Percentage of human immunodeficiency virus (HIV)–infected patients seeking primary care over time.
Two-way interaction effect between sex and CAGE score, which is an alcoholism screening questionnaire containing 4 structured questions.
Customize your JAMA Network experience by selecting one or more topics from the list below.
Samet JH, Freedberg KA, Stein MD, Lewis R, Savetsky J, Sullivan L, Levenson SM, Hingson R. Trillion Virion DelayTime From Testing Positive for HIV to Presentation for Primary Care. Arch Intern Med. 1998;158(7):734–740. doi:10.1001/archinte.158.7.734
Copyright 1998 American Medical Association. All Rights Reserved. Applicable FARS/DFARS Restrictions Apply to Government Use.1998
Human immunodeficiency virus (HIV)–infected individuals' initial presentation to medical care frequently occurs at a point of advanced immunosuppression.
To investigate the time between HIV testing and presentation to primary care. Also to examine factors associated with delayed presentation.
One hundred eighty-nine consecutive outpatients without prior primary care for HIV infection were assessed at 2 urban hospitals: Boston City Hospital, Boston, Mass, and Rhode Island Hospital, Providence. Sociodemographics, alcohol and drug use, social support, sexual beliefs and practices, and HIV testing issues were examined in bivariate and multivariate analyses for association with delay in presentation to primary care after positive test results for HIV.
Of 189 patients, 74 (39%) delayed seeking primary care for more than 1 year, 61 (32%) delayed for more than 2 years, and 35 (18%) for more than 5 years after an initial positive HIV serologic evaluation. The median CD4+ cell count of subjects was 0.28×109/L (range, 0.001-1.71×109/L). In multiple linear regression analysis the following characteristics were found to be associated with delayed presentation to primary care after HIV testing: history of injection drug use (P<.001); not having a living mother (P=.01); not having a spouse or partner (P=.08); not being aware of HIV risk before testing (P<.001); and being notified of HIV status by mail or telephone (P=.002). An interaction effect between sex and screening for alcohol abuse was significant (P=.03) and suggested longer delays for men with positive screening test results (CAGE [an alcoholism screening questionnaire containing 4 structured questions], 2+) compared with men without positive screening test results or women.
Patients with positive HIV test results often delay for more than a year before establishing primary medical care. Information readily available at the time of HIV testing concerning substance abuse, social support, and awareness of personal HIV risk status is useful in identifying patients who are at high risk of not linking with primary care. Patients who were notified of their HIV status by mail or telephone delayed considerably longer than those notified in person. Efforts to ensure primary care linkage at the time of notification of positive HIV serostatus are necessary to maximize benefits for both individual and public health and should be an explicit task of posttest counseling.
HUMAN immunodeficiency virus (HIV)–infected patients frequently obtain primary medical care to address their HIV disease long after initial infection.1 This is problematic because early presentation for medical care yields maximum potential benefits for both individual and public health. Individuals can live longer and with less morbidity with increasingly aggressive antiretroviral therapy and opportunistic infection prophylaxis.2,3 From a public health perspective, early presentation provides more of an opportunity to decrease transmission of the virus through patient education and treatment. We and others have reported the median CD4+ lymphocyte count among HIV-infected patients at initial presentation to medical care to be in the range of 0.3 to 0.45×109/L.1,4 This suggests that linkage with primary care happens after damage to the immune system has already occurred, likely years after seroconversion. After initial infection HIV continues to replicate at an estimated average rate of 2 billion virions each day.5 The active state of viral replication has led to calls for early aggressive therapy for this infection once the infected individual is engaged in medical care.6
Delay between the time of infection and initiating primary medical care can be conceptualized as 2 periods: (1) the time between acquisition of the virus and initial positive HIV test results and (2) the time between positive HIV test results and presentation to primary care. Although posttest HIV counseling is recommended in the United States, its execution and content is variable and no formal mechanism is required for linking infected individuals to care. Our objectives were to study the second part of the delay period, the time between positive HIV test results and presentation to primary care, and to examine characteristics associated with delayed presentation.
