Premium
Prediction models for the incidence and progression of periodontitis: A systematic review
Author(s) -
Du Mi,
Bo Tao,
Kapellas Kostas,
Peres Marco A
Publication year - 2018
Publication title -
journal of clinical periodontology
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 3.456
H-Index - 151
eISSN - 1600-051X
pISSN - 0303-6979
DOI - 10.1111/jcpe.13037
Subject(s) - checklist , periodontitis , scopus , medline , critical appraisal , meta analysis , medicine , data extraction , systematic review , incidence (geometry) , predictive modelling , computer science , machine learning , psychology , dentistry , alternative medicine , pathology , mathematics , geometry , political science , law , cognitive psychology
Aims To comprehensively review, identify and critically assess the performance of models predicting the incidence and progression of periodontitis. Methods Electronic searches of the MEDLINE via PubMed, EMBASE , DOSS , Web of Science, Scopus and ProQuest databases, and hand searching of reference lists and citations were conducted. No date or language restrictions were used. The Critical Appraisal and Data Extraction for Systematic Reviews of Prediction Modelling Studies checklist was followed when extracting data and appraising the selected studies. Results Of the 2,560 records, five studies with 12 prediction models and three risk assessment studies were included. The prediction models showed great heterogeneity precluding meta‐analysis. Eight criteria were identified for periodontitis incidence and progression. Four models from one study examined the incidence, while others assessed progression. Age, smoking and diabetes status were common predictors used in modelling. Only two studies reported external validation. Predictive performance of the models (discrimination and calibration) was unable to be fully assessed or compared quantitatively. Nevertheless, most models had “good” ability to discriminate between people at risk for periodontitis. Conclusions Existing predictive modelling approaches were identified. However, no studies followed the recommended methodology, and almost all models were characterized by a generally poor level of reporting.