Reliability of Sexual Behavior Self-Reports in Male Couples of Discordant HIV Status.

Given the presence of strong cultural taboos against the direct observation of people's sexual activities, most sex researchers rely heavily on self-reports (Catania et al., 1993). Therefore, the reliability of these measures is of crucial importance. Attempts to assess reliability have involved contrasting long- and short-term behavioral recalls (Kauth, St. Lawrence, & Kelly, 1991; McElrath, Chitwood, Griffin, & Comerford, 1994;), comparing testretest results (Mahler, Hull, Stebinger, Perry, & Frances, 1995; McKinnon et al., 1993; Saltzman, Stoddard, McCusker, Moon, & Mayer, 1987; Weatherburn, Hunt, Davies, Coxon, & McManus, 1991), and checking the congruence of data obtained by different procedures (Boekeloo et al., 1994; James, Bignell, & Gillies, 1991; Siegel, Krauss, & Karus, 1994; Weatherbum et al., 1991) --including relating reported frequency of condom use and presence of STDs (Zenilman et al., 1995) and comparing the separate reports of primary partners (Coates et al., 1986; Darke, Hall, Heather, Ward, & Wodak, 1991; Jacobson & Moore, 1981; Kinsey, Pomeroy & Martin, 1948; Kinsey, Pomeroy, Martin, & Gebhard, 1953; Levinger, 1966; Padian, 1990; Seage, Mayer, Horsburgh, Cai, & Lamb, 1992; Upchurch et al., 1991). The results of these studies have been mixed, some finding that the self-reports were reliable and others that they were not. Overall, the studies highlighted the influence of methodological and respondent variables on reliability (for a thorough review of methodological problems in sexual behavior studies see Catania, Gibson, Chitwood, & Coates, 1990, and Catania, Kegeles, & Coates, 1990.). Finally, some authors have altogether criticized quantitative investigations of sexual behavior and have advocated, instead, for qualitative inquiries (Parker & Carballo, 1990).

If sexual matters are frequently shrouded in secrecy, sexual behavior that carries the potential of HIV transmission is likely to be even more concealed. It may be particularly difficult for people in HIV serodiscordant relationships (where one partner is HIV-positive and the other HIV-negative) to report high risk behavior given the social pressure to avoid such acts. Reliable and valid assessments of HIV-risk behaviors are critical to identify specific populations in need of intervention and to test the efficacy of any prevention program. Since no measure can be valid if not reliable, and given the absence of any published reliability study focusing specifically on HIV serodiscordant male couples, we saw a need to fill in this gap.

We had the rare opportunity of working with a group of 75 HIV-discordant male couples. We thoroughly explored the reliability of the reports the couples' members made of their behavior with each other in two ways: (a) We compared self-report to partner's report of anal sex, and (b) we analyzed the test-retest reliability of individual reports collected with a one-week interval.

METHOD

Participants

Study participants were recruited in the New York Metropolitan area through leaflets distributed at gay organizations (HIV support groups, recreational clubs, etc.), advertisements in gay and AIDS-related newsletters and magazines, and by word of mouth. The eligibility criteria were (a) 18 years or older, (b) involved in a relationship with another man for at least four months, and (c) one partner being HIV-positive and the other one HIV-negative. A total of 75 male couples who fulfilled the eligibility criteria participated in the study. The participants' mean age was 36 years (SD = 8; range 22 - 66), their average education 15 years (SD = 3; range 9 - 20), and their mean annual income $37,300. Their racial/ethnic distribution was 63% Caucasian, 16% African American, and 21% Latino.

Assessment

The assessment was divided in two parts: individual and joint (couple) assessment. The individual part consisted of an interviewer-administered section, which included the Sexual Practices Assessment Schedule - Couples version [SPAS-C] (Carballo-Dieguez, Exner, Gruen, & Meyer-Bahlburg, 1994; available from the first author), and a self-administered section. The interviewers were men who had been rigorously trained in psychosexual assessment following procedures described elsewhere (McKinnon et al., 1993). Briefly, the training included desensitization both to sexual terms and to gay men's culture, mock interviews, interviews with community candidates similar to those of the target population, and finally, actual interviews with study participants. All interviews were audio-taped and randomly reviewed by a supervisor for quality control.

