Detecting the presence of vaginal or cervical infection based solely on signs and symptoms is one of the major challenges to the syndromic management of STDs. Most of the studies evaluate the ability of VD algorithms to detect cervical infections, specifically those caused by C trachomatis and N gonorrhoeae. Although some of the studies did report the efficacy of the algorithms in detecting vaginal infections (bacterial vaginosis, T vaginalis, and candidiasis), they are not the focus of this report and the performance of VD algorithms in the detection of these infections is not reported. The WHO has created algorithms for use in settings with and without the ability to perform either speculum examinations or laboratory diagnostics (Figure 3).3 In a situation where no speculum examination was performed, healthcare providers asked patients a series of questions as part of the algorithm and flowchart; an affirmative answer to these questions may indicate infection and prompt treatment. When a speculum examination is implemented, the findings of the examination are incorporated into the algorithm and flowchart and contribute to the treatment decision made. In addition, some of the studies incorporate an algorithm that considers the report of LAP, a symptom that is often reported by women with pelvic inflammatory disease. Etiologic agents of pelvic inflammatory disease include N gonorrhoeae and C trachomatis7 and, therefore, LAP may indicate infection with either of these organisms. Many of the VD algorithms require that a woman present with VD before she can be evaluated using an algorithm; in other words, VD is the primary entry point into the algorithm. Furthermore, some studies use the algorithms as a way to diagnose and treat persons presenting with reports of VD, whereas others use the algorithms to screen women who have come to the healthcare center for reasons other than urogenital symptoms. It is when an algorithm is being used as a screening test that algorithms have often not proved accurate enough to detecting cervical infection, especially among low‐risk populations.
To improve the sensitivity and specificity of the VD algorithms to detect cervical infection, risk determinants and risk scores have been incorporated into many algorithms (Tables 3 and 4). Risk determinants are characteristics that are most closely associated with cervical or vaginal infection in a specific population, such as specific demographic factors, sexual history, laboratory diagnostic results, and physical signs and symptoms. These characteristics are often determined by entering all potential risk factors into a multivariate statistical model and determining which are most highly correlated or associated with vaginal or cervical infection. Those characteristics that are significantly associated with infection are then assigned a numeric value (often the value of the odds ratio resulting from the multivariate analysis). A total risk score for a person is attained by summing the numeric value of each characteristic present for that person. A numeric threshold or cut‐off value is then determined for the total score (e.g., cut‐off value of 14 or 28) after which point treatment is indicated (Appendix). However, algorithms using only risk determinants do not assign a numeric score, but rather provide treatment if, for example, two of three risk characteristics are present in a person. Both risk scores and determinants identify women likely to be infected with N gonorrhoeae and C trachomatis by assessing how well these women fit the profile of persons at risk for cervical infection in the target population. Risk determinants vary among populations, and it is possible that a set of characteristics that indicate risk in one community will be different from those in another community.
Five studies were reviewed that implemented the algorithms for UD: three in Africa, one in Indonesia, and one in Brazil. Three of the studies evaluated algorithms with laboratory support, and four studies assessed the algorithm without laboratory assistance (Table 1). Four of the five studies were conducted in primary healthcare clinics, the last study was implemented in an STD clinic. The sensitivities for the studies ranged from 87% to 99%. Two studies used cure rates to evaluate the success of the algorithms; these values were equally high, ranging from 92% to 99%. The algorithm without laboratory support performed well in all four studies evaluating it, with sensitivities of more than 91%. Specificity levels varied from 7% to 90%, so treatment of patients without N gonorrhoeae or C trachomatis was not uncommon when using these algorithms. Specificity was lowest for chlamydia (7%) and highest for gonorrhea when symptoms were combined with Gram stain results (99.1%).9
The results of the studies evaluated indicate that the algorithms that use microscopy tend to have higher specificities and PPVs than those that do not. In the study by Moherdaui et al,9 the addition of gram staining increased the specificity of the algorithm for N gonorrhoeae by more than 80% while decreasing the sensitivity only slightly (2%); however, the addition did not improve the specificity in detecting C trachomatis. In addition, the PPV was high (99.1%) compared with no microscopy (57.5%). In contrast, the study by Alary et al10 did not detect as dramatic an increase in specificity (7%) when microscopy was added, nor did it overtly affect the sensitivity (4% decrease) or PPV (3% increase). The addition of laboratory support to the UD algorithm may slightly decrease the sensitivity of the test while increasing the specificity and PPV.
