|Home | About | Journals | Submit | Contact Us | Français|
We assessed agreement of reported gender of sex partners in 2 statewide HIV databases linked by client identifiers.
Counseling, testing, and referral (CTR) records on all men aged 18 to 30 years who tested newly positive for HIV in North Carolina between 2000 and 2005 were matched to data abstracted from partner counseling and referral services (PCRS) records. We compared client-reported gender of sex partners at the time of testing (CTR records) with those reported during postdiagnosis partner notification (PCRS records).
PCRS records appeared to be a more complete measure of the gender of sex partners. Of the 212 men who told their HIV test counselor that they had only had female sex partner or partners in their lifetime, 62 (29.2%) provided contact information for male sex partner(s) during partner notification.
During the test counseling risk assessment, many men did not fully report the gender of their sex partners; this suggests that CTR data may not fully capture clients’ risk behaviors.
To monitor trends among people testing for HIV and to inform prevention programs, the Centers for Disease Control and Prevention (CDC) funds the collection of individual-level data on all persons accessing counseling, testing, and referral (CTR) services in publicly funded test sites, including demographics, self-reported risk behaviors, and test results.1 CTR behavioral risk assessments are usually completed in face-to-face, in-depth interviews with a trained HIV counselor as part of a client-centered, prevention-counseling approach.2 Behaviors disclosed to the counselor inform the development of a behavioral change goal to reduce the client’s risk of HIV acquisition. Risk behaviors disclosed to providers influence the services offered to clients. In a study of clinics offering both confidential and anonymous HIV testing, providers admitted to “push[ing] those individuals who are at high risk of HIV infection to test confidentially,”3(p162) presumably to aid in reporting and partner notification. Among gay and bisexual men attending a sexually transmitted disease (STD) clinic, men who disclosed a high-risk behavior (anal sex) were more like to be tested for gonorrhea than men who failed to disclose.4
CTR data are recorded on standardized forms and submitted to the CDC quarterly for tracking of national statistics.1 Aggregated risk behavior data are used to evaluate the effectiveness of the CTR program in providing high-risk populations with testing opportunities,5 to guide the development of prevention interventions, and to inform allocation of funds. For clients testing positive for HIV, the risks recorded on the CTR form may be used to determine likely mode of transmission, along with medical record review.
It is likely that nonheterosexual behaviors are underreported during the CTR risk assessment. In a study of the completeness of sexual histories obtained during STD examinations, 22% of men who reported same-gender sex during a computer-assisted interviewed failed to disclose that information during a face-to-face clinician interview.6 Almost 40% of men surveyed in the New York City National HIV Behavioral Surveillance Project reported not disclosing same-gender sex to their health care providers.7 During HIV test counseling, clients may not completely report the gender of their sex partners (e.g., they may report only sex with women when they had sex with both men and women). To date, no study has quantified the completeness of disclosure of the gender of sex partners in a CTR database.
There is no “gold standard” for validating self-reported risk behaviors,8 but other behavioral databases may be more complete in measuring the gender of sex partners than a CTR database. In North Carolina, the partner counseling and referral services (PCRS) program is part of the state Department of Health and Human Services (NCDHHS) field services office. PCRS counselors conduct voluntary, postdiagnosis interviews with clients newly infected with HIV to assist with partner notification of past sex and needle-sharing partners, counsel clients on prevention of subsequent risk behaviors, and facilitate referrals for treatment and services.9,10 Often meeting with clients multiple times, PCRS counselors stress the importance of partner notification to provide persons exposed to HIV the opportunity to be tested, as well as to remind clients of communicable disease control laws that mandate partner notification. Consequently, PCRS interviews may be a more complete measure of the gender of sex partners than the HIV counseling risk assessment.
We linked CTR records to data abstracted from PCRS records on young men newly diagnosed with HIV. We compared the client-reported genders of sex partners at the time of HIV testing (CTR records) with those reported during postdiagnosis partner notification (PCRS records), calculating measures of agreement.
As part of an ongoing investigation of the HIV epidemic in young men,11–14 this study included all men aged 18 to 30 years newly diagnosed with HIV in North Carolina between 2000 and 2005. Since the CTR data set is limited to tests performed in clinics funded by the NCDHHS, the analysis of the disclosure of the gender of sex partners excludes men diagnosed by a private provider.
