Abstract
Patients presenting with symptoms suggestive of urinary tract infection were recruited in a general practice survey aimed at measuring the predictive value of symptoms, history and urine dipstick testing for diagnosing the presence of bacterial infection. Urine specimens were obtained from 87% of the 521 patients recruited. A diagnosis of infection was established by urine culture producing a colony count in a pure culture exceeding 100 000 organisms per ml or between 10 000 and 100 000 organisms per ml plus a minimum of 100 leucocytes per mm3.
Occurrence rates for symptoms and other items of information in infected and non-infected groups were used to derive their positive and negative predictive values in making the diagnosis. The predictive value of volunteered symptoms was compared with that of elicited and volunteered symptoms combined. The positive predictive value of symptoms was increased where elicited symptoms were included but this was achieved at the cost of diminishing the negative predictive value. The occurrence rates were used to derive a mathematical model for diagnosing infection. The symptoms-history-urinalysis (SHU) score generated in this model compared well with a computer predicted probability. Both were substantially better than the assessment and action (decision to prescribe an antibiotic) of the recording doctor.
The scoring method described has been demonstrated in urinary tract infection but may be applied to any symptom combination related to a diagnosis for which there is an agreed definition.
- © Journal of the Royal College of General Practitioners