Share this article on:

Nonfinancial Factors Associated With Decreased Plasma Viral Load Testing in Ontario, Canada

Raboud, Janet M PhD*†; Abdurrahman, Zainab B MSc‡; Major, Carol MSc§; Millson, Peggy MD†; Robinson, Greg MD, MHSc; Rachlis, Anita MD†∥; Bayoumi, Ahmed M MD†¶

JAIDS Journal of Acquired Immune Deficiency Syndromes: 1 July 2005 - Volume 39 - Issue 3 - pp 327-332
Epidemiology and Social Science

Objective: To examine whether individual characteristics were associated with differential use of viral load testing when testing is available without charge to all HIV-positive patients with provincial health insurance.

Methods: Individuals enrolled in the HIV Ontario Observational Database with complete medication records and health insurance numbers for linkage were studied. Generalized estimating equation regression models were used to examine the relationship between time-varying covariates such as plasma viral load levels, CD4 counts, and antiretroviral regimen characteristics and the number of days between viral load tests and the occurrence of an interval of ≥6 or 9 months between tests.

Results: A total of 1032 individuals were included in the analysis with a median follow-up of 4.6 years and a median of 18 viral load tests. In multivariate analyses, clinically important gaps in viral load testing were more likely among injection drug users (odds ratio [OR] = 1.86, P < 0.0001), in more recent years (P < 0.01) and for individuals not using antiretrovirals (OR = 1.70, P < 0.0001) and less likely among individuals using >4 antiretrovirals (OR = 0.62, P < 0.0001). Results were similar when the outcome was the number of days between tests.

Conclusions: Injection drug users, younger individuals, and residents of Toronto used fewer viral load tests than other individuals, even when financial barriers to testing were removed.

From the *Division of Infectious Diseases, University Health Network, and †Faculty of Medicine, University of Toronto, Ontario, Canada; ‡Department of Statistics and Actuarial Science, University of Waterloo, Ontario, Canada; §Ontario HIV Treatment Network, Toronto, Ontario, Canada; ∥Sunnybrook and Women's College Health Sciences Centre, Toronto, Ontario, Canada; ¶Inner City Health Research Unit, St. Michael's Hospital, Toronto, Ontario, Canada; and Community Linked Evaluation AIDS Resource, McMaster University, Hamilton, Ontario, Canada.

Received for publication November 3, 2003; accepted August 17, 2004.

Dr. Bayoumi is supported by a Career Scientist Award from the Ontario HIV Treatment Network. Dr. Raboud is supported by the Skate the Dream Fund, Toronto General and Western Hospital Foundation.

Reprints: Janet Raboud, Prosserman Center for Health Research, 60 Murray St., Toronto, Ontario M5G 1X5, Canada (e-mail:

Viral load testing is an essential component of HIV-positive patients' care in developed countries. For individuals not receiving antiretroviral therapy (ART), viral load testing offers unique prognostic information that helps to inform decisions regarding treatment initiation.1 For individuals receiving ART, viral load testing is important for monitoring the efficacy of ART and for the early identification of patients experiencing treatment failure.2-4 Although randomized clinical trials have not been conducted that demonstrate decreased morbidity or mortality with viral load testing, guidelines strongly recommend that individuals receive viral load testing at intervals of 3-4 months.1,5

Despite the standards recommended in HIV clinical practice guidelines, the delivery of services has often been inequitable. For example, several observational studies have demonstrated differential access to highly active ART for injection drug users (IDUs),6 women,7 and those of lower socioeconomic status.8-11 To date, most studies of access to HIV care have focused on a single event, such as the first use of highly active ART. In contrast, studying a process that should be repeated regularly, such as viral load testing, may yield important insights regarding patterns of care over several years. Furthermore, because viral load testing is indicated for all individuals, clinical decision making considerations-which may exert a large unmeasured confounding effect on the decision to initiate ART but are not easily captured in large databases-are of lesser concern.

