The revolution in the care of patients with HIV/AIDS caused by the advent of antiretroviral therapy (ART) in the mid 1990s [1,2], has only recently started to reach patients in resource-constrained settings . However, only 12% of the estimated 6 million people living in the developing world who need ART are receiving it [3,4].
A number of recent developments over the past 2 to 3 years have greatly improved the likelihood that ART coverage will increase significantly. These include newly-available financial support from bi and multi-lateral funding organizations [5–7]; initiatives like the World Health Organization's 3 by 5 Initiative ; and a rapid fall in drug prices . The groundwork in terms of funding for increased access to treatment has been laid . However, little evidence has been published on which are the most efficient and effective methods in which health systems in low and middle-income countries can roll out ART.
The objective of this paper is to review the Mexican experience of scaling-up access to ART, focusing on prescription and adherence practices and their cost implications. We examine data from 1997 to 2001, the period immediately prior to the Mexican Government officially committing itself to providing universal access to antiretrovirals in 2001. There have been major efforts to improve access since 2001, thus the data presented here do not necessarily reflect the current standard of care for people living with HIV/AIDS in Mexico. We argue that even though the speed at which Mexico started scaling-up access to treatment was impressive , the standard of care, as observed in ART prescription and adherence practices during the first few years of scaling-up did not consistently ensure minimum acceptable quality. This is particularly evident from the fact that improving access to treatment did not correlated with a lowering of the AIDS mortality rate as experienced by Brazil, another country that has rapidly expanded treatment coverage .
Data from 1997 to 2001 shows that Mexico started to scale-up access to treatment at tremendous financial cost and with a low standard of care. Guidelines for treatment were initially established in 1992 and have been constantly updated since then, but they were not mandated by law until 2004. By the time a policy commitment to universal access was made in 2001 , an adequate standard of care could have been provided at no increase in cost, at least for the sites included in this study.
Any replication of Mexico's experience without simultaneously ensuring the capacity of the health system to consistently ensure quality of care in a program of such scale, would have significant and unacceptably negative results, both in clinical and in resource expenditure terms. By documenting this experience, we hope to ensure that countries now facing ART scale-up recognize that in the case of ART, more so than for the great majority of health problems, there is a critical difference between coverage and effective coverage. Programs that increase the former rather than the latter do so at their peril.
The Mexican health system
Mexicans receive care from one of three health sub-systems depending on whether they are employed in the formal sector of the economy and on their economic status .
- Social Security, which offers services to formal sectors employees, public and private, and their families (around 50% of Mexicans). There are several social security institutions depending on the type of employer, the largest of which is the IMSS.
- Ministry of Health services which are open to all, but are targeted at those without social security coverage. Sliding-scale user fees are indexed to the patient's socio-economic status. Medications, laboratory tests and other material must be paid out-of-pocket.
- Private sector care is widely available and highly heterogeneous: services range from small, unregulated and unsupervised physician offices to highly specialized tertiary care facilities in the largest cities.
In 2002, the Mexican Government started the Seguro Popular public insurance program (SPS) – a program that seeks to provide universal social insurance independent of the employment status of population. The SPS offers a basic health package at Ministry of Health clinics for a monthly premium which is indexed to family income with the poorest 20% exempt from paying a premium. Currently, more than a million families are covered by the SPS, but none were enrolled in the study cities during the period of observation.
The Mexican ART program, 1997–2001
Between 116 000 and 177 000 Mexicans are estimated to be infected with HIV . Access to treatment has varied historically across socio-economic groups. Between 1992 and 1997, only social security beneficiaries had free access to ART and to HIV care from specialists.
Until very recently, ART remained out of reach for people seeking care at the Ministry of Health facilities. Starting in 1998, FONSIDA (National Fund for Persons Living with HIV/AIDS) provided drugs for a limited number, with priority given to pregnant women and children.
By 2000, the Ministry of Health announced increased resources for HIV/AIDS care. Official budget figures suggest that the purchase of antiretroviral drugss increased 1.200% between 2001 and 2004, making treatment available to more than 12 000 persons living with HIV/AIDS (PLHA). By 2003, the Ministry of Health announced that the coverage of ART for PLHA in need of treatment was 100% .
The Mexican Government has never purchased generic antiretroviral drugs as there are no domestic generic producers. In the context of Mexico's relatively modest epidemic there has been no significant public debate regarding generic importation or compulsory licensure of antiretrovirals invoking the ‘public health emergency’ provision of the TRIPS agreement . The Ministry of Health is responsible for antiretroviral importation and has participated in multiple rounds of price negotiations with pharmaceutical firms, most notably with Merck, and has achieved significant price discounts.