In the opening remarks of the SPAS-C the interviewer warns the respondent that he will be asked detailed questions about his sexual behavior, and that the respondent may feel uncomfortable discussing such issues with a stranger. Reassurance is provided concerning the confidentiality of the interview and its scientific purpose. A vocabulary elicitation phase follows during which various sexual practices are described in behavioral terms and the respondent's labels for such practices are elicited. For example, we ask, "When a man puts his penis in someone's anus, what do you call it?" If during this phase the respondent uses only formal language, we volunteer vernacular terms to indicate that it is acceptable to use them. For example, we say, "Some people call it `penetration,' others call it `fucking'." At the end of the word elicitation, we check with the respondent whether formal or vernacular language feels more comfortable to him, and we use his choice for the rest of the interview.

The SPAS-C requests participants to recall their sexual behavior over the past year. Although we are aware that shorter recall periods improve accuracy, we inquire about the past year to capture seasonal variations, particularly with Latino participants living in New York City who spend the summer in their Caribbean countries of origin, where sexual patterns may vary. To help participants to be precise about the time frame, we ask them to recall an event that happened the same month (or even the same day) the year before, and then to report all sexual activity from that date until the assessment date. When the number of sexual partners or sexual occasions is high, we help the participant calculate the monthly or weekly average and corroborate whether the numbers seem right to him. Next, we inquire about three different types of male sexual partners that the individual may have had during the previous year (lover, one night stand, and other). For each partner, we ask the frequency with which each of 35 sexual behaviors took place. All answers are given on a 5-point Likert scale ranging from never to always. When the section on male partners is complete, we inquire about female partners. The joint assessment, which takes place after the individual administration of the SPAS-C to both members of the couple, consists of open-ended questions that allow the interviewer to explore discrepancies in the types of behaviors that the primary partners reported having with each other.

Data Analysis

Due to space limitations, this report focuses exclusively on anal sex, the behavior that carries the highest risk of HIV transmission. First, we calculated the prevalence of protected and unprotected anal sex, with and without ejaculation. Second, we compared the report that one member of the couple gave concerning his anal sex behavior with his primary partner with the report of that primary partner. Third, we analyzed the level of agreement of a one week interval test-retest of 27 individuals--the first recruited.

Missing data. Important methodological issues need to be considered when assessing the reliability of a sexual questionnaire. The first one is how to handle data missing due to the skip patterns built into the instrument. To illustrate this problem, let us concentrate on the variables related to insertive anal sex and the question of intracouple reliability (though the same issues apply to the test-retest reliability analyses as well).

All participants were asked whether they had put their penis in their partner's rectum. They chose their answers from the 5-point scale (0 = never, 1 = infrequently, 2 = half the time, 3 =frequently, 4 = always). Those who responded never were not asked the next question regarding whether they used a condom when they had insertive anal sex. When the reliability of the condom question is assessed, one needs to decide whether to include these individuals. Some would argue that they should not be included in the reliability analysis since they were not asked the question. However, their answers can be inferred from their previous response. If a researcher wanted to assess, for example, correlates of unprotected insertive anal sex, those respondents who did not engage in insertive anal sex could be included in the analyses as never having had unprotected insertive anal sex. Furthermore, it is conceivable that respondent A could report engaging in insertive anal sex with B and consequently could be asked the condom question, while B could say he did not engage in receptive anal sex with A and not be asked the condom question. To exclude couples in this situation from the reliability analysis for the condom question might increase its apparent reliability but not fully reflect the available data. We consider that a more accurate test should include all participants in the reliability analyses regarding condom use. To do this we created a fifth category, so that responses to the question may range from 0 = never (had insertive anal sex but never used a condom) to 4 = always (had insertive anal sex and always used a condom) with an additional category, 5 = no unprotected insertive anal sex, because the respondent did not engage in insertive anal sex. This category was placed at the positive end of the response scale since that is consistent with the direction of the scale in terms of risk.

When the next question in the interview is considered (ejaculation without a condom), additional problems emerge. If a participant does not engage in insertive (or receptive) anal sex without a condom, he is not asked the question. If, following our previous reasoning, we want to include all participants in the reliability analyses, we need to create two additional categories. The highest risk corresponds to 4 = always (had unprotected insertive anal sex and always ejaculated) and the low end to 0 = never (had unprotected insertive anal sex and never ejaculated). The additional categories extend the low risk end of the scale with -1 = never ejaculated unprotected because a condom was always used, and -2 = never ejaculated unprotected because respondent did not engage in insertive anal sex.

