Predicting peri‐implant disease: Chi‐square automatic interaction detection (CHAID) decision tree analysis of risk indicators

• Published in: Journal of periodontology (1970) 2019-08, Vol.90 (8), p.834-846
• Main Authors: Atieh, Momen A., Pang, Ju Keat, Lian, Kylie, Wong, Stephanie, Tawse‐Smith, Andrew, Ma, Sunyoung, Duncan, Warwick J.
• Format: Article
• Language: English
• Subjects: Decision Trees; Dental Implants; Dentistry; Humans; Peri-Implantitis; Retrospective Studies; retrospective study; risk assessment; Risk Factors; Stomatitis
• ISSN: 0022-3492; 1943-3670
• DOI: 10.1002/JPER.17-0501

Background Further validation of the risk indicators / predictors for peri‐implant diseases is required to allow clinicians and patients to make informed decisions and optimize dental implant treatment outcomes. The aim of this study was to build prediction models, using Chi‐square automatic interaction detection (CHAID) analysis, to determine which systemic‐, patient‐, implant‐, site‐, surgical‐ and prostheses‐related risk indicators had more impact on the onset of peri‐implant diseases. Methods A retrospective analysis of 200 patients who received implant‐supported prostheses between 1998 and 2011 was conducted to evaluate the prevalences and risk indicators for peri‐implant mucositis and peri‐implantitis. The data were further analyzed using CHAID to produce two predictive models. Results The prevalence of peri‐implant mucositis was 20.2% and 10.2% for patients and implants, respectively, while the prevalence of peri‐implantitis was 10.1% at the patient level and 5.4% at the implant level. CHAID decision tree analysis identified three predictors (history of treated periodontitis, absence of regular supportive peri‐implant maintenance, and use of bone graft) for peri‐implant mucositis and three predictors (smoking, absence of regular supportive peri‐implant maintenance, and placement of ≥2 implants) for peri‐implantitis. Conclusions Within the limitations of this study, CHAID decision tree analysis identified the most plausible risk indicators and provided two predictive models for use in a particular university setting that would allow early detection and ensure appropriate care and maintenance of patients at high risk of peri‐implant diseases.