We studied patterns of viral load testing in Ontario, Canada. Ontario is an ideal location for such a study for 3 reasons. First, the province is home to more than half of Canada's HIV-positive population and supports an observational study that has collected longitudinal clinical data on a cohort of HIV-positive patients. Second, viral load testing is available without charge to all HIV-infected residents as part of the public universal health insurance plan. Accordingly, differential use of testing cannot be simply explained by financial factors at the point of care. Third, apart from a few tests conducted in clinical trials, all viral load tests are coordinated through the central governmental laboratory and results are recorded in one administrative database, resulting in a very high rate of data capture. We studied the determinants of regular viral load testing in Ontario and identified factors associated with a suboptimal frequency of testing.

Back to Top | Article Outline


We used 2 data sources for our study. The first, the HIV Ontario Observational Database (HOOD), is a voluntary longitudinal clinical cohort study of HIV-infected individuals that records a broad range of clinical information including clinical events, medication use, and laboratory results. The second, the Central Public Health Laboratory (PHL) of the Ontario Ministry of Health and Long-term Care, is the administrative database for viral load testing.

Back to Top | Article Outline

The HIV Ontario Observational Database

The HOOD project recruited participants from HIV specialty and primary care clinics throughout the province. More than 3400 individuals have enrolled in HOOD since 1994 and have ongoing data collection. All participants consented to have trained data extractors review their medical charts and extract detailed clinical, laboratory, and medication information. Additionally, participants completed a self-administered questionnaire at enrollment pertaining to demographic data and HIV testing and exposure history. The HOOD study is currently administered by the Ontario HIV Treatment Network, an independently incorporated, not-for-profit organization funded by the AIDS Bureau, Ontario Ministry of Health and Long-Term Care. HOOD received research ethics approval from the University of Toronto as well as from the ethics boards of several hospitals where patients were enrolled.

The HOOD population is generally representative of the Ontario population of HIV-infected individuals with respect to age, gender, and geographic location. However, the proportions of immigrants, IDUs, and recently infected individuals are significantly lower in HOOD than in the larger HIV-positive population in Ontario.12

To link the 2 databases, we focused on HOOD enrollees with complete medication histories who had voluntarily provided us with their health card numbers at enrollment for data linkage. Medication histories were sometimes incomplete because HOOD changed its medication data collection processes in September 2000 to collect more thorough information. The reabstraction process continues; in this report, we include all participants with complete data regarding use of antiretroviral medication as of March 2002.

Back to Top | Article Outline

The Central Public Health Laboratory

Viral load testing is performed at 5 regional laboratories in Ontario and results are entered into a centralized computerized database over a secure server. Almost all viral load tests were performed with the Chiron 2.0 assay (before mid-1998) or Chiron 3.0 assay (VERSANT(R) HIV-1 RNA 3.0 Assay [bDNA] Bayer, Tarrytown, NY, formerly Chiron Corp.) (after mid-1998). For quality assurance purposes, some individuals had >1 viral load measurement on the same plasma sample. In such cases, we averaged the available values, without adjustment for differences across assays, since <0.1% of samples had multiple measurements from different kits. The Ontario Ministry of Health and Long-Term Care approved the project. Data were deterministically linked using Ontario Health Card numbers.

Back to Top | Article Outline

Statistical Methods

Because viral load data were combined from the HOOD and PHL databases, we eliminated values that were likely to be duplicates. We designated viral load tests as duplicates if they were recorded as having been taken within 8 days of each other and the results were within 0.1 log10 copies/mL. In such cases, we kept the first viral load date and assigned it a value that was the average of the 2 absolute viral load measurements. We also manually checked all values that were in the HOOD database but not the PHL database and could not be easily identified as duplicate measurements. For example, HOOD includes viral load measurements conducted at non-PHL-affiliated laboratories for participants in clinical trials.

