IMPLEMENTATION
AND
OPERATIONAL RESEARCH: EPIDEMIOLOGY
AND
PREVENTION
Effect of Universal Access to Antiretroviral Therapy on HIV/AIDS Mortality in Mexico 1990–2011 Juan E. Hernández-Ávila, MsC, ScD,* Lina S. Palacio-Mejía, MA, PhD,* Alfonso Hernández-Romieu, MD, MPH,† Sergio Bautista-Arredondo, MsC,*,** Jaime Sepúlveda Amor, MD, MPH, ScD,‡ and Mauricio Hernández-Ávila, MD, ScD*,††
Introduction: Universal coverage of highly active antiretroviral therapy (HAART) for HIV/AIDS in Mexico was achieved in a staggered fashion. First, HAART was made available in 1997 for employees in the formal sector and subsequently, in 2003, for informal sector workers and the unemployed population. We evaluated the impact of this free universal HAART program on HIV/AIDS mortality in the Mexican population aged 15 years or older.
Methods: Time-series cross-sectional data analysis of standardized HIV/AIDS mortality rates (1990–2011) using marginal linear models with inflection points at 1994, 1997, and 2003. Results: Men employed in the formal sector experienced the greatest effect of access to HAART, with a 32% reduction in mortality from 1997 to 2011 (P , 0.01). For men in this sector, mortality increased from 1990 to 2003, then stabilizing at 7 deaths per 100,000. For women in the informal sector, mortality increased from 1990 to 2011. We found large between-state variability: mortality decreased by 59% in Mexico City while it increased by 245% in Tabasco during the study period. Six of 32 states achieved statistically significant reductions in mortality while 5 states continued to experience increasing mortality rates despite universal HAART. Conclusions: Access to universal HAART has had relative success in Mexico. The heterogeneity of HIV/AIDS mortality by employment status, gender, and state of residence highlight urgent needs to improve health equity with pro-poor and gender-responsive programs and a call for country-specific operational research in HIV/AIDS prevention and treatment. Our results may apply to other countries seeking to implement universal access to antiretroviral therapy. Key Words: highly active antiretroviral therapy, mortality, México (J Acquir Immune Defic Syndr 2015;69:e100–e108) Received for publication August 12, 2014; accepted February 16, 2015. From the *Centro de Información para Decisiones en Salud Pública, Instituto Nacional de Salud Pública, Cuernavaca, México; †Department of Epidemiology, Emory University, Atlanta, GA; ‡Global Health Sciences, University of California (UC) San Francisco, San Francisco, CA, USA and Emeritus Professor of Public Health, National Institute of Public Health, Mexico; **Centro de Investigación en Sistemas de Salud. Instituto Nacional de Salud Pública México; and ††Dirección General. Instituto Nacional de Salud Pública. Presented at the XV Congreso de Investigación en Salud Pública, at the Instituto Nacional de Salud Pública, March 5–8, 2013, Cuernavaca, México. The authors have no funding or conflicts of interest to disclose. Correspondence to: Mauricio Hernández-Ávila, MD, ScD, Instituto Nacional de Salud Pública, Avenida Universidad, #655, Colonia Santa Maria Ahuacatitlán, Cuernavaca, México (e-mail: [email protected]). Copyright © 2015 Wolters Kluwer Health, Inc. All rights reserved.
e100
| www.jaids.com
INTRODUCTION The advent of highly active antiretroviral therapy (HAART) has significantly reduced morbidity and mortality associated with the HIV. According to the United Nations Program on HIV/AIDS, an estimated 6.6 million deaths were averted from 1995 to 2012 worldwide, with 5.5 million deaths averted in low- and middle-income countries.1 Many of these low- and middle-income countries have achieved important declines in HIV-related mortality by developing policies and programs to provide universal access to HAART. In countries like Malawi, Thailand, Brazil, and South Africa,2–9 universal coverage of HAART has had a positive impact on reducing mortality resulting from HIV. In Mexico, universal access to HAART has been available since 2003; its impact on HIV mortality, however, remains understudied. Mexico is a middle-income country with 31 states and a federal district, and a population of 112.3 million. Approximately 170,000 individuals aged 15 years or older are estimated to be living with HIV/AIDS (prevalence 0.2%).10 The HIV epidemic is concentrated: the highest prevalence of HIV has been observed among men who have sex with men, estimated at 17% in 2011.11 Between 2008 and 2012, an estimated 85% of diagnosed HIV-infected individuals who were eligible for treatment were receiving HAART.12,13 The Mexican Health System comprised a set of public and private institutions. Public health care institutions are divided into 2 main groups. On the one hand, the social security institutions (SSIs) provide integral health care and social services to the people employed in the formal economy sector and government employees, as well as their families (39% of the total population in 2012). SSIs are statecontrolled organizations, funded by government, employers, and employees to finance; they organize and provide health care through their human resources and health facilities (SISs health system). Citizens in this system are eligible to receive all medically and hospital necessary services free at the point of delivery. On the other hand, the Ministry of Health & State Health Secretariats managed clinics and hospitals (MoH/SHS health system), which, before 2003, provided health care services to the population working in the informal economy sector, the self-employed, and the unemployed, charging outof-the-pocket recovery fees at the point of delivery. In 2003, the National System for Social Protection in Health, also known as Seguro Popular (SP), was created as a public health
J Acquir Immune Defic Syndr Volume 69, Number 3, July 1, 2015
Copyright © 2015 Wolters Kluwer Health, Inc. Unauthorized reproduction of this article is prohibited.
J Acquir Immune Defic Syndr Volume 69, Number 3, July 1, 2015
insurance program for low-income uninsured Mexicans. SP is funded jointly by federal and state governments to finance a limited health package consisting of 218 health interventions and 17 high-cost diseases,14 including HIV/AIDS. Most services financed by SP are provided by the MoH/SHS through their human resources and health facilities. By 2012, approximately 33% of the Mexican population was enrolled in SP; a remaining 28% of the population, although eligible for SP, was still uninsured and demand services from the private sector or the MoH/SHS systems with an out-of-pocket recovery fee. The implementation of the universal HAART program in Mexico was staggered and dependent on health insurance. Antiretroviral therapy first became available in 1997 for use in population affiliated to the public sector’s SSIs and in 2003 for population affiliated to SP. HIV/AIDS diagnosis on uninsured persons self-selects them to receive treatment through SP and they are therefore affiliated at diagnosis. This adverse selection by disease into SP makes it challenging to study these 2 last population groups separately. HIV/AIDS mortality in Mexico has not decreased as observed in other countries despite offering universal free access to HAART. In Spain, the mortality rate decreased from 14.9 in 1995 to approximately 5 per 100,000 in 1998,15 a 3-fold reduction in only 3 years. The low impact of the universal HAART program on HIV-related mortality in Mexico has raised questions about its implementation. It remains unclear whether differences in HIV mortality maybe associated with the institution providing health care or if there are other between-state variations that maybe contributing to the observed low decline in the rates. Finally, there maybe differences in program outcomes originated by gender inequalities. To address these questions, we assessed the effect of changes in policies regulating access to HAART in Mexico’s HIV/AIDS mortality rate trends from 1990 to 2011 and analyzed differences in mortality by institution providing health care, state of residence, and sex using government health records (“real world data”16).
Universal HAART and AIDS Mortality Mexico
The estimated individual probabilities were subsequently fed to a random number generator with a Bernoulli distribution to obtain estimated health coverage, which was then imputed to observations with missing data. Inclusion criteria were age at death 15 years or older; death occurring between 1990 and 2011 and HIV/AIDS as the underlying cause of death (ICD-9 codes18 279.5 and 279.6, before 1998; ICD-10 codes19 B20–B24).
Population Estimations made by the National Population Council (CONAPO) 1990–2030,20 based on the 2010 census,21 were used as denominators in the calculation of crude and agespecific mortality rates. The proportion of the population with health coverage was estimated using previous population projections,22 based on the second population count in 2005.23 To incorporate socioeconomic and demographic differences, the states and Federal District were characterized by the marginalization index published by the CONAPO,24 which was based on data collected in the 2010 census.21 This multidimensional index allows for differentiating among states based on the global impact of the shortcomings to which the population is exposed; these include lack of access to education, inadequate housing conditions, low income, and those related to residing in very small communities. The index is grouped into 5 categories ranging from very low (best) to very high (worse) marginalization. The standardization of mortality rates was conducted using the 2000–2020 world standard average population, with methods described by World Health Organization.25 For the purposes of this work, we named “population with social security” (PWSS) to people receiving health services from the SSIs. Because of the difficulties caused by the automatic affiliation to SP at the time of diagnosis with HIV/AIDS, we conflated uninsured population with SP affiliates into a group we named “population without social security” (PWOSS).
Statistical Models METHODS Data Sources Information from individual death certificates was obtained from databases provided by the Directorate General for Health Information17 at the MoH and included year of death, underlying cause of death (International Classification of Diseases ICD-coded18,19), age, sex, marital status, education, state of residence (n = 32), and institution providing health care, which we analyzed as a dichotomous variable: “1” indicating coverage by the public SSIs and “0” indicating no health coverage and/or—beginning in 2003—coverage by SP. Data on institution providing health care were missing in 8.1% (n = 7227) of the death certificate records. To address missing values, we fitted a logistic model to estimate the probability that an individual had health coverage by any SSI using age, sex, years of education, and state of residence as predictor variables, which are also found in the death certificate database. Copyright © 2015 Wolters Kluwer Health, Inc. All rights reserved.