Patients were enrolled from 2 sites: the HIV Diagnostic Evaluation Unit, Boston City Hospital (BCH), Boston, Mass, from February 1994 to April 1996, and HIV Clinic, Rhode Island Hospital (RIH), Providence, from December 1994 to March 1996. The HIV Diagnostic Evaluation Unit is a weekly clinic designed for the initial assessment and triage of all new patients with HIV infection entering the BCH system, except those who are pregnant.7 Referrals to both sites come from a wide variety of sources, including inpatient hospital services, hospital outpatient clinics, self-referrals, the emergency department and urgent care clinic, community health centers, drug treatment programs, HIV testing sites, and local correctional institutions. The RIH is a private, nonprofit urban hospital. The research study was approved by the institutional review boards of both institutions.
The subjects were patients who sought primary care for HIV for the first time. We used specific criteria to identify patients who had not received primary medical care for HIV infection. Specifically, new to primary care was defined as the following: (1) an initial positive HIV test result within 4 calendar months of the evaluation; or (2) an initial positive HIV test result more than 4 months before presentation and an absence of the following history, determined by medical record review or patient report: specific prior HIV primary care, past use of zidovudine or any other antiretroviral, or 2 or more prior CD4+ lymphocyte counts. The criterion of 2, rather than 1, prior CD4+ lymphocyte counts was used because some patients had 1 CD4+ lymphocyte count obtained at the time of HIV testing. Only patients fluent in English, Spanish, or Haitian Creole were eligible. Each patient provided written informed consent before entering the study.
Patients were asked to participate in this study after their initial clinical care was performed, including medical history taking, physical examination, and laboratory tests. At RIH this was at the initial encounter, and at BCH this was at a clinical appointment generally 1 week after the first. Those who met entry criteria and agreed to participate underwent a 60- to 90-minute standardized interview. One of 3 research associates carried out all interviews concerning behavioral, medical, and social history. Interviews were administered in Spanish or Haitian Creole when appropriate by interpreters working with the research associates. Spanish and Haitian Creole interview instruments had been translated into these languages, back-translated into English to check for accuracy, and then corrected.
The outcome of interest was "delay after HIV testing" measured in months. Delay is defined as the time between self-reported initial positive HIV test results and the date of the initial HIV primary care appointment. Since commercial testing for HIV infection became available in March 1985, the maximum delay for any individual was the difference in months between the date of presentation to primary care and March 1985.
Independent variables were selected to address the patient's status at the time of the interview or HIV testing. The following specific variables were examined by the questions included in Table 1: demographics (sex, age, race or ethnicity, birthplace, English as a first language, education, employment when tested positive, highest yearly income, health insurance, homelessness, time in current residence, and jail time in the past); drug and alcohol use (CAGE questionnaire, a screening test for alcohol problems8); social support; victimization history; sexual beliefs and practices; medical and psychiatric issues; and status regarding HIV risk awareness at the time of testing. We used the Centers for Disease Control and Prevention9 hierarchical classification of HIV risk factors with the exception of the combined category of men who have sex with men and injection drug use. If injection drug use was positive, this was the designated risk factor even in the presence of other risk factors.
Reviewing the clinical record we examined CD4 lymphocyte counts obtained within 3 calendar months of the initial medical evaluation of the subjects from each institution. When 2 initial CD4+ cell counts prior to antiretroviral therapy were available we used the mean count.
Descriptive statistics were generated and analyzed for all study variables. Particular attention was paid to the distributional properties of the dependent variable, delay after HIV testing. The log transformation of delay in months was also considered as an outcome variable to account for the skewness of the distribution because of subjects with prolonged delay. Bivariate analyses were conducted to assess the relationships between each independent variable and the dependent variable using analysis of variance and Pearson correlations for discrete and continuous variables, respectively. Variables that were either significant at the P<.10 level in the bivariate analyses or clinically significant (ie, were important to consider as covariates) were considered candidates for a multiple linear regression model. We also conducted bivariate analyses to assess relationships among independent variables using χ2 analyses, analysis of variance, and Pearson correlations. When variables were highly correlated with one another, for example, r>0.4, one was selected for inclusion in the multiple regression model to reduce colinearity.10
A direct-entry multiple linear regression model was developed relating the candidate variables to delay after HIV testing and to the log transformation of delay. A Cox proportional hazards model was also estimated relating the delay in months to candidate variables. The 3 models were consistent, and the results of the multiple linear regression analysis are presented herein. Data were collected at 2 sites, but sample size limitations did not allow for development of separate models within each site. Relationships between significant independent variables and the dependent variable were evaluated separately for each site. Since in all cases the direction and magnitude of the associations were consistent, the data were pooled. An indicator of site was included in the final model to account for the difference in the mean delay in seeking care between sites.
