Building team agreement on large population surveys through inter-rater reliability among oral health survey examiners

Background: Oral health surveys conducted on a very large population involve many examiners who must be consistent in scoring different levels of an oral disease. Prior to the oral health survey implementation, a measurement of inter-rater reliability (IRR) is needed to know the level of agreement among examiners or raters. Purpose: This study aimed to assess the IRR using consensus and consistency estimates in large population oral health surveys. Methods: A total of 58 dentists participated as raters. The benchmarker showed the clinical sample for dental caries and community periodontal index (CPI) score, with the raters being trained to carry out a calibration exercise in dental phantom. The consensus estimate was measured by means of a percent agreement and Cohen's Kappa statistic. The consistency estimate of IRR was measured by Cronbach's alpha coefficient and intraclass correlation. Results: The percent agreement is 65.50% for photographic slides of dental caries, 73.13% for photographic slides of CPI and 78.78% for calibration of dental caries using phantom. There were statistically significant differences between dental caries calibration using photographic slides and phantom (p<0.000), while the consistency of IRR between multiple raters is strong (Cronbrach's Alpha: >0.9). Conclusion: A percent agreement across multiple raters is acceptable for the diagnosis of dental caries. Consistency between multiple raters is reliable when diagnosing dental caries and CPI.