The (1990–2011) trend of the HIV/AIDS standardized mortality rate (SMR) in the population aged 15 years and above was studied with time-series cross-sectional data analysis, using marginal linear models.26–29 Such models are used in social sciences to study the effect of policy changes, where observation units are usually states or countries and cross-sectional correlations are expected between them through different measurements in time.26 The analyzed time series was divided into 4 periods according to the availability of antiretroviral treatment and the current access-to-treatment policy in Mexico: (1) 1990–1994, in which there was no treatment available; (2) 1994–1997, when antiretroviral treatment became available; (3) 1997–2003, in which treatment with highly active antiretroviral therapy (HAART) became available and was introduced to PWSS as part of an integral health plan; and (4) 2003–2011, in which the policy for universal access to treatment was expanded to PWOSS (universal HAART). Three models were adjusted independently, with the first at the nationwide level, adjusted for www.jaids.com |
e101
Copyright © 2015 Wolters Kluwer Health, Inc. Unauthorized reproduction of this article is prohibited.
J Acquir Immune Defic Syndr Volume 69, Number 3, July 1, 2015
Hernández-Ávila et al
social security status and sex. The second and third models were adjusted at the statewide level by health coverage status. The natural logarithm of the SMR was used as the outcome variable in all 3 models. To estimate changes in trends in mortality rates over the 4 periods analyzed, we incorporated a linear spline30 in each model, with 4 components, and knots at 1994, 1997, and 2003. The results of the statistical models estimated in this study are presented as the annual percent of change for each study period using the following relation: %Dy ¼ ebp 2 1 · 100; where bp is a regression coefficient or a linear combination of the coefficients. All analyses were performed using Stata software, Stata12 corporation.31
RESULTS Between 1990 and 2011, a total of 90,621 deaths caused by HIV/AIDS were reported. Of these, 98% (89,097) were people aged 15 years or older; 84.5% (75,280) were men, and 15.5% (13,817) were women. More than 67% of the deaths occurred among people aged between 25 and 44 years, with most deaths occurring in the 30–34 age group (20.1%). According to death certificates, 48.8% (43,488) of the deceased were PWOSS and 43.1% were PWSS. Of those missing health insurance coverage data in their death certificates (7227), 41% (2959) were classified as PWSS and 59% (4268) as PWOSS, using the imputation method described. The resulting proportions closely followed the distribution of social security status at the national level. Trends in the mean age at death from HIV/AIDS by sex and social security status are shown in Figure 1. Overall, the
mean age at death in PWSS was higher (37–43 years) compared with PWOSS (33–37 years) and higher among men than women. However, after stratifying by sex and social security status, we observed that women without social security had the lowest mean age at death of all groups. Women with social security fared better than both men and women without social security, and after 2010, they had the highest age of death of all groups. Figure 2 illustrates the SMRs for PWSS and PWOSS from 1990 to 2011; the annual percentages of change in HIV mortality by social security status and sex are shown in Table 1. The death rate from HIV/AIDS in PWSS saw its largest increase from 1990 to 1994, with an annual increase estimated at 24% (Fig. 2, Table 1). From 1994 to 1997 this positive trend was reversed, but the decrease observed was not statistically significant. After 1997, HIV/AIDS SMR in PWSS decreased by 4.3% annually until 2003; it continued to decrease by 1.8% per year until 2011, for a cumulative reduction of 32% during the period between 1997 and 2011 (Fig. 2, Table 1). Among PWOSS, we observed an annual percentage increase in the HIV/AIDS SMR of 17.0% during the period between 1990 and 1994. In contrast to PWSS, the death rate among PWOSS continued to increase by 4.2% annually from 1994 to 1997 and by 2.0% from 1997 to 2003 (Fig. 2, Table 1). Since 2003, when the universalization of HAART began, the annual percentage change in the mortality rate among PWOSS has remained close to zero, stabilizing at around 7.27 deaths per 100,000 people in 2011 (Fig. 2, Table 1). Overall sex-specific SMR curves and those stratified by social security status are shown in Figure 3. The overall annual percent change in the SMR in women did not decrease. In women without social security, we observed a statistically significant increase in HIV/AIDS SMR from
FIGURE 1. Differences in mean age at death by sex and social security status, Mexico 1990–2011. Women have benefited less from access to antiretroviral therapy, as age at death is consistently lower in women than men and the lowest age is in women without social security. This highlights a severe gap in access to HAART in women both with and without social security. Source: Prepared by authors using public data from Mortality database published by General Directorate of Health Information; MoH.17
e102
| www.jaids.com
Copyright © 2015 Wolters Kluwer Health, Inc. All rights reserved.
Copyright © 2015 Wolters Kluwer Health, Inc. Unauthorized reproduction of this article is prohibited.
J Acquir Immune Defic Syndr Volume 69, Number 3, July 1, 2015
Universal HAART and AIDS Mortality Mexico
FIGURE 2. HIV/AIDS observed SMRs and statistical model results, Mexico 1990–2011. The introduction of HAART in the PWSS in 1997 caused a steep decline in the HIV/AIDS mortality in the period 1997–2011, but the universalization policy started in 2003 did not cause the expected decline in the mortality of the PWOSS. Source: Prepared by authors using public data from Mortality database from General Directorate of Health Information; Minister of Health17 and populations estimations made by CONAPO 1990 -203020, population projections COLMEX-DGIS, 1990–2012.22
1990 to 2011 (Table 1). In contrast, among women with social security, an increasing death rate was observed from 1990 to 1994, after which it reached a plateau and remained constant. Among men without social security, the SMR
increased from 1990 to 2003 and reached a plateau from 2003 to 2011; among men with social security, we observed a sharp, statistically significant decrease in the SMR from 1997, which continued until 2011 (Fig. 3, Table 1).
TABLE 1. Annual Percent Change in the Standardized Mortality Rate by Sex and Social Security Status, Mexico 1990–2011 Health Coverage/Economy Sector
Sex
Study Period
Coefficient
Statistical Significance (P)
Annual Percent Change
95% Confidence Interval
Population without social security (informal sector)
Both
1990–1994 1994–1997 1997–2003 2003–2011 1990–1994 1994–1997 1997–2003 2003–2011 1990–1994 1994–1997 1997–2003 2003–2011 1990–1994 1994–1997 1997–2003 2003–2011 1990–1994 1994–1997 1997–2003 2003–2011 1990–1994 1994–1997 1997–2003 2003–2011
0.157 0.041 0.020 0.003 0.090 0.040 0.058 0.016 0.169 0.047 0.016 0.001 0.215 20.022 20.033 20.014 0.111 20.004 0.006 20.004 0.235 20.027 20.036 20.015
,0.001 0.002 ,0.001 0.529 ,0.001 0.032 ,0.001 0.013 ,0.001 ,0.001 0.003 0.763 ,0.001 0.249 ,0.001 0.032 ,0.001 0.846 0.446 0.535 ,0.001 0.178 ,0.001 0.034
17.0 4.2 2.0 0.3 9.4 4.1 6.0 1.6 18.4 4.8 1.6 0.1 24.0 22.1 23.2 21.4 11.8 20.4 0.6 20.4 26.5 22.7 23.5 21.5
14.3 to 19.7 1.6 to 6.9 0.9 to 3.2 20.6 to 1.2 5.8 to 13.2 0.3 to 8.0 4.3 to 7.7 0.3 to 2.9 15.9 to 21.0 2.4 to 7.4 0.5 to 2.7 20.7 to 1.0 20.2 to 28.0 25.7 to 1.5 24.9 to 21.6 22.7 to 20.1 8.3 to 15.4 23.9 to 3.3 21.0 to 2.3 21.6 to 0.9 22.2 to 30.9 26.5 to 1.2 25.3 to 21.8 22.9 to 20.1
Women
Men
Population with social security (formal sector)
Both
Women
Men
Source: Prepared by authors using public data from Mortality database published by the General Directorate of Health Information; MoH17; population estimations made by CONAPO 1990–2030,20 population projections COLMEX–DGIS-1990–2012.22
Copyright © 2015 Wolters Kluwer Health, Inc. All rights reserved.
www.jaids.com |
e103
Copyright © 2015 Wolters Kluwer Health, Inc. Unauthorized reproduction of this article is prohibited.
Hernández-Ávila et al
J Acquir Immune Defic Syndr Volume 69, Number 3, July 1, 2015
FIGURE 3. HIV/AIDS observed SMRs and statistical model results by sex and social security status, Mexico 1990–2011. Although the HIV/AIDS epidemic in Mexico is focalized in men, women have experienced fewer benefits after the introduction of HAART. Even when treatment was available for the PWSS in 1997, the mortality rate of women in this group continued to rise. The slope on the trend in women without social security is even larger. Source: Prepared by authors using public data from Mortality database published by General Directorate of Health Information; MoH17; population estimations made by CONAPO 1990– 2030,20 population projections COLMEX–DGIS-1990–2012.22
We observed large variability in the trend of HIV/ AIDS mortality rates among states. Taking into account the periods when HAART was made accessible to the population with (1997) and without (2003) social security, we grouped states into 3 categories: (1) those which saw a decrease in the HIV/AIDS mortality rate after the introduction of the intervention, (2) those in which there was at least a halt to the rising trend in the mortality rate, and (3) those which saw an upward trend in the mortality rate despite continuing interventions. The SMR in PWSS has declined steadily since 1997 in 5 states with different marginalization indexes: the Federal District (very low), Jalisco (low), Mexico State (low), Morelos (medium), and Puebla (high) (Fig. 4). Querétaro, Colima, and Yucatan showed a downward trend in the SMR in 1997–2003, although the results were not significant after 2003. Coahuila showed reductions in SMR during 1997–2003, but the trend was again positive after 2003. Veracruz, Chiapas, and Tamaulipas exhibited positive annual percentage changes in the SMR from 1997 to 2003; after 2003, the trend stabilized and was not statistically significant. The only states with an
upward trend from 2003 to 2011 were San Luis Potosi (6.17%) and Tabasco (4.75%), both with a high marginalization index. The analysis of the PWOSS demonstrated a reversal of the upward trend in the SMR for HIV/AIDS from 2003 in the states of Aguascalientes (low marginalization index), Federal District (very low), Guanajuato (medium), Guerrero (very high), Jalisco (low), and Puebla (high) (Fig. 4). Our results indicate that the trend in the Federal District in the SMR in PWOSS has been declining since 199432; in the state of Jalisco, the epidemic was contained in the same year. For the remaining 26 states, the annual percentage change was not statistically different from zero during the period 2003–2011. Of particular importance, however, is that in the states of Sonora (low) and Tabasco (high), the trend in the period 2003 to 2011 is still rising with statistical significance (Fig. 4).
e104
Copyright © 2015 Wolters Kluwer Health, Inc. All rights reserved.
| www.jaids.com
DISCUSSION We present SMR trends resulting from HIV/AIDS in the population aged 15 years and older in Mexico during the
Copyright © 2015 Wolters Kluwer Health, Inc. Unauthorized reproduction of this article is prohibited.
