Intended for healthcare professionals

CCBYNC Open access
Research

# Intervention strategies to reduce the burden of non-communicable diseases in Mexico: cost effectiveness analysis

BMJ 2012; 344 (Published 02 March 2012) Cite this as: BMJ 2012;344:e355
1. Joshua A Salomon, associate professor of international health1,
2. Natalie Carvalho, doctoral student2,
3. Cristina Gutiérrez-Delgado, deputy director general of economics and health3,
4. Ricardo Orozco, analyst4,
5. Anna Mancuso, clinical social worker5,
6. Daniel R Hogan, postdoctoral fellow6,
7. Diana Lee, doctoral student7,
8. Yuki Murakami, health economist/policy analyst8,
9. Lakshmi Sridharan, resident in internal medicine9,
10. María Elena Medina-Mora, director general4,
11. Eduardo González-Pier, director of finance10
1. 1Department of Global Health and Population, Harvard School of Public Health, 665 Huntington Avenue, Boston, MA 02115, USA
2. 2Harvard University, Cambridge, MA
3. 3Unidad de Análisis Económico, Secretaría de Salud, México DF, Mexico
4. 4Instituto Nacional de Psiquiatría Ramón de la Fuente Muñiz, México DF
5. 5Boston Center for Refugee Health and Human Rights, Boston Medical Center, Boston, MA
6. 6Harvard School of Public Health, Boston, MA
7. 7University of California, Berkeley, CA
8. 8Organisation for Economic Co-operation and Development, Paris, France
9. 9University of California, San Francisco, CA
10. 10Instituto Mexicano del Seguro Social, México DF
1. Correspondence to: J A Salomon jsalomon{at}hsph.harvard.edu
• Accepted 6 November 2011

## Abstract

Objective To inform decision making regarding intervention strategies against non-communicable diseases in Mexico, in the context of health reform.

Design Cost effectiveness analysis based on epidemiological modelling.

Interventions 101 intervention strategies relating to nine major clusters of non-communicable disease: depression, heavy alcohol use, tobacco use, cataracts, breast cancer, cervical cancer, chronic obstructive pulmonary disease, cardiovascular disease, and diabetes.

Data sources Mexican data sources were used for most key input parameters, including administrative registries; disease burden and population estimates; household surveys; and drug price databases. These sources were supplemented as needed with estimates for Mexico from the WHO-CHOICE unit cost database or with estimates extrapolated from the published literature.

### Data sources

We used Mexican data sources when possible, drawing extensively on four main local sources: administrative registries, population estimates, household surveys, and drug cost databases.

Information on age specific mortality and causes of death are compiled from death certificates by the Ministry of Health and the Instituto Nacional de Estadística y Geografía. Additionally, public institutions providing hospital services maintain discharge registries, which include patient level information on sociodemographic characteristics, causes of admission to hospital classified by ICD-10 codes (international classification of diseases, 10th revision), diagnostic and treatment procedures coded according to ICD-9-CM, result of procedures, mortality in hospital, reasons for discharge, number of hospital bed days, and insurance status. For the present analyses, we used mortality data from the year 2004. We used discharge registries from Ministry of Health hospitals for the years 2000–5 and from hospitals affiliated with the Instituto Mexicano del Seguro Social (IMSS) for the years 2004 and 2005.

#### Population estimates

Population estimates for 2000–5 were provided by the Consejo Nacional de Población (CONAPO), which develops yearly projections of population numbers by age, sex, state, and insurance status. Projections are made based on official surveys including the National Survey on Fertility and Health, censuses and special surveys relating to migration.

#### Population household surveys

The Encuesta Nacional de Salud y Nutrición 2005–6 (ENSANut) was a nationally and state representative survey that sampled 47 695 households and 206 700 individuals. Modules in ENSANut included information on household characteristics, health insurance, risk factors (smoking and alcohol use), health states, and use of services. Biomarkers were collected on concentrations of cholesterol, plasma glucose, and haemoglobin A1c and blood pressure. ENSANut 2005–6 was used in this analysis primarily to measure current coverage of selected interventions.

#### Drug cost database

Public institutions providing healthcare in Mexico are required to purchase only drugs included in the Cuadro Básico de Medicamentos (Mexican Positive List, MPL). Public purchasing regulations require public bidding for multiple source drugs, although there is no general consolidated purchase system for the whole public health sector, as all institutions are essentially independent. We used the 2005 purchase price database available from IMSS, which is the largest public purchaser in Mexico.

