C Kruse1,2, P Eiken3,4, P Vestergaard5,6. 1. Department of Endocrinology, Aalborg University Hospital, Hobrovej 19, 9100, Aalborg, Denmark. C.kruse@rn.dk. 2. Department of Clinical Medicine, Aalborg University Hospital, Aalborg, Denmark. C.kruse@rn.dk. 3. Department of Cardiology, Nephrology and Endocrinology, Nordsjaellands Hospital, Hilleroed, Denmark. 4. Faculty of Health and Medical Sciences, University of Copenhagen, Copenhagen, Denmark. 5. Department of Endocrinology, Aalborg University Hospital, Hobrovej 19, 9100, Aalborg, Denmark. 6. Department of Clinical Medicine, Aalborg University Hospital, Aalborg, Denmark.
Abstract
Clustering analysis can identify subgroups of patients based on similarities of traits. From data on 10,775 subjects, we document nine patient clusters of different fracture risks. Differences emerged after age 60 and treatment compliance differed by hip and lumbar spine bone mineral density profiles. INTRODUCTION: The purposes of this study were to establish and quantify patient clusters of high, average and low fracture risk using an unsupervised machine learning algorithm. METHODS: Regional and national Danish patient data on dual-energy X-ray absorptiometry (DXA) scans, medication reimbursement, primary healthcare sector use and comorbidity of female subjects were combined. Standardized variable means, Euclidean distances and Ward's D2 method of hierarchical agglomerative clustering (HAC), were used to form the clustering object. K number of clusters was selected with the lowest cluster containing less than 250 subjects. Clusters were identified as high, average or low fracture risk based on bone mineral density (BMD) characteristics. Cluster-based descriptive statistics and relative Z-scores for variable means were computed. RESULTS: Ten thousand seven hundred seventy-five women were included in this study. Nine (k = 9) clusters were identified. Four clusters (n = 2886) were identified based on low to very low BMD with differences in comorbidity, anthropometrics and future bisphosphonate compliance. Two clusters of younger subjects (n = 1058, mean ages 30 and 51 years) were identified as low fracture risk with high to very high BMD. A mean age of 60 years was the earliest that allowed for separation of high-risk clusters. DXA scan results could identify high-risk subjects with different antiresorptive treatment compliance levels based on similarities and differences in lumbar spine and hip region BMD. CONCLUSIONS: Unsupervised HAC presents a novel technology to improve patient characteristics in bone disease beyond traditional T-score-based diagnosis. Technological and validation limitations need to be overcome to improve its use in internal medicine. Current DXA scan indication guidelines could be further improved by clustering algorithms.
Clustering analysis can identify subgroups of patients based on similarities of traits. From data on 10,775 subjects, we document nine patient clusters of different fracture risks. Differences emerged after age 60 and treatment compliance differed by hip and lumbar spine bone mineral density profiles. INTRODUCTION: The purposes of this study were to establish and quantify patient clusters of high, average and low fracture risk using an unsupervised machine learning algorithm. METHODS: Regional and national Danish patient data on dual-energy X-ray absorptiometry (DXA) scans, medication reimbursement, primary healthcare sector use and comorbidity of female subjects were combined. Standardized variable means, Euclidean distances and Ward's D2 method of hierarchical agglomerative clustering (HAC), were used to form the clustering object. K number of clusters was selected with the lowest cluster containing less than 250 subjects. Clusters were identified as high, average or low fracture risk based on bone mineral density (BMD) characteristics. Cluster-based descriptive statistics and relative Z-scores for variable means were computed. RESULTS: Ten thousand seven hundred seventy-five women were included in this study. Nine (k = 9) clusters were identified. Four clusters (n = 2886) were identified based on low to very low BMD with differences in comorbidity, anthropometrics and future bisphosphonate compliance. Two clusters of younger subjects (n = 1058, mean ages 30 and 51 years) were identified as low fracture risk with high to very high BMD. A mean age of 60 years was the earliest that allowed for separation of high-risk clusters. DXA scan results could identify high-risk subjects with different antiresorptive treatment compliance levels based on similarities and differences in lumbar spine and hip region BMD. CONCLUSIONS: Unsupervised HAC presents a novel technology to improve patient characteristics in bone disease beyond traditional T-score-based diagnosis. Technological and validation limitations need to be overcome to improve its use in internal medicine. Current DXA scan indication guidelines could be further improved by clustering algorithms.
Authors: S I Ao; Kevin Yip; Michael Ng; David Cheung; Pui-Yee Fong; Ian Melhado; Pak C Sham Journal: Bioinformatics Date: 2004-12-07 Impact factor: 6.937
Authors: Chris Cotsapas; Benjamin F Voight; Elizabeth Rossin; Kasper Lage; Benjamin M Neale; Chris Wallace; Gonçalo R Abecasis; Jeffrey C Barrett; Timothy Behrens; Judy Cho; Philip L De Jager; James T Elder; Robert R Graham; Peter Gregersen; Lars Klareskog; Katherine A Siminovitch; David A van Heel; Cisca Wijmenga; Jane Worthington; John A Todd; David A Hafler; Stephen S Rich; Mark J Daly Journal: PLoS Genet Date: 2011-08-10 Impact factor: 5.917
Authors: Alexander L Hornung; Christopher M Hornung; G Michael Mallow; J Nicolás Barajas; Augustus Rush; Arash J Sayari; Fabio Galbusera; Hans-Joachim Wilke; Matthew Colman; Frank M Phillips; Howard S An; Dino Samartzis Journal: Eur Spine J Date: 2022-03-27 Impact factor: 2.721
Authors: Vikas Kumar; Christopher Roche; Steven Overman; Ryan Simovitch; Pierre-Henri Flurin; Thomas Wright; Joseph Zuckerman; Howard Routman; Ankur Teredesai Journal: Clin Orthop Relat Res Date: 2020-10 Impact factor: 4.755