AIDS is considered the most terrible public health catastrophe that humanity has experienced in recent years . UNAIDS estimated that, worldwide, 39.5 million people were living with HIV, 4.3 million became newly infected and an estimated 2.9 million lost their lives to AIDS in 2006 . In Latin America, approximately 140 000 people were newly infected with HIV in 2006, which increased the total number of people living with HIV to 1.7 million. This makes Latin America and the Caribbean the second most affected region in the world .
The economic impact of AIDS is severe because the disease primarily affects people in their most productive years. The effects of AIDS on families, health systems and nations are immense, frequently causing the loss of the families' main source of income, for example. In many countries, the number of patients who require care has surpassed the capacity of the healthcare system. To make matters worse, the most affected countries are also the poorest; approximately 90% of infected people live in developing countries .
Mexico has a relatively low prevalence of HIV, estimated at approximately 0.3% . Nevertheless, the economic impact of this disease is significant. According to Magis-Rodríguez et al. , expenditures in prevention and healthcare increased by 143% from 1997 to 2002. The provision of antiretroviral therapy (ART) constituted the majority of this increase . These results are consistent with another study carried out in Mexico . This increment in AIDS expenditure has not been accompanied by a proportional increase in the total national healthcare budget , suggesting that other health services have been cut or constrained in order to benefit the AIDS programme.
ART represents the highest percentage of the overall cost in AIDS patient care [3,4,6]. ART should not be withheld, because of its effectiveness in improving the quality of life and extending life expectancy but also as a result of ethical implications and the existing societal pressures to provide treatment [7–9]. These ethical implications are frequently taken into account by policy makers when deciding whether or not to provide ART, instead considering the results of economic analyses as a basis for making decisions about treatment provision. Recently, the price of ART has been lowered significantly, a consequence of public pressure, advocacy, competition from generic manufacturers and initiatives from pharmaceutical companies. Nevertheless, there is no systematic approach to pricing. Each company determines its own eligibility criteria for countries, sectors and institutions that may benefit from its reduced price.
Further price reductions have been motivated by an agreement between the International Dispensary Association Foundation and the Clinton Foundation, enabling developing countries to purchase high-quality AIDS medicines and diagnostic tests at the lowest available price. Under this agreement, for example, the most common first-line formulation costs US$132 per person per year , one-third to one-half of the lowest price that is otherwise available in most settings.
Although a significant number of Mexicans live in extreme poverty, the International Dispensary Association/Clinton Foundation agreement does not include Mexico, which is classified as a medium-income country. Mexico's current average annual healthcare cost for people living with HIV is approximately US$6000 . The Mexican government has extended the HAART programme to everyone who needs it, and included AIDS among the diseases covered by its new financial protection system, ‘Seguro Popular’. This universal health insurance programme is intended to cover previously uninsured people, more than half of Mexican citizens , and is aimed at the population, mostly poor, that does not qualify for social health insurance [12,13].
Despite improvements in the clinical and epidemiological course of the disease reported by many studies [14–28], the real effect of ART on survival is unknown in Mexico, where this type of research is scarce. Because of the large financial implications involved in ART provision, it is necessary to generate reliable information for making informed decisions and good policies.
Two studies in the same hospital in Mexico City have reported the benefits related to the improvement of medical treatment, including HAART [29,30]. Using data from these studies, this paper analyses the effects of HAART on a cohort of HIV/AIDS patients in Mexico and estimates the cost of extending a patient's life.
We conducted a retrospective longitudinal study focusing on the survival of patients living with HIV. The main objective of survival analysis is to follow subjects over a certain length of time and to observe at what point in time they experience the event of interest. With this method it was possible to examine how covariates affect the length of time between consecutive events [31,32].
This study followed a dynamic cohort of 797 HIV patients. Unlike a closed cohort in which the exposure groups are defined at the beginning of follow-up, with no movement of individuals between exposure groups during the follow-up, a dynamic cohort is a type of cohort study in which entrance and exit of new subjects during the follow-up period is permitted. Therefore, the number of study participants can change over time [33–35]. Patients were included in the dynamic cohort at any time during the 25 years of follow-up. When the 25 years of follow-up is divided into three decades, 10% of the cohort entered between 1982 and 1989, 49% entered between 1990 and 1999 and 41% were included between 2000 and 2006.
The participating patients received medical care from a typical public hospital in Mexico, from the time they were diagnosed (as early as 1982) to the time this study concluded in December 2006. Inclusion in the cohort was determined by the HIV diagnosis, regardless of when it was made. Patient information was collected by the local HIV/AIDS programme. From this database we obtained the following information about each patient: date of initial diagnosis, current health status, date when ART was initiated, age, sex, marital status, education, geographical location (urban or rural) and occupation. The database does not include information about healthcare utilization, the treatment used when the patient suffered from opportunistic infections, or the type of antiretroviral drugs used. A review of the patients' medical records was therefore conducted to determine utilization, carried out by authorized hospital personnel. Identifying data about the hospital have been omitted to ensure confidentiality.