### Estimation of intervention effects

#### Epidemiological estimates

For most interventions, current epidemiology of relevant diseases or injuries was defined based on results from the Mexican burden of disease analysis,6 expressed in terms of age and sex specific incidence, prevalence, case fatality, remission, and mortality. Vital registration data for 2004 were used to estimate mortality by age, sex, and cause. Total mortality figures by age and sex were adjusted with standard demographic techniques to account for under-recording of deaths at certain ages, misreporting of age on the death certificate, and migration.15 16 Estimates for causes of death were adjusted following standard algorithms for redistributing deaths coded to “ill defined” disease or injury categories, cancers of unknown sites, and cardiovascular disease, as well as miscoding of diabetes to cardiovascular or other chronic diseases.6 17 18 19 Estimates of the incidence of different diseases and their relevant sequelae by age and sex were derived from a combination of sources and imputation approaches applied to different groups of causes.6

#### Intervention definitions

The selection and specification of interventions, in cases where a WHO-CHOICE regional analysis was available, corresponded closely to the interventions defined in the prior analysis. In cases where no WHO-CHOICE regional analysis was available, the choice of interventions was guided by consultation with experts in Mexico or by existing norms for clinical practice in Mexico. Definition of interventions, in general, was as explicit as possible and specified both immediate components of intervention delivery as well as the types of actions that would be undertaken in response to downstream consequences of the disease process or of the intervention itself.

#### Intervention effectiveness

The definition of the current epidemiology, combined with information on current coverage and effectiveness of interventions, served as the analytic starting point for defining all intervention scenarios, including the null scenario. Following the standard WHO-CHOICE approach, we derived intervention effectiveness from meta-analyses and systematic reviews where these were available. For interventions that were examined previously in WHO-CHOICE regional analyses, we maintained consistent assumptions about intervention efficacy unless there was sufficient evidence to suggest different outcomes in Mexico. For incorporation into the population model described below, intervention effectiveness was translated into changes in relevant disease model parameters. For example, the effects of primary prevention were expressed as percent reductions in age and sex specific incidence rates from a particular condition, while treatment interventions were allowed to affect transitions to other disease states, remission rates, case fatality rates, or health state valuations. Details on the parameters of effectiveness for specific intervention analyses are provided in the technical appendix on bmj.com, including a full listing of data sources on effectiveness.

#### Population health outcomes

Population health outcomes under different intervention scenarios were modelled with the multistate population model PopMod20 (see general appendix on bmj.com) or analogous tools developed for this project where more detailed disease models were required. Interventions were compared with a null scenario, which was simulated by altering the baseline epidemiological parameters to remove the estimated effects of current intervention coverage. For intervention scenarios, effectiveness estimates were adjusted to account for target population coverage levels and provider and patient adherence to interventions. Each intervention scenario assumed implementation of the intervention for a 10 year period, but the population model captured all effects over a 100 year time horizon. For both the null and intervention scenarios, the inputs to the population model included rates of incidence, remission, and case fatality; estimates of prevalence; and health state valuations for relevant outcomes. Outputs from the model included estimates of the residence time in each disease state and total number of healthy life years lived by age, sex, and calendar year. These outputs were used to calculate intervention effectiveness in terms of changes in aggregate population level health outcomes. Where alternatives to PopMod were used for specific analyses, these alternative modelling approaches are described in the technical appendix on bmj.com.

### Estimation of intervention costs

We used an ingredients approach to costing, by which the quantities of inputs that are used in delivering a particular service or intervention are multiplied by their unit prices to obtain total costs. Following the standard WHO-CHOICE approach, we considered three broad categories of costs: patient costs, programme costs, and training costs.

#### Patient costs

Patient level costs included hospital bed days, hospital visits, health centre visits, ancillary care, laboratory and diagnostic tests, drugs, and other costs related to specific interventions. Resource quantities were derived through review of the published literature, and from practice guidelines. Drug prices were taken from the IMSS purchase price database. Unit prices for non-traded goods, including patient services such as hospital bed days, were standardised across intervention analyses and derived from the WHO-CHOICE price database, which includes country specific estimates based on an econometric analysis of multinational datasets on costs.21

#### Programme costs

We considered several key categories of programme activities. Basic administration includes planning and overhead costs, in addition to staff required to effectively monitor, evaluate, and supervise the programme. These costs depend on whether the intervention requires legislation, the level at which administration is required, and the complexity of monitoring and evaluation needed. Other categories of programme costs relevant to certain interventions included media campaigns, other information, education or communication activities, and law enforcement. For most interventions, we adopted estimates of resource use and prices for programme cost components from previous WHO-CHOICE regional analyses. Details on assumptions of programme costs for specific analyses are in the technical appendix (see bmj.com).

#### Training costs

Training costs, relevant for some interventions, depend on the length of training required, the number of supervisory visits needed a year, and the capacity for a single training session. All interventions in this study that required training costs were linked to previous WHO-CHOICE regional analyses, and we maintained the assumptions from these previous analyses.

### Estimation of cost effectiveness

We computed total costs for a given intervention as the sum of all patient, programme, and training costs. The null scenario by definition includes no costs, so costs of all other interventions can be interpreted as being incremental on the null. The total health benefits of an intervention were computed by comparing the number of healthy life years lived in the population in a particular intervention scenario with the total number of healthy life years lived in the population under the null scenario. In the base case, we used discounted age weighted DALYs as the unit of account for healthy life years, but we also conducted a sensitivity analysis comparing these results with those without age weighting.

