Golden, Matthew R. MD, MPH*†; Stekler, Joanne MD, MPH*†; Kent, James B. MS*; Hughes, James P. PhD†‡; Wood, Robert W. MD*†
PARTNER COUNSELING AND REFERRAL SERVICES (PCRS) are efforts made by public health authorities to assure that the sex and needle sharing partners of persons with sexually transmitted infections (STI) are notified, tested and treated. In 2001, US health departments provided PCRS to approximately one-third of persons with newly diagnosed HIV.1 However, in recent years, the Centers for Disease Control and Prevention (CDC) has placed greater emphasis on PCRS,2,3 and new CDC guidelines recommend that health departments link PCRS to HIV surveillance and provide the service to all persons with newly diagnosed HIV.4
In the United States, public health staff–usually Disease Intervention Specialists (DIS)–record the results of PCRS efforts using a standardized set of CDC disposition codes. For HIV, these codes capture the following information: (1) whether a partner was notified; (2) whether the partner had previously been diagnosed with HIV; (3) whether the partner HIV tested as a result of the case’s diagnosis; and (4) the results of that test.
Starting in 2001, Public Health–Seattle & King County (PHSKC) began expanding its HIV PCRS program with the goal of providing the service to all persons newly diagnosed with HIV in King County, WA.5 In developing the program, we identified 3 aspects of the disposition codes which we believed were problematic. First, the codes do not distinguish verified from unverified outcomes. Traditional CDC training dictates that DIS should record only outcomes that they can verify with medical providers. However, the failure to record patient reported (unverified) outcomes may result in an underestimation of PCRS success. Second, the disposition codes conflate notification, testing, and knowledge of test results. They are consequently complex and cannot capture all potential outcomes. Finally, the codes do not distinguish whether a partner was notified or tested before or after an index case received PCRS.
Here we evaluate our PCRS program using disposition codes that disaggregate the components of PCRS (notification, testing, and test results), distinguish verified and unverified outcomes, and differentiate outcomes that happen before and after DIS interviews.
The PHSKC HIV PCRS program identifies newly diagnosed cases of HIV when they are tested through health department-operated or -funded testing programs, through referral from private sector providers, via public health surveillance, and as part of a program designed to link persons with newly diagnosed HIV to ongoing medical care. Washington State law requires medical providers to report new cases of HIV to PHSKC, and requires laboratories to report the names of persons testing positive for antibodies to HIV, and persons tested for HIV RNA (positive or negative) or CD4+ lymphocytes.
If clients with newly diagnosed HIV are in the STD clinic, DIS interview them on-site. When PHSKC identifies a person with newly reported HIV via surveillance, DIS contact the reporting provider to obtain permission to contact their patients; Washington State law permits providers to assume responsibility for PCRS and to refuse the health department permission to communicate with their patients. After obtaining contact information, DIS attempt to interview cases, either by telephone or in person, using a structured interview record. This record captures information about the case’s sex and needle sharing partners during a defined interview period. We define the interview period as being 3 months before the case’s last negative HIV test until the time of their HIV diagnosis. If a case has never tested HIV negative, DIS ask them about any illness consistent with acute HIV and, if they identify such an illness, the interview period is defined as 3 months before that time until the time of HIV diagnosis. If the DIS cannot identify a potential syndrome of acute HIV, the interview period is defined as 2 years unless the case has no partners in that timeframe, in which case they are asked about their last partner.
DIS attempt to establish a means to notify and test each partner, and encourage cases to allow them to contact partners directly. DIS reinterview cases in the week after their initial interview and continue to contact cases until all identified partners have been notified and tested, or until the DIS determines that additional efforts are futile.
DIS record disposition codes when they complete the initial interview, and again when they close the case. These codes separately record whether a partner was notified, the partner’s previous HIV status, whether the partner had a new HIV test, the results of that test, and whether that result was verified. We consider test results verified only if DIS see a document with the result (e.g., laboratory report) or learn the result from a medical provider who tested the partner. We regard outcomes as unverified if they are reported by the case or their partner, but not otherwise confirmed.
