Risk assessment, symptoms, and signs as predictors of vulvovaginal and cervical infections in an urban US STD clinic: implications for use of STD algorithms.

OBJECTIVE: To identify clinical epidemiological correlates of cervical and vaginal infections and assess alternative algorithms, including two new reproductive tract infection (RTI) algorithms, for syndromic management of these infections. DESIGN, SETTING AND
SUBJECTS: We prospectively studied clinical manifestations and risk correlates of cervical and vaginal infections in a randomly sampled group of 779 female patients seeking evaluation for a new problem at a Seattle STD clinic.
METHODS: One experienced clinician performed standardised history, physical examination, and microscopy. Reference laboratories performed microbiological tests. Three levels of retrospective evaluation of algorithms included risk assessment and symptom review (RAS) alone; addition of speculum and bimanual examinations; and further addition of microscopy.
RESULTS: (1) Chief complaint of abnormal vaginal discharge predicted a significantly lower rate of gonorrhoea (GC) or chlamydial infection (CT) than rates observed with no complaint of vaginal discharge. Only the elicited symptom of yellow vaginal discharge (not the more common symptoms of increased or malodorous vaginal discharge) predicted GC or CT. Chief complaint of abnormal vaginal discharge itself predicted trichomoniasis (TV) and bacterial vaginosis (BV), not cervical infection. Candida albicans was strongly associated with the chief complaint of vulvar pruritus, not with the chief complaint of abnormal vaginal discharge. (2) Applying these algorithms in STD clinics only to women with the chief complaint of abnormal vaginal discharge, rather than to all women, decreases sensitivity for GC or CT, without increasing positive predictive value (PPV). Criteria for inclusion of patients have more effect on the performance of these algorithms than do the levels of evaluation used. (3) A modified World Health Organisation (WHO) algorithm applied only to patients with symptoms of vaginal discharge, involving treatment of RAS positives for cervical infection, followed by treatment of vaginal infections and cervicitis based on examination of RAS negatives and positives, had a sensitivity of 50% and PPV of 33% for cervical infection, and very low sensitivity for BV, TV, and for vulvovaginal candidiasis (VVC). (4) An RTI algorithm derived from these data, and applied to all STD patients, involving RAS and examination of all RAS negatives, provided treatment to all cases of BV and TV associated with symptoms of vaginal discharge; treatment of all VVC associated with symptoms of vulvar pruritus; treatment for GC and GT to all RAS positives (using easily elicited risk factors) and to RAS negatives with signs of cervicitis or PID. This algorithm had a sensitivity of 87% and a PPV of 33% for GC or CT in this population, with its 24% prevalence of GC or CT. The sensitivity for BV, TV, and VVC greatly exceeded that of the modified WHO algorithm. (5) A modified RTI algorithm, involving examination rather than treatment of RAS positive women, no examination of RAS negatives, decreased the sensitivity for cervical infection to 55% but increased the PPV to 51%.
CONCLUSIONS: Syndromic management of vaginal discharge offers relief of symptoms, prevention of transmission of trichomonas, and perhaps prevention of complications of BV. The 51% PPV of the modified RTI algorithm probably would warrant treatment and partner notification for GC and CT in settings with similar rates of GC and CT where more specific tests are lacking. However, as the prevalence of GC or CT decreases, the ratio of uninfected to infected who receive treatment with these algorithms would increase greatly, making the algorithms potential victims of their own success.