Genital Ulcer Disease
Five studies evaluated the algorithms for the detection of genital ulcers (Table 2). Only one study implemented the algorithm with laboratory support11; the others evaluated algorithms where no laboratory support was possible. Three studies were conducted in primary healthcare clinics, and two were conducted in STD clinics. The majority of the studies were conducted in Africa, and one study was conducted in Brazil.9 Three studies used cure rates as the evaluation measure and one study used the percentage of patients receiving adequate treatment; only one study reported the sensitivity of the algorithms. The ability of the GUD algorithms to detect syphilis and chancroid was high overall, with sensitivities ranging from 72% to 100%. The results were also high in the four studies that used cure rates and adequate treatment as the evaluation measures, ranging from 68% to 100%. Only one study evaluated the sensitivity of the WHO algorithm in detecting the various STDs that can lead to genital ulcers.11 In this study, the WHO algorithm with and without laboratory support was found to have high sensitivities in detecting syphilis and chancroid, but both algorithms performed poorly in detecting herpes (4.5%). To improve the sensitivity of algorithms for detecting genital herpes and lymphogranuloma venereum, Ye Htun et al28 used an algorithm that attempted to better identify these cases. Using this protocol, 90% of patients received appropriate treatment, but 57% of patients who did not have syphilis were treated for the disease. The study by Bogaerts et al11 was the only study that compared the algorithms with and without laboratory support. The results show that the sensitivity for syphilis or herpes did not change substantially when laboratory support was added to the algorithm; however, sensitivity for chancroid decreased by 28% when laboratory support was included. PPVs and specificities were not reported for any of the studies reviewed, so overtreatment is hard to evaluate. Clinical diagnosis was evaluated in two studies and was not effective in detecting the majority of GUD infections.
All studies evaluated some version of the WHO algorithm for VD. In these studies, the VD algorithms were used to detect cervical infection. For a person to be assessed using this algorithm, she must report symptoms of VD. Of the studies evaluating the VD algorithm, four of seven implemented the algorithm among women presenting to the clinic with symptoms of VD or other urogenital symptoms. The other three studies were implemented among a random sample of women at different clinic types (antenatal, family planning); women in this sample who reported having VD or urogenital symptoms were entered into the algorithm. Six of the seven studies were conducted in Africa, the remaining study was conducted in Jamaica.12 The majority of studies were conducted among antenatal and primary healthcare populations. Two studies also incorporated the symptom of LAP into the algorithm to help detect cervical infection.
The sensitivities reported by the studies evaluating the VD algorithms among symptomatic women are high ranging, from 73% to 93% (Table 5). Two studies only reported the percent of women who were cured of VD on a return visit; these values ranged from 86% to 96%. In the study by Alary et al, the algorithm incorporating a speculum examination increased the sensitivity of the algorithm slightly (6%), while decreasing specificity by 6%.10 La Ruche et al13 found that 34% of cases assessed using the algorithms without speculum examination received inadequate treatment as opposed to 6% using the algorithm with speculum examination. Further, the study by Alary et al incorporating LAP provided the highest combination of sensitivities and specificities of the algorithms that were evaluated among symptomatic women. PPVs were highest among the high‐risk clinic populations (STD clinics) and lowest among the low‐risk populations (primary healthcare centers).10
All women. When implementing the WHO algorithm for VD among women not attending a healthcare provider because of symptoms of VD, the sensitivity of the algorithm decreases (range, 29‐86%). Vuylsteke et al and Diallo et al14 found a major decrease in sensitivity when speculum examination was added to the algorithm. In the study by Vuylsteke et al, which was conducted among an antenatal population, the sensitivity dropped from 48% to 29% when speculum examination was added with only a slight increase (10%) in specificity.4 Diallo et al observed a similar decrease in sensitivity (44‐18%) when speculum examination was added as part of the diagnosis among a sex worker population.14 The study by Vuylsteke et al also incorporated LAP as part of their algorithm.4 Among low‐risk women, incorporating symptoms of LAP into the algorithm did not appear to increase the accuracy over those that did not incorporate LAP. Overall, the PPVs of the studies reviewed were low among the low‐risk populations (approximately 10%) and higher in the studies conducted among sex workers and in STD clinic (46‐67%). Therefore, among low‐risk populations, approximately 90% of patients were treated unnecessarily.
