Modeling the clinical assessment of men with suspected obstructed voiding using Bayes' Theorem

• Published in: Neurourology and urodynamics 2008-01, Vol.27 (8), p.797-801
• Main Authors: Caffarel, Jennifer, Griffiths, Clive, Pickard, Robert, Robson, Wendy, Drinnan, Michael
• Format: Article
• Language: English
• Subjects: Bayes Theorem; bladder outlet obstruction; Diagnostic Techniques, Urological; Health Status Indicators; Humans; Male; Models, Biological; Predictive Value of Tests; Pressure; pressure-flow studies; Urinary Bladder Neck Obstruction - diagnosis; Urinary Bladder Neck Obstruction - physiopathology; Urodynamics; uroflowmetry
• ISSN: 0733-2467; 1520-6777
• DOI: 10.1002/nau.20587

Aims Pressure‐flow studies (PFS) are the only reliable way to diagnose bladder outlet obstruction (BOO) in men with lower urinary tract symptoms (LUTS). However, in routine clinical practice, BOO is usually inferred by any of a number of tests (symptoms, flow rate, prostate size…). Bayes' Theorem provides a mathematical method, which may be similar to the process used by clinicians, for combining the results of multiple tests to reach a diagnosis. We have applied Bayes' Theorem to the results of several tests known weakly to predict BOO in men with LUTS to assess if they improve the diagnostic accuracy of a flow rate test which alone is known to predict obstruction moderately well. Methods We applied Bayes' Theorem to data from 50 patients using Qmax alone and with the inclusion of additional variables (IPSS, PSA, and residual urine), to establish individual probabilities of BOO. The chi‐squared statistic (with trend) was used to compare the relative diagnostic values, against the BOO index calculated from the results of subsequent PFS. Results The diagnostic value of Qmax alone (chi‐squared = 9.2, P = 0.002), was superior than that for the Bayesian model using the combination of tests available (chi‐squared = 4.9, P = 0.026). Conclusions Although in our sample relevant additional tests do not improve the diagnostic power of Qmax as a predictor of BOO, we believe the Bayesian approach is conceptually suited to modeling clinical decision making but may be better tested for a more clinically relevant outcome such as treatment response. Neurourol. Urodynam. 27:797–801, 2008, © 2008 Wiley‐Liss, Inc.