Although this scale modification allows all participants to be included and uses all available information, it is debatable how accurately it assesses the reliability of the ejaculation question since the majority of the participants in the study would not have been asked that question. It may be more relevant to know whether there is a high degree of agreement regarding ejaculation among those who had unprotected insertive anal sex.

To deal with these issues, we conducted the analyses in two different ways that address somewhat different reliability questions. First, reliability was assessed with the responses given by couples where both members were asked the same question (or, for the test-retest analyses, with the responses given by an individual to the same question in two occasions). Taking condom use during insertive anal sex as an example, this approach answers the question of agreement regarding condom use in couples where A reports having engaged in insertive anal sex with B and B reports having had receptive anal sex with A. Analyses involving the ejaculation question assess agreement regarding ejaculation within couples where A reports having engaged in unprotected insertive anal sex with B and B reports unprotected receptive anal sex with A. A subsequent set of analyses includes the total sample available using the coding scheme outlined above. These analyses assess agreement on the overall pattern of responses to the set of questions regarding the given sexual practice. For example, analyses of the ejaculation question are not limited to those who had unprotected sex but also assess whether the reason the question was not asked is based on agreement in responses to the two prior questions.

Weight issues. The unweighted kappa statistic we initially used for this study (Cohen, 1960; Fleiss, 1981) is an index of exact agreement between two answers adjusted for the amount of agreement that would be expected by chance. This kappa answers the question of whether responses agree or not, regardless of the extent of the disagreement. That is, the pair of responses Half the Time and Frequently, (one point away on the response scale) is treated equally to the pair Always and Never (five points away). A second option is to apply a disagreement weight to each pair of responses (Cohen, 1968; Fleiss & Cohen, 1973). In general, the weights simply reflect the distance between the responses on the response scale. Therefore, perfect agreements are given a disagreement weight of 0. The disagreement between responses such as Infrequently and Half the Time or Frequently and Always get a disagreement weight of 1, being one point away from each other on the response scale. The pair of responses such as Infrequently and Always receive a disagreement weight of 3, being three points away on the response scale. However, we had an additional consideration in determining the appropriate weights. We considered that there was a more substantive and meaningful disagreement when one response was that the behavior never occurred and the other response indicated at least some occurrence of the behavior than a disagreement between two responses indicating different frequencies in the behavior. To reflect this situation in the case of insertive or receptive sex, we designed a weight system by which, for example, the response pair of Never and Infrequently receives a disagreement weight of 3 and the response pair of Never and Always receives a disagreement weight of 6--three points for a discrepancy between Never and Infrequently and three additional points reflecting three more points separating the responses on the response scale (see Table 1).

Table 1. Values for Weighted Kappas for Frequency of Anal
Penetration

                Never   Infrequently   Half the Time

Never             0          3               4
Infrequently      3          0               1
Half the Time     4          1               0
Frequently        5          2               1
Always            6          3               2

                Frequently   Always

Never               5          6
Infrequently        2          3
Half the Time       1          2
Frequently          0          1
Always              1          0

For the condom use questions, this greater disagreement weight occurs between 4 (always used a condom) and 3 (frequently used a condom, which indicates some unprotected sex). For the ejaculation questions, the greater disagreement weight occurs between 0 (never ejaculated) and 1 (infrequently, ejaculated, which indicates unprotected ejaculation occurred). Therefore, unweighted kappas are a more conservative assessment requiring exact agreement, while weighted kappas take into consideration the quality and amount of any observed disagreements. We have used both unweighted and weighted kappas in the analyses reported in the results section.

Nomenclature

We followed the Landis and Koch's (1977) nomenclature to label kappa agreement points between 0 (no better than chance) and 1 (perfect), (Poor [is less than] 0.00; Slight = 0.00 - 0.20; Fair = 0.21 - 0.40; Moderate = 0.41 - 0.60; Substantial = 0.61 - 0.80; Almost perfect = 0.81 - 1.00).

RESULTS

Quantitative Results

To simplify the presentation of results, Table 2 shows the responses our participants gave to six questions concerning anal sex in a dichotomized fashion. The table shows that about half of the couples reported that they engaged in anal sex. Half of the HIV-positive men penetrated their HIV-negative partners. About half of the HIV-positive men reported being penetrated. Disagreement between partners on this issue was low, about one in ten couples. Unprotected anal sex was reported by 15% to 23% of the couples, and the HIV-positive partner was more likely to be receptive than insertive. In only one of the couples engaging in unprotected anal sex did one positive partner acknowledge having ejaculated in his partner's rectum, although this was denied by the partner. The converse case (HIV-negative ejaculating inside the HIV-positive) was more frequent.