We used logistic regression to compare the characteristics of persons who received PCRS to those who did not, Poisson regression in a log linear model to assess the association of case demographic characteristics and behaviors and PCRS procedures with the number of partners notified and diagnosed with HIV per case, and generalized estimating equations to assess the association of case and partnership characteristics with the likelihood that a specific described partner was notified or tested HIV positive. For multivariable models, we initially included variables with univariate P values of <0.1, eliminated nonsignificant variables stepwise, and retained only statistically significant variables (P <0.05) in our final models. We compared numbers of partners notified and newly diagnosed with HIV among men who have sex with men (MSM), women, and heterosexual men using the Kruskal-Wallis test.
The University of Washington Institutional Review Board approved this analysis. We performed all statistical analyses using the SAS system (Cary, NC).
Between October 2005 and September 31, 2007, PHSKC identified 890 persons age >14 initially thought to have newly diagnosed HIV. DIS investigations resulted in closure of 231 (26%) cases without interview. These cases were closed for the following reasons: previous HIV diagnosis (n = 112), anonymous HIV test (n = 68), residence outside King County (n = 19), case deceased or too ill to be interviewed (n = 13), diagnosis as part of a study (n = 10), duplicate report (n = 5), interviewed by DIS working in a detention facility (n = 4). Of the remaining 659 cases, DIS interviewed 427 (65%). Four additional cases completed partial interviews. The remaining 228 cases were not interviewed for the following reasons: case not located (n = 89), provider refused to allow PCRS (n = 61), case refused interview or did not respond to contact attempts (n = 61), foreign language for which translation was unavailable (n = 3), and other reasons (n = 4). DIS did not record a reason for not interviewing 10 cases.
On multivariate analysis, women, persons diagnosed with HIV through PHSKC, and cases reported more quickly after their diagnosis were more likely to receive PCRS (Table 1). The likelihood of receiving PCRS was not significantly associated with age or race/ethnicity.
The 427 persons interviewed for PCRS included 307 (73%) MSM, 20 (5%) bisexual men, 44 (10%) heterosexual men, and 51 (12%) women (DIS did not record a sexual orientation for five men). The population was 64% white, with a median age of 34 years (range 16–75).
A total of 382 (89%) of the 427 interviewed cases answered questions about their total number of partners during their interview period. These 382 persons reported having 4254 partners, but provided specific information about only 766 (18%). Among interviewed cases, 369 (86%) provided sufficient information about at least 1 partner to allow DIS to assign that partner a disposition.
Table 2 presents data on PCRS outcomes, comparing verified and unverified results, and outcomes based on whether partners were notified or HIV tested before or after an initial DIS interview. Based on either verified or unverified dispositions, among the 766 partners, 138 (18%) had previous HIV diagnoses. Of the remaining 628, DIS classified 410 (65%) as notified, 295 (47%) as tested, and 35 (6%) as newly diagnosed with HIV. DIS verified outcomes for 183 (24%) of the 766 partners. Among partners who HIV tested, the proportion testing HIV positive was similar among partners with verified and unverified results (14% vs. 11%, P = 0.34). The number of index cases DIS needed to interview [number needed to interview (NNTI)] to identify 1 new case of HIV varied based on whether we defined NNTI to included unverified outcomes and partners diagnosed with HIV before cases received PCRS. Including all partners diagnosed with HIV (i.e., including unverified outcomes and partners tested before PCRS), the NNTI was 12.2. Excluding partners diagnosed before DIS interviewed cases increased the NNTI to 30.5. Further limiting the number of partners defined as newly diagnosed to include only persons for whom DIS verified an HIV diagnosis increased the NNTI to 47.4.
Table 3 separately presents data on PCRS outcomes among MSM, women, and heterosexual men. Compared with women and heterosexual men, MSM acknowledged having significantly more partners and provided DIS with information about more partners per case interviewed. However, MSM described a significantly smaller proportion of their total acknowledged partners to DIS than did women or heterosexual men. The NNTI was higher for MSM and for heterosexual men than for women, though this difference was not statistically significant. Overall, 138 (32%) of 427 people with HIV had at least 1 partner with previously or newly diagnosed HIV.