J Acquir Immune Defic Syndr Volume 69, Number 3, July 1, 2015
Universal HAART and AIDS Mortality Mexico
FIGURE 4. Annual percent change in the SMR by state and social security status, Mexico 1990–2011. Note: Light-shaded cells indicate a statistically significant increasing annual percent change in the mortality rate. Dark-shaded cells indicate a statistically significant decreasing annual percent change in the mortality rate. Unshaded cells are not statistically significant. Source: Prepared by authors using public data from Mortality database from General Directorate of Health Information; MoH17; population estimations made by CONAPO 1990–2030,20 population projections COLMEX–DGIS-1990–201222; marginalization index estimation by CONAPO.24
period 1990–2011 to determine the effectiveness of the national implementation of universal access policies to HAART. During this period, there were no other major changes in policy or access to health care that could influence HIV/AIDS mortality beside the introduction of HAART in 1997 and its universalization through SP in 2003. Copyright © 2015 Wolters Kluwer Health, Inc. All rights reserved.
Mean age at death in the HIV/AIDS PWSS increased from 37.7 in 1990 to 43.3 in 2011 in women and from 36.5 to 41.7 in men. The mean age at death among those without formal employment (ie, the PWOSS) was lower, increasing from 30.2 to 35.9 in women and from 33.6 to 37.8 in men, between 1990 and 2011. www.jaids.com |
e105
Copyright © 2015 Wolters Kluwer Health, Inc. Unauthorized reproduction of this article is prohibited.
Hernández-Ávila et al
J Acquir Immune Defic Syndr Volume 69, Number 3, July 1, 2015
Our results show important reductions in AIDS mortality associated with the introduction of HAART but with differing levels of effectiveness in the population groups analyzed. First, after the introduction of HAART for PWSS in 1997, a cumulative reduction of 32% was observed in the SMR (Fig. 2); however, its effect was not equal for men and women. In the female PWSS, the trend was not reverted: the SMR was only stabilized around 1.5 per 100,000 (Fig. 3). Following the introduction of HAART for PWOSS as part of the SP implementation in 2003, the SMR trend did not revert: it was only stabilized around 7.27 per 100,000. Again, women benefited less from treatment availability, as the SMR kept its upward trend until 2011 when it reached 2.7 per 100,000. Our results also show important gaps between men and women by type of insurance. The SMR of male PWSS was substantially lower (8 per 100,000) than male PWOSS (12.5) in 2011. The female PWOSS SMR was almost twice as high as their PWSS counterparts in that year. Nonetheless, even with these shortcomings, we estimate that from 1997 to 2011, close to 179,000 deaths were prevented. Our results show wide variations in the impact of the introduction and subsequent universalization of HAART by state. The Federal District, Jalisco, and Puebla were the only states that achieved sustained reductions in the HIV/AIDS SMR in PWSS and PWOSS after the introduction of HAART in the respective populations. The States of Mexico and Morelos achieved sustained reductions in PWSS while Tabasco and San Luis Potosi are as yet unable to stop the SMR increasing trend in PWSS. Guanajuato and Guerrero have achieved reductions in the SMR in PWOSS after introduction of HAART in 2003, but Sonora and Tabasco have not stopped the upward trend of the SMR in this segment of the population. People in formal employment (the PWSS) are in general more financially secure and have achieved higher levels of education. In Mexico, having formal employment is a key marker of one’s socioeconomic position and access to quality medical care. The observed differences in the effect of access to HART in the studied populations maybe indicative of different phenomena. For example, PWSS were more affected during the early days of the HIV/AIDS epidemic in Mexico because of the role of sexual contact with foreign nationals during travels to other countries,33 as they were more likely to travel. Later in the epidemic, the dissimilar mortality trends between employment groups maybe explained by differences in HIV/AIDS medical care between insurance programs: while formal-sector employees receive integrated care provided by infectious diseases specialists that covers HAART, HIV complications, and adequate referrals to specialized levels of care, people in the informal sector receive HAART through HIV Specialized Outpatient Clinics coordinated by general practitioners, with limited coverage for disease complications and ill-defined mechanisms for referral to specialized care. Furthermore, although there is scarcity of data regarding the extent of late initiation of HAART in Mexico, it has been documented that people living with HIV in the informal sector tend to be late testers and late HAART initiators.34 Data from one tertiary care hospital indicate that close to 75% of those with HIV initiated
HAART with a CD4+ count ,200 cells per microliter or with an AIDS-defining illness.34 Although this problem has not been properly documented in PWSS, it could possibly explain the differential mortality trends observed between the 2 population segments. This suggests that coverage of HAART may need to improve not only in the number of people receiving treatment but also in the number of people who start treatment in the early phases of the disease. Other flaws in the cascade of HIV care have not been documented; however, data from patients receiving care through the MoH/SHS health system suggest that serious problems with linkage to care, retention, and adherence exist.35 Additionally, a recent analysis on the variability of mortality trends within the MoH/SHS health system suggests that clinics where health providers have more training and experience, a stronger reference network, and greater ability to make decisions related to the provision of care, tend to have lower mortality rates.35 We cannot overlook certain selection bias by which people who lose their jobs in the formal employment seek medical coverage through SP. This is a strong possibility, given the frequent movement in and out of the formal sector,36 and such bias would increase the mortality rate among the PWOSS and lower mortality rates for the PWSS. Finally, the observed difference in the post-HAART era could also reflect that those in the formal employment sector—those who are more educated—received more information regarding risk factors that promoted earlier testing and initiated treatment earlier in the disease’s progression. Our results indicate that the greatest impact of access to HAART was observed in men with formal-sector employment (Table 1; Fig. 3), where there was an overall decline of 32% between 1997 and 2011. In contrast, for women receiving care by the same providers, only a paused stabilization in the mortality rate was achieved. For men outside the formal employment sector, the rising trend paused and transited into a plateau while a continuing increase in the mortality rate was observed among women in this same sector. This finding raises concern that even with universal access to HAART, mortality rates among women have not shown similar reductions as those among men despite documented evidence of women visiting health services more often37,38 and presenting to care with higher CD4+ levels.34,39 There is no clear explanation for the differences found in the mortality curves between men and women in the postHAART era. It has been documented, however, that women on HIV/AIDS treatment in Mexico City’s MoH health system have poorer results because many women living with HIV are subject to stigma, discrimination, and a higher risk of suffering violence.40 Furthermore, although our results show that despite an overall increase in age at death in both genders, the age of death in women with and without social security exhibits a much larger gap compared to men when stratified by sex and employment sector (Fig. 1). Reportedly, in Mexico, women are infected by a stable partner, and the HIV diagnosis frequently occurs after a partner/child is diagnosed with AIDS while a small proportion were diagnosed by screening during pregnancy.40,41 This suggests that although women may
e106
Copyright © 2015 Wolters Kluwer Health, Inc. All rights reserved.
| www.jaids.com
Copyright © 2015 Wolters Kluwer Health, Inc. Unauthorized reproduction of this article is prohibited.
J Acquir Immune Defic Syndr Volume 69, Number 3, July 1, 2015
become infected at a later age on average than men, they benefit less than men in survival. Mortality trends by state of residence (n = 32) indicate strong heterogeneity in the post-HAART period for both populations with and without social security. The marginalization index of the state of residency was not an important predictor for mortality for PWOSS. However, for PWSS, lower impacts of HAART were documented for states with a high marginalization index, with only 6 states (18.7%) achieving statistically significant reductions in mortality following universalization of HAART. For PWSS, only 5 states (15.6%) have achieved sustained reduction in mortality from 1997, the year that HAART became available to this population. These results suggest that state programs for prevention and control of HIV/AIDS should be reviewed to identify areas of opportunity to improve and achieve results similar to those that have sustained reduction in mortality. The most important next step is to boost the HIV/AIDS operational research agenda to identify good practices that help achieve consistent results throughout the country. The focus should be directed toward early diagnosis and earlier initiation of HAART, to improve quality of care to increase survival rates and decrease AIDS complications, and to reduce stigma and discrimination to improve early access to testing and treatment. The states of Tabasco, Sonora, San Luis Potosi, Veracruz, and Quintana Roo require particular attention because they show the largest gap compared to the national results. The Federal District is the only state that has reversed the upward trend in mortality in the PWOSS since 1997: this maybe attributed to resources invested by the local government to increase early diagnosis and treatment and the specialized health care offered by the SSIs and the MoH in this area. The fact that most states have succeeded in stabilizing mortality makes evident the effect of universal access; however, without proper monitoring and evaluation at the state level, we may not adequately understand the current gaps in health services. There are limitations in the information systems used in this study. In the case of death registration, potential classification and coding errors in the cause of death, and failure to record social security status, maybe present. With these errors, the possibility of underestimation exists. Moreover, the difficulty in establishing the insured population (denominators) according to the various SSIs is a challenge that could also affect our results. These, however, are random errors that would make detected associations less apparent. In our data, we cannot differentiate whether the observed increase in the mean age at death is an indicator that access to HAART is prolonging the lives of people with HIV/AIDS or that they are becoming infected at older ages, as a combined result of behavioral changes and a reduced community viral load. In conclusion, type of employment, gender, age, and state of residence are social determinants that contribute to HIV/AIDS mortality in Mexico. Although the legislation of universal access to antiretroviral therapy has had relative success in Mexico, the heterogeneity and gaps in HIV/AIDS mortality highlight the need to boost country-specific operational research in HIV/AIDS prevention and treatment. Our findings may apply to other low- and middle-income countries seeking to maximize the effects of universal access to antiretroviral therapy. Copyright © 2015 Wolters Kluwer Health, Inc. All rights reserved.
