Diagnostic process

The diagnostic process relates to how diagnosis is carried out in practice, how doctors achieve a final diagnosis, and how they decide which tests to order, if any. The diagnostic process starts with the demographic features of patients, followed by clinical history and examination and then leads on to further testing. Diagnostic accuracy studies tend to focus on further testing with little work having been done on the demographic and clinical investigations. One aspect that this project is interested in is this initial clinical diagnosis. This can also be thought about in terms of the pre-test probability of disease, which is determined by patients' demographic and clinical features.

Current projects

Patient journeys

Although diagnosis is often understood as a decontextualised activity taking place within the individual clinician's mind, we assume that the diagnostic process is also a social one; that reaching a diagnosis takes place not only in the clinician's head, but also in the interactions between clinicians and patients, clinicians and laboratory specialists, and between clinicians.

Our aim is to understand the process of diagnosis as it is currently practised in a real world clinical setting with specific focus on multiple sclerosis. We will use a qualitative approach to follow patients through the neurology clinic from time of first referral to final diagnosis. We will use observations, audio recording of clinic interactions and interviews. We intend to use our findings to develop clinical guidelines that will enable quantitative information to be understood and applied in context (e.g. results of a systematic review may need to be modified to fit in with the human factors involved in practice) and are of practical use to clinicians in their day-to-day work.

Systematic reviews methodology

Methods for the conduct of systematic reviews of diagnostic accuracy studies are still at a relatively early stage compared to methods for the conduct of systematic reviews of intervention studies. The Cochrane Collaboration is currently in the process of preparing to accept diagnostic systematic reviews and is providing guidelines on how these should be conducted. Members of the Evidence Based Diagnosis team are closely involved in this process. There is the potential to further develop these methods. Work is required on methods of analysis, in particular the different methods for pooling results and investigating heterogeneity. The most useful ways to present results of reviews, in particular the role of graphs and diagrams, is another area where little work has been done. This relates to the third theme on which this project will focus, the communication of diagnostic information.

Current projects

Systematic review of the accuracy of MRI for the diagnosis of MS

We are conducting a systematic review on the accuracy of MRI for the diagnosis of MS. This will provide us with a data-set which we can use to investigate some of the issues discussed above. In particular, this review raises interesting issues regarding how to synthesise data from studies that are very heterogeneous in terms of the threshold used to define a positive test result. It will also tie in with the qualitative work on "patient journeys". This work should provide the clinical context in which to interpret the results of the review. Understanding why neurologists order MRI scans will provide an insight into how the results of the review can be of benefit to clinical practice.

Meta-analysis methods

Several different methods have been proposed for meta-analysis of summary data from diagnostic studies. These differ in their assumptions, complexity and outputs. We are comparing four of these methods in order to:

• clarify and compare their assumptions.
• compare their results when applied to example meta-analyses.
• gain experience in fitting these models in different software.

Possible future areas of work

• how diagnostic algorithms can be developed from the results of a diagnostic systematic review.
• the role of "indeterminate" test results.
• the conduct of systematic reviews of comparative diagnostic accuracy studies.
• how to incorporate data on observer variability into a systematic review.
• the role of individual patient data me ta-analyses in diagnostic systematic reviews.

Communication diagnostic information

The measures commonly used to summarise diagnostic accuracy such as sensitivity, specificity, likelihood ratios and diagnostic odds ratios, are generally poorly understood. Some of the graphical methods used to portray these measures, in particular ROC plots, can also be difficult to understand. This theme will focus on examining which measures are currently used to present diagnostic accuracy information, how these measures are presented graphically, and how they are interpreted in practice. This work will be extended to try and improve the presentation and communication of diagnostic information.

Current projects

Presentation of diagnostic information

This project consists of two parts. The first will aim to identify all the different graphical and pictorial methods used to present diagnostic accuracy data. The second involves carrying out a methodological review of primary studies and systematic reviews to determine which of these methods they have used to present results, and to relate these to the measures of diagnostic accuracy which they report.

Primary diagnostic accuracy study methodology

Although more work has been done on the conduct of primary diagnostic accuracy studies than has been done on systematic reviews of such studies, there are still a number of areas where further work is required. The first is in the area of clinical prediction rules, or risk scoring schemes or prevalence functions as they are also known. A second area with links to this is the analysis and presentation of continuous test results. Lastly, we will consider the role of RCTs in the area of diagnosis. Some people argue that the only true way to determine whether a test should be used in practice is to examine its effect on patient outcome. This can be done using the RCT design by randomising patients to receive different testing strategies and then to compare patient outcomes between the groups. However, in many situations it is not practical to carry out such studies. We could look at situations in which it is not practical to carry out RCTs, and compare this to situations in which RCTs may be informative. The aim would be to produce guidelines on when RCTs should be carried out in the area of diagnosis.

Current projects

Continuous test results

Standard methods for analysing and presenting the results of diagnostic accuracy studies assume that continuous test results will ultimately be dichotomised into positive and negative results by choosing a suitable threshold. This has the disadvantage of losing of important information. For example, someone with a very high result on a particular test is much more likely to have the target condition than someone who has a result very close to the threshold. In addition if the threshold is chosen based on the results of the study the accuracy of the test may be overstated. We are exploring methods of analysis and presentation that avoid this dichotomisation.