Confidential CTR services are provided in all 100 counties in North Carolina.9 CTR sites are primarily county health departments and outreach venues, but they may include county jails if the jail health service is provided by the health department. Data, which are collected at CTR sites via a Scantron form with a unique CTR identifier, document individual-level demographics, including a standardized risk assessment. The CTR data collection form was revised in July 2005 to include additional client identifiers (name and address) and a slightly modified risk assessment. Data are stored in an electronic database at the NCDHHS.8
During the study period, risk assessments were conducted by HIV test counselors trained in a state-sponsored curriculum.15,16 NCDHHS counselors are trained to ask clients the gender of their sex partners, using the question, “Do you have sex with men, women or both?” (with data captured as yes or no for each gender), with additional open-ended probes. Before July 2005, the Scantron documented the gender of the client’s sex partners “since 1978,” which for this young population denotes lifetime. Beginning in July 2005, the time frame used to assess the risk period during CTR was changed to “in the last year.”
Since 1989, the NCDHHS has offered PCRS to all newly reported cases of HIV. A review of the program indicated that 90% of clients testing positive in a CTR site were interviewed by PCRS counselors.10 Time between the CTR test and interview varies by ability to locate client, but the field investigation begins within 24 hours of the case being assigned to a PCRS counselor. PCRS records document information gathered from providers and clients in electronic and hard copy records.9 In our study, trained research assistants, using a case abstraction form, abstracted data from the standardized fields and written narratives of the PCRS records and entered them into an Access database (Microsoft Corporation, Redmond, WA). Variables abstracted included the patient’s unique state identification number, client demographics, and risk behaviors.
PCRS counselors document the gender of sex partners during the likely infection period (for chronic infections, the standardized assessment period is 1 year). The gender of sex partners is deduced through partner notification efforts, where clients provide counselors with the names and contact information for past and current sex partners.
The study sample of newly diagnosed young men was identified through PCRS chart abstraction. Using the unique state identification number only, we linked PCRS records to North Carolina’s HIV/AIDS Reporting System, which contains the unique state identification number, client name, diagnosis facility, and, for clients testing in a publicly funded test site, the unique CTR identifier for their reactive test. For clients testing in a publicly funded test site (identified by diagnosis facility) whose CTR identifier was not found in the HIV/AIDS Reporting System, we abstracted the CTR identifier from the CTR test site medical record using patient name (n = 267). The PCRS abstraction data set was then merged with the CTR electronic database by CTR identifier. We compared client demographics (race and gender) and date of diagnosis between databases to ensure correct matches and conducted a sensitivity analysis by repeating all analyses restricted only to observations that matched on all 3 variables.
On the basis of the reported gender of sex partner, men were categorized as men who had sex only with men (MSM), men who had sex only with women (MSW), men who had sex with men and women (MSMW), and men who reported no sex partners. The statistical analysis was limited by the risk assessment time frame in each data set. For example, a man may have accurately reported to his test counselor that he had had sex with both men and women in his lifetime (the CTR risk assessment period), but may have only had sex with women in the last year (the PCRS risk assessment period) and accurately provided contact information for only female sex partners to the PCRS counselor. Therefore, some men may be falsely classified as inaccurately disclosing the gender of their sex partner to their test counselor when differences are due solely to the assessment time frame. Using cases whose CTR data were recorded on the revised Scantron, which assessed risk in the past year, providing perfect time frame overlap with the PCRS assessment period as a validation sample (n = 75), we conducted a probabilistic analysis to account for misclassification in the full data set by reclassifying men on the basis of distributions in the validation sample.17
Agreement of reported gender of sex partners was calculated with the Cohen κ.18 As we hypothesized that the PCRS database would be more complete, we calculated conditional κ,19 sensitivity, and specificity using PCRS as the “gold standard.” Conditional κ calculates agreement conditional on an affirmative response in the PCRS data (e.g., MSM = yes in the PCRS report). Measures of agreement for each classification (MSM, MSW, and MSMW) are reported for the original data set and the reclassified data set as described earlier in this section. We conducted stratified analysis by race, as we hypothesized that disclosure patterns during test counseling may be different because of different social norms around sexual orientation.20,21 To quantify how differences in report of gender of sex partners may alter aggregate CTR statistics, we assigned risk categories based on the CDC’s hierarchy of risk1 using reported gender of sex partners in (1) only the CTR database and (2) only the PCRS database. Analysis was completed in SAS version 9.13 (SAS Institute Inc, Cary, NC) and Excel (Microsoft Corporation, Redmond, WA).