Table 2. Interpartner Agreement to Anal Sex Questions

                                          Binary Scale (Yes/No)

Sexual Activity                       n(a)    Both "Yes"   Both "No"

Does the HIV+ respondent penetrate    75          32          35
 his partner?
Does the HIV+ respondent penetrate    40(b)       11          27
 his partner without a condom?
Does the HIV+ respondent ejaculate    12(c)        0          11
 in his partner's rectum?
Does the partner penetrate the        75          35          30
 HIV+ respondent?
Does the partner penetrate the HIV+   45(b)       17          19
 respondent without a condom?
Does the partner ejaculate in the     19(c)        4           7
 HIV+ respondent's rectum?

                                       No Agreement

Sexual Activity                       + Yes   - Yes

Does the HIV+ respondent penetrate      6       2
 his partner?
Does the HIV+ respondent penetrate      1       1
 his partner without a condom?
Does the HIV+ respondent ejaculate      1       0
 in his partner's rectum?
Does the partner penetrate the          4       6
 HIV+ respondent?
Does the partner penetrate the HIV+     4       5
 respondent without a condom?
Does the partner ejaculate in the       5       3
 HIV+ respondent's rectum?

Note. n = 75 mixed HIV-status couples. + Yes = HIV-positive partner replied yes, while HIV-negative partner said no. - Yes = HIV-negative partner replied yes, while HIV-positive partner said no.

(a) Inconsistencies in N due to missing data. (b) Includes couples where one or both members reported engaging in penetration. (c) Includes couples where one or both members reported engaging in unprotected anal penetration.

Table 2 also shows that in response to each of questions on anal sex there were some couples who offered conflicting information. Below, we examine this issue in detail and calculate reliability coefficients using the full Likert scale to increase the strictness of our analyses. Due to space limitations, only partial data is presented in tabular form for illustrative purposes.

Tables 3 and 4 show the interpartner agreement or disagreement in response to anal sex questions for the case of the HIV-positive partner penetrating the HIV-negative partner. The numbers in each cell represent the number of couples with a particular pattern of responses. For example, in the first line of Table 3 we see that in 35 couples both members agreed that the HIV-positive man never penetrated his HIV-negative partner during the twelve months prior to the interview; in that same line we see that there was one couple where the HIV-negative man said that he was always penetrated by his HIV-positive lover, while the lover said that this never happened. The diagonal cells in Tables 3 and 4 correspond to couple agreements.

Table 3. Congruence of Interpartner Responses to the
Question "Did the HIV+ Respondent Penetrate his
Partner?"

                    According to Negative Partner

According to       Never   Infrequently   Half the
Positive Partner                            Time

Never               35           1
Infrequently         5          11
Half the Time                    2           2
Frequently                                   1
Always               1

                   According to Negative Partner

According to            Frequently   Always
Positive Partner

Never                                   1
Infrequently                1           1
Half the Time               1
Frequently                  6           2
Always                      2           3

Table 4. Congruence of Interpartner Responses to the Question "Did the HIV+ Respondent Penetrate his Partner With a Condom?"

                                 According to Negative Partner

                                 Condom Use During Penetration

                                                       Half the
According to Positive Partner   Never   Infrequently     time

Condom use
  Never                           1
  Infrequently                    1          2
  Half the Time
  Frequently                                              1
  Always
  No penetration(a)

According to Positive Partner   Frequently   Always

Condom use
  Never                             1
  Infrequently
  Half the Time
  Frequently                        5
  Always                            1          20
  No penetration(a)                             2

According to Positive Partner   No penetration(a)

Condom use
  Never
  Infrequently
  Half the Time
  Frequently                           1
  Always                               5
  No penetration(a)                   35

(a) Data excluded from the calculation of unweighted and weighed kappas that are restricted to actual answers.

The additional values computed following the application of the modified response codings previously described are indicated by the "No penetration" section. These data are excluded from the calculation of unweighted and weighted kappas that are restricted to actual answers, that is, to answers not inferred from previous answers. A similar procedure was employed to analyze the question "Did the HIV-positive positive respondent ejaculate in his partner's rectum?"