We evaluated 3 measures of viral load testing frequency. First, we calculated the annual rate of viral load testing by dividing the number of tests for an individual by their duration of follow-up. Second, we examined the time between successive test intervals. Third, we determined the proportion of individuals who had a suboptimal interval between tests, which we defined as ≥6 months. This interval corresponds to at least 2 missed or delayed tests if the optimal frequency between tests is 3 months. As a form of sensitivity analysis, we repeated the analysis defining a suboptimal interval as ≥9 months, corresponding to at least 3 missed or delayed tests if the optimal frequency between tests is 3 months. We also analyzed the number of suboptimal intervals experienced by participants.

We compared the annual rate of viral load testing and the number of suboptimal intervals experienced by gender, risk factor, education, age, place of residence, CD4 count, and characteristics of the antiretroviral regimen with the Wilcoxon rank sum test. The relationship between the viral load intertest interval and individual characteristics was examined using a generalized estimating equation (GEE) regression model,13 which is similar to a regular regression model but accounts for correlation among repeated observations within individuals. An exchangeable correlation structure was used for this model, which assumed that any pair of measurements from a single individual had the same correlation. We treated plasma viral load levels, CD4 counts, and characteristics of the antiretroviral regimen as time-varying covariates in all regression models. We used GEE logistic regression models to determine factors associated with the occurrence of a suboptimal testing interval.

Back to Top | Article Outline


Of the 3460 HOOD participants, we were able to deterministically link 1945 with the PHL viral load database. From January 1996 to January 2002, these individuals had 31,183 viral load measurements recorded in the PHL database. We excluded data from 913 participants because of missing or incomplete medication data for those individuals and we averaged multiple measurements performed on some plasma samples. After eliminating duplicate values in HOOD and PHL, we found 124 viral load measurements recorded only in HOOD, for a total of 18,312 viral load measurements from 1032 subjects.

The typical HOOD participant was a well-educated white man (Table 1). More than half of all participants lived in Toronto, Ontario's largest city, and about two-thirds of participants were men who acquired HIV through having sex with another man. Compared with all HOOD participants, the study population was similar with regard to gender, education, and Toronto residence but contained proportionately more men who have sex with men (MSM).

Among the study sample, the median number of viral load tests per individual was 18 (interquartile range [IQR] 13, 22) and the median length of follow-up was 4.6 years (IQR 4.1, 4.8). The median rate of testing was 4.2 viral load measurements per year (IQR 3.5, 4.7). Eighty-five percent of the population had >3 tests per year and 58% had >4 tests per year. In univariate analyses, we found higher rates of viral load testing among individuals with more than high school education, MSM, and individuals with CD4 counts <300 cells/mm3 (Table 2). We observed the greatest difference in rates between MSM and IDUs, with the latter getting 0.6 fewer tests per year than the former.

We calculated the intertest interval for each pair of viral load measurements per individual. Overall, the median intertest interval was 84 days (IQR 56, 111; 5th, 95th percentiles 24, 185). Of 17,344 sequential pairs of viral load measurements, 9712 (56%) were within 90 days. The intertest interval increased with calendar year; we observed a median of 81 days between viral load measurements in 1997 and 90 days between measurements in 2001. The median intertest interval was 86 days for patients not taking any antiretroviral medications and 91, 90, 86, 76, and 65 days for individuals on 1, 2, 3, 4, and ≥5 antiretroviral medications, respectively.

We analyzed independent predictors of increased intertest intervals using a GEE linear regression model (Table 3). Longer intervals between viral load measurements were associated with injection drug use, younger age, urban residence, lower plasma viral load, later calendar year, and less than high school education. The intertest interval decreased with the number of antiretroviral medications in an individual's regimen; the intervals of individuals not taking any antiretroviral medications were approximately 3 weeks longer than those of individuals receiving ≥4 antiretrovirals.