Risk scores for vaginal discharge. Thirteen studies incorporated risk scores into the algorithms for VD. The majority of studies were conducted in Africa, whereas other study sites included Haiti and Jamaica (Table 3). A large proportion of the studies were conducted among low‐risk populations, such as in antenatal clinics or family planning clinics. Overall, sensitivities for the detection of N gonorrhoeae and C trachomatis ranged from 10% to 98%, specificities ranged from 15% to 95%, and PPVs ranged from 5% to 48%; sensitivities increased as the corresponding specificities decreased. The determinants used in the risk scores varied substantially among studies (Table 6).
Sensitivities did not increase dramatically when speculum examination was added to the risk algorithms. At a cut‐off risk score of 8, Gertig et al15 observed a sensitivity and specificity of 98% and 15% with speculum examination, respectively. Without the examination, the corresponding sensitivity and specificity were 97% and 17%, and the PPV was 5% in both cases.15 When Behets et al16 added speculum examination to their algorithm, the sensitivity rose from 71.1% to 89.2% while the specificity dropped from 61.5% to 42.7% at a risk‐score of 4; the PPV decreased from 21.5 to 18.8%. Mayaud et al17 observed an increase in sensitivity of approximately 5% when speculum examination was added to the algorithm, while specificities and PPVs remained fairly constant or decreased slightly. In contrast to the studies that observed an increase in sensitivity, Diallo et al observed a small decrease in sensitivity (4%) when speculum examination was added, but detected an increase in specificity (32% to 54%).14
Several risk algorithms provided moderate to high sensitivities and specificities in detecting cervical infection. Vuylsteke et al produced the highest combination of sensitivities and specificities using a risk‐score algorithm that used a leukocyte‐esterase dipstick test; this algorithm provided a sensitivity of 72% and a specificity of 73.5% among an antenatal clinic population, though the PPV remained low (15.8%).4
Seven studies used a number of specified risk factors or determinants in conjunction with the algorithms (Table 4). These studies ranged in location from the United States to Africa, Turkey, Brazil, and the Philippines. Three of the seven studies were conducted among low‐risk populations; the remaining four were conducted among higher‐risk groups. Those algorithms implemented among higher‐risk populations had higher sensitivities, specificities, and PPVs than those conducted among lower‐risk groups. The majority of sensitivities for the high‐risk populations ranged from approximately 40% to 70%, whereas sensitivities for the low‐risk populations ranged from 0% to 80%. Using an algorithm that did not use VD as the primary entry point, Ryan et al18 observed a sensitivity of 76% and a specificity of 57% without speculum examination and a slightly higher PPV of 36%. With speculum examination, the sensitivity of the algorithm increased by approximately 10%, and the specificity decreased by approximately 15%.
