Secondary Logo

Journal Logo

Epidemiology and Social Science

Loss to Follow-Up of Adults in Public HIV Care Systems in Central Mozambique: Identifying Obstacles to Treatment

Micek, Mark A MD, MPH*†; Gimbel-Sherr, Kenneth MPH*†; Baptista, Alberto João MD; Matediana, Eduardo MD; Montoya, Pablo MD, MPH*†; Pfeiffer, James PhD, MPH*†; Melo, Armando MD§; Gimbel-Sherr, Sarah RN, MPH*†; Johnson, Wendy MD, MPH*†; Gloyd, Stephen MD, MPH*†

Author Information
JAIDS Journal of Acquired Immune Deficiency Syndromes: November 2009 - Volume 52 - Issue 3 - p 397-405
doi: 10.1097/QAI.0b013e3181ab73e2



Over the last several years, access to antiretroviral (ARV) medications has expanded dramatically in resource-limited countries, where the majority of people with HIV live. From December 2003 to December 2007, the number of people on antiretroviral treatment (ART) in low- and middle-income countries increased from 400,000 to almost 3 million people, with a concurrent 20-fold increase in sub-Saharan Africa alone.1 Mozambique is one high HIV prevalence country that has greatly expanded ARV coverage to its citizens. HIV services were initially provided in “HIV Care Networks,” a model of care that linked HIV testing services, ART centers, and community-based groups. Because free ARVs became available in the public sector in June 2004, the number of people on ART increased from <2000 in 2003 to over 125,000 in December 2008.2

Despite increased access to ARVs, only about 30% of those estimated to need ARV therapy are on treatment.1,3 In areas where services exist, poor service utilization and high rates of loss to follow-up can dramatically limit the population coverage of HIV treatment programs. Although prior studies have identified problems in utilization and follow-up in components of HIV care, such as HIV testing or follow-up after starting ART,4-7 few have systematically described HIV care utilization and follow-up at the population level through all steps in HIV care.8

The aim of this article is to systematically examine the flow of people from HIV testing to treatment in 2 urban HIV Care Networks in central Mozambique, to identify the bottlenecks that contribute to low service utilization and loss to follow-up. This approach could lead to interventions to improve program effectiveness and may also be applied to similar resource-limited settings where rapid and equitable expansion of ARV treatment is a priority.


Setting and Data Collection

This study used routinely collected data from HIV testing and treatment facilities in 2 HIV Care Networks in the cities of Beira and Chimoio in central Mozambique. Beira, in Sofala Province, is the second largest city in Mozambique (population approximately 430,000) with one of the highest adult HIV seroprevalence rates in the country at 29% in 2007.9 Chimoio, in Manica Province, has a population of about 240,000 and an adult HIV seroprevalence rate of 25%.9

This study used data from a retrospective cohort of people testing for HIV between July 1, 2004, and June 30, 2005, the first complete 12-month period after free ARVs became widely available in the public sector. During the analysis period, the HIV Care Networks in both cities included multiple HIV testing centers, but only 1 primary public ART clinic that provided free HIV clinical services, including CD4 testing and ART. These 2 ART clinics were the only source of public sector ART in their respective provinces. The ART clinics enrolled an average of 353 (Beira) and 219 (Chimoio) new patients per month and started an average of 66 (Beira) and 47 (Chimoio) new patients on ART per month, which was well below the supply of ARV medications at these sites. Although Beira city did have one other small non-governmental organization-run treatment clinic, this clinic was not included as it was segregated from the larger public system and provided services predominately to pregnant women receiving prenatal care at that site.

This study utilized routinely collected data from HIV testing centers and ART clinics. These data included information entered into paper registers or computerized databases by facility or non-governmental organization personnel for program monitoring purposes. From HIV testing centers, 2 of the following data sources were used: (1) monthly reports of the number tested and testing HIV positive, compiled from anonymous site-based paper registers, and (2) computerized databases that captured information from testing registration forms, including a unique HIV test code, sociodemographic information, reasons for testing, sexual practices, and test results. At ART clinics, computerized databases were used that captured sociodemographic and clinical information from medical records, including information about age, sex, site and date of HIV testing, HIV test code, CD4 testing, and ART. National identification numbers were not recorded at HIV testing sites to maintain anonymity and were not captured by ART clinic databases.


To examine the flow of people through these HIV Care Networks, we identified 5 key steps (the “HIV treatment cascade”) that were required to identify and treat patients with HIV: (1) test for HIV, (2) enroll for care at the local ART clinic, (3) undergo CD4 testing, (4) start ART (if eligible), and (5) adhere to ART (Fig. 1). The first step (test for HIV) occurred in HIV testing centers, whereas the remaining steps occurred in ART clinics. These steps are similar to the 4 steps identified by the Centers for Disease Control and Prevention's Serostatus Approach to Fighting the HIV Epidemic,8,10 although our model included an additional step within utilization of care and treatment services (CD4 testing) to better measure ART-related aspects of care.