Figure 1 presents the mean number of partners notified and newly diagnosed with HIV, stratified by the time between the case’s diagnosis and their receipt of PCRS; these indices are further divided by whether partners were notified and tested before or after the case first received PCRS. Results here, as well as throughout the remainder of the results section, include both verified and unverified outcomes. The mean number of partners notified was higher among persons interviewed 15 to 28 days after diagnosis than among those interviewed sooner or later. Comparing cases interviewed in the first 28 days after diagnosis with those interviewed later, those interviewed within 28 days of their diagnosis were significantly more likely to have at least 1 partner notified directly by DIS (27% vs. 16%, P = 0.007), notified more partners after DIS interview (mean 0.27 vs. 0.10, P = 0.004), and notified more total partners (mean 1.2 vs. 0.85, P = 0.04). Cases interviewed within 28 days of diagnosis had more partners diagnosed after their initial DIS interview than those receiving PCRS later (mean 0.07 vs. 0.01, P = 0.01). However, although the total number of partners newly diagnosed with HIV per case was higher among persons interviewed in the 28 days after their HIV diagnosis than in persons interviewed later, this difference was not statistically significant (mean 0.11 vs. 0.06, P = 0.19).
On multivariate analysis, notifying more partners was significantly associated with age <25, having more sex partners, partner notification interview in the 15 to 58 days after HIV diagnosis (vs. >58 days after diagnosis), testing HIV negative in the preceding year, and participation in a program designed to link the case to HIV services (Table 4). Only age <25 was significantly associated with a higher HIV case-finding yield on multivariate analysis. This association with age remained significant [odds ratio (OR) 3.6, 95% confidence interval (CI) 1.2–10.7] in a model using number of partners diagnosed with HIV after PCRS as an outcome, controlling for time between HIV diagnosis and receipt of PCRS.
Among the 410 partners without a previous HIV diagnosis about whom cases gave specific information, the following factors were associated with successful notification on multivariate analysis: meeting a partner through friends (vs. through any other means) (OR 7.0, 95% CI 2.4–20), reporting that a partner had told the case that they were HIV negative (vs. not knowing a partner’s HIV status) (OR 11.3, 95% CI 5.1–25.1), and duration of relationship with partner of ≤1 day or 2 to 364 days (vs. ≥365 days) (OR 0.16, 95% CI 0.07–0.37 and OR 0.35, 95% CI 0.15–0.81, respectively). Factors associated with a partner testing HIV positive on multivariate analysis included: unprotected anal or vaginal intercourse (vs. oral sex only or protected intercourse only) (OR 9.6, 95% CI 1.2–77.5), reporting that a partner was HIV negative (vs. unknown HIV status) (OR 3.6, 95% CI 1.1–11.2), and duration of relationship with partner of ≤1 day or 2 to 364 days (vs. ≥365 days) (OR 0.14, 95% CI 0.03–0.67 and OR 0.28, 95% CI 0.08–0.99, respectively). None of the 107 persons with a sex partner with a previous HIV diagnosis had another partner newly diagnosed with the infection. Neither successful notification nor a partner’s risk of testing HIV positive was independently associated with the case’s age, race, sexual orientation, number or partners, methamphetamine use or sildenafil use; time between HIV diagnosis and PCRS interview; participation in a program to link the case to ongoing HIV care; or allowing DIS to notifying a partner directly. Successful notification was also not associated with the type of sex the case had with partner (i.e., oral vs. anal/vaginal) or condom use.
Using newly developed disposition codes, we evaluated a population-based PCRS program in King County, WA. Measures of PCRS case-finding effectiveness were highly dependent on how we defined partner dispositions, but our findings generally suggest that the effectiveness of PCRS has been overestimated in at least some previous analyses. PCRS success was greatest among young people, those who tested HIV negative within the last year, persons interviewed 15 to 58 days after their HIV diagnoses, and persons who participated in a program designed to link them with ongoing HIV care.