Effective STD‐control programs are essential in developing countries to limit the transmission of STDs, including HIV. With or without microscopy, the WHO algorithms for UD were able to detect gonorrhea or chlamydial infection with a sensitivity greater than 90% in almost all of the studies evaluated, indicating that a large number of persons with UD were effectively identified and treated. The majority of the studies reviewed here were conducted without microscopy, primarily because UD alone is a good indicator for N gonorrhoeae and C trachomatis infection in men. The WHO algorithm for UD without microscopy consistently had the highest sensitivity rates (91‐99%) (Table 1). Low specificity rates for chlamydia detection were observed by Moherdaui et al, indicating that inappropriate treatment was prescribed; patients may have been either uninfected or had another infection that caused UD. None of the three studies that reported specificities observed overly high rates, but because of the high sensitivities and relatively high PPVs, the algorithms reviewed were successful in identifying and treating persons with gonorrhea and chlamydia without heavily overtreating uninfected persons. The study by Moherdaui et al was the exception to this finding, and had a low specificity for chlamydia. Although adding gram staining to the algorithm increased the specificity in detecting gonorrhea, it did not improve the specificity for chlamydia.9 This may be because rapid detection tests are fairly insensitive in detecting chlamydia.8
Syndromic‐treatment algorithms for GUD showed a range in cure rates, from 100% in Cote d'Ivoire to 68.4% among men in Zambia. The algorithms were most effective in identifying syphilis and chancroid. Bogaerts et al found that adding simple laboratory tests to the algorithm decreased the ability of the algorithm to detect chancroid. The authors speculate that this is because when an rapid plasma reagin test result is positive, a person only receives treatment for syphilis; therefore, persons who are dually infected with chancroid and syphilis are missed or latent syphilis in persons who are currently infected with chancroid is detected. The authors recommend that if rapid plasma reagin tests are used, persons with a positive test result should be treated for both syphilis and chancroid, and persons with a negative results should be treated for chancroid.11 GUD algorithms were not effective at identifying herpes, as demonstrated by the 4.5% sensitivity observed by Bogaerts et al. The algorithm created by Ye Htun et al28 to detect herpes and lymphogranuloma venereum was 90% sensitive; however, it resulted in a large number of uninfected patients being treated for primary syphilis. It is not clear exactly why the algorithm used by Ye Htun et al was more effective in detecting more herpes cases than the algorithm used by Bogaerts et al, but it may be that the prevalence of infection was higher in Lesotho than in Rwanda or that providers were more sensitized to recognizing and managing herpes in Lesotho. In the recent mass‐treatment trial in Rakai, Uganda, 40% of all GUDs were attributable to herpes simplex virus type 2, which is not responsive to current syndromic‐management treatment regimes.19 As the role that herpes may play in increased transmission of HIV is better understood, herpes detection and suppression may become more important. These results indicate that the use of syndromic algorithms alone is a highly effective method for the detection and treatment of syphilis and chancroid, but not as effective in detecting herpes and other organisms that cause GUD.
The success of the VD algorithms for the detection of cervical infection in women varied substantially in the studies reviewed, depending greatly on the population in which the algorithm was implemented. Sensitivities ranged from 29.3% among low‐risk women visiting an antenatal clinic in Zaire to 93.3% among symptomatic women in a primary healthcare center in Benin. The majority of studies evaluated here focus on the ability of algorithms to detect cervical infection with gonorrhea or chlamydia because most infections are asymptomatic, and without treatment the negative sequelae are serious and costly. Syndromic algorithms for detecting cervical infections also have not been highly successful to date. Furthermore, more biologic evidence suggests that cervical infection attracts CD4 cells, the target cells for HIV, to the endocervix.1 Nevertheless, given the association that has been documented between vaginal infections, such as bacterial vaginosis and trichomoniasis, and the increased risk of HIV transmission (odds ratios, 3.7 and 1.9‐4.7, respectively),1 the importance of detecting and treating vaginal infections should not be overlooked or underemphasized.
The addition of speculum examination to an algorithm may increase or decrease the likelihood that a person will be diagnosed as having a cervical infection. If the provider observes on speculum examination, for example, cervical motion tenderness or mucopus coming from the cervix, then treatment of N gonorrhoeae and C trachomatis is recommended by the WHO.3 Nevertheless, many times these infections are asymptomatic and speculum examinations may find none of these signs present, which may result in providers not prescribing treatment to asymptomatic women. In the studies evaluated here, the WHO algorithms for VD without speculum examination resulted in moderate sensitivities. The addition of speculum examination to the algorithms tended to slightly increase the sensitivity of an algorithm to detect cervical infections; five of seven algorithms comparing algorithms with and without speculum examination found a slight increase with speculum examination. Nevertheless, Diallo et al14 and Vuylsteke et al4 observed that when implementing the algorithm to a mixed population of infected and noninfected women, the sensitivity of the algorithms decreased with the addition of the speculum examination, suggesting that signs suggestive of cervical infection (e.g., cervical mucopus) are not always sensitive enough to detect gonorrhea and chlamydia. Furthermore, speculum examination significantly increases the time and cost of STD treatment while only moderately increasing sensitivity.16–18
Because algorithms for VD produce a range of sensitivities, risk scores are a means to improve the detection of women with true‐positive results. Many studies focused on the challenge of detecting cervical infection among low‐risk populations. The use of risk scores tends to increase both the sensitivity and specificity of the algorithm to detect cervical infections in all populations. In many studies, the higher the cut‐off point the lower the sensitivity and higher the specificity; this is the result of requiring more risk factors to be present before a person is given treatment.
