Introduction: Risk prediction models are used as part of formal risk assessment for disease and health events in UK primary care. To improve the accuracy of risk prediction, new risk factors are being added to established risk prediction models. However, current methods used to evaluate the added value of these new risk factors have shown to be limited. These limitations can be addressed using health economic methodology, but is yet to be used to evaluate and compare risk prediction models by means of their effectiveness and cost. Methods: A cost effectiveness analysis was performed using a decision tree framework. The decision tree was populated risk model effects and cost measures. The cost-effectiveness analysis derived the incremental cost effectiveness ratio (ICER) using the Youden Index and Harrell’s C-Index performance measures, and the net monetary benefit (INB). A probabilistic sensitivity analysis was performed, based on 10,000 iterations. A range of £0-£100,000 was used for the willingness to pay (WTP), which when combined with the INB, provided the probability the new risk factor was cost effective. This method was applied in two exemplar prospective cohort studies; adding family history (FH) to cardiovascular disease (CVD) risk prediction; and bone mineral density (BMD) to fracture risk prediction. Results: A cost-effectiveness analysis using a decision tree framework was shown to be an effective way of evaluating the added value of the new risk factor. Adding FH to standard CVD risk factors produced an ICER of £799.91 (-£5,962.15 to £5,968.22) and £7,788.76 (-£42,760.16 to £48,962.39) per percentage unit increase in the Youden Index and Harrell’s C-Index, respectively. The maximum probability of FH being cost effective is 0.7, with a minimum WTP of £15,000 (Youden Index). Further, treating low risk patients with statin therapy incorrectly was less costly (£788.40) than not treating them (£916.16). Adding continuous BMD measurement to standard fracture risk factors produced an ICER of £367.25 (-£4,241.88 to £4,828.50) and £4,480.54 (-£22,816.84 to £22,970.55) per percentage unit increase in the Youden Index and Harrell’s C-Index, respectively. The maximum probability BMD being cost-effective is 0.8, with a minimum WTP of £32,500 (Youden Index). Further, using BMD in a binary format to indicate osteoporotic patients, did not improve Harrell’s C-Index of standard fracture risk prediction (∆C-Index=-0.62%). Conclusion: A cost-effectiveness analysis was a novel method to compare two risk prediction models; and to evaluate the added value of a new risk factor. It identifies the added value of a new risk factor; encompassing the statistical and clinical improvement, and cost consequences when using the new risk factor in an established risk prediction model. Based on the added value of FH and BMD, there is a good evidence base to add these risk factors into routine risk assessment of the respective conditions. Increased use of this method could help standardise risk prediction and increase comparability of risk prediction models within diseases; producing a league table approach to evaluate, appraise and identify beneficial new risk factors and better risk prediction models.
| Identifer | oai:union.ndltd.org:bl.uk/oai:ethos.bl.uk:680268 |
| Date | January 2015 |
| Creators | Dhiman, Paula |
| Publisher | University of Nottingham |
| Source Sets | Ethos UK |
| Detected Language | English |
| Type | Electronic Thesis or Dissertation |
| Source | http://eprints.nottingham.ac.uk/30666/ |
Page generated in 0.0021 seconds