Universal HAART and AIDS Mortality Mexico
ACKNOWLEDGMENTS The authors acknowledge the General Directorate for Information in Health for making the 1990–2011 mortality dataset available. REFERENCES 1. Joint United Nations Programme on HIV/AIDS (UNAIDS). Geneva, Switzerland: UNAIDS/World Health Organization; 2013. 2. Badri M, Maartens G, Mandalia S, et al. Cost-effectiveness of highly active antiretroviral therapy in South Africa. PLoS Med. 2006;3:e4. 3. Chasombat S, McConnell MS, Siangphoe U, et al. National expansion of antiretroviral treatment in Thailand, 2000-2007: program scale-up and patient outcomes. J Acquir Immune Defic Syndr. 2009;50:506–512. 4. Ferradini L, Jeannin A, Pinoges L, et al. Scaling up of highly active antiretroviral therapy in a rural district of Malawi: an effectiveness assessment. Lancet. 2006;367:1335–1342. 5. Todd J, Glynn JR, Marston M, et al. Time from HIV seroconversion to death: a collaborative analysis of eight studies in six low and middleincome countries before highly active antiretroviral therapy. AIDS. 2007; 21(suppl 6):S55–S63. 6. Teixeira PR, Vitoria MA, Barcarolo J. Antiretroviral treatment in resource-poor settings: the Brazilian experience. AIDS. 2004;18(suppl 3): S5–S7. 7. Marins JR, Jamal LF, Chen SY, et al. Dramatic improvement in survival among adult Brazilian AIDS patients. AIDS. 2003;17:1675–1682. 8. World Health Organization. Global HIV/AIDS response—epidemic update and health sector progress towards Universal Access. In: WHO, ed. Global HIV/AIDS Response. Geneva, Switzerland: WHO; 2011:233. 9. Alonso Gonzalez M. Antiretroviral Treatment in the Spotlight: A Public Health Analysis in Latin America and the Caribbean 2013. Washington, DC: PAHO; 2013. 10. Gutiérrez JP, Sucilla-Pérez H, Conde-González CJ, et al. Seroprevalencia de VIH en población mexicana de entre 15 y 49 años: resultados de la Ensanut 2012. Salud Publica Mex. 2014;56:323–332. 11. Bautista-Arredondo S, Colchero MA, Romero M, et al. Is the HIV epidemic stable among MSM in Mexico? HIV prevalence and risk behavior results from a nationally representative survey among men who have sex with men. PLoS One. 2013;8:e72616. 12. Centro Nacional para la prevención y control de VIH SIDA. El VIH/SIDA en México 2011. Numeralia epidemiológica. México: Centro Nacional para la prevención y control de VIH SIDA; 2011. 13. Centro Nacional para la prevención y control de VIH SIDA. Vigilancia Epidemiológica de casos de VIH/SIDA en México, Registro Nacional de Casos de SIDA, Actualización al 30 de Septiembre del 2012. Ciudad de México, México: Centro Nacional para la prevención y control de VIH SIDA (CENSIDA); 2012. 14. Comisión Nacional de Protección Social en Salud. Catálogo Universal de Servicios de Salud. In: Secretaría de Salud, ed. Causes 2014. Mexico: Secretaría de Salud; 2014. Available at: http://www.seguro-popular. salud.gob.mx/images/Contenidos/CAUSES%202014.pdf Accessed on April 24, 2014. 15. Área de vigilancia de VIH y conductas de riesgo. Mortalidad por VIH/Sida en España, año 2010. Evolución 1981-2010. Madrid, Spain: Centro Nacional de Epidemiología/Secretaría del Plan Nacional Sobre el Sida; 2012. 16. Garrison LP Jr, Neumann PJ, Erickson P, et al. Using real-world data for coverage and payment decisions: the ISPOR Real-World Data Task Force report. Value Health. 2007;10:326–335. 17. Defunciones 1979 a 2011 (Bases de datos) Dirección General de Información en Salud. 2012. Available at: http://www.sinais.salud.gob. mx/basesdedatos/std_defunciones.html. Accessed March 1, 2012. 18. Organización Mundial de la Salud. Clasificación Internacional de las Enfermedades 9 revisión. Geneva, Switzerland: World Health Organization; 1977. 19. Organización Mundial de la Salud. Clasificación internacional de enfermedades CIE 10. Geneva, Switzerland: World Health Organization; 1992. 20. Consejo Nacional de Población. Proyecciones de la población 1990-2030. 2012. Available at: http://www.portal.conapo.gob.mx/index.php?option=com_ content&view=article&id=125&Itemid=203. Accessed November 2012.
www.jaids.com |
e107
Copyright © 2015 Wolters Kluwer Health, Inc. Unauthorized reproduction of this article is prohibited.
Hernández-Ávila et al
J Acquir Immune Defic Syndr Volume 69, Number 3, July 1, 2015
21. Censo de Población y Vivienda 2010. INEGI; 2010. Available at: http:// www.inegi.org.mx/est/contenidos/proyectos/ccpv/cpv2010/Default.aspx. Accessed January 14, 2014. 22. Estimaciones de población 1990-2012, COLMEX (Base de datos en linea). Sistema Nacional de Información en Salud (SINAIS), Secretaría de Salud; 2010. Available at: http://www.sinais.salud.gob.mx/ basesdedatos/poblacion_colmex.html. Accessed March 27, 2012. 23. Conteo de Población y Vivienda 2005. INEGI; 2005. Available at: http:// www.inegi.org.mx/est/contenidos/proyectos/ccpv/cpv2005/Default.aspx. Accessed January 14, 2014. 24. Consejo Nacional de Población. Indices de Marginación por Estado. 2010. Available at: www.conapo.gob.mx. Accessed April 24, 2014. 25. Ahmad OB-PC, Lopez AD, Murray CJL, et al. Age Standardization of Rates: A New WHO Standard. Geneva, Switzerland: World Health Organization; 2001; GPE Discussion Paper No. 31. 26. Rabe-Hesketh S, Skrondal A. Multilevel and Longitudinal Modeling Using Stata. 3rd ed. College Station, TX: Stata Press Publication; 2012. 27. Greene WH. Econometric Analysis. 7th ed. Boston, MA: Prentice Hall; 2012. 28. Beck N, Katz JN. What to do (and not to do) with time-series crosssection data. Am Polit Sci Rev. 1995;89:634–647. 29. Wooldridge JM. Introductory Econometrics: A Modern Approach. 4th ed. Mason, OH: South Western, Cengage Learning; 2009. 30. Gould WW. Linear Splines and Piecewise Linear Functions. Stata Technical Bulletin. Vol 3: College Station, TX: Stata Press; 1993. 31. Stata Statistical Software: Release 12 [computer program]. College Station, TX: StataCorp LP; 2011. 32. Soler-Claudín C. Acceso universal al Programa de VIH/SIDA de la Ciudad de México: resultados a seis años [in Spanish]. Salud Publica Mex. 2009;51:26–33.
33. Valdespino-Gomez JL, Garcia-Garcia Mde L, del Rio-Zolezzi A, et al. The epidemiology of AIDS/HIV in Mexico: from 1983 to March 1995 [in Spanish]. Salud Publica Mex. 1995;37:556–571. 34. Crabtree-Ramírez B, Caro-Vega Y, Shepherd BE, et al. Cross-sectional analysis of late HAART initiation in Latin America and the Caribbean: late testers and late presenters. PLoS One. 2011;6:e20272. 35. Bautista-Arredondo S, Servan-Mori E, Silverman-Retana O, et al. AIDS mortality in Mexico: exploring the role of supply constraints, management, training and experience (Manuscript under review in SPM. 2014) [in Spanish]. Salud Publica Mex. 36. Perry GE, Maloney WF, Arias OS, et al. Informality: Exit and Exclusion World Bank Latin America and Caribbean Studies. Washington, DC: The World Bank; 2007. 37. Guarnizo-Herreno CC, Agudelo C. Gender-related equity/inequity in gaining access to health services [in Spanish]. Rev Salud Publica (Bogota). 2008;10(suppl):44–57. 38. Rubio-Mendoza ML. Equity in gaining access to health services and equity in funding being attended by them in Bogota [in Spanish]. Rev Salud Publica (Bogota). 2008;10(suppl):29–43. 39. Hernandez-Romieu AC, Del Rio C. Evolution of HIV disease under the universal antiretroviral treatment program in Mexico: analysis of an open cohort, 2004–2011. International HIV Treatment as Prevention Workshop; April 22–25, 2013; Vancouver, Canada. 40. Bautista-Arredondo S, Servan-Mori E, Beynon F, et al. A tale of two epidemics: gender differences in socio-demographic characteristics and sexual risk behaviours in HIV positive individuals in Mexico City (Manuscript under review in Int J Equity Health, 2014). 41. Onraet MA, Gonzalez A, Guerrero-Almeida M, et al. Missed opportunities and late diagnosis in HIV infected Mexican women. J Int AIDS Soc. 2013;16(2 suppl 1):18720.
e108
Copyright © 2015 Wolters Kluwer Health, Inc. All rights reserved.
| www.jaids.com
Copyright © 2015 Wolters Kluwer Health, Inc. Unauthorized reproduction of this article is prohibited.
