Publications in relation to CHOICE
Prediction of imminent osteoporotic fracture risk in Danish postmenopausal women—can the addition of self-reported clinical risk factors improve the prediction of the register-based FREM algorithm?
Emilie Rosenfeldt Christensen, Kasper Westphal Leth, Frederik Lykke Petersen, Tanja Gram Petersen, Sören Möller, Bo Abrahamsen, Katrine Hass Rubin
Abstract:
Summary
Obtaining accurate self-reports on clinical risk factors, such as parental hip fracture or alcohol and tobacco use, limits the utility of conventional risk scores for fracture risk. We demonstrate that fracture-risk prediction based on administrative health data alone performs equally to prediction based on self-reported clinical risk factors.
Background
Accurate assessment of fracture risk is crucial. Unlike established risk prediction tools that rely on patient recall, the Fracture Risk Evaluation Model (FREM) utilises register data to estimate the risk of major osteoporotic fracture (MOF). We investigated whether adding self-reported clinical risk factors for osteoporosis to the FREM algorithm improved the prediction of 1-year fracture risk by comparing three approaches: the FREM algorithm (FREMorig), clinical risk factors (CRFonly), and FREM combined with clinical risk factors (FREM-CRF).
Method
Clinical risk factor information was obtained through questionnaires sent to women aged 65–80 years living in the Region of Southern Denmark in 2010, who participated in the Risk-stratified Osteoporosis Strategy Evaluation study. Register data was obtained through national health registers and linked to the survey data. Positive and negative predictive values and concordance statistics were calculated for the performance of each approach using logistic regression and Cox proportional hazards models.
Results
Of the 18,605 women included, 280 sustained a MOF within 1 year. All three approaches performed similarly in 1-year fracture risk prediction for low- and high-risk individuals. However, the FREMorig and FREM-CRF approach slightly overestimated fracture risk for medium-risk individuals.
Conclusion
Adding self-reported clinical data to FREM did not increase precision in predicting 1-year MOF risk. The discrimination of FREMorig was similar to that of CRFonly, suggesting it may be possible to estimate fracture risk with the same precision by using register data instead of self-reported risk information. Register-based prediction models may be applicable in individualised risk monitoring or large-scale osteoporosis screening programmes.
An enhanced version of FREM (Fracture Risk Evaluation Model) using national administrative health data: analysis protocol for development and validation of a multivariable prediction model
Simon Bang Kristensen, Anne Clausen, Michael Kriegbaum Skjødt, Jens Søndergaard, Bo Abrahamsen, Sören Möller, Katrine Hass Rubin
Abstract:
Background
Osteoporosis poses a growing healthcare challenge owing to its rising prevalence and a significant treatment gap, as patients are widely underdiagnosed and consequently undertreated, leaving them at high risk of osteoporotic fracture. Several tools aim to improve case-finding in osteoporosis. One such tool is the Fracture Risk Evaluation Model (FREM), which in contrast to other tools focuses on imminent fracture risk and holds potential for automation as it relies solely on data that is routinely collected via the Danish healthcare registers. The present article is an analysis protocol for a prediction model that is to be used as a modified version of FREM, with the intention of improving the identification of subjects at high imminent risk of fracture by including pharmacological exposures and using more advanced statistical methods compared to the original FREM. Its main purposes are to document and motivate various aspects and choices of data management and statistical analyses.
Methods
The model will be developed by employing logistic regression with grouped LASSO regularization as the primary statistical approach and gradient-boosted classification trees as a secondary statistical modality. Hyperparameter choices as well as computational considerations on these two approaches are investigated by an unsupervised data review (i.e., blinded to the outcome), which also investigates and handles multicollinarity among the included exposures. Further, we present an unsupervised review of the data and testing of analysis code with respect to speed and robustness on a remote analysis environment. The data review and code tests are used to adjust the analysis plans in a blinded manner, so as not to increase the risk of overfitting in the proposed methods.
Discussion
This protocol specifies the planned tool development to ensure transparency in the modeling approach, hence improving the validity of the enhanced tool to be developed. Through an unsupervised data review, it is further documented that the planned statistical approaches are feasible and compatible with the data employed.
