Diagnostic research in daily practice: methodological challenges and applications

The ultimate goal in the evaluation of a diagnostic test, model or strategy, is whether it improves the health of the targeted patients. The striking difference between therapeutic interventions and diagnostic strategies is that diagnostic strategies generally have indirect effects on patient’s prognosis via generating information that can be used by the physician to guide their patient management, which in turn may improve a patient’s prognosis. This stepwise, and interlinked, diagnostic and treatment decision making process provides in fact two different levels to evaluate a diagnostic strategy. The first focuses on the examination of the quality or rather accuracy of the information generated by a diagnostic strategy and the second level the consequences of the results of the diagnostic strategy on therapeutic management decisions and on the actual prognosis of the patients. In the first part of this thesis are two methodological challenges in diagnostic research discussed: the use of a panel diagnosis or composite reference standards as possible solutions to deal with the absence of a single error-free reference standard, and the advantages of multinomial logistic regresion modeling when dealing with multiple alternative diagnoses. In the second part of thesis are the results of a diagnostic screening study. The screening strategy yielded a considerable proportion of new daignoses of heart failure and COPD. However, the influence of the new diagnoses according to the strategy on the changes in patent management was only marginal. These findings illustrate the importance of gathering follow-up data on the impact of a (new) diagnostic strategy on changes in patient management. The studies presented in this thesis underline the difficulties in diagnostic research and illustrate that diagnostic accuracy is better placed as an intermediate outcome in the evaluation of a diagnostic test, instead of a formal outcome. From this perspective, the results of diagnostic accuracy studies reflect the potential impact of a diagnostic strategy on the prognosis of patients through changes in therapy. To enhance such inferences of diagnostic accuracy studies, the design, analyses and reporting of current studies can be optimized.