AND
OPERATIONAL RESEARCH: EPIDEMIOLOGY
AND
PREVENTION
Effect of Universal Access to Antiretroviral Therapy on HIV/AIDS Mortality in Mexico 1990–2011 Juan E. Hernández-Ávila, MsC, ScD,* Lina S. Palacio-Mejía, MA, PhD,* Alfonso Hernández-Romieu, MD, MPH,† Sergio Bautista-Arredondo, MsC,*,** Jaime Sepúlveda Amor, MD, MPH, ScD,‡ and Mauricio Hernández-Ávila, MD, ScD*,††
Introduction: Universal coverage of highly active antiretroviral therapy (HAART) for HIV/AIDS in Mexico was achieved in a staggered fashion. First, HAART was made available in 1997 for employees in the formal sector and subsequently, in 2003, for informal sector workers and the unemployed population. We evaluated the impact of this free universal HAART program on HIV/AIDS mortality in the Mexican population aged 15 years or older.
Methods: Time-series cross-sectional data analysis of standardized HIV/AIDS mortality rates (1990–2011) using marginal linear models with inflection points at 1994, 1997, and 2003. Results: Men employed in the formal sector experienced the greatest effect of access to HAART, with a 32% reduction in mortality from 1997 to 2011 (P , 0.01). For men in this sector, mortality increased from 1990 to 2003, then stabilizing at 7 deaths per 100,000. For women in the informal sector, mortality increased from 1990 to 2011. We found large between-state variability: mortality decreased by 59% in Mexico City while it increased by 245% in Tabasco during the study period. Six of 32 states achieved statistically significant reductions in mortality while 5 states continued to experience increasing mortality rates despite universal HAART. Conclusions: Access to universal HAART has had relative success in Mexico. The heterogeneity of HIV/AIDS mortality by employment status, gender, and state of residence highlight urgent needs to improve health equity with pro-poor and gender-responsive programs and a call for country-specific operational research in HIV/AIDS prevention and treatment. Our results may apply to other countries seeking to implement universal access to antiretroviral therapy. Key Words: highly active antiretroviral therapy, mortality, México (J Acquir Immune Defic Syndr 2015;69:e100–e108) Received for publication August 12, 2014; accepted February 16, 2015. From the *Centro de Información para Decisiones en Salud Pública, Instituto Nacional de Salud Pública, Cuernavaca, México; †Department of Epidemiology, Emory University, Atlanta, GA; ‡Global Health Sciences, University of California (UC) San Francisco, San Francisco, CA, USA and Emeritus Professor of Public Health, National Institute of Public Health, Mexico; **Centro de Investigación en Sistemas de Salud. Instituto Nacional de Salud Pública México; and ††Dirección General. Instituto Nacional de Salud Pública. Presented at the XV Congreso de Investigación en Salud Pública, at the Instituto Nacional de Salud Pública, March 5–8, 2013, Cuernavaca, México. The authors have no funding or conflicts of interest to disclose. Correspondence to: Mauricio Hernández-Ávila, MD, ScD, Instituto Nacional de Salud Pública, Avenida Universidad, #655, Colonia Santa Maria Ahuacatitlán, Cuernavaca, México (e-mail: [email protected]). Copyright © 2015 Wolters Kluwer Health, Inc. All rights reserved.
e100
| www.jaids.com
INTRODUCTION The advent of highly active antiretroviral therapy (HAART) has significantly reduced morbidity and mortality associated with the HIV. According to the United Nations Program on HIV/AIDS, an estimated 6.6 million deaths were averted from 1995 to 2012 worldwide, with 5.5 million deaths averted in low- and middle-income countries.1 Many of these low- and middle-income countries have achieved important declines in HIV-related mortality by developing policies and programs to provide universal access to HAART. In countries like Malawi, Thailand, Brazil, and South Africa,2–9 universal coverage of HAART has had a positive impact on reducing mortality resulting from HIV. In Mexico, universal access to HAART has been available since 2003; its impact on HIV mortality, however, remains understudied. Mexico is a middle-income country with 31 states and a federal district, and a population of 112.3 million. Approximately 170,000 individuals aged 15 years or older are estimated to be living with HIV/AIDS (prevalence 0.2%).10 The HIV epidemic is concentrated: the highest prevalence of HIV has been observed among men who have sex with men, estimated at 17% in 2011.11 Between 2008 and 2012, an estimated 85% of diagnosed HIV-infected individuals who were eligible for treatment were receiving HAART.12,13 The Mexican Health System comprised a set of public and private institutions. Public health care institutions are divided into 2 main groups. On the one hand, the social security institutions (SSIs) provide integral health care and social services to the people employed in the formal economy sector and government employees, as well as their families (39% of the total population in 2012). SSIs are statecontrolled organizations, funded by government, employers, and employees to finance; they organize and provide health care through their human resources and health facilities (SISs health system). Citizens in this system are eligible to receive all medically and hospital necessary services free at the point of delivery. On the other hand, the Ministry of Health & State Health Secretariats managed clinics and hospitals (MoH/SHS health system), which, before 2003, provided health care services to the population working in the informal economy sector, the self-employed, and the unemployed, charging outof-the-pocket recovery fees at the point of delivery. In 2003, the National System for Social Protection in Health, also known as Seguro Popular (SP), was created as a public health
J Acquir Immune Defic Syndr Volume 69, Number 3, July 1, 2015
Copyright © 2015 Wolters Kluwer Health, Inc. Unauthorized reproduction of this article is prohibited.
J Acquir Immune Defic Syndr Volume 69, Number 3, July 1, 2015
insurance program for low-income uninsured Mexicans. SP is funded jointly by federal and state governments to finance a limited health package consisting of 218 health interventions and 17 high-cost diseases,14 including HIV/AIDS. Most services financed by SP are provided by the MoH/SHS through their human resources and health facilities. By 2012, approximately 33% of the Mexican population was enrolled in SP; a remaining 28% of the population, although eligible for SP, was still uninsured and demand services from the private sector or the MoH/SHS systems with an out-of-pocket recovery fee. The implementation of the universal HAART program in Mexico was staggered and dependent on health insurance. Antiretroviral therapy first became available in 1997 for use in population affiliated to the public sector’s SSIs and in 2003 for population affiliated to SP. HIV/AIDS diagnosis on uninsured persons self-selects them to receive treatment through SP and they are therefore affiliated at diagnosis. This adverse selection by disease into SP makes it challenging to study these 2 last population groups separately. HIV/AIDS mortality in Mexico has not decreased as observed in other countries despite offering universal free access to HAART. In Spain, the mortality rate decreased from 14.9 in 1995 to approximately 5 per 100,000 in 1998,15 a 3-fold reduction in only 3 years. The low impact of the universal HAART program on HIV-related mortality in Mexico has raised questions about its implementation. It remains unclear whether differences in HIV mortality maybe associated with the institution providing health care or if there are other between-state variations that maybe contributing to the observed low decline in the rates. Finally, there maybe differences in program outcomes originated by gender inequalities. To address these questions, we assessed the effect of changes in policies regulating access to HAART in Mexico’s HIV/AIDS mortality rate trends from 1990 to 2011 and analyzed differences in mortality by institution providing health care, state of residence, and sex using government health records (“real world data”16).
Universal HAART and AIDS Mortality Mexico
The estimated individual probabilities were subsequently fed to a random number generator with a Bernoulli distribution to obtain estimated health coverage, which was then imputed to observations with missing data. Inclusion criteria were age at death 15 years or older; death occurring between 1990 and 2011 and HIV/AIDS as the underlying cause of death (ICD-9 codes18 279.5 and 279.6, before 1998; ICD-10 codes19 B20–B24).
Population Estimations made by the National Population Council (CONAPO) 1990–2030,20 based on the 2010 census,21 were used as denominators in the calculation of crude and agespecific mortality rates. The proportion of the population with health coverage was estimated using previous population projections,22 based on the second population count in 2005.23 To incorporate socioeconomic and demographic differences, the states and Federal District were characterized by the marginalization index published by the CONAPO,24 which was based on data collected in the 2010 census.21 This multidimensional index allows for differentiating among states based on the global impact of the shortcomings to which the population is exposed; these include lack of access to education, inadequate housing conditions, low income, and those related to residing in very small communities. The index is grouped into 5 categories ranging from very low (best) to very high (worse) marginalization. The standardization of mortality rates was conducted using the 2000–2020 world standard average population, with methods described by World Health Organization.25 For the purposes of this work, we named “population with social security” (PWSS) to people receiving health services from the SSIs. Because of the difficulties caused by the automatic affiliation to SP at the time of diagnosis with HIV/AIDS, we conflated uninsured population with SP affiliates into a group we named “population without social security” (PWOSS).
Statistical Models METHODS Data Sources Information from individual death certificates was obtained from databases provided by the Directorate General for Health Information17 at the MoH and included year of death, underlying cause of death (International Classification of Diseases ICD-coded18,19), age, sex, marital status, education, state of residence (n = 32), and institution providing health care, which we analyzed as a dichotomous variable: “1” indicating coverage by the public SSIs and “0” indicating no health coverage and/or—beginning in 2003—coverage by SP. Data on institution providing health care were missing in 8.1% (n = 7227) of the death certificate records. To address missing values, we fitted a logistic model to estimate the probability that an individual had health coverage by any SSI using age, sex, years of education, and state of residence as predictor variables, which are also found in the death certificate database. Copyright © 2015 Wolters Kluwer Health, Inc. All rights reserved.