Risk scores offer the opportunity to tailor questions in a culturally relevant way. Nevertheless, not all risk assessments will uncover the true risk factors within a population, and depending on the population different risk factors will be more important than others. When Thomas et al20 applied the algorithms used by Vuylsteke et al4 in Zaire and by Mayaud et al in Tanzania17,21 to a study population in Kenya, few of the demographic or behavioral risk factors identified by other researchers were associated with cervical infections among their study population. Mayaud et al21 recommend that variables in a risk score should be piloted before their widespread introduction.
Risk factors may not accurately identify persons at risk for STDs within a community, and such indicators may not be culturally appropriate and could contribute to stigmatization by labeling persons as high‐risk or promiscuous. Mayaud et al17 observed that among the antenatal setting, “very few pregnant women gave an affirmative response to the questions on sexual partners.” Kapiga et al22 emphasize that although many women in their study did not report high‐risk behavior, the risk of STDs and HIV was determined by their male partners' sexual behavior. To identify women at risk for STDs effectively, survey tools must incorporate information such as condom usage, whether the partner is a migrant worker, etc. These intensely personal questions must be asked in a nonthreatening and nonstigmatizing way, and privacy and confidentiality must be ensured during visits, especially given the high rate of false‐positive results that often occur when using VD algorithms. Creating a comfortable environment is the key to encouraging honest answers from women at risk for STDs.
One of the key findings reported by the majority of researchers is that VD as the primary entry point for VD algorithms is not a good indicator of cervical infection.18,21,22 This is especially true if the algorithm is to be used as a screening tool because many women are asymptomatic. Entering all women into the algorithm rather than restricting the algorithm to symptomatic women tends to produce higher sensitivities and lower specificities. Whether to use VD algorithms as screening or case management tools depends on a number of factors, such as tolerance for false‐positive results in low‐risk populations, STD and HIV prevalence and incidence, and other population characteristics.
Strengthening the diagnosis and treatment of genital ulcers for both men and women is also another area that STD programs should focus on, given the fairly high accuracy of syndromic management to detect syphilis and chancroid and the substantially increased risk of HIV transmission in the presence of genital ulcers. Because of the difficulties inherent in detecting cervical infections in women (especially among low‐risk populations) and given the ability of algorithms to better detect these infections in men, one strategy to decrease infection rates is to focus on treating men and male partners. In many developing countries, migrant labor is an integral part of the economy. Therefore, targeted interventions to reduce the STD burden include providing education, prevention, and treatment facilities on work premises; ensuring that employees have the opportunity to be treated before returning home; and providing mobile services in male hostels.
