This seminar introduces for discussion the idea of patients being given a summary in writing of how their individual symptoms, signs and test results have been used to arrive at a diagnosis and to choose treatment that will be of benefit to them. It also describes how predictive combinations of these symptoms, signs and test results can be indentified from research for use as evidence for diagnoses and benefit from treatment in the summaries.

The evidence-based summaries first outline the current problems as perceived by the patient. The various diagnoses are then listed and numbered. Each diagnosis is followed (e.g. in brackets) by its outline evidence: the presenting finding, the latest marker of progress and the types of finding used to confirm the diagnosis (e.g. ‘clin’ for clinical, ‘haem’ for haematological, etc,) with dates allowing a reader to look up the results if necessary. The tests to be done, the treatments to be stopped, continued and started are then listed followed in brackets by the numbered diagnostic indication. This information can be included as an ‘up-to-the minute’ past medical history in hospital referral letters and discharge summaries and updated subsequently in primary and secondary care. The evidence used to justify the diagnoses and treatments is arrived at as follows:

Postulate (or use established) possible differential diagnoses of presenting finding (e.g. albumin in the urine). Postulate (or use established) diagnostic criteria for each of the differential diagnoses. Postulate a finding (or a result or score that combines different results) that reflects disease severity (e.g. the albumin excretion rate – AER). Document these findings in a group of patients to be recruited into a RCT to compare the efficacy of a treatment (e.g. an angiotensin receptor blocker-ARB). Fit logistic calibrated regression functions to the data to create curves displaying the probabilities of an outcome (e.g. ‘nephropathy’) for a test result (e.g. AER) that reflects different degrees of disease severity. Assess the probabilities at each point on the curve for use in shared decision making. Identify the probability below which no one would opt for the treatment. This can be used a severity threshold in a diagnostic criterion (instead of upper 2 standard deviations of some reference population). The process can be repeated for different tests (e.g. albumin creatinine ratio) and other treatments (e.g. ACE inhibitors). A number of ‘sufficient’ diagnostic and treatment indication criteria can be formulated in this way to provide a choice for the diagnostician.

There will be other risk factors for ‘nephropathy’ (e.g. BP, HbA1c, etc) that cause treatment heterogeneity. Some will be modified by the same treatment (e.g. an ARB causes BP to fall too) and some not (e.g. an ARB does not reduce the HbA1c), these preliminary causal inferences being important when designing studies to estimate their combined effects. The latter can be explored by designing a multiple risk score (e.g. based on the AER, BP and HbA1c) and calculating this before randomization and analysing the result in the same was as described for the AER above.

Objectives The objective of this seminar is to discuss the idea of patients being given a summary in writing of how their individual symptoms, signs and test results have been used to arrive at a diagnosis and to choose treatment that will be of benefit to them. It also describes how predictive combinations of these symptoms, signs and test results can be identified from research for use as evidence for diagnoses and benefit from treatment in the summaries.

Method The evidence-based summaries first outline the current problems as perceived by the patient. The various diagnoses are then listed and numbered. Each diagnosis is followed (e.g. in brackets) by its outline evidence: the presenting finding, the latest marker of progress and the types of finding used to confirm the diagnosis (e.g. ‘clin’ for clinical, ‘haem’ for haematological, etc, with dates) allowing a reader to look up the results if necessary. The tests to be done, the treatments to be stopped, continued and started are then listed followed in brackets by the numbered diagnostic indication. This information can be included as an ‘up-to-the minute’ past medical history in hospital referral letters and discharge summaries and updated subsequently in primary and secondary care. The evidence suitable to justify the diagnoses and treatments is described under results.

Results Example results from an RCT to assess the efficacy of irbesartan in preventing nephropathy will be presented at each of the following steps: The possible differential diagnoses of presenting finding (e.g. albumin in the urine) and the diagnostic criteria for each of the differential diagnoses. A finding (or a result or score that combines different results) that reflects disease severity (e.g. the albumin excretion rate – AER). The result of fitting logistic calibrated regression functions to the data to create curves displaying the probabilities of an outcome (e.g. nephropathy) for a test result (e.g. AER) that reflects different degrees of disease severity. The probabilities at each point on the curve for use in shared decision making and the probability below which no one would opt for the treatment for use as a threshold for use in a diagnostic criterion.

Conclusions The process can be repeated for different tests (e.g. albumin creatinine ratio) and other treatments (e.g. ACE inhibitors). A number of ‘sufficient’ diagnostic and treatment indication criteria can be formulated in this way to provide a choice for the diagnostician. There will be other risk factors for ‘nephropathy’ (e.g. BP, HbA1c, etc) that cause treatment heterogeneity. Some will be modified by the same treatment (e.g. an ARB causes BP to fall too) and some not (e.g. an ARB does not reduce the HbA1c), these preliminary causal inferences being important when designing studies to estimate their combined effects. The latter can be explored by designing a multiple risk score (e.g. based on the AER, BP and HbA1c) and measuring this before randomisation and analysing the result in the same was as described for the AER above.