The (1990–2011) trend of the HIV/AIDS standardized mortality rate (SMR) in the population aged 15 years and above was studied with time-series cross-sectional data analysis, using marginal linear models.26–29 Such models are used in social sciences to study the effect of policy changes, where observation units are usually states or countries and cross-sectional correlations are expected between them through different measurements in time.26 The analyzed time series was divided into 4 periods according to the availability of antiretroviral treatment and the current access-to-treatment policy in Mexico: (1) 1990–1994, in which there was no treatment available; (2) 1994–1997, when antiretroviral treatment became available; (3) 1997–2003, in which treatment with highly active antiretroviral therapy (HAART) became available and was introduced to PWSS as part of an integral health plan; and (4) 2003–2011, in which the policy for universal access to treatment was expanded to PWOSS (universal HAART). Three models were adjusted independently, with the first at the nationwide level, adjusted for www.jaids.com |
e101
Copyright © 2015 Wolters Kluwer Health, Inc. Unauthorized reproduction of this article is prohibited.
J Acquir Immune Defic Syndr Volume 69, Number 3, July 1, 2015
Hernández-Ávila et al
social security status and sex. The second and third models were adjusted at the statewide level by health coverage status. The natural logarithm of the SMR was used as the outcome variable in all 3 models. To estimate changes in trends in mortality rates over the 4 periods analyzed, we incorporated a linear spline30 in each model, with 4 components, and knots at 1994, 1997, and 2003. The results of the statistical models estimated in this study are presented as the annual percent of change for each study period using the following relation: %Dy ¼ ebp 2 1 · 100; where bp is a regression coefficient or a linear combination of the coefficients. All analyses were performed using Stata software, Stata12 corporation.31
RESULTS Between 1990 and 2011, a total of 90,621 deaths caused by HIV/AIDS were reported. Of these, 98% (89,097) were people aged 15 years or older; 84.5% (75,280) were men, and 15.5% (13,817) were women. More than 67% of the deaths occurred among people aged between 25 and 44 years, with most deaths occurring in the 30–34 age group (20.1%). According to death certificates, 48.8% (43,488) of the deceased were PWOSS and 43.1% were PWSS. Of those missing health insurance coverage data in their death certificates (7227), 41% (2959) were classified as PWSS and 59% (4268) as PWOSS, using the imputation method described. The resulting proportions closely followed the distribution of social security status at the national level. Trends in the mean age at death from HIV/AIDS by sex and social security status are shown in Figure 1. Overall, the
mean age at death in PWSS was higher (37–43 years) compared with PWOSS (33–37 years) and higher among men than women. However, after stratifying by sex and social security status, we observed that women without social security had the lowest mean age at death of all groups. Women with social security fared better than both men and women without social security, and after 2010, they had the highest age of death of all groups. Figure 2 illustrates the SMRs for PWSS and PWOSS from 1990 to 2011; the annual percentages of change in HIV mortality by social security status and sex are shown in Table 1. The death rate from HIV/AIDS in PWSS saw its largest increase from 1990 to 1994, with an annual increase estimated at 24% (Fig. 2, Table 1). From 1994 to 1997 this positive trend was reversed, but the decrease observed was not statistically significant. After 1997, HIV/AIDS SMR in PWSS decreased by 4.3% annually until 2003; it continued to decrease by 1.8% per year until 2011, for a cumulative reduction of 32% during the period between 1997 and 2011 (Fig. 2, Table 1). Among PWOSS, we observed an annual percentage increase in the HIV/AIDS SMR of 17.0% during the period between 1990 and 1994. In contrast to PWSS, the death rate among PWOSS continued to increase by 4.2% annually from 1994 to 1997 and by 2.0% from 1997 to 2003 (Fig. 2, Table 1). Since 2003, when the universalization of HAART began, the annual percentage change in the mortality rate among PWOSS has remained close to zero, stabilizing at around 7.27 deaths per 100,000 people in 2011 (Fig. 2, Table 1). Overall sex-specific SMR curves and those stratified by social security status are shown in Figure 3. The overall annual percent change in the SMR in women did not decrease. In women without social security, we observed a statistically significant increase in HIV/AIDS SMR from
FIGURE 1. Differences in mean age at death by sex and social security status, Mexico 1990–2011. Women have benefited less from access to antiretroviral therapy, as age at death is consistently lower in women than men and the lowest age is in women without social security. This highlights a severe gap in access to HAART in women both with and without social security. Source: Prepared by authors using public data from Mortality database published by General Directorate of Health Information; MoH.17
e102
| www.jaids.com
Copyright © 2015 Wolters Kluwer Health, Inc. All rights reserved.
Copyright © 2015 Wolters Kluwer Health, Inc. Unauthorized reproduction of this article is prohibited.
J Acquir Immune Defic Syndr Volume 69, Number 3, July 1, 2015
Universal HAART and AIDS Mortality Mexico
FIGURE 2. HIV/AIDS observed SMRs and statistical model results, Mexico 1990–2011. The introduction of HAART in the PWSS in 1997 caused a steep decline in the HIV/AIDS mortality in the period 1997–2011, but the universalization policy started in 2003 did not cause the expected decline in the mortality of the PWOSS. Source: Prepared by authors using public data from Mortality database from General Directorate of Health Information; Minister of Health17 and populations estimations made by CONAPO 1990 -203020, population projections COLMEX-DGIS, 1990–2012.22
1990 to 2011 (Table 1). In contrast, among women with social security, an increasing death rate was observed from 1990 to 1994, after which it reached a plateau and remained constant. Among men without social security, the SMR
increased from 1990 to 2003 and reached a plateau from 2003 to 2011; among men with social security, we observed a sharp, statistically significant decrease in the SMR from 1997, which continued until 2011 (Fig. 3, Table 1).
TABLE 1. Annual Percent Change in the Standardized Mortality Rate by Sex and Social Security Status, Mexico 1990–2011 Health Coverage/Economy Sector
Sex
Study Period
Coefficient
Statistical Significance (P)
Annual Percent Change
95% Confidence Interval
Population without social security (informal sector)
Both
1990–1994 1994–1997 1997–2003 2003–2011 1990–1994 1994–1997 1997–2003 2003–2011 1990–1994 1994–1997 1997–2003 2003–2011 1990–1994 1994–1997 1997–2003 2003–2011 1990–1994 1994–1997 1997–2003 2003–2011 1990–1994 1994–1997 1997–2003 2003–2011
0.157 0.041 0.020 0.003 0.090 0.040 0.058 0.016 0.169 0.047 0.016 0.001 0.215 20.022 20.033 20.014 0.111 20.004 0.006 20.004 0.235 20.027 20.036 20.015
,0.001 0.002 ,0.001 0.529 ,0.001 0.032 ,0.001 0.013 ,0.001 ,0.001 0.003 0.763 ,0.001 0.249 ,0.001 0.032 ,0.001 0.846 0.446 0.535 ,0.001 0.178 ,0.001 0.034
17.0 4.2 2.0 0.3 9.4 4.1 6.0 1.6 18.4 4.8 1.6 0.1 24.0 22.1 23.2 21.4 11.8 20.4 0.6 20.4 26.5 22.7 23.5 21.5
14.3 to 19.7 1.6 to 6.9 0.9 to 3.2 20.6 to 1.2 5.8 to 13.2 0.3 to 8.0 4.3 to 7.7 0.3 to 2.9 15.9 to 21.0 2.4 to 7.4 0.5 to 2.7 20.7 to 1.0 20.2 to 28.0 25.7 to 1.5 24.9 to 21.6 22.7 to 20.1 8.3 to 15.4 23.9 to 3.3 21.0 to 2.3 21.6 to 0.9 22.2 to 30.9 26.5 to 1.2 25.3 to 21.8 22.9 to 20.1
Women
Men
Population with social security (formal sector)
Both
Women
Men
Source: Prepared by authors using public data from Mortality database published by the General Directorate of Health Information; MoH17; population estimations made by CONAPO 1990–2030,20 population projections COLMEX–DGIS-1990–2012.22
Copyright © 2015 Wolters Kluwer Health, Inc. All rights reserved.
www.jaids.com |
e103
Copyright © 2015 Wolters Kluwer Health, Inc. Unauthorized reproduction of this article is prohibited.
Hernández-Ávila et al
J Acquir Immune Defic Syndr Volume 69, Number 3, July 1, 2015
FIGURE 3. HIV/AIDS observed SMRs and statistical model results by sex and social security status, Mexico 1990–2011. Although the HIV/AIDS epidemic in Mexico is focalized in men, women have experienced fewer benefits after the introduction of HAART. Even when treatment was available for the PWSS in 1997, the mortality rate of women in this group continued to rise. The slope on the trend in women without social security is even larger. Source: Prepared by authors using public data from Mortality database published by General Directorate of Health Information; MoH17; population estimations made by CONAPO 1990– 2030,20 population projections COLMEX–DGIS-1990–2012.22
We observed large variability in the trend of HIV/ AIDS mortality rates among states. Taking into account the periods when HAART was made accessible to the population with (1997) and without (2003) social security, we grouped states into 3 categories: (1) those which saw a decrease in the HIV/AIDS mortality rate after the introduction of the intervention, (2) those in which there was at least a halt to the rising trend in the mortality rate, and (3) those which saw an upward trend in the mortality rate despite continuing interventions. The SMR in PWSS has declined steadily since 1997 in 5 states with different marginalization indexes: the Federal District (very low), Jalisco (low), Mexico State (low), Morelos (medium), and Puebla (high) (Fig. 4). Querétaro, Colima, and Yucatan showed a downward trend in the SMR in 1997–2003, although the results were not significant after 2003. Coahuila showed reductions in SMR during 1997–2003, but the trend was again positive after 2003. Veracruz, Chiapas, and Tamaulipas exhibited positive annual percentage changes in the SMR from 1997 to 2003; after 2003, the trend stabilized and was not statistically significant. The only states with an
upward trend from 2003 to 2011 were San Luis Potosi (6.17%) and Tabasco (4.75%), both with a high marginalization index. The analysis of the PWOSS demonstrated a reversal of the upward trend in the SMR for HIV/AIDS from 2003 in the states of Aguascalientes (low marginalization index), Federal District (very low), Guanajuato (medium), Guerrero (very high), Jalisco (low), and Puebla (high) (Fig. 4). Our results indicate that the trend in the Federal District in the SMR in PWOSS has been declining since 199432; in the state of Jalisco, the epidemic was contained in the same year. For the remaining 26 states, the annual percentage change was not statistically different from zero during the period 2003–2011. Of particular importance, however, is that in the states of Sonora (low) and Tabasco (high), the trend in the period 2003 to 2011 is still rising with statistical significance (Fig. 4).
e104
Copyright © 2015 Wolters Kluwer Health, Inc. All rights reserved.
| www.jaids.com
DISCUSSION We present SMR trends resulting from HIV/AIDS in the population aged 15 years and older in Mexico during the
Copyright © 2015 Wolters Kluwer Health, Inc. Unauthorized reproduction of this article is prohibited.