Having accurate diagnostic tools to successfully treat STDs is important, but without appropriate provider training and support and adequate infrastructure, syndromic management will not be effective. Other priorities in the improvement of STD detection, treatment, and prevention include access to effective drugs and supplies, proper training of healthcare workers and alternative care providers in syndromic management, and continued and improved patient education. Drug compliance improves with single‐dose, directly observed therapy and will likely improve treatment success and, therefore, be a more cost‐effective means to treating many STDs.7,23
The role that syndromic management may have to play in increasingly drug‐resistant organisms has not yet been quantified, but must continue to be monitored when large numbers of uninfected patients are being treated. Although antibiotic resistance to the currently recommended drugs for gonorrhea infection has not yet been observed in sub‐Saharan Africa, emerging resistance in Southeast Asia indicates that Africa might not be far behind.24
The studies evaluated here show that syndromic management is effective in treating UD in men and in treating syphilis and chancroid in men and women when laboratory diagnostic techniques are not possible. Focusing and strengthening programs to improve detection and treatment of male gonococcal infections and male and female genital ulcers is important; however, management of cervical infection using VD algorithms is problematic. Because the success of algorithms is highly dependent on the prevalence of infection within the target population, using algorithms to treat cervical infection among high‐risk populations and symptomatic persons may prove more successful than using VD algorithms as screening tools for detection of cervical infection, especially among low‐risk populations. The use of risk scores can improve the effectiveness of algorithms in certain risk groups. Further study will determine whether rapid diagnostic tests, such as urine‐based leukocyte‐esterase dipstick tests, in conjunction with risk scores may help improve the sensitivity of algorithms to detect cervical and vaginal infection. Accurately assessing the prevalence of specific STDs within a population and determining what risk factors are associated with infection may help create more accurate algorithms. Problems that may further decrease the effectiveness of algorithms in the field include clinicians' resistance to implementing syndromic management, incorrect drug treatment and prescription, inadequate counseling regarding STDs, (e.g., partner notification, condom usage), poor provider training, lack of supplies, and clinic populations who either are lacking prevention knowledge or are unable to protect themselves. Although syndromic management of STDs may not offer the most sensitive or specific means to diagnose STDs, especially cervical infections in women, it is currently the most cost‐effective approach to treat populations that do not have access to more advanced diagnostic techniques. For vaginal and cervical infections, the development of rapid, accurate diagnostic tests is the priority for improved care in the future.
Outline of Selected Studies Using Risk Scores and a Brief Description of How the Algorithms Were Developed and Applied
Mayaud et al5
Risk score 1. Score based on sociodemographic factors equal or exceeding a determined cut‐off value.
Risk score 2. Sociodemographic variables plus reported symptoms.
Optimal. An optimal score was obtained from a logistic regression model and the coefficients that were still significant on multivariate analysis were incorporated in the score.
Simplified. A more practical procedure for healthcare workers in the field, a score of 1 was allotted for the presence of each characteristic represented in the optimal score.
Behets et al12,16
A questionnaire was administered among the study population to obtain demographic data, sexual history, and any symptoms. Using logistic regression, risk factors for gonococcal or chlamydial infections were determined. The odds ratios obtained from the logistic regression for the risk factors for infection were used to create a risk score. If the sum of the risk score was equal to or greater than a certain cut‐off point, the woman was treated.
Gertig et al15
Women were interviewed about sociodemographic characteristics, type of contraceptive use, sexual history and other potential risk factors for HIV and STDs. The authors evaluated an algorithm developed by Vuylsteke et al14 to diagnose chlamydia using variables they had collected and examination variables. If the sum of the risk score was greater than a certain cut‐off value, treatment was given.
Kapiga et al22
Women were interviewed regarding sociodemographic characteristics, obstetric history, contraceptive practice, sexual behavior, and detailed medical history. In addition, a gynecologic examination was performed. Multiple logistic regression was used to determine associations between characteristics and symptoms and signs suggestive of STDs and simple laboratory tests. All variables associated with N gonorrhoeae or C trachomatis infection were used to create a risk score. Scores were developed by using the odds ratio from the logistic regression model. Each woman's score was determined by summing the odds ratios of each variable to create a total score. Treatment was given if a score was equal to or above a certain cut‐off value.
Schneider et al25
Sociodemographic variables, potential risk factors, and symptoms and signs in various combinations were tested for association with cervical infection. Each characteristic was assigned a value equal to the odds ratio from a logistic regression model. Each person is given a total score based on the number of characteristics present and the total sum of the values of each characteristic. If the score was equal to or greater than a certain cut‐off score, the patient was offered treatment.
Mayaud et al17
Using logistic regression, risk factors for cervical infection for persons were entered into the model. Significant factors from the model were used to construct a risk score. Factors with an odds ratio of less than 10 were given a score of 1, whereas persons with an odds ratio greater than 10 were given a score of 2. At a certain cut‐off point, treatment was given.
Vuylsteke et al4
Using logistic regression models, a nonhierarchical quantitative algorithm using risk factors, symptoms, signs, and simple laboratory tests related to cervical infection were created. Coefficients from the logistic regression model were multiplied by 10 and used as scores in this system. The total score was the sum of the points for each variable present in a person. If the score was above a certain cut‐off value, the person was considered positive for infection.
