|
.gif)
|
|
The sensitivity and
positive predictive value of STI algorithms in use at a female sex
worker clinic in Abidjan, Côte d'Ivoire, have decreased since they
were originally validated.
Assessment of the
validity of and adherence to sexually transmitted infection algorithms at
a female sex worker clinic in Abidjan, Côte D'Ivoire.
Vuylsteke BL, Ettiegne-Traore V, Anoma CK,
Bandama C, Ghys PD, Maurice CE, Van Dyck E, Wiktor SZ, Laga M.
Sexually
Transmitted Diseases 2003;30;284-291
Summary:
Question
How well do the algorithms used at a female sex worker clinic in Abidjan,
Côte d'Ivoire, which were designed in 1993, currently perform for the
diagnosis of vaginal and cervical infections when applied to returning
clients, and how accurately are they used by the health workers?
Design
The performance of algorithms designed in 1993 and used for sexually
transmitted infection (STI) case management at a female sex worker (FSW)
clinic in Abidjan, Côte d'Ivoire, were evaluated in a cross-sectional
study conducted from 1999 to 2000 among FSWs attending as returning
clients. The adherence of health workers to the application of the
algorithms was also assessed.
Participants
Eight hundred forty-two female sex workers attending the Clinique de
Confiance in Abidjan, Côte d'Ivoire, as returning clients were tested.
Women younger than 18 years, attending for the first time, or who had
already participated once were excluded. The median age was 25 years
(range = 18-53 years), the median duration of sex work was 16 months, and
94% reported using a condom with all their clients during the past working
day.
Description of Tests and Diagnostic
Standard
Physicians applied one of three STI algorithms to women who spontaneously
declared an STI symptom, including 1) women with vaginal discharge,
dysuria, or itching, 2) women with lower abdominal pain, and 3) women with
both symptoms. A gynecological examination with speculum and vaginal
and/or cervical sampling were performed according to the algorithm.
Cervical smears were examined for the number of leukocytes per microscopic
field, and vaginal smears for the presence of T. vaginalis and
candida. Sociodemographic, behavioral, and other information were
collected on a standardized questionnaire and used to calculate a risk
score based on age, money received for last sex act, and duration of sex
work.
A first-void urine sample and two
self-collected vaginal swabs were obtained for laboratory testing. The
urine was tested by PCR for N. gonorrhoeae and C. trachomatis
DNA (Amplicor, Roche Diagnostic Systems, Branchburg, NJ). A Gram stain was
prepared from one of the vaginal swabs and examined for the presence of
yeast and clue cells, and scored for bacterial vaginosis organisms
according to the Nugent method. The same swab was used to inoculate medium
for culture of T. vaginalis. The second swab was shipped dry and
frozen to a laboratory in Belgium and tested for T. vaginalis using
two in-house PCR assays. Women were considered positive for C.
trachomatis or N. gonorrhoeae if the PCR assay was positive;
for T. vaginalis if the culture or if both of the PCR assays were
positive; for yeast infection if mycelia were seen on the vaginal smear
and abnormal vaginal discharge was observed; for bacterial vaginosis if
the Nugent score was >6 or clue cells were present on the vaginal
smear.
Main Outcome Measures
The sensitivity, specificity, and positive predictive value (PPV) of the
algorithms were determined by comparing the outcome of the algorithm as
noted by the physician with the gold standard laboratory diagnosis.
Adherence of the health staff to the algorithm was evaluated by
calculating the proportion of women for whom all steps of the algorithms
were followed correctly.
Main Results
The rates of STI determined by laboratory diagnosis were 8.2% N.
gonorrhoeae and/or C. trachomatis, 16.7% T. vaginalis,
62.3% bacterial vaginosis, and 6.2% yeast. An alternative risk assessment,
taking into account age, prompted symptoms, and time since last visit was
significantly associated with and highly predictive of gonococcal and/or
chlamydial infection. Vaginal discharge was predictive of yeast infection
and was the only symptom predictive of any infection. The performance of
the algorithms for diagnosis of STI as determined by the laboratory
analyses is shown in the table. Correct case management by health workers
using the algorithms was lowest at the first step and highest at the last
step of the algorithms. The overall rate of correct case management was
30%.
Authors' Conclusions
The sensitivity and PPV of the algorithms
currently in use in a FSW clinic in Abidjan decreased over time. The
sensitivity and PPV for N. gonorrhoeae and/or C. trachomatis
were 79% and 48% five years ago compared to 20% and 14% in the current
study. Several factors could explain the difference including a change in
sociodemographic and risk behavior, the use of dynamic rather than static
factors in the risk assessment, a lower prevalence of N. gonorrhoeae
and C. trachomatis, the restriction of subjects to returning
clients, and the restricted criteria for entry into the algorithm. Poor
adherence to the algorithms further reduced the validity. Although the
performance of the algorithms was poor, they remain the only practical
tool available for the diagnosis of STI in FSWs in Abidjan, and should be
periodically evaluated and adapted to the changing characteristics of the
population.
Source of funding: Wellcome Trust,
London, UK and the Belgian International Cooperation (DGCI).
For correspondence: Bea Vuylsteke,
Project RETRO-CI, BP 1712 Abidjan 01, Côte d'Ivoire. E-mail address:
bbv1@cdc.gov.
|