Table 5 presents kappa values for the interpartner agreement to the six questions under study. It can be observed that all but one of the coefficients is significant at the p [is less than] .001 level, and that, following Landis and Koch (1977), most coefficients fall into the moderate to almost perfect agreement categories. The sole exception is the restricted kappa value for the ejaculation question, undoubtedly influenced by the small number of men answering it.

Table 5. Kappa Coefficients of Interpartner Agreement

                                      Kappa

                                    Unweighted

                               Restricted(a)   All(b)
Question                            (n)         (n)

A. Did the HIV+ respondent         .636         .636
penetrate his partner?             (75)         (75)

B. Did the HIV+ respondent         .768         .741
penetrate his partner with a       (32)         (75)
condom?

C. Did the HIV+ respondent         NC(c)        .774
ejaculate in his partner's         (10)         (74)
rectum?

D. Did the partner penetrate       .513         .513
the HIV+ respondent?               (75)         (75)

E. Did the partner penetrate       .519         .579
the HIV+ respondent with a         (35)         (75)
condom?

F. Did the partner ejaculate       .281(ns)     .625
in the HIV+ respondent's           (15)         (68)
rectum?

                                       Kappa

                                     Weighted

                                Restricted      All
Question                            (n)         (n)

A. Did the HIV+ respondent         .760         .760
penetrate his partner?             (75)         (75)

B. Did the HIV+ respondent         .763         .841
penetrate his partner with a       (32)         (75)
condom?

C. Did the HIV+ respondent         NC(c)        .741
ejaculate in his partner's         (10)         (74)
rectum?

D. Did the partner penetrate       .699         .699
the HIV+ respondent?               (75)         (75)

E. Did the partner penetrate       .739         .700
the HIV+ respondent with a         (35)         (75)
condom?

F. Did the partner ejaculate       .428         .656
in the HIV+ respondent's           (15)         (68)
rectum?

Note. All kappas were significant at p < .001, with the exception of restricted unweighted kappa for question F (ns).

(a) Restricted to cases where both partners answered contingent question. (b) All cases included (except those with real missing data), (c) Not Calculated. Negative partners responses were a constant, all answered never.

The results of the test-retest analyses showed that weighted kappas involving the total sample ranged from .83 to 1.00. As in the intracouple reliability results, unweighted kappas were somewhat lower, although still quite high (.69 to 1.00).

Qualitative Results

As stated in the Methods section, the lack of concordance between partners in their reports of sexual behavior with each other was explored with open-ended questions formulated by the interviewer in the presence of both members of the couple. The main reasons for interpartner discrepancies were poor recall of infrequent incidents, inaccurate restriction of the response to the given time frame, different views on what constitutes sex despite tight and detailed definitions in the assessment ("Maybe he counts when he puts movies on and I'm there"), and technical differences ("We don't use condoms for foreplay but [we use them] when one is about to come"). Finally, some men simply could not reach an agreement in the report of the sexual behavior they had with each other.

DISCUSSION

Approximately half of the men in this sample of couples of mixed HIV-status practiced anal sex, and between half and one quarter of those men did not use protection consistently (more detailed results on sexual behavior within these couples are reported in Remien, Carballo-Dieguez, & Wagner, 1996). Our results support the impression that, overall, the retrospective 12-month self-reports of anal sex in male couples of mixed HIV status are reliable. Primary partners confirmed, for the most part, the report of index respondents. Although initially we failed to find a significant agreement--beyond chance--in the responses to the question on whether the HIV-negative partner ejaculated inside the rectum of his HIV-positive partner, when the responses were weighted the values fell in the moderate agreement range.

The qualitative results show that the most frequent reasons for interpartner discrepancies are poor recall of isolated incidents, inaccuracy in limiting responses to the given time frame, different interpretations of what constitutes sex, and, occasionally, plain lack of agreement about what happened during the sexual encounter. These results call for the construction of assessments with improved accuracy. Perhaps further probing could be used in the case of absence of reporting of certain behaviors. For example, if someone said he penetrated his partner unprotected but never ejaculated inside his partner's rectum, the interviewer could ask "Not even once?" in an attempt to elicit isolated incidents. This type of probing should be consistent throughout the questionnaire for any denial of risky behavior.