FREM predicts 10-year incident fracture risk independent of FRAX® probability: a registry-based cohort study
William D. Leslie, Sören Möller, Michael K. Skjødt, Lin Yan, Bo Abrahamsen, Lisa M. Lix, Eugene V. McCloskey, Helena Johansson, Nicholas C. Harvey, John A. Kanis, Katrine Hass Rubin
Abstract:
Summary
The Danish Fracture Risk Evaluation Model (FREM) was found to predict fracture risk independent of 10-year fracture probability derived with the FRAX® tool including bone mineral density from DXA.
Introduction
FREM was developed from Danish public health registers without DXA information to identify high imminent risk of major osteoporotic fracture (MOF) and hip fracture (HF), while FRAX® estimates 10-year fracture probability from clinical risk factors and femoral neck bone mineral density (BMD) from DXA. The FREM algorithm showed significant 1- and 2-year fracture risk stratification when applied to a clinical population from Manitoba, Canada. We examined whether FREM predicts 10-year fracture risk independent of 10-year FRAX probability computed with BMD.
Methods
Using the Manitoba BMD Program registry, we identified women and men aged ≥ 45 years undergoing baseline BMD assessment. We calculated FREM and FRAX scores, and identified incident fractures over 10 years. Hazard ratios (HRs) for incident fracture were estimated according to FREM quintile, adjusted for FRAX probability. We compared predicted with observed 10-year cumulative fracture probability estimated with competing mortality.
Results
The study population comprised 74,446 women, mean age 65.2 years; 7945 men, mean age 67.5 years. There were 7957 and 646 incident MOF and 2554 and 294 incident HF in women and men, respectively. Higher FREM scores were associated with increased risk for MOF (highest vs middle quintile HRs 1.49 women, 2.06 men) and HF (highest vs middle quintile HRs 2.15 women, 2.20 men) even when adjusted for FRAX. Greater mortality with higher FREM scores attenuated its effect on 10-year fracture probability. In the highest FREM quintile, observed slightly exceeded predicted 10-year probability for MOF (ratios 1.05 in women, 1.49 in men) and HF (ratios 1.29 in women, 1.34 in men).
Conclusions
Higher FREM scores identified women and men at increased fracture risk even when adjusted for FRAX probability that included BMD; hence, FREM provides additional predictive information to FRAX. FRAX slightly underestimated 10-year fracture probability in those falling within the highest FREM quintile.
Validation of the Fracture Risk Evaluation Model (FREM) in predicting major osteoporotic fractures and hip fractures using administrative health data
Michael K. Skjødt, Sören Möller, Nana Hyldig, Anne Clausen, Mette Bliddal, Jens Søndergaard, Bo Abrahamsen, Katrine Hass Rubin
Abstract:
Background
Prevention of osteoporotic fractures remains largely insufficient, and effective means to identify patients at high, short-term fracture risk are needed. The FREM tool is available for automated case finding of men and women aged 45 years or older at high imminent (1-year) risk of osteoporotic fractures, based on administrative health data with a 15-year look-back. The aim of this study was to validate the performance of FREM, and the effect of applying a shorter look-back period. We also evaluated FREM for 5-year fracture risk prediction.
Methods
Using Danish national health registers we generated consecutive general population cohorts for the years 2014 through 2018. Within each year and across the full time period we estimated the individual fracture risk scores and determined the actual occurrence of major osteoporotic fractures (MOF) and hip fractures. Risk scores were calculated with 15- and 5-year look-back periods. The discriminative ability was evaluated by area under the receiver operating curve (AUC), and negative predictive value (NPV) and positive predictive value (PPV) were estimated applying a calculated risk cut-off of 2% for MOF and 0.3% for hip fractures.
Results
Applying a 15-year look-back, AUC was around 0.75–0.76 for MOF and 0.84–0.87 for hip fractures in 2014, with minor decreases in the subsequent fracture cohorts (2015 to 2018). Applying a 5-year look-back generated similar results, with only marginally lower AUC. In the 5-year risk prediction setting, AUC-values were 0.70–0.72 for MOF and 0.81–0.84 for hip fractures. Generally, PPVs were low, while NPVs were very high.
Conclusion
FREM predicts the 1- and 5-year risk of MOF and hip fractures with acceptable vs excellent discriminative power, respectively, when applying both a 15- and a 5-year look-back. Hence, the FREM tool may be applied to improve identification of individuals at high imminent risk of fractures using administrative health data.
