies (PMSs) are increasingly used in oral health to predict disease risk and treatment outcomes.
• Challenges such as risk of bias and insufficient reporting impede the reproducibility and implementation of these models.
• PROBAST (Prediction Model Risk of Bias Assessment Tool) and TRIPOD (Transparent Reporting of a Multivariable Prediction Model for Individual Prognosis or Diagnosis) are guidelines to address these issues, but their application is limited.

Purpose of the Study:

• To assess the risk of bias and transparency of reporting in oral health prediction models using PROBAST and TRIPOD.
• To identify common biases and reporting deficiencies in published oral health PMSs.

Main Methods:

• A systematic literature search was conducted in PubMed for oral health PMSs.
• Included studies were assessed for risk of bias using PROBAST and reporting transparency using TRIPOD.
• 34 studies (58 models) were analyzed from 2,881 identified papers.

Main Results:

• 75% of studies were susceptible to at least 4 of 20 bias sources, including measurement error, handling of missing data, variable selection, overfitting, and lack of performance assessment.
• 95% of studies inadequately reported at least 5 essential items based on TRIPOD, such as sampling and model-building procedures.
• Common outcomes investigated were periodontal diseases (42%) and oral cancers (30%).

Conclusions:

• Oral health prediction models exhibit significant deficiencies in transparent reporting and bias identification.
• Adherence to PROBAST and TRIPOD recommendations is crucial for enhancing the quality, reproducibility, and clinical applicability of oral health PMSs.