Clinically relevant 2-way interaction terms were considered for inclusion in the multiple regression model and were entered in a stepwise fashion after inclusion of the main effects. The interaction terms considered included sex/CAGE, site/injection drug use, site/awareness of HIV risk status, site/victimization history, and site/spouse or partner with HIV or the acquired immunodeficiency syndrome.
A 2-tailed P<.05 was considered statistically significant in multivariate analyses. Data were analyzed using SAS statistical software.11
Enrolled patients represented 189 (72%) of all 262 eligible patients presenting for initial primary care for HIV infection from both sites during the period of study. This represents 144 (67%) of 216 patients at BCH and 45 (98%) of 46 patients at RIH. At the BCH site, 72 patients were not enrolled in this study, including 37 who refused to participate, 25 who agreed to participate but never returned for the initial interview, and 10 who were never contacted. There were no significant differences between patients who enrolled in the study at BCH (n=144) and those who were not enrolled (n=72) with respect to age, sex, and HIV risk group category. There was a significant difference with respect to race (P<.05). Disproportionately fewer Haitians (5 [19%] of 27) and more whites (44 [81%] of 54) enrolled in the study compared with Hispanics or Puerto Ricans and African Americans.
We examined a total of 189 patients, 76% of whom were from BCH and 24% from RIH. The mean age was 36 years. Other characteristics of the study population are described in Table 2. More than one quarter of the study patients were women and a majority of patients were nonwhite. The 2 most common primary HIV risk factors were injection drug use and heterosexual intercourse. The median CD4 lymphocyte count of the subjects was 0.28×109/L (range, 0.001-1.71×109/L) with less than a quarter presenting with CD4 lymphocyte counts greater than 0.5×109/L. Ninety-one (48%) of 189 subjects had 2 or more positive responses to the CAGE questionnaire, an alcohol screening test response highly suggestive of present or prior alcohol problems.12
The median delay from HIV testing to initial presentation for primary care was 3 months (mean delay, 24 months). Of 189 patients, 74 (39%) delayed seeking primary care for more than 1 year, 35 (32%) delayed for more than 2 years, and 61 (18%) for more than 5 years (Figure 1). For 3 subjects we considered the delay period shorter than reported, since according to these patients' reported history, testing had occurred before March 1985.
In the bivariate analysis the following factors were associated (P<.10) with increased delay (Table 3): male sex, living in residence less than 6 months, prior time in jail, not having a living mother, not having a spouse or partner, a score higher than 2 on the CAGE, injection drug use, unaware of their HIV risk status at the time of testing, and being notified of HIV status by mail or telephone. In addition, a history of victimization was associated with a shorter delay. Variables not found to be significantly associated with increased delay included age, health insurance at the time of positive test results, and the primary risk factor of men who have sex with men.
The impact of the time of enrollment in the study on delay in seeking care was examined since many advances in HIV care were made during the time of the study. We compared subjects who enrolled before March 1995 with those enrolled after this time. Our analysis showed no difference in the average delay for those who enrolled before (20 months) compared with those who enrolled after this time (27 months) (P=.19).
Prior time in jail was significantly associated with injection drug use; therefore, we eliminated prior time in jail to reduce colinearity. We considered the remaining variables as well as patient age, a potentially important covariate, in a multiple linear regression model. We found the following characteristics significantly (P<.05) associated with delay: history of injection drug use, not having a living mother, not aware of HIV risk status at testing, and being notified of HIV status by mail or telephone (Table 4). The mean additional delay in months attributable to the variables listed in Table 4 is reported as the parameter estimate and varies between 8.6 months and 30.4 months. The interaction between sex and CAGE score reached statistical significance (P<.05) and was retained in the model. This interaction term suggested longer delays in men abusing alcohol compared with men not abusing alcohol and women. Holding all other characteristics constant, men abusing alcohol delayed 14.6 months longer than men who did not. There was no significant difference in delay between women abusing and not abusing alcohol. However, men abusing alcohol delayed 28.5 months longer than women abusing alcohol (Figure 2).
Initial primary care presentation of patients with HIV infection generally occurs years after acquisition of the virus.1 The portion of the delay in medical care attributable to the period between positive HIV test results and initiating primary care has not been well defined. We found that although a majority of patients (n =189) make the initial linkage to medical care within 3 months, a substantial proportion (74 ) delayed for more than 1 year, approximately a third (61 ) delayed for more than 2 years, and nearly one fifth (35 ) waited more than 5 years. Considering the replication rate of HIV and the mean 24-month delay between notification of positive HIV test results and presentation to primary care, the average delay to primary care allows for more than 1 trillion replications of the HIV virus.