Diallo et al14
Algorithm A was created by adding significantly associated risk factors into a risk score. Factors with an odds ratio of 1 or 2 were given 1 point, and those with an OR between 2 and 3 were given 2 points. A cut‐off value was then chosen for the sum of the points at which point treatment would be administered to a person. In algorithm C, women with high risk scores were considered infected, women with low risk scores were considered noninfected, and women with intermediate scores were considered possibly infected and would be examined before any decision regarding treatment was made. Cited Here...
1. Fleming D, Wasserheit J. From epidemiological synergy to public health policy and practice: the contribution of other sexually transmitted diseases to sexual transmission of HIV infection. Sex Transm Infect 1999; 75:3–17.
2. Holmes KK, Ryan C. STD care management. In: Holmes KK, Sparling PF, Mardh P-A, et al., eds. Sexually Transmitted Diseases. New York: McGraw-Hill, 1999.
3. World Health Organization. Management of patients with sexually transmitted diseases. World Health Organ Tech Rep Ser 1991; 810:1–103.
4. Vuylsteke B, Laga M, Alary M, et al. Clinical algorithms for the screening of women for gonococcal and chlamydial infection: evaluation of pregnant women and prostitutes in Zaire. Clin Infect Dis 1993; 17:82–88.
5. Mayaud P, Grosskurth H, Changlucha J, et al. Risk assessment and other screening options for gonorrhoea and chlamydial infections in women attending rural Tanzanian antenatal clinics. Bull World Health Organ 1995; 73:621–630.
6. Last J. A Dictionary of Epidemiology. New York: Oxford University Press, 1995.
7. World Health Organization. Management of Sexually Transmitted Diseases. Geneva: UNAIDS, World Health Organization, 1997.
8. Martin D, Bowie W. Urethritis in men. In: Holmes KK, Sparling PF, Mardh P-A, et al., eds. Sexually Transmitted Diseases. New York: McGraw-Hill, 1999.
9. Moherdaui F, Vuylsteke B, Siqueira L, et al. Validation of national algorithms for the diagnosis of sexually transmitted diseases in Brazil: results from a multicentre study. Sex Transm Infect 1998; 74(Suppl 1):S38-S43.
10. Alary M, Baganizi E, Guedeme A, et al. Evaluation of clinical algorithms for the diagnosis of gonococcal and chlamydial infections among men with urethral discharge or dysuria and women with vaginal discharge in Benin. Sex Transm Infect 1998; 74(Suppl 1):S44-S49.
11. Bogaerts J, Vuylsteke B, Martinez Tello W, et al. Simple algorithms for the management of genital ulcers: evaluation in a primary health care centre in Kigali, Rwanda. World Health Organ Bull OMS 1995; 73:761–767.
12. Behets F, Williams Y, Brathwaite A, et al. Management of vaginal discharge in women treated at a Jamaican sexually transmitted disease clinic: use of diagnostic algorithms versus laboratory testing. Clin Infect Dis 1995; 21:1450–1455.
13. La Ruche G, Lorougnon F, Digbeu N. Therapeutic algorithms for the management of sexually transmitted diseases at the peripheral level in Côte d'Ivoire: assessment of efficacy and cost. Bull World Health Organ 1995; 73:305–313.
14. Diallo M, Ghys P, Vuylsteke B, et al. Evaluation of simple diagnostic algorithms for Neisseria gonorrhoeae
and Chlamydia trachomatis
cervical infections in female sex workers in Abijan, Cote d'Ivoire. Sex Transm Infect 1998; 74(Suppl 1):S106-S111.
15. Gertig D, Kapiga S, Shao J, et al. Risk factors for sexually transmitted diseases among women attending family planning clinics in Dar-es-Salaam, Tanzania. Genitourin Med 1997; 73:39–43.
16. Behets F, Desormeaux J, Joseph D, et al. Control of sexually transmitted diseases in Haiti: Results and implications of a baseline study among pregnant women living in Cité Soleil shantytowns. J Infect Dis 1995; 172:764–771.