The HIV treatment cascade representing the steps required to identify and treat patients with HIV in central Mozambique. The first step (test for HIV) occurs in HIV testing centers, whereas the remaining steps occur in ART clinics.

We used routine data to separately calculate the number and proportion of people passing through each step, following the cohort tested for HIV between July 2004 and June 2005. We restricted analyses to adults 15 years and older because initial ARV expansion efforts focused on adults and pediatric ARV preparations were not available until mid-2005. The procedure for calculating flow through each of these 5 steps includes the following:

Step 1: Test for HIV

We determined the monthly number of adults tested for HIV and HIV positive, at 3 types of public HIV testing facilities between July 2004 and June 2005: voluntary counseling and testing (VCT), prevention of mother-to-child transmission (pMTCT), and youth VCT centers. VCT centers primarily served people wishing to know their status on their own initiative, although they also provided testing for outpatients referred by clinicians. pMTCT and youth VCT centers offered “targeted” testing for pregnant women and youth. Although pMTCT centers also provided testing to infants and male partners of pregnant women, we only included women testing at these sites because program reports focused on this population. All testing centers performed pre- and post-test counseling by trained counselors and used rapid HIV tests (Determine and Unigold for confirmation) with results given the same day. The majority (97%) of people tested at these sites were informed of their results. Although HIV testing also occurred at hospitals, recording at these sites was not standardized and therefore not included in these analyses. Routine HIV testing of adults did not occur at ART clinics.

As described above, 2 sources of data were available from HIV testing sites: monthly reports including numbers of people testing and testing positive, for HIV, and computerized databases that also included the age of testees. Data from monthly reports were generally considered more accurate for total numbers tested because this only required abstraction of data from in-clinic registers without individual-level data entry. To assess the comparability of these data sources, we compared cumulative and monthly figures during the 12-month study period for the 2 indicators (testing and HIV positive) obtained from the databases and monthly reports. In only 4 sites were cumulative figures similar (within ±5%) for both indicators, and in no sites were all 12 individual monthly figures similar (within ±5%), with discrepancies most commonly due to incomplete data entry into computerized databases. After review of the data and consultation with local program managers, 2 strategies were used to determine the number of adults tested at these sites. For 2 sites, we used only the information on testing and age from the computerized databases: 1 site where discrepancies were low (<1% cumulative with a maximum individual month discrepancy of 5.6%) and 1 site where monthly report figures for 2 months seemed to be duplicated with a subsequent year and where discrepancies for the remaining 10 months were low (<1% cumulative with a maximum individual month discrepancy of 7.5%). In the remaining 9 sites, overall and/or individual month discrepancies were high (>15%) most frequently due to incomplete data entry into computerized databases and figures from monthly reports were therefore used. For these sites, the number of adults 15 years and older was estimated by multiplying monthly report figures by the overall proportion of adults testing (and HIV positive) at that site, as calculated from demographic information in their respective computerized database.

Step 2: Enroll for Care at the Local ART Clinic

HIV-positive persons identified at testing sites included in this study should have been referred by testing counselors to the local ART clinic as they were the only public clinics offering free HIV treatment services in the respective provinces. The number and proportion of HIV-positive adults who enrolled for care within 30 days after their HIV test was used as a measure of successful referral to the ART clinic. The 30-day period allowed a time-defined calculation and included most (95%) of those eventually enrolling for care. The numerator was determined using information on age (and gender for those tested at pMTCT sites), date of ART clinic enrollment, and date/site of HIV testing from the ART clinic database, and the denominator was the monthly totals of HIV-positive adults calculated in step 1. Although the HIV testing and ART clinic databases were theoretically linked by an HIV test code, formal linking was not necessary for our calculations and capture of this code was often incomplete and nonstandardized precluding its use.

Step 3: Undergo CD4 Testing

During the analysis period, CD4 tests in the public sector were only available at ART clinics. The number and proportion of ART clinic enrollees who had CD4 tests drawn within 30 days of their enrollment and whose results were received and registered at the ART clinic was used as the measure of CD4 testing. The date recorded in the database and used for this analysis was the date CD4 tests were drawn. As above, the 30-day period allowed a time-defined calculation and also included most (91%) of those who eventually underwent CD4 testing. Although the number of CD4 tests “lost” between the laboratory and ART clinic and the lag time for CD4 results to become available at ART clinics also contribute to patient flow, these data were not collected routinely. In Beira, CD4 tests were drawn daily and run in an on-site laboratory and results took several days to weeks to arrive at the ART clinic because of batch testing and logistical delays. In Chimoio, CD4 tests were initially shipped to an off-site central laboratory and blood was only drawn several days per week and results often took several weeks to arrive at the ART clinic. In mid-October 2004, Chimoio acquired on-site CD4 testing that allowed daily blood draws and shortened the time to receive results.

