Skip navigation

Diagnostic strategies

The Panzer reference

Bandolier 28 examined the use of likelihood ratios (LR) for use with diagnostic tests as an alternative to figures on sensitivity and specificity. One of our eminent readers sent an email with "the Panzer reference" which he thought would be useful for readers to know of because it has many examples of diagnostic strategies using LR methods.

The book [1] is published by the American College of Physicians. As the foreword says "We are moving, somewhat haltingly, from intuitive or informal clinical judgements based on the experience of individual practitioners to a more formal process in which evidence from studies of groups of similar patients augments the judgement of practitioners. The change portends an exciting new era in medicine as it harnesses, for clinical care, currently underused information about diagnostic tests derived from intensive biomedical research".
The book first examines general issues on the characteristics of diagnostic tests, about where to find information and how to use it but then moves quickly into the specific - nearly 50 short chapters examining diagnostic strategies. Each is packed with useful information and insight - a guide book rather than a rule book.

Most doctors and many scientists involved with diagnostic procedures will find this a worthwhile read. A note of caution, though, because it isn't all as clear as it could be, and there seem to be a few leaps of faith, though the general approach is excellent.

A Bandolier prize (signed copy of Bandolier - the first 20 issues ) for the first person to write with an explanation of how pre-test probabilities of hypothyroidism are arrived at in chapter 39.


RJ Panzer, ER Black, PF Griner (Eds). Diagnostic Strategies for Common Medical Problems. 1991. American College of Physicians. 550pp. ISBN 0-943126-20-7. US$38 (in the USA).

LR for urine dipsticks for UTI

The use of dipstick testing for urinary tract infections (UTI) was mentioned in Bandolier 27 . The paper examined how sensitivity and specificity change with different levels of clinical suspicion [1]. We can use data from that paper to see how LRs might work.

Of 366 patients with suspected UTI, 72 were positive by laboratory urine culture with >100,000 CFU/mL as a threshold. This prevalence of 20% is similar to that of 25% found in a UK general practice study [2]. This forms the pre-test probability.

However, both reports showed that using a combination of frequencies of symptoms (dysuria, nocturia, frequency) and signs (pelvic tenderness, costovertebral angle tenderness) it was possible to increase the probability of UTI to more than 50%. The UK study used an explicit scoring system.

With overall sensitivity of 0.83 and specificity of 0.71, the positive LR was 3 and the negative LR was 0.2.

The nomogram shows the post-test probabilities for a patient without symptoms with a positive test (Pos - about 55%) and one with symptoms with a positive test (S&S - about 90%). The post-test probability for a patient without symptoms and a negative test (Neg) was about 5%.

One of the papers [2] indicated that a probability of above 50% represented a suitable level for initiating antibiotic treatment in cases of suspected UTI. So using symptoms plus test to generate pre-test probability and likelihood ratios respectively could be useful, and much of the data needed could be generated from audit.

Bandolier would like to hear from readers using clinical scoring symptoms for common conditions like UTI or thyroid disease, especially in combination with tests.


  1. MS Lachs, I Nachamkin, PH Edelstein et al. Spectrum bias in the evaluation of diagnostic tests: lessons from the rapid dipstick test for urinary tract infection. Annals of Internal Medicine 1992 117: 135-40.
  2. FF Dobbs, DM Fleming. A simple scoring system for evaluating symptoms, history and urine dipstick testing in the diagnosis of urinary tract infection. Journal of the Royal College of General Practitioners 1987 37: 100-4.

previous or next story in this issue