To deal with respondents' inaccuracy in limiting the period covered by their response to the one established in the assessment, shorter time frames may be utilized, as other authors have proposed (Kauth, St. Lawrence, & Kelly, 1991). This, however, does not solve the problem of seasonal variations in sexual behavior, especially in the case of populations that migrate to a different geographical area for vacations or work.

The persistence of idiosyncratic interpretations to our sexual questionnaire showed the personal constructions respondents make of their sexual behavior. Confronted with the question "How frequently did you put the head of your penis in your partner's anus without a condom?" some individuals responded "never when I was about to ejaculate" but "frequently during foreplay." One could try to deal with this issue by stressing at the beginning of the interview that questions are not to be qualified by "ifs and buts," but a certain degree of personal interpretation is probably unavoidable. Those personal interpretations may constitute the most fertile ground to focus on during the development of primary prevention interventions.

The test-retest of participants provided additional support to the reliability of anal sex self-reports. Finally, the admittance of engaging in behaviors that are socially condemned further leads us to believe in the honesty of the reports, although, of course, we had no way of detecting couples where both participants may have failed to report their actual risky sexual behavior.

One obvious limitation of our study is that while our analyses have focused on the reliability of sexual self-reports, no claims can be made concerning their validity-the extent to which the measures actually reflect behavior (Catania et al., 1993; Zenilman, 1995). Other limitations include potential sample bias given our small number of participants, and the use of paid volunteers who may constitute an especially motivated group. Also, the participants' awareness that their partners were going to be asked the same questions and that discrepancies would be discussed may have further motivated respondents to provide accurate information. Finally, face-to-face interviews have resulted in reports of less risk behavior in some studies (Boekeloo et al., 1994). Within these constraints, however, our assessment offers satisfactory reliability.

REFERENCES

Boekeloo, B. O., Schiavo, L., Rabin, D. L., Conlon, R. T., Jordan, C. S., & Mundt, D. J. (1994). Self-reports of HIV risk factors by patients at a sexually transmitted disease clinic: Audio vs written questionnaires. American Journal of Public Health, 84, 754-761.

Carballo-Dieguez, A., Exner, T., Gruen, R., & Meyer-Bahlburg, H. (1994). Sexual Practices Assessment Schedule-Couples Version (SPAS-C).

Catania, J. A., Coates, T. J., Peterson, J., Dolcini, M. M., Kegeles, S., Siegel, D., Golden, E., & Fulllilove, M. T. (1993). Changes in condom use among Black, Hispanic, and White heterosexuals in San Francisco: The AMEN cohort study. The Journal of Sex Research, 30, 121-128.

Catania, J. A., Gibson, D. R., Chitwood, D. D., & Coates, T. J. (1990). Methodological problems in AIDS behavioral research: Influences on measurement error and participation bias in studies of sexual behavior. Psychological Bulletin, 108, 339-362.

Catania, J., Kegeles, S., & Coates, T. (1990). Towards an understanding of risk behavior: An AIDS risk reduction model (ARRM). Health Education Quarterly, 17, 381-399.

Coates, R. A., Soskolne, C. L., Calzavara, L., Read, S. E., Fanning, M. M., Shepherd, F. A., Klein, M. M., & Johnson, J. K. (1986). The reliability of sexual histories in AIDS-related research: Evaluation of an interview-administered questionnaire. Canadian Journal of Public Health, 77, 343-348.

Cohen, J. (1960). A coefficient of agreement for nominal scales. Educational and Psychological Measurement, 20, 37-46.

Cohen, J. (1968). Weighted kappa: Nominal scale agreement with provision for scaled disagreement or partial credit. Psychological Bulletin, 70, 213-220.

Darke, S., Hall, W., Heather, N., Ward, J., & Wodak, A. (1991). The reliability and validity of a scale to measure HIV risk-taking behavior among intravenous drug users. AIDS, 5, 181-185.

Fleiss, J. L. (1981). Statistical methods for rates and proportions (2nd ed.). New York: John Wiley and Sons.

Fleiss, J. L., & Cohen, J. (1973). The equivalence of weighted kappa and the intraclass correlation as measures of reliability. Education and Psychological Measurement, 33, 613-619.

Jacobson, N. S., & Moore, D. (1981). Spouses as observers of the events in their relationships. Journal of Consulting and Clinical Psychology, 49, 269-277.

James, N. J., Bignell, C. J., & Gillies, P. A. (1991). The reliability of self-reported sexual behavior. AIDS, 5, 333-336.

