OBJECTIVE: To evaluate the performance of the WHO algorithm for the detection of cervical infection in women presenting with vaginal discharge and modify the risk assessment score for optimum effectiveness in Malawi.

METHODS: 550 consecutive women presenting with non-ulcerative genitourinary complaints were interviewed and examined. Cervical infection was defined as presence of Neisseria gonorrhoeae on culture and/or Chlamydia trachomatis by EIA. Other laboratory investigations included wet mount microscopy, serology for syphilis and HIV, LED testing of cervical and vaginal secretions, and pH testing of vaginal fluid. Sensitivity, specificity, and positive predictive values (PPV) of different algorithms were determined in the analysis.

RESULTS: Cervical infection was identified in 19.5% of women (17.1% gonorrhoea, 3.7% chlamydial infection). The sensitivity/specificity/PPV of the WHO risk assessment were 43%/73%/28%, respectively by history and 62%/61%/27% with the addition of speculum examination. Using Malawi results to modify the risk assessment improved the performance to 61%/68%/31% respectively by history alone, which increased to 73%/64%/33% with bimanual examination and 72%/56%/29% with speculum examination.

CONCLUSION: The sensitivity of the WHO risk assessment is low for the detection of cervical infection in Malawi. Although the Malawi risk assessment performed somewhat better on history alone, this study identified external and bimanual examination variables that improved the diagnostic performance of the algorithm in settings where speculum examination is not possible. Although the PPVs of the algorithms are low, country specific risk assessments can provide a framework for management until simple, affordable diagnostic tests for the definitive diagnosis of cervical infection are available.