J Acquir Immune Defic Syndr Volume 69, Number 3, July 1, 2015
Universal HAART and AIDS Mortality Mexico
FIGURE 4. Annual percent change in the SMR by state and social security status, Mexico 1990–2011. Note: Light-shaded cells indicate a statistically significant increasing annual percent change in the mortality rate. Dark-shaded cells indicate a statistically significant decreasing annual percent change in the mortality rate. Unshaded cells are not statistically significant. Source: Prepared by authors using public data from Mortality database from General Directorate of Health Information; MoH17; population estimations made by CONAPO 1990–2030,20 population projections COLMEX–DGIS-1990–201222; marginalization index estimation by CONAPO.24
period 1990–2011 to determine the effectiveness of the national implementation of universal access policies to HAART. During this period, there were no other major changes in policy or access to health care that could influence HIV/AIDS mortality beside the introduction of HAART in 1997 and its universalization through SP in 2003. Copyright © 2015 Wolters Kluwer Health, Inc. All rights reserved.
Mean age at death in the HIV/AIDS PWSS increased from 37.7 in 1990 to 43.3 in 2011 in women and from 36.5 to 41.7 in men. The mean age at death among those without formal employment (ie, the PWOSS) was lower, increasing from 30.2 to 35.9 in women and from 33.6 to 37.8 in men, between 1990 and 2011. www.jaids.com |
e105
Copyright © 2015 Wolters Kluwer Health, Inc. Unauthorized reproduction of this article is prohibited.
Hernández-Ávila et al
J Acquir Immune Defic Syndr Volume 69, Number 3, July 1, 2015
Our results show important reductions in AIDS mortality associated with the introduction of HAART but with differing levels of effectiveness in the population groups analyzed. First, after the introduction of HAART for PWSS in 1997, a cumulative reduction of 32% was observed in the SMR (Fig. 2); however, its effect was not equal for men and women. In the female PWSS, the trend was not reverted: the SMR was only stabilized around 1.5 per 100,000 (Fig. 3). Following the introduction of HAART for PWOSS as part of the SP implementation in 2003, the SMR trend did not revert: it was only stabilized around 7.27 per 100,000. Again, women benefited less from treatment availability, as the SMR kept its upward trend until 2011 when it reached 2.7 per 100,000. Our results also show important gaps between men and women by type of insurance. The SMR of male PWSS was substantially lower (8 per 100,000) than male PWOSS (12.5) in 2011. The female PWOSS SMR was almost twice as high as their PWSS counterparts in that year. Nonetheless, even with these shortcomings, we estimate that from 1997 to 2011, close to 179,000 deaths were prevented. Our results show wide variations in the impact of the introduction and subsequent universalization of HAART by state. The Federal District, Jalisco, and Puebla were the only states that achieved sustained reductions in the HIV/AIDS SMR in PWSS and PWOSS after the introduction of HAART in the respective populations. The States of Mexico and Morelos achieved sustained reductions in PWSS while Tabasco and San Luis Potosi are as yet unable to stop the SMR increasing trend in PWSS. Guanajuato and Guerrero have achieved reductions in the SMR in PWOSS after introduction of HAART in 2003, but Sonora and Tabasco have not stopped the upward trend of the SMR in this segment of the population. People in formal employment (the PWSS) are in general more financially secure and have achieved higher levels of education. In Mexico, having formal employment is a key marker of one’s socioeconomic position and access to quality medical care. The observed differences in the effect of access to HART in the studied populations maybe indicative of different phenomena. For example, PWSS were more affected during the early days of the HIV/AIDS epidemic in Mexico because of the role of sexual contact with foreign nationals during travels to other countries,33 as they were more likely to travel. Later in the epidemic, the dissimilar mortality trends between employment groups maybe explained by differences in HIV/AIDS medical care between insurance programs: while formal-sector employees receive integrated care provided by infectious diseases specialists that covers HAART, HIV complications, and adequate referrals to specialized levels of care, people in the informal sector receive HAART through HIV Specialized Outpatient Clinics coordinated by general practitioners, with limited coverage for disease complications and ill-defined mechanisms for referral to specialized care. Furthermore, although there is scarcity of data regarding the extent of late initiation of HAART in Mexico, it has been documented that people living with HIV in the informal sector tend to be late testers and late HAART initiators.34 Data from one tertiary care hospital indicate that close to 75% of those with HIV initiated
HAART with a CD4+ count ,200 cells per microliter or with an AIDS-defining illness.34 Although this problem has not been properly documented in PWSS, it could possibly explain the differential mortality trends observed between the 2 population segments. This suggests that coverage of HAART may need to improve not only in the number of people receiving treatment but also in the number of people who start treatment in the early phases of the disease. Other flaws in the cascade of HIV care have not been documented; however, data from patients receiving care through the MoH/SHS health system suggest that serious problems with linkage to care, retention, and adherence exist.35 Additionally, a recent analysis on the variability of mortality trends within the MoH/SHS health system suggests that clinics where health providers have more training and experience, a stronger reference network, and greater ability to make decisions related to the provision of care, tend to have lower mortality rates.35 We cannot overlook certain selection bias by which people who lose their jobs in the formal employment seek medical coverage through SP. This is a strong possibility, given the frequent movement in and out of the formal sector,36 and such bias would increase the mortality rate among the PWOSS and lower mortality rates for the PWSS. Finally, the observed difference in the post-HAART era could also reflect that those in the formal employment sector—those who are more educated—received more information regarding risk factors that promoted earlier testing and initiated treatment earlier in the disease’s progression. Our results indicate that the greatest impact of access to HAART was observed in men with formal-sector employment (Table 1; Fig. 3), where there was an overall decline of 32% between 1997 and 2011. In contrast, for women receiving care by the same providers, only a paused stabilization in the mortality rate was achieved. For men outside the formal employment sector, the rising trend paused and transited into a plateau while a continuing increase in the mortality rate was observed among women in this same sector. This finding raises concern that even with universal access to HAART, mortality rates among women have not shown similar reductions as those among men despite documented evidence of women visiting health services more often37,38 and presenting to care with higher CD4+ levels.34,39 There is no clear explanation for the differences found in the mortality curves between men and women in the postHAART era. It has been documented, however, that women on HIV/AIDS treatment in Mexico City’s MoH health system have poorer results because many women living with HIV are subject to stigma, discrimination, and a higher risk of suffering violence.40 Furthermore, although our results show that despite an overall increase in age at death in both genders, the age of death in women with and without social security exhibits a much larger gap compared to men when stratified by sex and employment sector (Fig. 1). Reportedly, in Mexico, women are infected by a stable partner, and the HIV diagnosis frequently occurs after a partner/child is diagnosed with AIDS while a small proportion were diagnosed by screening during pregnancy.40,41 This suggests that although women may
e106
Copyright © 2015 Wolters Kluwer Health, Inc. All rights reserved.
| www.jaids.com
Copyright © 2015 Wolters Kluwer Health, Inc. Unauthorized reproduction of this article is prohibited.
J Acquir Immune Defic Syndr Volume 69, Number 3, July 1, 2015
become infected at a later age on average than men, they benefit less than men in survival. Mortality trends by state of residence (n = 32) indicate strong heterogeneity in the post-HAART period for both populations with and without social security. The marginalization index of the state of residency was not an important predictor for mortality for PWOSS. However, for PWSS, lower impacts of HAART were documented for states with a high marginalization index, with only 6 states (18.7%) achieving statistically significant reductions in mortality following universalization of HAART. For PWSS, only 5 states (15.6%) have achieved sustained reduction in mortality from 1997, the year that HAART became available to this population. These results suggest that state programs for prevention and control of HIV/AIDS should be reviewed to identify areas of opportunity to improve and achieve results similar to those that have sustained reduction in mortality. The most important next step is to boost the HIV/AIDS operational research agenda to identify good practices that help achieve consistent results throughout the country. The focus should be directed toward early diagnosis and earlier initiation of HAART, to improve quality of care to increase survival rates and decrease AIDS complications, and to reduce stigma and discrimination to improve early access to testing and treatment. The states of Tabasco, Sonora, San Luis Potosi, Veracruz, and Quintana Roo require particular attention because they show the largest gap compared to the national results. The Federal District is the only state that has reversed the upward trend in mortality in the PWOSS since 1997: this maybe attributed to resources invested by the local government to increase early diagnosis and treatment and the specialized health care offered by the SSIs and the MoH in this area. The fact that most states have succeeded in stabilizing mortality makes evident the effect of universal access; however, without proper monitoring and evaluation at the state level, we may not adequately understand the current gaps in health services. There are limitations in the information systems used in this study. In the case of death registration, potential classification and coding errors in the cause of death, and failure to record social security status, maybe present. With these errors, the possibility of underestimation exists. Moreover, the difficulty in establishing the insured population (denominators) according to the various SSIs is a challenge that could also affect our results. These, however, are random errors that would make detected associations less apparent. In our data, we cannot differentiate whether the observed increase in the mean age at death is an indicator that access to HAART is prolonging the lives of people with HIV/AIDS or that they are becoming infected at older ages, as a combined result of behavioral changes and a reduced community viral load. In conclusion, type of employment, gender, age, and state of residence are social determinants that contribute to HIV/AIDS mortality in Mexico. Although the legislation of universal access to antiretroviral therapy has had relative success in Mexico, the heterogeneity and gaps in HIV/AIDS mortality highlight the need to boost country-specific operational research in HIV/AIDS prevention and treatment. Our findings may apply to other low- and middle-income countries seeking to maximize the effects of universal access to antiretroviral therapy. Copyright © 2015 Wolters Kluwer Health, Inc. All rights reserved.