Kauth, M. R., St. Lawrence, J. S., & Kelly, J. A. (1991). Reliability of retrospective assessments of sexual HIV risk behavior: A comparison of biweekly, three-month, and twelve-month self-reports. AIDS Education and Prevention, 3, 207-214.

Kinsey, A., Pomeroy, W., & Martin, C. (1948). Sexual behavior in the human male. Philadelphia: Saunders Co.

Kinsey, A., Pomeroy, W., & Martin, C. (1953). Sexual behavior in the human female. Philadelphia: Saunders Co.

Landis, J. R., & Koch, G. G. (1977). The measurement of observer agreement for categorical data. Biometrics, 33, 159-174.

Levinger, G. (1966). Systematic distortion in spouses' reports of preferred and actual sexual behavior. Sociometry, 29(3), 291-299.

Mahler, J., Hull, J., Stebinger, A., Perry, S., & Frances, R. (1995). Reliability of sexual risk behavior interviews on an alcohol rehabilitation unit. Psychiatric Services, 46(2), 115-117.

McElrath, K., Chitwood, D. D., Griffin, D. K., & Comerford, M. (1994). The consistency of self-reported HIV risk behavior among injection drug users. American Journal of Public Health, 84, 1965-1970.

McKinnon, K., Couruos, F., Meyer-Bahlburg, H. F. L., Guido, J. R., Carballo, L. R., Margoshes, E. S., Herman, R., Gruen, R. S., & Exner, T. M. (1993). Reliability of sexual risk behavior interviews with psychiatric patients. American Journal of Psychiatry, 150, 972-974.

Padian, N. S. (1990). Sexual histories of heterosexual couples with one HIV-infected partner. American Journal of Public Health, 80, 990-991.

Parker, R. G., & Carballo, M. (1990). Qualitative research on homosexual and bisexual behavior relevant to HIV/AIDS. The Journal of Sex Research, 27, 497-525.

Remien, R. H., Carballo-Dieguez, A., & Wagner, G. (1996). Prevalence and determinants of sexual risk behavior in male couples of mixed HIV status. Poster session (MOD. 1711) presented at the XIth International Conference on AIDS, Vancouver, Canada.

Saltzman, S. P., Stoddard, A. M., McCusker, J., Moon, M. W., & Mayer, K. H. (1987). Reliability of self-reported sexual behavior risk factors for HIV infection in homosexual men. Public Health Reports, 102, 693-697.

Seage, G. R., Mayer, K. H., Horsburgh, C. R., Cai, B., & Lamb, G. A. (1992). Corroboration of sexual histories among male homosexual couples. American Journal of Epidemiology, 135, 79-84.

Siegel, K., Krauss, B. J., & Karus, D. (1994). Reporting recent sexual practices: Gay men's disclosure of HIV risk by questionnaire and interview. Archives of Sexual Behavior, 23, 217-230.

Upchurch, D. M., Weisman, C. S., Shapherd, M., Brookmeyer, R., Fox, R., Celentanto, D. D., Colletta, L., & Hook, E. W. (1991). Interpartner reliability of reporting of recent sexual behaviors. American Journal of Epidemiology, 134, 1159-1166.

Weatherburn, P., Hunt, A. J., Davies, P. M., Coxon, A. P., & McManus, T. J. (1991). Condom use in a large scale cohort of homosexually active men in England. AIDS Care, 3, 31-41.

Zenilman, J. M., Weisman, C. S., Rompalo, A. M., Ellish, N., Upchurch, D. M., Hook, E. W., & Celentano, D. (1995). Condom use to prevent incident STDs: The validity of self-reported condom use. Sexually Transmitted Diseases, 22, 15-21.

Manuscript accepted September, 16, 1998

Alex Carballo-Dieguez, Robert H. Remien, Curtis Dolezal, and Glenn Wagner HIV Center for Clinical and Behavioral Studies, New York State Psychiatric Institute, New York City, NY and Columbia University.

The authors would like to thank Fernando Carnavali, Arthur Fox, and Ray Smith for their collaboration as research assistants; the administrative support of Roberta Leftenant; and the study participants who candidly offered the intimate information that made this study possible.

Address correspondence to Alex Carballo-Dieguez, Ph.D., 1051 Riverside Drive, Unit 15, New York, NY 10032; e-mail: [e-mail]