Data collection and background study
This analysis is part of a large study documenting the Mexican experience in scaling-up ART from 1997 through 2001 . The study focused on the patterns of utilization of services and on the costs associated with the introduction of HAART. The data were collected from 902 patients’ records in 11 facilities (clinics, hospitals) across three different public-sector health systems. The 11 sites were selected by convenience from three of the urban centers with the largest prevalence of patients needing HIV/AIDS care in Mexico. The number of patients selected from each site was proportional to the total number of HIV/AIDS patients in the site.
Even though the sample was not designed to be representative at any level, there are two related points to highlight: first, we compared some characteristics of patients included in the study with those reported at the national level and found no significant differences ; and second, the sample was designed to include a broad range of institutions with respect to patient volume, level of care and location.
Utilization data from patient records were summarized by month. The utilization data were collected from the patient's records at each site. While we are certain that patients received some level of care at other facilities, it was beyond the scope of this study to access records from other facilities. Micro-costing analysis was conducted to estimate unit costs of the drugs dispensed [For further information on the methodology used by the study, please refer to the report. Available from URL: http://www.phrplus.org/Pubs/Tech020_fin.pdf].
Monthly patient data were aligned by two methods.
- By calendar year; that is, from 1997 to 2001.
- By month of initiation of triple antiretroviral therapy. Thus Year 1 (Yr+1) refers to the 12-month period beginning with that month and Year –1 (Yr–1) is the 12-month period prior to initiation. We refer to these as ART-years.
We use 2001 drug prices throughout this analysis regardless of the prices actually paid in each year, because our focus is on comparing the impact of prescription decisions over time, not on revealing secular trends in drug prices. Given that in 2001 major price negotiation took place between the Ministry of Health and some pharmaceutical companies, these were the lowest prices that obtained during the period of study. Unit costs for all ART medications were obtained from the National AIDS Program.
In order to understand the cost implications of prescription decisions, we present four scenarios assuming different patterns of prescriptions and levels of adherence.
The Observed Scenario was an accurate snapshot of as-observed medical records. This means only chart-confirmed months of medications supplied to patients were included.
Scenario 1 was an adaptation of the previous scenario in light of Mexican practice patterns. Although physicians at the studied facilities can only prescribe drugs to patients for a single month, the policy of many hospitals is to allow patients to refill their prescriptions at the hospital pharmacy without an in-patient visit over the subsequent 2 months. Scenario 1 assumes that all patients act upon their right to refills for a maximum of two additional months, or a total of 3 months including the month in which the medications were prescribed.
Scenario 2 assumed a 100% level of adherence by patients to their prescriptions until the prescription changes. Thus patients continue to take ART prescribed in their previous consultation without interruption every month until their next registered consultation, the end of the study period, or death. The scenarios thus assume progressively higher levels of adherence, moving from the least adherent scenario to the perfectly adherent one.
The following two scenarios were developed assuming not only perfect adherence, but also that ART prescription are in line with Official Therapy Guidelines (OTG) . We used the OTG published by the National AIDS Program in 2000 as the reference standard for evaluating recommended ART combinations and instructions on when to start/modify therapy. The two Recommended Scenarios were based on the selection of two officially recommended ART combinations with full adherence. In order to present the consequences, in terms of annual costs, of the prescription decisions made for these patients, we selected the lowest-price combination recommended for each of the two categories of patients with most advanced disease (Table 1).
Table 2 shows the 10 most frequently prescribed ART combinations during the period of study as well as the five most common and the distribution of mono, double, triple, etc. therapy by year. The second column of the table shows the percentage of patient-months to which the combination was prescribed. In the third column we present the monthly cost per patient.
The 10 combinations account for over 63% of the total monthly prescriptions registered. Four of these 10 combinations are among those recommended by the OTG for first-line treatment (Table 1). However these four combinations account for less than 29% of all patient-months for which a prescription was registered.
Furthermore, we found that whereas 27 of the 34 combinations recommended by the OTG were prescribed during the time period covered by the study, those 27 combinations were prescribed to only 37% (or 28% if we do not consider double therapies) of patient-months. The remaining 63% were prescribed one of the 282 other combinations found –302 different ART combinations prescribed to 902 patients in an average period of 2 years of follow-up per patient.
Table 2 also shows that mono therapies were the second and third most prescribed combinations in 1997, second in 1998, and surprisingly, the most frequently prescribed in 1999, 2000 and 2001, well after the literature and even the National AIDS Program recommended against its use. Additionally, the proportion of mono therapies remained constant in all years despite the fact that the five most prescribed combinations each year account for an increasingly small proportion of patient-months each year; 81% in 1997 but only 45% by 2001.