Step 4: Start ART (if Eligible)

The clinical criteria for starting ART in Mozambique were similar to those recommended by the World Health Organization (WHO) for resource-limited settings11 and included all patients with CD4 counts <200 cells per cubic millimeter, patients with CD4 counts <350 cells per cubic millimeter who were also in WHO clinical stage 3 or pregnant, and all patients in clinical stage 4 regardless of CD4 count. The process for starting ART in Mozambique followed Ministry of Health guidelines and included (1) clinical visits to determine ART eligibility, start co-trimoxazole prophylaxis, and treat opportunistic infections and (2) a goal of 3 visits with an ART counselor to address psychosocial issues, help patients identify an ART treatment partner, and counsel patients on ART side effects and adherence; together, these visits were expected to take several weeks or months to complete. We determined initial ART eligibility for those patients who successfully passed through steps 1-3 above, using initial CD4 counts and WHO clinical staging information through 30 days after enrollment. For those eligible for ART, we calculated the proportion of patients who started ART within 90 days after their CD4 test. A 90-day period was used to allow a time-defined calculation and because CD4 reporting and pre-ART preparation were expected to take less than this period.

Step 5: Adhere to ARV Medications

Adherence to ARV medications is critical to achieve the benefits of ART and prevent the emergence of resistance.12-15 We estimated adherence to ARV medications through 180 days after ART initiation using pharmacy refill data in the ART clinic database. Adherence was defined as the number of medication days dispensed between the date of ART initiation and the first refill at least 180 days later, divided by the number of days during the same period. Although pharmacy refill data cannot determine whether patients took their medications,16 pharmacy-based estimates of adherence correlate with self-reported adherence17-19 and clinical outcomes.16,17,20 For adherence calculations, we excluded patients who left the clinic before 180 days for reasons unrelated to adherence (ie, died, transferred to another clinic, or suspended ART by a clinician). Patients otherwise lost to follow-up before 180 days were included in adherence calculations and assigned an adherence rate of “0.” We then dichotomized patients as having either good (≥90%) or bad (<90%) adherence.


Trends in HIV testing over time were determined using Pearson correlation between study month and the monthly number tested and HIV positive. Differences in the proportions of patients moving through subsequent steps based on type of HIV testing center, city, and over time were determined using χ2 tests for independence and trend. Multivariate logistic regression was then used to determine the independent relationships between the proportion passing through each step (dependent variable) and city and type of HIV testing center (independent variables). Although differences in flow based on type of HIV testing center may seem most relevant to steps 1 and 2, we analyzed flow through subsequent steps based on this variable to determine if the different populations coming from various HIV testing centers had different follow-up rates even after ART clinic enrollment.

Using the results of the calculations for each step, a model was then constructed to determine the flow of people through all steps of the HIV treatment cascade. This model was used to determine the number of people lost at each step (the difference between those eligible for and those successfully passing through each step). The model was also used to calculate the additional number of people who would complete all 5 steps if each step was individually improved, assuming that drop-offs for all other steps remained constant and that the rates of ART eligibility were similar among those who did and did not pass through each step. For step 1, we calculated the extra number completing all 5 steps if the HIV testing rate was independently doubled; for steps 2-5, we calculated the extra number completing all 5 steps if each drop-off was independently eliminated (reduced to 0%).

Statistical analyses were performed with SPSS for Windows statistical software (Standard Version 13.0; SPSS, Inc, Chicago, IL), and the flow model was produced using Microsoft Office Excel (Edition 2003; Microsoft Corporation, Redmond, WA).


At the beginning of the study period, the HIV Care Network in Beira included 6 HIV testing centers (3 VCT, 2 pMTCT, and 1 youth VCT) and 1 ART clinic; in Chimoio, there were 5 HIV testing centers (1 VCT, 3 pMTCT, and 1 youth VCT) and 1 ART clinic. By the end of the study period, 1 additional HIV testing center was added to these HIV Care Networks-a pMTCT center in Beira that began referring patients to the Beira ART clinic in June 2005. Because only this site entered the Care Network in the final month under consideration, it was not included in the analysis.

Step 1: Learn of HIV Status

An estimated 23,430 adults were tested for HIV during the study period, and 7005 (29.9%) were HIV positive. Seropositivity rates were higher at VCT centers (44.7%) compared with pMTCT (15.5%, χ2P < 0.001) and youth VCT (13.7%, χ2P < 0.001) centers. Although there was no significant increase in the total number of patients tested for HIV per month over the study period (r = 0.37, P = 0.23), there was a significant increase in the number of HIV-positive people identified per month (r = 0.75, P = 0.005, Table 1). The number tested and HIV positive increased significantly in VCT centers in Beira, although testing reduced in pMTCT centers overall and particularly in Chimoio.