Variables of interest
The primary variable is the patients' survival time, measured in years: the time elapsed from the date of diagnosis (starting event) to the date of death (final event). The censored variable was defined as the patients who did not present the interest event (did not die) during the follow-up period. We defined as censored observations the follow-up that could not be continued (patients who discontinued care in the hospital and patients whose medical records were lost), patients who had died but whose date of death was not in the database and patients who were still alive at the end of the study. The absolute survival or crude survival variable corresponds to the number of years lived from diagnosis to death, whether or not the patient was receiving ART. We also differentiated survival time according to the type of treatment received: no treatment (those who never took ART), always HAART (those who received HAART from the beginning of their treatment); HAART as second line (those who started with another therapy, mono or dual therapy, and then changed to HAART); and others, (those who were treated with only monotherapy or only dual therapy or those who went from monotherapy to dual therapy or vice versa, but never took HAART). The drugs available in the hospital were: indinavir, saquinavir, ritonavir, atazanavir, nelfinavir, Kaletra (lopinavir/ritonavir), Combivir (lamivudine/zidovudine), abacavir, efavirenz, nevirapine, amprenavir, emtricitabine, tenofovir, lamivudine, zalcitabine, zidovudine, didanosine, stavudine and Trizivir (abacavir/zidovudine/lamivudine). The drugs were received every month during the monitored outpatient visit.
We used a non-parametric approach. The Kaplan–Meier method  was used to estimate the probability of survival to 1, 5, 10, 15 and more years from the dates of diagnoses, and the corresponding curves of estimated overall survival were also generated. The estimator was stratified by type of ART therapy: (i) no therapy; (ii) always HAART; (iii) initial dual therapy and subsequent HAART; and (iv) other treatment schemes (monotherapy or dual therapy). Survival functions were compared with the log-rank test for equality of survival functions. A Cox proportional hazard regression model was fitted to assess differences in HIV/AIDS mortality by ART therapy scheme, age and sex, adjusting for place of residence, occupation and certain elements considered by experts as risk factors related to HIV infection, such as sexual preference, blood transfusion, haemophilia and risky sexual activity. The assumption of proportional hazard was verified by inspecting log estimates against log time plots. All tests were two-tailed and P values less than 0.05 were considered statistically significant.
A life table was also constructed, which summarizes the collected information and contains data that provide a general view of mortality, overall survival, risk and other patient characteristics for each interval. Data analysis was undertaken using the statistical package Stata 9.2 .
The direct costs were measured from the perspective of the healthcare provider, which means that only the costs that were incurred by the hospital were included. The non-medical costs incurred by the hospital (e.g. per diem paid to patients from other cities) were not considered. Costs incurred by the patients (medical or non-medical) were not considered. Societal costs (i.e. productivity losses) were also not included in this analysis. All costs were considered and calculated in Mexican pesos (Mex$) at prices current in 2006. The final amounts were converted to US dollars. In 2006, the average Mexican peso to US dollar exchange rate was Mex$10.8 per US$1 .
The average total cost was calculated for the two possible healthcare services that patients can use: ambulatory and hospitalization. Ambulatory services included outpatient visits, ART drugs and other medications used to treat or prevent opportunistic infections or general discomfort. We also included in this category laboratory tests and any other studies needed to support the medical diagnoses. Hospitalization activities included inpatient days, drugs, laboratory tests and radiological or surgical procedures carried out during the hospitalization.
The cost of an outpatient visit and inpatient day cost were estimated using microcosting methodology [39,40]. Three categories of costs were identified: labor, recurrent and capital costs. The labor cost was calculated using salaries and all other monetary benefits paid to the medical and administrative personnel who participated in these activities. The recurrent costs included the costs for needles and syringes, gloves, cotton, soap, etc., and the cost of such as building maintenance, electricity, water, telephone, gas/oil and waste disposal. We determined the proportion of these aforementioned maintenance and utility services that are used for the HIV programme. Capital costs, which included the equipment and buildings, were annualized for the original capital expenditure over the useful life of the asset . Using this method we calculated the equivalent annual cost. The replacement value of each item and a discount rate of 5% were used. The cost of medicines was not included in the cost of an outpatient visit or in the cost of a hospital day.