17. Mayaud P, Uledi E, Cornelissen J, et al. Risk scores to detect cervical infections in urban antenatal clinic attenders in Mwanza, Tanzania. Sex Transm Infect 1998; 74(Suppl 1):S139-S146.
18. Ryan C, Courtois B, Hawes S, et al. Risk assessment, symptoms, and signs as predictors of vulvovaginal and cervical infections in an urban US STD clinic: implications for use of STD algorithms. Sex Transm Infect 1998; 74(Suppl 1):S59-S76.
19. Wawer M, Sewankambo N, Serwadda D, et al. Control of sexually transmitted diseases or AIDS prevention in Uganda: a randomised community trial. Lancet 1999; 353:525–535.
20. Thomas T, Choudhri S, Kariuki C, et al. Identifying cervical infection among pregnant women in Nairobi, Kenya: limitations of risk assessment and symptom-based approaches. Genitourin Med 1996; 72:334–338.
21. Mayaud P, Ka-Gina G, Grosskurth H. Effectiveness, impact and cost of syndromic management of sexually transmitted diseases in Tanzania. Int J STD AIDS 1998; 9:11–14.
22. Kapiga SH, Vuylsteke B, Lyamuya E, et al. Evaluation of sexually transmitted diseases diagnostic algorithms among family planning clients in Dar es Salaam, Tanzania. Sex Transm Infect 1998; 74(Suppl 1):S132-S138.
23. Haddix A, Hillis S, Kassler W. The cost effectiveness of azithromycin for Chlamydia trachomatis
infections in women. Sex Transm Dis 1995; 22:274–280.
24. Van Dyck E, Crabbe F, Nzila N, et al. Increasing resistance of Neisseria gonorrheae
in West and Central Africa: consequences on therapy of gonococcal infection. Sex Transm Dis 1997; 24:32–37.
25. Schneider H, Coetzee D, Fehler H, et al. Screening for sexually transmitted diseases in rural South African women. Sex Transm Infect 1998; 74(Suppl 1):S147-S152.
26. Djajakusumah T, Sudigdoadi S, Keersmaekers K, et al. Evaluation of syndromic patient management algorithm for urethral discharge. Sex Transm Infect 1998; 74(Suppl 1):S29-S33.
27. Hanson S, Sunkutu RM, Kamanga J, et al. STD care in Zambia: an evaluation of the guidelines for case management through a syndromic approach. Int J STD AIDS 1996; 7:324–332.
28. Ye Htun, Morse S, Dangor Y, et al. Comparison of clinically directed, disease specific, and syndromic protocols for the management of genital ulcer disease in Lesotho. Sex Transm Infect 1998; 74(Suppl 1):S23-S28.
29. Seck K, Samb N, Tardy M, et al. Disappointing results of the evaluation of the syndromic approach for the diagnosis of STDs in pregnant women in Dakar, Senegal (abstract no. O104). In: Proceedings of the International Congress of Sexually Transmitted Diseases. Seville, Spain, 1997.
30. Bourgeois A, Henzel D, Malonga-Mouelet G, et al. Clinical algorithms for the screening of pregnant women for STDs in Libreville, Gabon: which alternatives? Sex Transm Infect 1998; 74:35–39.
31. Meda N, Sangaré L, Lankoandé S, et al. Pattern of sexually transmitted diseases among pregnant women in Burkina Faso, West Africa: potential for a clinical management based on simple approaches. Genitourin Med 1997; 73:188–193.
32. Ronsmans C, Bulut A, Yolsal N, et al. Clinical algorithms for the screening of Chlamydia trachomatis
in Turkish women. Genitourin Med 1996; 72:182–186.
33. Wi T, Mesola V, Manalastas R, et al. Syndromic approach to detection of gonococcal and chlamydial infections among female sex workers in two Philippine cities. Sex Transm Infect 1998; 74(Suppl 1):S118-S122.
34. Miller W, Hoffman I, Leone P, et al. A comparison of selective screening algorithms for chlamydial infection (abstract no. O108). In: Proceedings of the International Congress of Sexually Transmitted Diseases. Seville, Spain, 1997.
© Copyright 2000 American Sexually Transmitted Diseases Association
This article has been cited