Step 1: Average Monthly Number of People Tested for HIV in VCT, PMTCT, and Youth VCT Sites in Beira and Chimoio

Step 2: Enroll for Care at the Local ART Clinic

Of the estimated 7005 HIV-positive adults passing through step 1, 3956 (56.5%) enrolled at the local ART clinic within 30 days of their HIV test. In bivariate analyses, successful referral rates were higher for people tested at VCT centers (67.0%) compared with pMTCT (26.4%, χ2P < 0.001) and youth VCT centers (23.5%, χ2P < 0.001) and higher in Chimoio compared with Beira (62.4% vs. 53.3%, χ2P < 0.001). In multivariate analyses, there was a significant interaction between city and type of HIV testing center, prompting stratified analyses. The relationship between referral and type of HIV testing center was similar when stratified by city (Fig. 2). When stratified by type of HIV testing center, referral rates were higher in Chimoio for VCT centers (78.0% in Chimoio vs. 61.3% in Beira, χ2P < 0.001), higher in Beira for pMTCT centers (29.3% in Beira vs. 21.7% in Chimoio, χ2P = 0.001), and similar in both cities from youth VCT centers (23.6% in Beira vs. 23.4% in Chimoio, χ2P = 0.97). In only 2 sites (1 VCT site and 1 pMTCT site from Chimoio), referral rates significantly increased over time (χ2 test for trend P < 0.05).

Step 2: Successful referral rates from VCT, pMTCT, and youth VCT centers in Beira and Chimoio, Mozambique. The proportion of HIV-positive people who enrolled at an ART clinic within 30 days of HIV testing differed among those tested at the various types of HIV testing centers in Beira (left section of figure) and Chimoio (right section of figure).

Step 3: Undergo CD4 Testing

Of the 3956 adults passing through step 2, 6 enrollees were already taking ART at the time of their enrollment and were excluded from further analyses. Of the remaining 3950, 3046 (77.1%) had a CD4 test done within 30 days after ART clinic enrollment. In bivariate analyses, there were no significant differences in CD4 testing between those tested for HIV at VCT centers (77.4%), pMTCT centers (75.2%), or youth VCT centers (73.3%). The rate of CD4 testing was higher in Beira than in Chimoio (81.5% vs. 70.2%, χ2P < 0.001), although in Chimoio, the rate of CD4 testing significantly increased over time (χ2 test for trend P < 0.001) likely due to the availability of on-site CD4 testing in mid-October 2004. Multivariate analyses were similar to bivariate analyses, demonstrating a higher rate of CD4 testing in Beira compared with Chimoio [odds ratio (OR) 1.9, 95% confidence interval (CI): 1.6 to 2.2; P < 0.001] with no significant differences by type of HIV testing center.

Step 4: Start ART (if Eligible)

Of the 3046 HIV-positive adults passing through step 3, 1506 (49.4%) were eligible for ART based on information in the database. However, eligibility could not be definitively determined for 735 patients (24%) because of missing WHO staging information; although these patients were classified as ineligible, this classification may have changed with additional staging information. The proportion of those eligible for ART was significantly higher among those tested in VCT centers (51.7%) compared with pMTCT (31.4%, χ2P < 0.001) and youth VCT (34.1%, χ2P = 0.02) centers.

Of the 1506 adults eligible for ART, 735 (48.8%) started ART before July 1, 2006 (the censorship date of our database), at a median time of 71 days after the initial CD4 test and 471 (31.3%) started ART within 90 days of CD4 testing. In bivariate analyses, the proportion starting ART within 90 days of CD4 testing was higher in Chimoio compared with Beira (36.3% vs. 28.0%, P = 0.001) and was lower among those tested at pMTCT centers (19.8%) compared with VCT (32.0%, χ2P = 0.01) and youth VCT centers (40.0%, Fisher exact P = 0.10). The results of multivariate analyses were similar, with a higher rate of starting ART in Chimoio (OR 1.4, 95% CI: 1.2 to 1.8; P = 0.001) and a lower rate among those tested at pMTCT compared with VCT centers (OR 0.5, 95% CI: 0.3 to 0.9; P = 0.02). There was also a significant trend for a higher proportion of patients starting ART over time in each city (χ2 test for trend: Beira, P = 0.02; Chimoio, P = 0.001).

Step 5: Adhere to ARV Medications

Of the 471 patients passing through step 4 above, 89 patients (18.9%) left the clinic before 180 days for reasons unrelated to adherence and were therefore excluded from adherence calculations. Reasons for exclusion included death (n = 75), transfer to another facility (n = 13), and suspension of medications by a clinician (n = 1). Chimoio had a higher proportion of deaths (19.6%) compared with Beira (12.8%, P = 0.045).

For the remaining 382 patients not excluded from the adherence analysis, 317 (83.0%) had adherence rates ≥90%. No significant differences in adherence rates were noted in bivariate or multivariate analyses between cities, type of HIV testing centers, or over time.

Overall Flow of People Through the HIV Treatment Cascade

Using data from the preceding analyses, Figure 3 provides a summary of the overall flow of patients through the HIV care systems in Beira and Chimoio for the 12-month study period. Based on the model, the greatest number of people were lost at steps 1 (348,614 people) and 2 (3049 people). The greatest increase in people completing all 5 steps would occur if the drop-off at step 4 was eliminated (697 people), although substantial increases would also be seen if step 1 was doubled (317 people) or if the drop-off in step 2 was eliminated (244 people). Similar relationships were observed when the model was stratified by city and type of HIV testing site (data not shown).

