USING INVERSE DECISION THEORY TO DETERMINE A C/B RATIO FOR SCREENING AND DIAGNOSIS OF CERVICAL INTRAEPITHELIAL NEOPLASIA (CIN)

To Register SMDM Homepage

Sunday, 17 October 2004

This presentation is part of: Poster Session - Public Health; Methodological Advances

USING INVERSE DECISION THEORY TO DETERMINE A C/B RATIO FOR SCREENING AND DIAGNOSIS OF CERVICAL INTRAEPITHELIAL NEOPLASIA (CIN)

Richard J. Swartz, PhD¹, Dennis Cox, PhD², Scott B. Cantor, PhD³, Kalatu Davies², and Michele Follen, MD, PhD⁴. (1) The University of Texas M. D. Anderson Cancer Center, Department of Behavioral Science, Houston, TX, (2) Rice University, Department of Statistics, Houston, TX, (3) The University of Texas M. D. Anderson Cancer Center, Biostatistics & Applied Mathematics, Houston, TX, (4) The University of Texas M. D. Anderson Cancer Center, Biomedical Engineering Center, Houston, TX

Purpose: We propose a new method, Inverse Decision Theory, to determine bounds on relative costs and losses for medical decisions. These bounds can be used to compare different treatment strategies and to determine a range for the cost-benefit (C/B) ratio (Metz CE. SEM. NUC. MED.: 1978: 8: 283-298). Methods: We used Bayesian sequential decision theory to model the sequence of tests used to screen and diagnose cervical intraepithelial neoplasia (CIN) (It includes a Papanicolaou smear followed by a colposcopic exam, followed by a biopsy. Positive results indicate progression to the next test, with positive biopsy indicating treatment.) We assumed this current standard of care (SOC) for CIN was optimal. We identified operating characteristics for the tests from the literature and a sample of 624 women. We then worked backwards through the sequential decision problem and solved for the costs (both monetary and patient outcomes associated with performing a test) and losses (both monetary and patient outcomes associated with a treatment decision) associated with the SOC. This yielded bounds on each of the costs and losses considered. Furthermore, the bounds identified conditions for situations when competing strategies would perform better than the current SOC and these boundaries were linear constraints on the costs and losses. We used linear programming to determine bounds on the C/B ratio. Results: We found that treating someone based on a positive colposcopic result (forgoing the biopsy) is better than the SOC when the net loss for treating non-diseased women is less than 1.4 times the incremental cost of biopsy. The C/B ratio was found to have a lower bound of 0.0862. Conclusions: The lower bound suggests that the C/B ratio for screening and diagnosing CIN is higher than previously reported and that the current standard of care has a high degree of benefit relative to its cost. IDT is an effective method for characterizing the costs and losses associated with established decision rules such as a screening and diagnosis care setting.

This research was supported in part by The National Cancer Institute grant 2PO1-CA82710 and in part by a fellowship supported by the National Cancer Institute grant R25 CA57730, Robert M. Chamberlain, Ph.D., Principal Investigator.

See more of Poster Session - Public Health; Methodological Advances
See more of The 26th Annual Meeting of the Society for Medical Decision Making (October 17-20, 2004)