Detection of HIV infection through testing without linkage to primary medical care is problematic for several reasons. First, the individual is unable to benefit from therapeutic advances. These benefits include effective antiretroviral therapy, prophylaxis of opportunistic infections, immunizations, behavioral interventions, and future potential therapeutic vaccines.2,3 Aside from decreasing the morbidity of HIV disease, antiretroviral therapy has been clearly shown to decrease mortality in patients with CD4 lymphocyte counts higher than the median of those presenting for medical care.13,14 These benefits can only occur if HIV-infected patients are engaged in medical care. Patients may also benefit materially from social services gained as a result from linkage with medical care and from social support of the medical staff.
Second, the potential public health benefits of linking HIV-infected patients to medical care are missed. Medical therapy provided to individual patients has significant public health impact. Treatment of tuberculosis-exposed individuals prevents infection in the person receiving treatment and prevents spread to the 2 to 3 other people who would be exposed on average by 1 active case of tuberculosis.15 Antiretroviral therapy has clear implications for vertical HIV transmission, reducing spread to children from pregnant mothers.16 Antiretroviral therapy has been hypothesized to reduce infectivity, potentially affecting sexual transmission.17 In addition, results from the Mwanza trial18 indicate that treating sexually transmitted diseases results in reduced sexual transmission, which cannot occur unless the infected individual seeks medical care. Beyond the therapeutic interventions, health care providers' discussions with patients about appropriate behaviors provide opportunities to influence sexual behaviors and substance abuse. These tangible benefits are missed if medical care is delayed. Efforts to link patients newly diagnosed as having HIV to primary care should be a major explicit task of required HIV posttest counseling as described in the Centers for Disease Control and Prevention technical guidance on HIV counseling.19
THE RESULTS of this study provide insight into which patients are at increased risk for delayed linkage to primary medical care. Three major characteristics of patients were associated with significant delay: history of substance abuse, poor social support, and being unaware of their HIV risk status at the time of testing. An unexpected finding was that a patient's lack of awareness of risk for HIV at the time of testing was associated with an additional delay of 18 months in linking with primary medical care. The patient's element of surprise and the associated delay in medical linkage suggest that HIV testing for this group initiates the process of seeking help. This is in contradistinction to those who are aware of their risk, for whom HIV testing is the first demonstration of readiness to receive care after possibly long consideration of the multitude of issues that accompany seeking help for HIV. The assessment of the patient's perspective on this issue at the time of testing may reveal his or her tendency for increased delay in linking to medical care.
The finding that injection drug users delay, on average, 19 months longer than those without a history of such behavior supports our a priori hypothesis that substance abuse is a substantial barrier to medical care. Extensive efforts have been made to reach populations involved in substance abuse to provide HIV counseling and testing. Recognition that injection drug users are at great risk of becoming infected by the virus and transmitting HIV to their partners, and in fact account for 35% of new cases of acquired immunodeficiency syndrome diagnosed since 1990,9 has propelled these outreach efforts. Ensuring that patients who undergo HIV testing and are found to be infected successfully link to primary medical care is a crucial step in the individual and public health objectives of reaching this population. Including the arrangement of a primary care appointment as a standard during posttest counseling would build a care system that directly addresses this problem.
Interestingly, there is a significant interaction effect of sex and alcohol abuse. It suggests that a history of alcohol abuse in men, but not in women, is a risk factor for delayed presentation. In fact, men abusing alcohol delayed 14.6 months longer than men not abusing alcohol. Women abusing and those not abusing alcohol were similar with respect to delay. An explanation for the different results for men and women is not evident. However, this finding suggests the implementation of an intervention focused specifically at a population of HIV-infected men abusing alcohol at the time of the receipt of positive HIV test results.
Two factors reflecting poor social support, not having a living mother and not having a partner or spouse, were each independently associated with delayed presentation to primary care. Other markers of social support were not found to be significant: having a father or children, spending most free time alone, having close friends (0-1 vs >1), and close relationships with mother or any child. Again, patient characteristics concerning the presence of a mother and a partner or spouse are easily ascertainable at the time of HIV testing and could be used to determine which patients are at a higher risk for not presenting to primary medical care.