Summary of the flow of people through the HIV care system and major bottlenecks in care. The flow of people through the HIV care systems in Beira and Chimoio is depicted moving from left to right, with the percent drop-off at each step calculated using the denominator from the previous box. Green boxes represent the 5 steps in the HIV treatment cascade. Yellow boxes represent denominator adjustments based on the results of HIV testing and other relevant clinical information. Pink boxes indicate (1) the number lost at each step and (2) the number of additional people who would complete all 5 steps of the HIV treatment cascade if the drop-off in each step was individually eliminated (or, in the case of HIV testing, if testing rates were doubled) while holding the drop-offs in all other steps constant. As the figure depicts, the greatest numbers lost are at steps 1 (348,614 people) and 2 (3049 people), while the greatest additional number starting ART (697 people) could be achieved by individually eliminating the drop-off at step 4. *Total adult population of Beira and Chimoio cities estimated using 2007 census figures; the total population was multiplied by the estimated proportion of adults in Mozambique 15 years and older (55.1%). †ART eligible based on Mozambique guidelines at the time of this analysis (see text). §Not excluded before 180 days for reasons unrelated to adherence (ie, death, transfer, or ART suspension by a clinician). ΔART adherence calculated as the proportion with ≥90% adherence based on pharmacy refill data through 180 days after ART; patients lost to follow-up assigned an adherence of 0%. ^Number lost at each step calculated as the difference between those who completed the step and the denominator from the prior box. ¥Calculated as the number of additional people who would complete all 5 steps of the HIV treatment cascade if the drop-off in each step was individually eliminated (or, in the case of HIV testing, if testing rates were doubled) while holding the drop-offs in all other steps constant.


This analysis demonstrates the feasibility of applying a systematic approach to identify bottlenecks within HIV care systems in resource-limited settings. Based on our analyses, it is apparent that each step in the HIV treatment cascade is associated with an attrition of people. First, we found that a relatively low proportion of people were tested for HIV at the HIV testing center included in this analysis (step 1 of our framework). An expansion of testing centers would likely improve coverage of testing; indeed, over 20 new testing sites have been opened since July 2005, and HIV testing is being integrated into tuberculosis and outpatient facilities. However, obstacles to testing in existing centers are likely also present, for reasons such as those identified in developed and resource-limited settings related to education and knowledge of HIV, proximity to facilities, confidentiality, and stigma.4,5,21,22 The declines we noted in testing women in pMTCT centers may also be related to other factors, such as staff availability or the use of opt-in (vs. opt-out) testing.

We also observed substantial drop-offs in the referral of HIV-positive people to ART clinics (step 2), particularly from pMTCT and youth VCT sites. Referral variations could be related to a number of factors, including the reasons for HIV testing, the quality of counseling, community perceptions of ART clinics and ARV medications, the distance from people's homes to ART clinics, and differences in the demographic characteristics of people tested at these sites. The consistently lower rates of referral we observed from pMTCT centers may also result from testing relatively healthy populations not actively seeking HIV testing or stigma associated with the social vulnerability of women. Other studies of service utilization among HIV-positive pregnant women in pMTCT programs found that frequently less than half receive simple ARV prophylaxis regimens to prevent mother-to-child transmission of HIV23-26 and that issues of stigma, lack of social support, negative interactions with program staff, lack of preparation for HIV testing, and inability to afford transportation costs may contribute to this low rate of service uptake.27,28 Different strategies may be needed to ensure that women tested at these sites understand their results and are offered sufficient support to access treatment.

Our results also indicate that even among those who were identified as HIV positive, arrived to the HIV clinic, and underwent CD4 testing within the defined periods for these steps, less than 50% of ART eligible patients started treatment and less than 33% started treatment within 90 days after CD4 testing. Similar to step 2, loss was most pronounced among women tested for HIV at pMTCT centers. Again, however, this drop-off is common. In one study from Côte d'Ivoire, only 34% of patients screened in an ARV program did not return for a follow-up visit and only 51% of those returning and eligible for ARVs actually started treatment.29 Although other studies from resource-limited settings report higher rates of starting ART,30,31 these studies do not specifically report denominators of all determined to be eligible within the relevant HIV care system and may therefore not be comparable to our figures. In studies in developed countries, uptake of ART among eligible patients is also not universal, but rather varies between clinics32 and provider types,33 and is particularly low among vulnerable groups such as women and those with competing subsistence needs.34-38 Follow-up with chronic care, which is strongly associated with starting ART in developed countries,35,39,40 is also influenced by the perceived need for treatment, satisfaction with care, and the quality of provider-patient relationships.41 Additional factors such as stigma, an unfamiliarity with chronic care, limited human resource capacity, a cumbersome pre-ART preparation process, and transportation difficulties may also contribute to the low rates of starting ART that we observed. These issues suggest potential interventions that could improve follow-up in care, such as improved counseling, rapid CD4 testing, reduced pre-ART preparation visits, decentralization of care (including CD4 testing and ART), and nutritional and transportation subsidies.