To estimate the cost of ambulatory services, we identified (from the medical record) the number of outpatient visits and multiplied them by the aforementioned unit costs that were previously computed. From the medical record we also extracted the laboratory and clinical diagnosis tests and medications used. Each amount was multiplied by its corresponding price. We added all of these products to obtain the total ambulatory cost. The same procedure was followed to estimate the hospitalization costs. The prices of medications, laboratory and clinical diagnosis tests were obtained from the hospital records.
We calculated the accumulated cost for 1, 5, 10 and 15 years and each year thereafter for patients undergoing HAART. To calculate the monetary resources needed for each time period, the accumulated cost estimate was multiplied by the survival probability for each period.
Several parameters are uncertain, including medications, patients and health systems factors. Ideally, a sensitivity analysis should take into account indicators that are related to the clinical situation at the moment of starting HAART treatment, adherence, monitoring, and the type of cocktail used for the first treatment. Nevertheless, we only had access to hospital data about the type of treatment used in naive patients. For this reason our sensitivity analysis only included the type of first treatment used and the prices of the antiretroviral medications, which is another important variable that can influence healthcare costs. As a result of the fact that HAART is the recommended therapy , we limited the sensitivity analysis to the group of always HAART. For this analysis we only have three measures (0–1 year, 1–5 years and 5–10 years), which correspond to the amount of time that HAART has been used. We considered two scenarios, a best scenario in which patients are treated with the recommended medication offered at a lower cost, or the current scenario in which patients are not always given the appropriate medication at a higher cost.
The original sample included 184 women (23.1%) and 613 men (76.9%) for whom information on the time of HIV diagnosis was retrieved from the medical records. Information on the time of death or last contact was not retrievable for 81 of these subjects, who were therefore excluded from the survival analysis. The average age of the men was 37.6 years with a standard deviation of 12.1 years; the average age of the women was 38.7 years with a standard deviation of 12.7 years. A total of 138 women and 465 men had never taken ART. For those who had had some ART, 14 of the women and 66 of the men had had access to HAART from the start of treatment. Another 22 women and 60 men were treated with a different therapy scheme before taking HAART. Ten women and 22 men used another type of therapy that did not include HAART. This information is summarized in Table 1.
Table 2 shows the survival probabilities at different time periods for the four treatment categories. These results show that the number of patients in the cohort decreased from 546 after 1 year to 80 after 10 years. Overall, 89% of deaths occurred within the first 5 years after diagnosis.
The Kaplan–Meier estimator is shown in Fig. 1. This graph shows that the probability of survival after diagnosis without ART was approximately 0.75 during the first year, 0.35 during the fifth year and 0.26 during the tenth year. This probability does not show any important changes during the fifteenth year. For each year thereafter the survival probability suffered a slight decrease (0.25).
Good responses were always obtained when patients were treated with HAART, but when they received HAART from the beginning of their treatment the results significantly exceeded the results of patients who did not. During the first 5 years the survival probability was more than 0.9. In year 10 this probability reached 0.85 and stabilized at 0.81 during the rest of the follow-up. Those cases in which the patients used HAART as a second-line therapy also showed very good responses; the survival rate was higher than 50% after year 15. As expected, the worst scenario under ART treatment corresponded to the patients who never received HAART and who were only treated with monotherapy or dual therapy. These cases had a survival probability of approximately 0.25 after the fifth year.
Equation (Uncited)Image Tools
The results of the Cox regression model are presented in Table 3. After adjusting for covariates, we found that those who received no treatment were 7.1 and 4.0 times more likely to die than those who either: 1) started and continued with HAART and 2) those who eventually received HAART, respectively. The difference between those always on HAART and those who used HAART as a second line therapy was not statistically significant at 0.05 (P = 0.12). On the contrary, we found that subjects who received either mono or dual therapy experienced no reduction in the risk of mortality compared with those who received no therapy. The results suggested a higher survival probability in the group of women. This can be explained, however, by the small sample size; only 23% were women. Age was also related to the survival probability and the risk of death; the probability of death increasing 1% with each additional year of age.
Equation (Uncited)Image Tools
Table 4 shows the average annual treatment cost per patient. Care costs were higher for patients who had access to ART but who never took HAART. In contrast with what was expected, there was little difference between the cost of care of patients without ART and those patients who use triple therapy from the beginning of their treatment. This result may, however, be due to the difference in sample size (603 patients who never received ART against 80 patients who always used HAART). The low cost of care corresponds to the group of patients who started with mono or dual therapy but never used triple therapy.
Equation (Uncited)Image Tools
Figure 2 describes the cost treatment components of each of the treatment options. We found that the largest proportion corresponded to antiretroviral drug costs representing nearly 90% of the average annual care cost in the ‘always HAART’ group and the ‘HAART as second line’ group. The cost of care of the patients who never took ART was mostly inpatient days and other medicines used to control and treat opportunistic infections, representing together almost 85% of the total. Regarding the patients who were treated with mono or dual therapy (but never took HAART), we found that the largest proportion of cost of care consumption was attributable to inpatient days, representing nearly 70% of the total cost of care.