We next analyzed how likely individuals were to experience at least one gap of ≥6 months between tests using a multivariate GEE logistic regression model (Table 4). Of the 17,344 pairs of viral load measurements, 899 (5%) were >6 months apart and 270 (1.5%) were >9 months apart. The likelihood of an intertest interval of >6 months was higher in recent calendar years, among IDUs, and among individuals receiving no antiretroviral medication. In contrast to the finding that individuals with higher viral loads had shorter intervals between viral load measurements on average, this model demonstrated that individuals with higher viral loads were more likely to experience a 6-month gap between measurements than individuals with lower viral loads. Results were similar when we defined a clinically important gap in viral load measurement as ≥9 months.

Back to Top | Article Outline


In our observational database in Ontario, Canada, 15% of patients had <3 viral load tests per year, indicating that decreased use of this service was significantly prevalent despite free and universal health insurance. In multivariate analyses, we identified 5 demographic factors (lower education, a history of injection drug use, age, residence in Toronto, and calendar year) and 2 clinical factors (the last viral load level and the number of drugs in the last regimen) that were associated with decreased rates of viral load testing. Although these comparisons were statistically significant, the absolute differences were small; the biggest delays were observed for IDUs and individuals not receiving medications, each of which was associated with an increase of about 10 days between tests. The regression analysis indicates that some individuals with >1 factor could have more prolonged delays (eg, IDUs residing in Toronto and not on medications have average delays of almost 4 weeks).

To more adequately address whether some factors were associated with clinically meaningful delays in testing, we assessed which factors were associated with prolonged gaps between tests. Multivariate analyses identified only 2 demographic factors-a history of injection drug use and calendar year of testing-and 2 clinical factors-the most recent viral load and fewer drugs in a regimen-to be significantly associated with gaps in testing.

The finding that IDUs receive fewer services despite adequate health insurance has been observed in several settings.6,14,15 Similar to our results, a study of IDUs receiving ART in New York State who were enrolled in Medicaid found that nearly half the cohort did not have adequate viral load testing, defined as ≥2 viral load tests in 1 year.14 We hypothesize several nonexclusive explanations for such findings. Some patient groups may use services less. For example, in Vancouver, Canada, IDUs have been shown to be less likely to use free ART if they are not enrolled in an addiction treatment program, are female, are younger, or are cared for by a physician with little experience treating HIV-infected individuals.16 A second explanation may stem from stigma or discrimination from health care providers. For example, some physicians have been reluctant to prescribe antiretroviral agents to IDUs due to concerns about poor adherence and unstable lifestyles.15,17 Such reluctance may not be well founded, however; homelessness and a history of imprisonment have been found to be unrelated to either acceptance of or adherence to ART.16,18 A third explanation relates to how health care is delivered. Structuring care for IDUs, eg, by combining daily observed ART and methadone maintenance programs, has enabled more drug users to effectively receive ART.18,19 Finally, poor understanding of their own disease may play a role in how IDUs access interventions. In one setting, fewer than half of the cohort of IDUs had a basic understanding of CD4 cell counts and viral load.18 HIV-infected IDUs have been shown to be more likely to be admitted to hospital than other HIV-infected individuals with similar health status,20 indicating that while IDUs may be using fewer resources at earlier stages of disease, they require more intense resources later in disease.

In studies to date, the strongest indicator of acceptance and adherence to ART among IDUs is participation in an addiction treatment program.16,18 While methadone treatment has been available in Canada and covered by provincial health insurance plans since the 1960s,21 <10% of IDUs accessed methadone treatment until the mid-1990s. In the past decade, the availability of methadone treatment has increased approximately 2-fold, yet there is still much unmet demand for addiction treatment.22 Our study is supportive of arguments to increase access to such programs. Other efforts that may improve access for IDUs include educating health providers about the possibilities of successful treatment of HIV-infected IDUs and reducing their reluctance to prescribe ART to this population and facilitating care and treatment of HIV-infected IDUs through daily observed therapy, flexible clinic hours, and more convenient locations of clinics.

