Skip navigation

Making sense of testing

An immediate response to problems of diagnostic testing strategies is that of entrenched routine. 'We always do this' is a powerful argument, especially when no-one can remember why it is done. Who wants to rock the boat, after all? So examples of how to make some sense of a testing strategy should be cherished. A nice example comes from Groote Schuur in Cape Town.

What's the problem?

The questions being asked were whether routine full blood counts (FBC) on admission to a medical inpatient ward made sense. Historically 96% of patients had the test done, but how often did the results of the test influence management?


For 165 consecutive admissions an FBC was requested, which included haemoglobin, white cell count, platelets, mean corpuscular volume and haematocrit. Results were made known to the registrar, but to the consultant only after she/he had completed a questionnaire. This included whether, after a history and examination, an FBC was needed for the medical management of the patient. The result was then made known and the consultant had to indicate whether the result influenced patient management.

In order to assess agreement between consultants, two independent consultants examined the results of the first 50 patients in the same way, relying on clinical notes to see if they agreed with the original decision.

Results: top docs in kappa shock

There was moderate agreement between consultants and independent assessment on whether tests were needed (kappa 0.49 - 0 is complete disagreement, 1 is complete agreement; see Bandolier 43 . There was not much agreement, though, on whether the test assisted in patient care (kappa 0.06).

Consultants' clinical judgement was good. When they had a high suspicion of abnormality on clinical grounds, the test was abnormal 80% of time. Where there was no clinical indication for an FBC, there was a low (2%) probability that the result would assist patient management. That 2% came from a single case, and there was disagreement even over that as to whether there was actually any real effect on patient management.

The bottom line was that 25% of FBCs ordered were unnecessary, and that avoiding them by changing policy could save Groote Schuur about £3,500 a year.


Small beer, to some extent. Groote Schuur is a big hospital, and £3,500 and a few tests saved doesn't add up to much. But perhaps the main lessons lie elsewhere - in the excellent clinical skills of their consultants in judging when tests were needed, but their inability to agree when results of tests affect patient management. This may be one of those abstruse definitional things, but lots of disagreement indicates an area where further evidence or work is needed.

At the very least, this paper gives an interesting exemplar of how any institution might begin to assess how to make the best use of its diagnostic test budgets. A rolling program starting with highest volume or cost tests would be educational. Are there other examples we should know of?


  1. MS King, N White. The influence of the full blood count on medical inpatient management. South African Medical Journal 1997 87: 734-7.

previous or next story in this issue