The healthcare costs of patients with HIV sampled in the follow-up periods are shown in Table 5. During the first year of follow-up, the average cost of treating a patient with HIV in this hospital was more than US$5600. The next 5 years represented an approximate cost of US$32 400, and the cost of taking care of the survivors for an additional 5 years surpassed US$65 000, meaning that every survival year after diagnosis has an approximate cost of US$6000. During the last year of life this pattern changed significantly, as the average cost increased substantially, corresponding to the diminishing survival probability. According to the study results, extending the length of life beyond 15 years after diagnosis represents an accumulated cost of more than US$280 000 per individual.
Figure 3 shows the results of the sensitivity analysis. The current treatment costs for each of the three time periods is at least three times higher than in the best scenario.
The results obtained in this study show that HAART represents a very effective way to extend the lives of HIV-infected individuals. Success in extending a patient's life is attained under certain conditions, such as adequate compliance to treatment. Freedberg et al.  assessed the cost-effectiveness of a nursing intervention on antiretroviral compliance and found that compliance interventions are likely to provide long-term survival benefits. A satisfactory level of compliance requires patient cooperation. In addition, the health system must be committed and able to ensure that drugs are administered in correct quantities and in a timely manner. Health personnel must also communicate effectively so that patients can understand the importance of taking the drugs correctly. We did not have access to specific patients' compliance information, but this hospital has had a history of supply problems. As a result, patients are likely to have compliance problems, and it is important to remember that even moderate treatment interruptions or modifications favour the rapid replication of the virus, causing viral resistance, which in turn influences patients' health status , and consequently their probability of survival.
Another indispensable element in patients' care is adequate monitoring of their clinical and immunological status with two laboratory tests: viral load and CD4 cell counts. The first measures the virus count present in the blood and the second provides information about the immunological situation. In the first years of the period studied, the hospital did not offer the CD4 cell count test because of its expense; therefore the patients' immunological situation at the beginning of their ART was unknown. As a result, it is difficult to know whether the patients' health improvement was actually related to the drugs added to their treatment. Fang et al.  identified the patients' clinical condition at the initiation of ART, and their results showed important differences compared with those of this study. For example, they found that the 5-year survival rate was 58% in patients who had already developed AIDS at the date of diagnosis (AIDS group) and 89% in those who had not (non-AIDS group). Our study was unable to make that type of distinction.
Equation (Uncited)Image Tools
Although HAART is a very effective way to increase the life expectancy of people with AIDS, it is necessary to look at the costs incurred. For example, the extension of life after 18 years, regardless of the type of therapy used, represents a cost of US$15 000; therefore strategies must be developed to control costs. Government authorities need to find ways to negotiate lower ART costs, the main element of AIDS patients' care costs. Moreover, it is essential to improve and create new infection-prevention strategies that reduce the prevalence rate, which has stabilized over the past years. It is also imperative to carry out strategies that guarantee adherence to treatment regimens. Furthermore, monitoring of patients' clinical and immunological status is necessary, because it helps identify the appropriate time to initiate ART and could reduce the waste of resources. Knowing the patients' immunological condition can also help prevent opportunistic infections through prophylaxis, which is cheaper than treatment. These strategies must all be considered if Mexican policymakers want to control the cost of treatment and save resources that are needed to provide care in other areas.
Equation (Uncited)Image Tools
The present study has some important limitations such as the absence of data on patients' clinical status at initiation of treatment, as well as adherence information. Nevertheless, these results are a very useful tool for approximating both the survival time and the costs associated with it. These data can give health authorities more accurate expectations of ART results, very important in the framework of the Mexican health system's financial reforms and in the context of the government's commitment to universal access to ART.
Equation (Uncited)Image Tools
This study focuses on survival rates under HAART treatment; but it is important to take into account not only the length of life but also the quality of life. A number of studies in other countries have found a positive relationship between the correct use of ART and patients' quality of life [44–47]. There have been no such studies in Mexico on this topic to date; studies are needed to explore the quality of life of patients taking HAART in Mexico.
Equation (Uncited)Image Tools
Equation (Uncited)Image Tools
The authors would like to thank Dr Stefano Bertozzi for his constructive contributions to the methodology. They would also like to thank Carola Muñoz, Christine Dipboye, Veronika Wirtz, Ana Solares, Ivonne Flores and David Washburn for their contributions to editing.
Conflicts of interest: None.
Equation (Uncited)Image Tools