Our results also indicate that for those successfully passing through steps 1-4 of the HIV treatment cascade, mortality rates are still high within the first 180 days, likely due to delays in HIV testing and moving through the required steps before starting ART. However, among those who did not die, transfer to another facility, or suspend ART for clinical reasons before 180 days, the short-term ART adherence rates were relatively high, similar to results from other programs in sub-Saharan Africa.42 The ART clinics in this study utilized several adherence support strategies, such as requiring several pre-ART counseling visits, involving peer counselors and community-based groups in adherence support, encouraging patients to identify treatment partners before ART, providing modified directly observed therapy, and tracing patients who did not refill medications. However, the high adherence rates we observed may also be because the proportion of eligible patients who actually started ART was quite low, such that those able to follow-up with stringent pre-ART requirements may be those who can also overcome barriers to ART adherence. The high adherence rates may additionally reflect our inclusion of only patients who successfully navigated the previous steps within defined periods; future comparisons of outcomes to other groups of patients may help clarify this issue. Further operations research is also needed to determine if reducing pre-ART care requirements may increase ART initiation and reduce mortality, without reducing adherence rates. In addition, assessments of adherence beyond 6 months are required to ensure that capacity exists to support long-term adherence, particularly as loss to follow-up may be substantial over time.7

Although a number of factors may contribute to the drop-offs we observed in each of the steps in the HIV treatment cascade, many may be related to the centralized model of HIV care that was initially implemented in these sites. Progressively decentralizing HIV services into primary health care facilities, as is currently being done in Mozambique and other resource-limited settings, may improve capacity and patient flow by increasing the number of health workers involved in HIV care and locating services closer to people's homes. In other areas, additional operations research may be necessary to specifically identify the causes of drop-offs before developing interventions to improve flow. For example, research to understand the reasons for high attrition among women tested at pMTCT centers could point to improved staffing or counseling at these sites, community-based interventions to improve readiness for testing, or strategies to improve the understanding and perception of care at treatment facilities. In all cases, the effectiveness of interventions should be evaluated for their ability to improve flow through the entire system while avoiding negative consequences, such as health worker overload and burnout, poor adherence levels, and poor quality of care.

Equally important, however, is determining the priorities for achieving the greatest improvement in treatment access over the short term and long-term. While efforts should be made to improve follow-up at all steps, increasing the flow of patients into ART clinics without improving systems to handle increased workload may oversaturate the clinics and worsen their efficiency. Our data suggest that improving the efficiency of step 4-starting eligible patients on ART-would have a substantial and immediate impact to increase the number of people on ART. Further research is urgently needed to better identify the causes of poor follow-up among ART eligible patients and to develop and test strategies to increase ART initiation without reducing post-ART follow-up and adherence.

This study has several limitations. First, our use of routine data raises the possibility of poor data quality and also limits our analyses to events captured in our data systems. Our routine data systems, for example, do not evaluate whether people received care outside the public system we evaluated, whether they were tested for HIV multiple times, or why they may have dropped out of care. However, we found that data quality was acceptable and produced relatively consistent measures between and within sites over time. Prior evaluations of the reliability of the ART clinic databases also showed high concordance between the databases and patient records.43,44 Although our analyses may have misclassified people who sought care at other locations as lost to follow-up, the care systems we evaluated had only 1 public free ART clinic in the province to which HIV-positive people could be easily referred. When the ART databases were used to detect transfers between Beira and Chimoio for our study population, we identified only 12 HIV-positive adults who tested in one city but enrolled in the ART clinic of the other city ≤30 days after testing, and only 24 adults who transferred cities after ART clinic enrollment (inclusion of all data for these 24 patients would not have affected any time-defined analyses presented here). Although developing a research cohort may help reduce data errors and increase the number of measured variables, this method also has limitations inherent in sampling, the introduction of bias caused by study team follow-up procedures, the need for repetition to evaluate different sites and trends over time, and the diversion of resources away from improving monitoring systems. Our use of routine data allowed a simple and operational assessment of the key steps in HIV care that could be easily replicated in other care systems and followed over time. More detailed studies could then be undertaken as necessary to better understand and solve the bottlenecks identified in these analyses.

Second, we focused our framework to identify obstacles for one component of HIV care: that of providing ART. Although other aspects of HIV care are also important, we focused on the system fundamental to ensure that people have access to appropriate treatment. Although other frameworks for providing ART are possible, our steps represented distinct events relevant to the care system in Mozambique that could be discreetly measured using routine data collection systems.

Third, our use of time-defined measures to reflect flow means that those passing through the steps but outside the defined period are categorized as “lost.” However, as our intent was to evaluate the efficiency of the HIV care system, we used indicators that incorporated some measure of time. The time definitions we chose were not intended to represent an optimally efficient system; indeed, an optimal system may be one that facilitates ART clinic enrollment on the day of HIV testing, obtains CD4 testing on the day of (or before) ART clinic enrollment, and starts ART on eligible patients within days or weeks of eligibility. Rather, our time definitions reflect a conservative estimate at which point all patients should have passed. Future analyses focused on individual steps and including all data regardless of time are necessary to evaluate additional variations in flow and their effects on patient outcomes.