Even though the prescription of mono and double therapies decreased over the study period, they were still prescribed to a significant share of the total patient-months: 89% in 1997 and 60% in 2001. The number of patient-months on triple therapy increased from 11% in 1997 to only 32% in 2001. Table 2 also shows the monthly cost of the combinations. Seven of the 10 most-prescribed are outside the OTG's 10 most-recommended combinations and are also more expensive.
Figure 1 shows the number of ART combinations prescribed per patient. Between four and thirteen different ART combinations were prescribed to 60% of patients during the period of observation.
Table 3 shows the distribution of the different combinations prescribed to our sample of patient-months by ART-year. It also shows the proportion of patient-months (per ART-year) in which OTG-recommended combinations were prescribed. For Yr+1, Yr+2 and Yr+3 it also shows these proportions excluding double therapies.
By definition, there is no combination with three or more drugs during ART-years Yr–3 to Yr–1. However, despite initiation of triple therapy, patients continue to receive mono and double therapies during ART-years Yr+1 to Yr+3. This is only possible if the patients went back to mono and/or double therapy after initiating triple therapy. This occurred to 50% of all patients who ever initiated triple therapy. Although the OTG recommended 32 different combinations (including both double and triple therapies), we encountered 239 different combinations prescribed during Yr+1 alone.
Additionally, the OTG recommended only four antiretroviral combinations with more than three different drugs. These are the four most expensive combinations and are recommended only for patients in the most advanced stage of disease (Table 1). In ART-years Yr+1 to Yr+3, only 1.3% of the patient-months using the 112 combinations with four or more drugs were among those recommended by the OTG. In ART-years Yr+1, Yr+2 and Yr+3, 9.7, 9.8 and 12.1% of patient-months, respectively, received expensive combinations with more than four antiretrovirals.
Table 3 also shows that only approximately 35% of patient-months were prescribed OTG-recommended combinations after HAART initiation, and the percentage of patient-months drops to approximately 23% if OTG-recommended double therapies are excluded. All patients who received ART in the years preceding initiation of triple therapy are also included in the Yr+1 population (patients who never received triple therapy are not presented in this table). Thus, the difference in the number of patient-months between Yr–1 and Yr+1 (2195 = 3288 – 1093) are patients who were ART-naive when they initiated triple therapy.
The cost consequences of prescribing practices
Figure 2 presents the average annual costs per patient using three adherence scenarios (Observed, 1 and 2), differentiated by calendar year and by ART-year. The left panel shows a significant increase in the average annual costs for ART in each calendar year. It ranged from US$ 1390 to US$ 2520 in the Observed Scenario, and from US$ 2330 to US$4270 in Scenario 1. Given that we are using constant prices for individual drugs, these cost increases over time reflect transitions to more expensive combinations – and/or higher levels of adherence in the case of the Scenarios Observed and 1.
The right panel shows that the introduction of HAART (between Yr–1 and Yr+1) significantly increases the average cost per patient. The average cost for Yr+1 increases 161% in the Observed Scenario and 140% in Scenario 1, relative to Yr–1.
The scenarios differ in degree of adherence: in Yr+1, Scenario 1 reflects 12% non-adherence and Observed Scenario reflects 44%. These are calculated by the difference between Scenario 2 (perfect adherence) and the respective scenarios. Even the 12% figure, despite being highly optimistic, is far greater than the maximum 5% called for in recent WHO report on adherence . Moreover the three scenarios also represent different levels of expenditure. Scenarios Observed and 1 represent a plausible range of the average annual cost per patient the Mexican government actually incurred, whereas Scenario 2 represents the annual cost that the government would have incurred had patients fully adhered to the prescribed treatment.
The large difference between the annual costs per patient actually incurred and the annual costs that should have been incurred shows that there was a large difference between the number of patients on treatment and the number of patients adherent to treatment. This is a critical issue because of the non-linear relationship between ART adherence and health benefit. By 2001 the levels of non-adherence were 27% (Scenario 1) to 58% (Observed), higher than in Yr+1. Such high levels of non-adherence suggest that a significant proportion of patients were receiving no benefit from their treatment and may have been better off with no ART until the system is able to ensure better adherence. In other words, treating large numbers of patients at low levels of adherence is far less effective and less cost-effective than treating smaller numbers with high levels of adherence. Efforts to scale-up access to treatment must consider treatment effectiveness and the prevention of viral resistance as key programme objectives.