Our basic case-finding index, a NNTI of 12.2, is similar to that reported in previous studies. A national survey of HIV PCRS outcomes found that 13.8 cases needed to receive PCRS to diagnose 1 new case of HIV,1 whereas a summary of published PCRS program evaluations reported a NNTI of 9.1.6 However, the procedures used to define PCRS dispositions in most of these evaluations were ill defined. Recently, the San Francisco Department of Health STD Program evaluated their PCRS efforts. Their analysis included both verified and unverified outcomes, and the authors restricted the definition of new HIV diagnoses to partners tested after case’s receipt of PCRS. They reported a NNTI of 26,7 though this index would have been lower had the program included diagnoses made before case’s receipt of PCRS.7,8 Our observations are consistent with these findings and suggest that within a population-based PCRS program, half or more of all partners of persons receiving PCRS who are newly diagnosed with HIV are tested before the case receives the intervention. Also, it seems likely that at least some people who notified partners after receiving PCRS would have done so in the absence of the service. These observations demonstrate how PCRS disposition codes can overestimate program success.
Disposition codes may also underestimate success. At least in King County, DIS verify a small minority of outcomes. To the extent that this is true elsewhere, relying only on verified outcomes underestimates the numbers of partners who are notified, tested, and newly diagnosed with HIV. Also, our DIS indicated that 20% of the partners described by cases were notified, but that PHSKC could not determine if they had tested for HIV. At least some of these partners almost certainly tested.
We identified several program and healthcare-related factors associated with PCRS success, including recent HIV testing, the time interval between HIV diagnosis and receipt of PCRS, and participation in a program designed to transition persons with newly diagnosed HIV into ongoing medical care. The association between the timing of PCRS and PCRS success was complex, with lower notification and case-finding indices in people interviewed ≤2 weeks after their diagnosis, higher indices in people receiving PCRS 15 to 28 days after diagnosis, and lower indices when DIS interviewed cases >28 days after diagnosis. This finding merits confirmation, but it may be that immediate PCRS is not optimal, that some people with HIV need time to adjust before trying to notify partners.
We also found that PCRS case-finding yields were higher among persons age <25 than in older persons, but were similar in MSM and heterosexuals. Some previous studies have suggested that PCRS is less effective in MSM.1,10 Although we did not observe this, MSM reported notifying a much smaller proportion of their total partners than heterosexuals, a finding that is consistent with some studies of partner notification among MSM with bacterial STD.11,12 These findings suggest that although PCRS may be comparably effective in identifying new cases of HIV and STD among MSM and heterosexuals, it may be a less effective public health intervention for preventing HIV/STD transmission in MSM.
Our evaluation of a single HIV PCRS program may not be generalizable to other areas, and, insofar as we included unverified outcomes, some of our dispositions may be inaccurate. Also, because PCRS identified a small number of new cases, we had limited power to identify factors associated with case finding success. Finally, we cannot say with confidence that the associations we observed between PCRS success and testing frequency, the timing of PCRS and participation in a program to link patients to ongoing care were causal.
In summary, we found that our PCRS program identified a potential source patient in only approximately one-third of cases, and that measurements of HIV PCRS success are highly dependent on how one defines PCRS dispositions. Our findings highlight the need to develop and disseminate a more explicit and nuanced approach to PCRS program assessment nationally. They also suggest that support for PCRS, long dampened by concerns related to privacy and individual liberty,14–16 may now exceed what is justified by existing data. Very few controlled studies have evaluated HIV PCRS,10,17 and better PCRS evaluations, particularly randomized controlled trials, may be warranted. However, rigorously controlled data will likely take years to generate and may never be available. Based on that reality and the paucity of alternative public health measures proven to prevent HIV, we do not believe that limitations in what we know about PCRS should prompt health departments to delay implementing CDC’s new guidelines. Rather, our findings should provide an additional impetus to HIV test persons at high risk frequently, to ensure that PCRS is timely, and to develop and widely institute programs that link persons diagnosed with HIV with appropriate medical follow-up.