Despite these limitations, efforts to systematically analyze the flow of people through all components of HIV care is a feasible and important step to evaluate the effectiveness of care systems in resource-limited settings. Although previous efforts have focused on examining isolated steps in care, simultaneously analyzing all relevant steps can help identify additional programmatic bottlenecks that limit the population effectiveness of HIV treatment programs. It can also serve to help prioritize interventions and policies to overcome the bottlenecks within systems of care and follow the effects of these interventions over time. In Mozambique, many interventions have occurred since 2005 to address the bottlenecks identified in this analysis, including the expansion of HIV testing locations, changing from opt-in to opt-out testing at pMTCT centers, decentralizing ART care to primary health care clinics, and increasing community involvement in follow-up. Future analyses, similar to those presented here and based on simple and routine data systems, could help evaluate changes in patient flow over time in Mozambique and elsewhere, to ensure that expanded availability of ARVs reaches the people who need it.


1. WHO/UNAIDS/UNICEF. Towards Universal Access: Scaling up Priority HIV/AIDS Interventions in the Health Sector: Progress Report 2008. Geneva, Switzerland: World Health Organization; 2008.
2. National ART Data for December 2008. Mozambique Ministry of Health; 2009. Available at: Accessed April 1, 2009.
3. Mozambique National AIDS Council. Universal Declaration of Commitment on HIV and AIDS: Mozambique Progress Report for the United Nations GeneralAssembly Special Session on HIV and AIDS, 2006-2007. Maputo, Mozambique: Mozambique National AIDS Council; 2008.
4. Hutchinson PL, Mahlalela X. Utilization of voluntary counseling and testing services in the Eastern Cape, South Africa. AIDS Care. 2006;18:446-455.
5. Sherr L, Lopman B, Kakowa M, et al. Voluntary counselling and testing: uptake, impact on sexual behaviour, and HIV incidence in a rural Zimbabwean cohort. AIDS. 2007;21:851-860.
6. Temmerman M, Quaghebeur A, Mwanyumba F, et al. Mother-to-child HIV transmission in resource poor settings: how to improve coverage? AIDS. 2003;17:1239-1242.
7. Rosen S, Fox MP, Gill CJ. Patient retention in antiretroviral therapy programs in sub-Saharan Africa: a systematic review. PLoS Med. 2007;4:e298.
8. Giordano TP, Suarez-Almazor ME, Grimes RM. The population effectiveness of highly active antiretroviral therapy: are good drugs good enough? Curr HIV/AIDS Rep. 2005;2:177-183.
9. National Directorate of Medical Care, National STD/HIV/AIDS Control Program, Mozambique Ministry of Health: HIV Epidemiological SurveillanceRound, 2007. Maputo, Mozambique; 2008.
10. Janssen RS, Holtgrave DR, Valdiserri RO, et al. The Serostatus Approach to Fighting the HIV Epidemic: prevention strategies for infected individuals. Am J Public Health. 2001;91:1019-1024.
11. WHO. Scaling up Antiretroviral Therapy in Resource-Limited Settings: Treatment Guidelines for a Public Health Approach, 2003 Revision. Geneva, Switzerland: World Health Organization; 2004.
12. Mannheimer S, Friedland G, Matts J, et al. The consistency of adherence to antiretroviral therapy predicts biologic outcomes for human immunodeficiency virus-infected persons in clinical trials. Clin Infect Dis. 2002;34:1115-1121.
13. Haubrich RH, Little SJ, Currier JS, et al. The value of patient-reported adherence to antiretroviral therapy in predicting virologic and immunologic response. California Collaborative Treatment Group. AIDS. 1999;13:1099-1107.
14. Paterson DL, Swindells S, Mohr J, et al. Adherence to protease inhibitor therapy and outcomes in patients with HIV infection. Ann Intern Med. 2000;133:21-30.
15. Press N, Tyndall MW, Wood E, et al. Virologic and immunologic response, clinical progression, and highly active antiretroviral therapy adherence. J Acquir Immune Defic Syndr. 2002;31(Suppl 3):S112-S117.
16. Harrigan PR, Hogg RS, Dong WW, et al. Predictors of HIV drug-resistance mutations in a large antiretroviral-naive cohort initiating triple antiretroviral therapy. J Infect Dis. 2005;191:339-347.
17. Fairley CK, Permana A, Read TR. Long-term utility of measuring adherence by self-report compared with pharmacy record in a routine clinic setting. HIV Med. 2005;6:366-369.
18. Altice FL, Mostashari F, Friedland GH. Trust and the acceptance of and adherence to antiretroviral therapy. J Acquir Immune Defic Syndr. 2001;28:47-58.
19. Walsh JC, Horne R, Dalton M, et al. Reasons for non-adherence to antiretroviral therapy: patients' perspectives provide evidence of multiple causes. AIDS Care. 2001;13:709-720.