Some of our findings are reassuring. Most prominently, the intercept from our multivariate analysis was exactly 90 days; this can be interpreted as an adjusted average intertest interval for an individual with reference values for the model parameters, ie, a homosexual man aged 39 years with less than a high school education living outside of Toronto with a viral load level of 2.7 log10 copies/mL who was using 3 antiretroviral drugs seeking a new viral load measurement in 1999. This interval coincides with the interval recommended in practice guidelines. Furthermore, although we found that lower socioeconomic status (assessed here by education level) and geographic location, factors identified in previous studies with decreased access to HIV services, were associated with lower testing rates, a more stringent classification of decreased testing-a gap of at least 6 months-indicated that these factors were not associated with prolonged delays. Furthermore, other factors previously associated with diminished access, such as gender and race, were not significant in any of our analyses.

Our data demonstrated that while higher viral load levels were associated with fewer days between viral load measurements on average, individuals with high viral loads were also more likely to have gaps between tests of >6 months. This observation suggests that a group of HIV-infected individuals in Ontario accesses care intermittently and suboptimally. This interpretation is supported by the finding that other characteristics that may be determinants or markers of inconsistent health care delivery, including injection drug use and nonreceipt of antiretroviral medications, were also independently associated with an increased frequency of prolonged testing gaps.

Our finding that late calendar year and fewer antiretroviral medications were each independently associated with an increased frequency of prolonged testing gaps should be interpreted cautiously. These findings may indicate that providers and patients have become more familiar with how to use and interpret viral load testing over time and are comfortable with prolonging intervals for individuals who are not using antiretroviral drugs. If these individuals are at low risk for disease progression, a decreased testing rate may be appropriate and cost-effective. Alternatively, these findings may represent complacency and inappropriate risk taking. Studies that assess the association between rates of viral load testing and clinical outcomes will help to determine the amount of concern these findings should elicit.

This study has several limitations. First, as discussed above, we examined a process variable-the use of viral load testing-rather than clinical outcomes. Nevertheless, examination of such processes may be more instructive than studies assessing outcomes when assessing access to care issues, given the difficulties of case mix adjustment. Second, our study may have a selection bias. We studied only individuals receiving regular care in clinics who had consented to participate in our study. As such, our study may be biased against identifying factors associated with prolonged use in the entire population of people living with HIV in Ontario and may have underestimated the effects observed. Third, other unmeasured confounders may have accounted for the observed differences. Most significantly, our study is unable to ascertain the contribution of decreased patient adherence to laboratory testing or the quality of care provided in clinics. Fourth, our study classified individuals as IDUs only on the basis of their self-reported risk for acquiring HIV and is therefore prone to misclassification bias (if individuals were not entirely truthful or if individuals started using injection drugs after enrollment in HOOD but before their first viral load test). Furthermore, we were unable to discriminate between individuals with ongoing, intermittent, and past drug use. Fifth, we were unable to identify which viral load measurements were taken as part of regular care and which were taken as a result of participation in a clinical trial. While we expect the magnitude of the bias to be small, it is possible that associations between variables such as injection drug use and age with the rate of viral load testing may be due in part to differential access to clinical trials.

Overall, 3 important conclusions relevant to access to care for HIV services merit emphasis. First, our study highlights the importance of continuing to monitor access to care for current HIV services, particularly in light of our finding that use patterns changed over time. Second, our good news conclusions, that average viral load testing rates correspond closely with guideline recommendations, should be recognized as a successful example of efficient HIV care provision on a large scale. Third, this study stresses the importance of factors other than insurance or out-of-pocket payments for health services that can be barriers to accessing adequate care. In particular, the care of individuals with a history of injection drug use seems to pose special challenges across health care systems.

Back to Top | Article Outline


The authors thank the people living with HIV who have volunteered their data for the HOOD project.

