From the clinical information, several symptom keywords coded predicted the presence or absence of E. coli along with pyuria. The specific mention of 'absence of dysuria' predicted no growth with no pyuria. On removal of specimens that mentioned dysuria, the strong predictors of UTI included blood, smell, loin pain and fever. Other words commonly associated with UTI were frequency and suprapubic pain; these were not are not associated with E. coli and pyuria. When coding for combined words of culture results for diagnosis of UTI, it was found that isolation of coliforms, Staphylococcus saprophyticus or Proteus, had high odds ratios for clinical diagnosis of UTI. With the help of new gold standards, it can be established that clinical details have high sensitivity (92%) but low specificity (52%). Culture has moderate sensitivity (77%) and specificity (93%). Pyuria has high specificity (99%) but low sensitivity (53%). (Figure 1). 

 

In conclusion, the diagnosis of UTI is incomplete with urine culture information alone. The presence or absence of pyuria and E. coli colonisation if unreported may lead to unnecessary use of antibiotics. 

 

Figure 1: Performance of clinical and lab gold standards