Predictive values instead of normal ranges: less data, more information

Diagnostic strategies can have various goals at two levels: to facilitate the diagnostic process on the cognitive level, and to serve considerations on the level of the doctor–patient relationship. Requests for laboratory tests could be intended to exclude a disease or to affirm the presence of disease. Thirdly, tactical motives to smoothen the negotiations between doctor and patient probably seem to be important as well. These three intentions differ in prior probability, should lead to different sets of tests, and to different interpretations. Even the cut-off points should differ. This leads to three different decision strategies, both at requesting, as at interpreting the results. Following this line of thought, post-test probabilities are more suitable than normal ranges. Excluding strategy: this is the most prevalent. However, the disadvantage of an excluding strategy (prior 1–5%) is a false-positive result. A positive test result should lead to follow-up by wait and see or by repeated testing. More extensive testing usually is not a very sensible strategy. In practice, physicians simply ignore slightly abnormal values. Mentally they put the cut-off points for normality more broader. The number of tests is small. Confirmative strategy: the disadvantage of a confirmative intention (prior 10–30%) is a false-negative result. Follow-up without testing, repeated testing, or even accepting marginal normal results as abnormal is a proper strategy. The number of tests is moderate to high. Tactical strategy: the tactical intention strategy to reassure the patient – or avoid referrals – could lead to ignoring all slightly positive test results by choosing a higher cut-off point. Actually, considering the usual insignificant diagnostic gain when testing for tactical reasons, all test results are clinically insignificant, unsuspected outliers excluded. Here, a very limited set of tests should be chosen. The laboratory test is the currency in mutual trading medical expectations and relationship considerations between doctor and patient. The number of tests is minimal. If the physician chooses a strategy, a limited range of prior probability is chosen. Then a possibly computerized algorithm produces a “Value (posterior probability)” as test result, replacing “Value (normal ranges)”. Thus one number less on the lab form, yielding more significant information.Presented at the 10th Conference Quality in the Spotlight, March 2005, Antwerp, Belgium