Back to Top | Article Outline


1. Panel on Clinical Practices for Treatment of HIV Infection. Guidelines for the use of antiretroviral agents in HIV-infected adults and adolescents. July 14, 2003. Available at: Accessed September 22, 2004.
2. Hammer SM. Advances in antiretroviral therapy and viral load monitoring. AIDS. 1996;3:S1-S11.
3. Holodniy M. Viral load monitoring in HIV infection. Curr Infect Dis Rep. 1999;1:497-503.
4. Mylonakis E, Paliou M, Rich JD. Plasma viral load testing in the management of HIV infection. Am Fam Physician. 2001;63:483-490, 495-496.
5. Carpenter CC, Cooper DA, Fischl MA, et al. Antiretroviral therapy in adults: updated recommendations of the International AIDS Society-USA Panel. JAMA. 2000;283:2936-2937.
6. Strathdee SA, Palepu A, Cornelisse PG, et al. Barriers to use of free antiretroviral therapy in injection drug users. JAMA. 1998;280:547-549.
7. Anderson KH, Mitchell JM. Differential access in the receipt of antiretroviral drugs for the treatment of AIDS and implications for survival. Arch Intern Med. 2000;160:3114-3120.
8. McKinney MM, Marconi KM. Delivering HIV services to vulnerable populations: a review of CARE Act-funded research. Public Health Rep. 2002;117:99-113.
9. Andersen R, Bozzette S, Shapiro M, et al. Access of vulnerable groups to antiretroviral therapy among persons in care for HIV disease in the United States. HCSUS Consortium. HIV Cost and Services Utilization Study. Health Serv Res. 2000;35:389-412.
10. Shapiro MF, Morton SC, McCaffrey DF, et al. Variations in the care of HIV-infected adults in the United States: results from the HIV Cost and Services Utilization Study. JAMA. 1999;281:2305-2315.
11. Smith SR, Kirking DM. Access and use of medications in HIV disease. Health Serv Res. 1999;34:123-144.
12. Forte T, Bayoumi AM, Remis RS, et al. Representativeness of the HIV Ontario Observational Database (HOOD). Paper presented at: Twelfth Annual Canadian Conference on HIV/AIDS Research; April 10-13, 2003; Halifax, Nova Scotia. Abstract 337.
13. Liang K-Y, Zeger SL. Longitudinal data analysis for discrete and continuous outcomes. Biometrics. 1986;42:121-130.
14. Laine C, Zhang D, Hauck WW, et al. HIV-1 RNA viral load monitoring in HIV-infected drug users on antiretroviral therapy. J Acquir Immune Defic Syndr. 2002;29:270-274.
15. Carrieri MP, Moatti JP, Vlahov D, et al. Access to antiretroviral treatment among French HIV infected injection drug users: the influence of continued drug use. J Epidemiol Community Health. 1999;53:4-8.
16. Strathdee SA, Palepu A, Cornelisse P, et al. Barriers to use of free antiretroviral therapy in injection drug users. JAMA. 1998;280:547-549.
17. Loughlin A, Metsch L, Gardner L, et al. Provider barriers to prescribing HAART to medically-eligible HIV infected drug users. AIDS Care. 2004;16:485-500.
18. Clarke S, Delamere S, McCullough L, et al. Assessing limiting factors to the acceptance of antiretroviral therapy in a large cohort of injecting drug users. HIV Med. 2003;4:33-37.
19. Clarke SM, Mulcahy FM. Antiretroviral therapy for drug users. Int J STD AIDS. 2000;11:627-631.
20. Weber AE, Yip B, O'Shaughnessy MV, et al. Determinants of hospital admission among HIV positive people in British Columbia. CMAJ. 2000;162:783-786.
21. Fischer B. Prescriptions, power and politics: the turbulent history of methadone maintenance in Canada. J Public Health Policy. 2000;21:187-210.
22. Fischer B, Rehm J, Blitz-Miller T. Injection drug use and preventive measures: a comparison of Canadian and Western European jurisdictions over time. CMAJ. 2000;162:1709-1713.

viral load measurement; HIV; access to health care

© 2005 Lippincott Williams & Wilkins, Inc.