One patient characteristic found to be associated with earlier presentation for care in bivariate analysis was a reported history of victimization. This association may seem counterintuitive. One hypothesis we had at the initiation of the study was that a woman might delay acting on a positive HIV test result because it might end up in her being victimized by a partner upset by her serostatus, particularly if that woman had a history of victimization. This hypothesis is not borne out by the data. Although the bivariate result was the opposite of our hypothesis, the association disappears in the multivariate analysis and is probably not real.
One aspect related to the testing process itself was associated with significant delay in presentation for primary medical care. If the patient was not told in person of the test result (ie, by mail or telephone), then the linkage with care was, on average, 2.5 years later than for patients told in person. While our numbers are small, this represents the first published data of which we are aware that assesses the impact of the test result notification method on medical care–seeking behavior. Since the time of our study, newly licensed home HIV testing technologies have become available in which individuals learn of their HIV serostatus by telephone. Home collection test systems include protocols for referrals of those with positive test results to additional services, specifically medical care. It is essential to understand the possible implications of these alternative notification mechanisms even though they include posttest counseling.20 Although our results raise concerns with regard to home HIV testing, patients not notified in person were unlikely to have received posttest counseling and probably represent a different population than those now using home HIV testing. Patients in the past who were notified by mail or telephone were often those who obtained HIV testing in nonmedical settings, such as the military, at immigration, or as part of insurance applications. In our study, those who were notified of their positive test results by mail or telephone obtained HIV testing in a clinic, in prison, or as part of insurance applications. These individuals were first tested between March 1985 and September 1994 with 58% having been tested since 1992. These data illustrate the importance of closely assessing this issue prospectively among patients being informed of their positive HIV test results via the home testing notification systems and controlling for known factors associated with delayed presentation for medical care.
Limitations of this study deserve discussion. First, the population sampled was drawn from medical centers located in urban northeastern United States and may not reflect those patients receiving HIV test results at sites in other geographic regions within and beyond the United States. However, the initial presentation for medical care for HIV frequently occurring at the time of advanced immunosuppression is a phenomenon described at other sites in the United States and abroad.4,21,22 It would not be surprising if a portion of the delay to medical care at other sites was a result of the delay after HIV testing.
Second, we are relying on patient recall of the date of the first positive HIV test results and this date was not confirmed. This is a methodological limitation that is difficult to overcome because many such tests are performed in anonymous testing sites where confirmation is not possible. Nevertheless, we believe that this approach is reasonable since the receipt of a positive HIV test result is a major life event that is not generally forgotten or even recalled with uncertainty. An alternative study approach to assess this issue is a prospective design, obtaining a group of individuals at the time of notification of HIV-positive status. This approach would not be confounded by recall bias but may suffer from a Hawthorne effect, the study itself being an unavoidable intervention. Additionally, the prospective design would require substantial follow-up time to detect the long-term nonpresenters. It is not evident that any differences in conclusions would result from a prospective approach. Finally, our sample may not be representative of Haitians, a group of patients found previously to present with marked immunosuppression, as enrollment of these patients was disproportionately low.
In conclusion, 39% of HIV-infected patients at 2 urban hospitals delayed more than 1 year between the time they received positive HIV test results and the time they sought primary care. This represents an extended period of risk of further immune dysfunction, development of opportunistic infections, and transmission of HIV to other individuals. Injection drug users, male alcohol abusers, and those who were unaware of their HIV risk at the time of HIV testing were the most likely to delay presentation to medical care after receiving positive test results. Not being told one's test result in person, not having a living mother, and not having a spouse or partner were each significantly associated with delay. Information readily available at the time of HIV testing related to substance abuse, awareness of HIV risk, and social support is useful in identifying patients who are at high risk of not linking with medical care. Increased efforts need to ensure that patients who receive positive HIV test results are quickly and successfully linked to primary medical care.
Accepted for publication August 17, 1997.
Dr Samet was a Generalist Physician Faculty Scholar, Robert Wood Johnson Foundation, Princeton, NJ.
We appreciate the contributions of our clinical staffs, Colleen LaBelle, RN, Susan Hart, MSW, and Kristin Jhamb, MD, as well as other contributors to the project: Jason Hammond, Margaret Marisi, and Kate Karter.
Reprints: Jeffrey H. Samet, MD, MA, MPH, Section of General Internal Medicine, Research Unit, 91 E Concord St, Suite 200, Boston Medical Center, Boston, MA 02118 (e-mail: email@example.com).