20. Low-Beer S, Yip B, O'Shaughnessy MV, et al. Adherence to triple therapy and viral load response. J Acquir Immune Defic Syndr. 2000;23:360-361.
21. Kalichman SC, Simbayi LC. HIV testing attitudes, AIDS stigma, and voluntary HIV counselling and testing in a black township in Cape Town, South Africa. Sex Transm Infect. 2003;79:442-447.
22. Fortenberry JD, McFarlane M, Bleakley A, et al. Relationships of stigma and shame to gonorrhea and HIV screening. Am J Public Health. 2002;92:378-381.
23. UNICEF. Evaluation of United Nations-Supported Pilot Projects for the Prevention of Mother-to-Child Transmission of HIV. New York, NY: United Nations Children's Fund (UNICEF); 2003.
24. Dabis F, Ekpini ER. HIV-1/AIDS and maternal and child health in Africa. Lancet. 2002;359:2097-2104.
25. Malonza IM, Richardson BA, Kreiss JK, et al. The effect of rapid HIV-1 testing on uptake of perinatal HIV-1 interventions: a randomized clinical trial. AIDS. 2003;17:113-118.
26. Perez F, Mukotekwa T, Miller A, et al. Implementing a rural programme of prevention of mother-to-child transmission of HIV in Zimbabwe: first 18 months of experience. Trop Med Int Health. 2004;9:774-783.
27. Painter TM, Diaby KL, Matia DM, et al. Women's reasons for not participating in follow up visits before starting short course antiretroviral prophylaxis for prevention of mother to child transmission of HIV: qualitative interview study. BMJ. 2004;329:543.
28. Bwirire LD, Fitzgerald M, Zachariah R, et al. Reasons for loss to follow-up among mothers registered in a prevention-of-mother-to-child transmission program in rural Malawi. Trans R Soc Trop Med Hyg. 2008;102:1195-1200.
29. Djomand G, Roels T, Ellerbrock T, et al. Virologic and immunologic outcomes and programmatic challenges of an antiretroviral treatment pilot project in Abidjan, Cote d'Ivoire. AIDS. 2003;17(Suppl 3):S5-S15.
30. Stringer JS, Zulu I, Levy J, et al. Rapid scale-up of antiretroviral therapy at primary care sites in Zambia: feasibility and early outcomes. JAMA. 2006;296:782-793.
31. Lawn SD, Myer L, Harling G, et al. Determinants of mortality and nondeath losses from an antiretroviral treatment service in South Africa: implications for program evaluation. Clin Infect Dis. 2006;43:770-776.
32. Sackoff JE, McFarland JW, Shin SS. Trends in prescriptions for highly active antiretroviral therapy in four New York City HIV clinics. J Acquir Immune Defic Syndr. 2000;23:178-183.
33. Landon BE, Wilson IB, McInnes K, et al. Physician specialization and the quality of care for human immunodeficiency virus infection. Arch Intern Med. 2005;165:1133-1139.
34. 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;35:389-416.
35. Gebo KA, Fleishman JA, Conviser R, et al. Racial and gender disparities in receipt of highly active antiretroviral therapy persist in a multistate sample of HIV patients in 2001. J Acquir Immune Defic Syndr. 2005;38:96-103.
36. Strathdee SA, Palepu A, Cornelisse PG, et al. Barriers to use of free antiretroviral therapy in injection drug users. JAMA. 1998;280:547-549.
37. Celentano DD, Galai N, Sethi AK, et al. Time to initiating highly active antiretroviral therapy among HIV-infected injection drug users. AIDS. 2001;15:1707-1715.
38. Cunningham WE, Andersen RM, Katz MH, et al. The impact of competing subsistence needs and barriers on access to medical care for persons with human immunodeficiency virus receiving care in the United States. Med Care. 1999;37:1270-1281.
39. Giordano TP, White AC Jr, Sajja P, et al. Factors associated with the use of highly active antiretroviral therapy in patients newly entering care in an urban clinic. J Acquir Immune Defic Syndr. 2003;32:399-405.
40. Celentano DD, Vlahov D, Cohn S, et al. Self-reported antiretroviral therapy in injection drug users. JAMA. 1998;280:544-546.
41. Griffin SJ. Lost to follow-up: the problem of defaulters from diabetes clinics. Diabet Med. 1998;15(Suppl 3):S14-S24.
42. Mills EJ, Nachega JB, Buchan I, et al. Adherence to antiretroviral therapy in sub-Saharan Africa and North America: a meta-analysis. JAMA. 2006;296:679-690.
43. Stubbs B, Micek MA, Pfeiffer J, et al. Understanding the effect of the treatment partner program on adherence to HAART in central Mozambique. 2008. In press.
44. Gimbel-Sherr K, Micek MA, Pfeiffer J, et al. Task shifting to mid-level clinical health providers: an evaluation of quality of ART provided by tecnicos de medicina and physicians in Mozambique. Paper presented at: XVII International AIDS Conference; August 3-8, 2008; Mexico City, Mexico.

HIV care utilization; HIV treatment; loss to follow-up; resource-limited settings

© 2009 Lippincott Williams & Wilkins, Inc.