How much would it have cost to provide satisfactory treatment to all these patients in terms of both appropriate prescription and adherence in Mexico? Here we interpreted ‘satisfactory’ in two ways – prescription practices in line with the OTG (clinical adequateness and effective levels of adherence) and prescribed with consideration to costs (economic adequateness). Table 4 shows the average annual cost per patient for four scenarios: 1, 2, Recommended A and Recommended B, for 1999–2001. For the recommended scenarios we used: Recommended A, zidovudine, didanosine and efavirenz; Recommended B, zidovudine, didanosine and saquinavir, assuming 12 months of treatment.
In year 2001, the annual cost of Recommended A was lower than the average annual costs of both Scenarios 1 and 2. The annual cost of Recommended B was lower than the average annual cost of Scenario 2. In other words, the same number of patients could have received guideline-recommended triple therapy at 100% adherence for less money. This savings relative to Scenario 2 are US$ 3170 per patient-year for Recommended A and US$ 365 for Recommended B. This difference is even more striking if we look at the costs incurred by the most expensive 25% of patients from Scenarios 1 and 2.
Mexico achieved remarkably rapid scaling-up of ART. However, the initial scale-up was suboptimal because of prevailing prescription behavior and lack of interventions to improve adherence. For example, only three of the 10 most prescribed combinations were recommended by the OTG. Furthermore, whereas 27 of the combinations prescribed were recommended by the guidelines, only 37% of all patient-months were treated using such combinations; the remaining 63% received 282 different non-OTG-recommended combinations. Although prescription of double therapies reduced from 1997 to 2001, mono therapy continued to be prescribed consistently. Sixty percent of the sample's patients were prescribed four or more different therapies in the period of study and 50% of patients experienced treatment transition from triple therapy to mono or double therapies.
The combination of low levels of adherence combined with obsolete drug combinations was surely associated with high levels of acquired resistance, the extent of which should be documented in the same population in future studies. This acquired resistance will translate into more costly future treatment and increased mortality.
The average annual cost per patient in constant prices increased from 1997 to 2001 reflecting the transition from double to triple therapies, the increasing use of newer, higher-priced drugs, and some increased compliance. By 2001, the average annual cost of ART per patient within the Mexican health system was between US$ 2430 and US$ 4270. However, had those same patients been 100% adherent to OTG-recommended prescriptions, the system should have incurred an annual cost between US$ 2640 to US$ 5445 per patient, in comparison with a figure as high as US$ 5810 that would have been incurred per patient had they been fully adherent to their existing prescriptions.
A World Bank study published recently modeled the potential costs and benefits of three alternative HIV/AIDS treatment policies in India. The findings suggest that suboptimal policy implementation results in severely inadequate treatment of patients to the point that the benefits of care provision are completely nullified . The authors argue that the decision to provide universal access to ART must be accompanied by regulation to ensure optimal prescription of antiretroviral medication, support for adherence and the linking of treatment with prevention programs.
The lessons we draw from these Mexican data are similar to those modeled for India. Inadequate training of doctors will lead to inadequate prescription practices. Regular monitoring should be conducted of physicians prescribing practices and of pharmacy dispensing of antiretrovirals. These drugs should only be available through pharmacies that can be monitored to both ensure uninterrupted supply and to regulate the dispensing of non-recommended combinations.
Given the exquisite sensitivity of ART effectiveness to adherence, the development of adherence interventions and the training of health-care providers in adherence support are critical to program success. The extraordinary inefficiency associated with providing expensive treatment that is ineffective due to low levels of adherence justifies investment of an important proportion of total program resources to this end.
There are a number of other factors which can be related to suboptimal standard of care in Mexico. Interruption in drug supply is one crucial issue which could explain some of the results encountered in this study. Patients having to buy their medications and laboratory tests could be a significant factor in Mexico's experience before 2002. An important issue for consideration but that cannot be answered by this study is the trade-off between HIV/AIDS care provision in specialized centers versus the distribution of patients more widely across the health system. In addition, the influx of patients from private sector clinics could affect our results, especially if the quality of private care is low.
Although we are certain that the quality of care varied across our sample of facilities, due to sample size problems we are unable to compare quality of care at the facility level. Continued follow-up of the cohort may make this possible in the future. Another factor could relate to inadequacy of the treatment guidelines, Mexico has made major efforts in that sense and the new treatment guidelines are not only state of the art, but also legally mandated – an important step to improve physicians’ practices.
The Mexican experience between 1997 and 2001 suggests that well intentioned efforts to rapidly scale-up access to ART can provide less benefit to patients at greater cost in comparison to a slower expansion of coverage that ensures supply of drugs, appropriate prescription, and support for and monitoring of adherence to therapy. International funding for scaling-up ART in developing countries must provide adequate support to adherence interventions, training, monitoring and evaluation if the promised benefits of treatment are to be realized.
Sponsorship: USAID through ABT Associates funded the data collection. The Mexican National Institute of Public Health funded the analysis.