Universal HAART and AIDS Mortality Mexico
ACKNOWLEDGMENTS The authors acknowledge the General Directorate for Information in Health for making the 1990–2011 mortality dataset available. REFERENCES 1. Joint United Nations Programme on HIV/AIDS (UNAIDS). Geneva, Switzerland: UNAIDS/World Health Organization; 2013. 2. Badri M, Maartens G, Mandalia S, et al. Cost-effectiveness of highly active antiretroviral therapy in South Africa. PLoS Med. 2006;3:e4. 3. Chasombat S, McConnell MS, Siangphoe U, et al. National expansion of antiretroviral treatment in Thailand, 2000-2007: program scale-up and patient outcomes. J Acquir Immune Defic Syndr. 2009;50:506–512. 4. Ferradini L, Jeannin A, Pinoges L, et al. Scaling up of highly active antiretroviral therapy in a rural district of Malawi: an effectiveness assessment. Lancet. 2006;367:1335–1342. 5. Todd J, Glynn JR, Marston M, et al. Time from HIV seroconversion to death: a collaborative analysis of eight studies in six low and middleincome countries before highly active antiretroviral therapy. AIDS. 2007; 21(suppl 6):S55–S63. 6. Teixeira PR, Vitoria MA, Barcarolo J. Antiretroviral treatment in resource-poor settings: the Brazilian experience. AIDS. 2004;18(suppl 3): S5–S7. 7. Marins JR, Jamal LF, Chen SY, et al. Dramatic improvement in survival among adult Brazilian AIDS patients. AIDS. 2003;17:1675–1682. 8. World Health Organization. Global HIV/AIDS response—epidemic update and health sector progress towards Universal Access. In: WHO, ed. Global HIV/AIDS Response. Geneva, Switzerland: WHO; 2011:233. 9. Alonso Gonzalez M. Antiretroviral Treatment in the Spotlight: A Public Health Analysis in Latin America and the Caribbean 2013. Washington, DC: PAHO; 2013. 10. Gutiérrez JP, Sucilla-Pérez H, Conde-González CJ, et al. Seroprevalencia de VIH en población mexicana de entre 15 y 49 años: resultados de la Ensanut 2012. Salud Publica Mex. 2014;56:323–332. 11. Bautista-Arredondo S, Colchero MA, Romero M, et al. Is the HIV epidemic stable among MSM in Mexico? HIV prevalence and risk behavior results from a nationally representative survey among men who have sex with men. PLoS One. 2013;8:e72616. 12. Centro Nacional para la prevención y control de VIH SIDA. El VIH/SIDA en México 2011. Numeralia epidemiológica. México: Centro Nacional para la prevención y control de VIH SIDA; 2011. 13. Centro Nacional para la prevención y control de VIH SIDA. Vigilancia Epidemiológica de casos de VIH/SIDA en México, Registro Nacional de Casos de SIDA, Actualización al 30 de Septiembre del 2012. Ciudad de México, México: Centro Nacional para la prevención y control de VIH SIDA (CENSIDA); 2012. 14. Comisión Nacional de Protección Social en Salud. Catálogo Universal de Servicios de Salud. In: Secretaría de Salud, ed. Causes 2014. Mexico: Secretaría de Salud; 2014. Available at: http://www.seguro-popular. salud.gob.mx/images/Contenidos/CAUSES%202014.pdf Accessed on April 24, 2014. 15. Área de vigilancia de VIH y conductas de riesgo. Mortalidad por VIH/Sida en España, año 2010. Evolución 1981-2010. Madrid, Spain: Centro Nacional de Epidemiología/Secretaría del Plan Nacional Sobre el Sida; 2012. 16. Garrison LP Jr, Neumann PJ, Erickson P, et al. Using real-world data for coverage and payment decisions: the ISPOR Real-World Data Task Force report. Value Health. 2007;10:326–335. 17. Defunciones 1979 a 2011 (Bases de datos) Dirección General de Información en Salud. 2012. Available at: http://www.sinais.salud.gob. mx/basesdedatos/std_defunciones.html. Accessed March 1, 2012. 18. Organización Mundial de la Salud. Clasificación Internacional de las Enfermedades 9 revisión. Geneva, Switzerland: World Health Organization; 1977. 19. Organización Mundial de la Salud. Clasificación internacional de enfermedades CIE 10. Geneva, Switzerland: World Health Organization; 1992. 20. Consejo Nacional de Población. Proyecciones de la población 1990-2030. 2012. Available at: http://www.portal.conapo.gob.mx/index.php?option=com_ content&view=article&id=125&Itemid=203. Accessed November 2012.
www.jaids.com |
e107
Copyright © 2015 Wolters Kluwer Health, Inc. Unauthorized reproduction of this article is prohibited.
Hernández-Ávila et al
J Acquir Immune Defic Syndr Volume 69, Number 3, July 1, 2015
21. Censo de Población y Vivienda 2010. INEGI; 2010. Available at: http:// www.inegi.org.mx/est/contenidos/proyectos/ccpv/cpv2010/Default.aspx. Accessed January 14, 2014. 22. Estimaciones de población 1990-2012, COLMEX (Base de datos en linea). Sistema Nacional de Información en Salud (SINAIS), Secretaría de Salud; 2010. Available at: http://www.sinais.salud.gob.mx/ basesdedatos/poblacion_colmex.html. Accessed March 27, 2012. 23. Conteo de Población y Vivienda 2005. INEGI; 2005. Available at: http:// www.inegi.org.mx/est/contenidos/proyectos/ccpv/cpv2005/Default.aspx. Accessed January 14, 2014. 24. Consejo Nacional de Población. Indices de Marginación por Estado. 2010. Available at: www.conapo.gob.mx. Accessed April 24, 2014. 25. Ahmad OB-PC, Lopez AD, Murray CJL, et al. Age Standardization of Rates: A New WHO Standard. Geneva, Switzerland: World Health Organization; 2001; GPE Discussion Paper No. 31. 26. Rabe-Hesketh S, Skrondal A. Multilevel and Longitudinal Modeling Using Stata. 3rd ed. College Station, TX: Stata Press Publication; 2012. 27. Greene WH. Econometric Analysis. 7th ed. Boston, MA: Prentice Hall; 2012. 28. Beck N, Katz JN. What to do (and not to do) with time-series crosssection data. Am Polit Sci Rev. 1995;89:634–647. 29. Wooldridge JM. Introductory Econometrics: A Modern Approach. 4th ed. Mason, OH: South Western, Cengage Learning; 2009. 30. Gould WW. Linear Splines and Piecewise Linear Functions. Stata Technical Bulletin. Vol 3: College Station, TX: Stata Press; 1993. 31. Stata Statistical Software: Release 12 [computer program]. College Station, TX: StataCorp LP; 2011. 32. Soler-Claudín C. Acceso universal al Programa de VIH/SIDA de la Ciudad de México: resultados a seis años [in Spanish]. Salud Publica Mex. 2009;51:26–33.
33. Valdespino-Gomez JL, Garcia-Garcia Mde L, del Rio-Zolezzi A, et al. The epidemiology of AIDS/HIV in Mexico: from 1983 to March 1995 [in Spanish]. Salud Publica Mex. 1995;37:556–571. 34. Crabtree-Ramírez B, Caro-Vega Y, Shepherd BE, et al. Cross-sectional analysis of late HAART initiation in Latin America and the Caribbean: late testers and late presenters. PLoS One. 2011;6:e20272. 35. Bautista-Arredondo S, Servan-Mori E, Silverman-Retana O, et al. AIDS mortality in Mexico: exploring the role of supply constraints, management, training and experience (Manuscript under review in SPM. 2014) [in Spanish]. Salud Publica Mex. 36. Perry GE, Maloney WF, Arias OS, et al. Informality: Exit and Exclusion World Bank Latin America and Caribbean Studies. Washington, DC: The World Bank; 2007. 37. Guarnizo-Herreno CC, Agudelo C. Gender-related equity/inequity in gaining access to health services [in Spanish]. Rev Salud Publica (Bogota). 2008;10(suppl):44–57. 38. Rubio-Mendoza ML. Equity in gaining access to health services and equity in funding being attended by them in Bogota [in Spanish]. Rev Salud Publica (Bogota). 2008;10(suppl):29–43. 39. Hernandez-Romieu AC, Del Rio C. Evolution of HIV disease under the universal antiretroviral treatment program in Mexico: analysis of an open cohort, 2004–2011. International HIV Treatment as Prevention Workshop; April 22–25, 2013; Vancouver, Canada. 40. Bautista-Arredondo S, Servan-Mori E, Beynon F, et al. A tale of two epidemics: gender differences in socio-demographic characteristics and sexual risk behaviours in HIV positive individuals in Mexico City (Manuscript under review in Int J Equity Health, 2014). 41. Onraet MA, Gonzalez A, Guerrero-Almeida M, et al. Missed opportunities and late diagnosis in HIV infected Mexican women. J Int AIDS Soc. 2013;16(2 suppl 1):18720.
e108
Copyright © 2015 Wolters Kluwer Health, Inc. All rights reserved.
| www.jaids.com
Copyright © 2015 Wolters Kluwer Health, Inc. Unauthorized reproduction of this article is prohibited.