Prediction of imminent fracture risk in Canadian women and men aged 45 years or older: external validation of the Fracture Risk Evaluation Model (FREM)
Sören Möller, Michael K. Skjødt, Lin Yan, Bo Abrahamsen, Lisa M. Lix, Eugene V. McCloskey, Helena Johansson, Nicholas C. Harvey, John A. Kanis, Katrine Hass Rubin, William D. Leslie
Abstract:
Summary
The Fracture Risk Evaluation Model (FREM) identifies individuals at high imminent risk of major osteoporotic fractures. We validated FREM on 74,828 individuals from Manitoba, Canada, and found significant fracture risk stratification for all FREM scores. FREM performed better than age alone but not as well as FRAX® with BMD.
Introduction
The FREM is a tool developed from Danish public health registers (hospital diagnoses) to identify individuals over age 45 years at high imminent risk of major osteoporotic fractures (MOF) and hip fracture (HF). In this study, our aim was to examine the ability of FREM to identify individuals at high imminent fracture risk in women and men from Manitoba, Canada.
Methods
We used the population-based Manitoba Bone Mineral Density (BMD) Program registry, and identified women and men aged 45 years or older undergoing baseline BMD assessment with 2 years of follow-up data. From linked population-based data sources, we constructed FREM scores using up to 10 years of prior healthcare information.
Results
The study population comprised 74,828 subjects, and during the 2 years of observation, 1612 incident MOF and 299 incident HF occurred. We found significant fracture risk stratification for all FREM scores, with AUC estimates of 0.63–0.66 for MOF for both sexes and 0.84 for women and 0.65–0.67 for men for HF. FREM performed better than age alone but not as well as FRAX® with BMD. The inclusion of physician claims data gave slightly better performance than hospitalization data alone. Overall calibration for 1-year MOF prediction was reasonable, but HF prediction was overestimated.
Conclusion
In conclusion, the FREM algorithm shows significant fracture risk stratification when applied to an independent clinical population from Manitoba, Canada. Overall calibration for MOF prediction was good, but hip fracture risk was systematically overestimated indicating the need for recalibration.
A New Fracture Risk Assessment Tool (FREM) Based on Public Health Registries
Katrine Hass Rubin, Sören Möller, Teresa Holmberg, Mette Bliddal, Jens Søndergaard, Bo Abrahamsen
Abstract:
Some conditions are already known to be associated with an increased risk of osteoporotic fractures. Other conditions may also be significant indicators of increased risk. The aim of the current study was to identify conditions for inclusion in a fracture prediction model (fracture risk evaluation model [FREM]) for automated case finding of high-risk individuals of hip or major osteoporotic fractures (MOFs). We included the total population of Denmark aged 45+ years (N = 2,495,339). All hospital diagnoses from 1998 to 2012 were used as possible conditions; the primary outcome was MOFs during 2013. Our cohort was split randomly 50/50 into a development and a validation dataset for deriving and validating the predictive model. We applied backward selection on ICD-10 codes (International Classification of Diseases and Related Health Problems, 10th Revision) by logistic regression to develop an age-adjusted and sex-stratified model. The FREM for MOFs included 38 and 43 risk factors for women and men, respectively. Testing FREM for MOFs in the validation cohort showed good accuracy; it produced receiver-operating characteristic (ROC) curves with an area under the ROC curve (AUC) of 0.750 (95% CI, 0.741 to 0.795) and 0.752 (95% CI, 0.743 to 0.761) for women and men, respectively. The FREM for hip fractures included 32 risk factors for both genders and showed an even higher accuracy in the validation cohort as AUCs of 0.874 (95% CI, 0.869 to 0.879) and 0.851 (95% CI, 0.841 to 0.861) for women and men were found, respectively. We have developed and tested a prediction model (FREM) for identifying men and women at high risk of MOFs or hip fractures by using solely existing administrative data. The FREM could be employed either at the point of care integrated into electronic patient record systems to alert physicians or deployed centrally in a national case-finding strategy where patients at high fracture risk could be invited to a focused DXA program. © 2018 The Authors. Journal of Bone and Mineral Research published by Wiley Periodicals, Inc. on behalf of American Society for Bone and Mineral Research (ASBMR).
DOI: 10.1002/jbmr.3528