Of the 1450 men aged 18 to 30 years who were newly diagnosed in North Carolina between 2000 and 2005 with available PCRS records, 673 (46.4%) were diagnosed in a CTR site. Compared with men testing in private facilities, men diagnosed at CTR sites were younger, had been tested for HIV previously, and reported more risk behaviors, such as using recreational drugs (Table 1). A total of 641 of the records of men testing in a publicly funded facility were successfully linked between the PCRS and CTR databases. Of the 32 records that did not match, 16 PCRS records did not have a locatable CTR identifier and 16 PCRS records had a CTR identifier that could not be linked to the CTR database. The 32 unmatched records were not statistically different from matched cases by client demographics or region of diagnosis. Nonmatched clients were more likely to have been previously incarcerated or to have a history of injection drug use, but the low prevalence of nonmatched cases (less than 5%) should have a minimal effect on the analysis. Comparing the demographics reported in the PCRS and CTR records, we found that 82% of the records matched on race, gender, and date of diagnosis and 100% matched on at least 1 of the variables. For the subset of CTR records that contained the patient’s name (those after July 2005), 100% of records matched on name. All matched records were included in the analyses, and a sensitivity analysis using only the records matching completely showed no substantive differences in results.
Overall agreement of reported gender of sex partners was low (κ = 0.44; 95% confidence interval [CI] = 0.39, 0.49). Of the 212 men who told their HIV test counselor that they had only had female sex partner(s) in their lifetime, 62 (29.2%) provided contact information for male sex partner(s) during partner notification. Of 25 men who reported during test counseling that they had never had sex in their lifetime, 22 (88.0%) gave contact information for at least 1 sex partner during postdiagnosis interviews. A majority of the men (83%; 373 of 449) who named a male sex partner during PCRS interviews had disclosed MSM behavior to their CTR counselor. Less than half of the men (54 of 110) who provided both male and female contacts during PCRS interviews had disclosed having sex with both men and women to their CTR test counselor. Although small sample sizes reduced the power to detect statistical differences, there appeared to be some differences in congruence of MSMW disclosure by race, with PCRS and CTR data agreeing for White men in over 76% of cases and for Black men in less than 45% of cases (P = .08).
When we assumed that the PCRS records were a “gold standard,” the sensitivity of the CTR record was lowest for nonheterosexual sexual behaviors (Table 2). When we assigned a risk category using the CDC hierarchy of risk for CTR,1 32.8% of men were assigned “heterosexual” based on CTR data compared with 23.7% based on PCRS data (P<.01; Table 3).
We compared the reported gender of sex partners in 2 statewide HIV databases that contained individual-level risk behavior data on newly diagnosed young men in North Carolina. PCRS records, which document postdiagnosis interviews for partner notification, provided higher levels of self-report of nonheterosexual behaviors than CTR data collected at the time of testing.
The difference may be partially attributed to the different rationales for the PCRS and CTR risk assessments. During PCRS counseling, complete risk assessment information is needed so that all sex partners can be notified. Alternatively, CTR counselors use the risk assessment for prevention counseling and may not focus on obtaining complete information on sex partners. As a result, PCRS and CTR counselors use different techniques and may have different levels of persistence in obtaining sensitive information from clients. PCRS counselors also have the opportunity to meet with clients multiple times in different locations, whereas CTR counselors usually have limited time for test counseling. Additionally, at the time of the CTR risk assessment, clients may perceive themselves as HIV uninfected, whereas at the time of the PCRS interview they have been diagnosed. The knowledge of the infection likely affects clients’ reflections on their past behaviors and may influence disclosure.
Social exchange theory proposes that persons choose when, and with whom, to disclose sensitive or stigmatizing information on the expected benefits and anticipated costs of disclosure.22 As accurate disclosure during partner notification helps ensure that all past partners can be tested and receive treatment if infected, clients may feel that it is socially desirable to provide accurate information on all sex partners or feel legally obligated to disclose. Alternatively, for some men, the perceived costs of complete disclosure might be higher during PCRS counseling (e.g., fear of partner retribution23) than during CTR counseling (e.g., perceived as more anonymous).
MSMW behavior had the lowest level of agreement between the databases and the lowest sensitivity measure in the CTR database. Patterns of disclosure were varied, with some men naming sex partners of both genders during PCRS and disclosing only MSM behavior to their CTR counselor. This suggests that for some men, disclosing MSM behavior is different from disclosing MSMW behavior. Differences in disclosure patterns of MSMW behavior by race may indicate different social norms around reporting these behaviors.
This analysis was limited to young men. Patterns of disclosure of the gender of sex partners may vary for women and older men. In this sample, men diagnosed at CTR sites differed from men testing in private facilities by demographics and reported risk behaviors. It is possible that they also differed in the disclosure patterns of those behaviors. The analysis was limited to cases that were reported to the NCDHHS and entered into the PCRS system. However, with over 90% of all cases contacted, the PCRS is the most comprehensive statewide behavioral database of HIV-infected persons.10 Data missing on specific variables, including cases that could not be linked between PCRS and CTR, were less than 5% and should have a minimal impact on the results. Additionally, it is theoretically possible that a client changed behaviors during the time between the CTR session and the PCRS interview (e.g., he reported no sex partners, but had sex immediately after the test counseling session); unfortunately, we were not able to assess this possibility.
Although the risk assessment time frame differed between the data sets (past year versus lifetime), resulting in possible misclassification of noncongruence of reported gender of sex partners, we adjusted agreement statistics using a sample with perfect assessment period overlap. The majority of estimates changed only slightly, suggesting that nonconcordance was probably not influenced by assessment period issues. The validation sample was not random, but rather based on the last 6 months of the study period when the risk assessment period had changed. Consequently, the validity of our corrected estimates rests on the assumption that the validation sample represents the true congruence between the data sets over the entire study period.
After comparing the gender of sex partners disclosed by clients during HIV test counseling and during postdiagnosis interviews, we observed that many clients disclosed the gender of sex partners differently, which suggests that self-reported behavioral data should be interpreted with caution. In this sample, CTR data on the gender of sex partners appears less complete than PCRS data. The inaccurate disclosure of the gender of sex partners during testing may affect the efficacy of prevention counseling. A male client may have unprotected sex with men and women, but only disclose that he has sex with women. In this case, there is a missed opportunity to provide risk education and develop a risk reduction plan specific to same-gender sexual contact (e.g., the differential risks related to insertive versus receptive sex).
At a population level, the misreporting of risks affects the evaluation of the CTR programs, as the database may not accurately describe the population testing at CTR sites. In this study, 28% of the men classified as heterosexual on the basis of CTR data were reclassified when PCRS data were used to assign risk categories. If CTR data are used exclusively to inform mode of transmission, surveillance data in the HIV/AIDS Reporting System may be skewed as well. Our findings also suggest the ineffectiveness of using a risk assessment to screen for testing. For example, the 25 men who reported no lifetime sex partners may not have been tested and subsequently diagnosed.
Although the CDC no longer recommends pretest counseling,24 prevention counseling is encouraged for high-risk clients, such as patients in STD clinics,25 and current HIV-testing guidelines for NCDHHS-funded clinics require that a risk assessment be documented for each test.16 NCDHHS guidelines suggest that the assessment can be performed in a variety of ways, including a self-administered questionnaire. 16 One method that may allow for more complete risk disclosure is audio- and computer-assisted self-interviews. In a survey of blood donors using audio- and computer-assisted self-interviews, 67% said they were more truthful than in face-to-face interviews and thought the methods were clear (91.8%) and private (92.3%).26 Among clients using audio-and computer-assisted self-interviews in an STD clinic, 56% reported preference for audio- and computer-assisted self-interviews compared with face-to-face interviews and 82% reported more honest responses.6
This analysis highlights the differences between client-reported gender of sex partners during the CTR and PCRS risk assessments. Further research is needed to understand barriers and facilitators to disclosure of risk behaviors. Investigation of the feasibility and cost-effectiveness of widespread use of audio- and computer-assisted self-interviews for risk assessments during HIV test counseling in publicly funded clinics may be warranted.
This study was funded by the University of North Carolina at Chapel Hill Centers for AIDS Research (grant P30 AI50410).
Reprints can be ordered at http://www.ajph.org by clicking the ”Reprints/Eprints” link.
ContributorsE. A. Torrone originated the study, conducted the analyses, and led the writing. J. C. Thomas and L. B. Hightow-Weidman supervised all aspect of the study, including synthesis of analyses and interpretation of findings. J. S. Kaufman provided guidance on the analysis. P. A. Leone facilitated access to data and assisted with interpretation of findings. All authors helped to conceptualize the study, reviewed drafts of the article, and approved the final version.
Human Participant Protection
Data analysis for this manuscript was approved by the institutional review board of the University of North Carolina at Chapel Hill.