For all interventions, we report annualised quantities for both total costs and total benefits. For total costs, these were derived simply by dividing the costs over the 10 year intervention period by 10. For total benefits, these were derived by taking the full difference in health effects over the 100 year modelled period and dividing this by 10. In this way, the annualised costs and benefits can be interpreted as the costs and benefits associated with a single year of intervention. We report average cost effectiveness ratios for interventions, ordered by increasing overall effects. Average cost effectiveness ratios are interpretable as the net costs per unit of net benefit associated with delivering the intervention, compared with doing nothing. We also report incremental cost effectiveness ratios where these measures are relevant—that is, in evaluating mutually exclusive interventions that represent competing choices. Incremental cost effectiveness ratios were computed for an intervention with respect to the next most effective alternative after eliminating strategies that were dominated (that is, those that were more costly and less effective than other options) or those that were weakly dominated (that is, had higher cost effectiveness ratios than more effective options).

Cervical cancer—Analyses for cervical cancer, similarly to the breast cancer analyses, evaluated strategies for treatment without screening or treatment combined with routine population screening according to the Mexican norm. The overall findings for cervical cancer interventions mirrored those for breast cancer, with treatment at earlier stages (including the precancerous stage of cervical intraepithelial neoplasia, grade II or III) contributing greater benefits than treatment at later stages. Even the relatively more costly and less effective treatments of later stages of cancer were found to be cost effective components of the overall treatment strategy, based on falling below the threshold of per capita GDP for each DALY averted. For cervical cancer, screening was among the most efficient strategies, increasing overall benefits more than 10-fold. The incremental cost effectiveness ratio for screening and treatment, compared with treatment without screening, was around $Int5600 per DALY averted, which implied that screening would be highly cost effective. Chronic obstructive pulmonary disease—Current interventions for chronic obstructive pulmonary disease (COPD) are aimed at slowing the progression of the decline in lung function associated with the disease. We evaluated five main interventions: intensive smoking cessation programme for current smokers with a diagnosis of COPD; influenza vaccination of COPD patients aged ≥65; inhaled bronchodilator for those with mild COPD; inhaled bronchodilator and corticosteroid for those with moderate to severe COPD; long term oxygen treatment (in addition to bronchodilator and corticosteroid) for those with severe COPD; and treatment of severe COPD exacerbations. We found that treatment of severe exacerbations associated with COPD averted the smallest number of DALYs, followed by long term oxygen treatment, inhaled bronchodilator for mild COPD, and then an inhaled bronchodilator plus inhaled corticosteroid for moderate to severe COPD. Influenza vaccine for people with COPD and an intensive smoking cessation programme for those diagnosed with COPD had the largest benefits in terms of DALYs averted and were less expensive than interventions directed solely at patients with later stages of disease. These were also the only two interventions with average cost effectiveness ratios below three times GDP per capita in Mexico (both having ratios between$Int2500 and $Int5000). Given the focus of the different interventions on different target populations, we did not conduct an incremental analysis treating the interventions as competing choices. Cardiovascular disease—We evaluated a wide range of interventions for primary prevention, treatment, and secondary prevention for cardiovascular disease. Prevention interventions included non-personal interventions involving health education through mass media programmes, legislation or voluntary agreements with the food industry, as well as personal health service interventions including detection and treatment of high risk individuals based on blood pressure, serum cholesterol, and absolute risk thresholds. The absolute risk approach estimates the combined risk of a cardiovascular event over the next decade above a given threshold, based on relative risk estimates of modelled risk factors. We also assessed 30 single interventions and combinations of interventions for treatment and secondary prevention relating to acute myocardial infarction, stroke, and congestive heart failure. All primary prevention interventions were found to be highly cost effective according to international benchmarks. Among population level (non-personal) interventions, those aimed at reducing blood pressure and cholesterol, and salt reduction through legislation, had the lowest costs per DALY. Individual primary prevention interventions resulted in much greater effectiveness, although they were more costly per unit of health benefit than the population-wide strategies. In incremental analyses, population salt reduction, a comprehensive population combination intervention, and absolute risk threshold approaches dominated all individual strategies focusing on either hypertension or cholesterol alone. The most aggressive strategy based on absolute risk thresholds, focusing on all patients with risks above 5%, was the most effective but at an incremental cost effectiveness ratio that was well above the benchmark of three times GDP per capita. All treatment interventions for myocardial infarction were found to be highly cost effective or cost effective in comparing average cost effectiveness ratios to GDP based benchmarks, with cardiac rehabilitation producing the most attractive cost effectiveness ratios because of the relatively low cost of this intervention combined with moderate population level health effects. For stroke, only post-acute stroke interventions were found to be highly cost effective, while interventions targeting the acute period resulted in low health gains at significantly higher costs. All heart failure interventions were highly cost effective, with diuretics being the most cost effective in the group. Because the set of interventions were not mutually exclusive, we have not reported on a full set of incremental comparisons for the cluster of interventions relating to cardiovascular disease. Diabetes—We evaluated four main interventions for secondary prevention of type 2 diabetes: blood pressure control, lipid control, and conventional or intensive glycaemic control. With a high overall prevalence of diabetes in Mexico, the largest population benefits would be realised through glycaemic control. The average cost effectiveness ratios for conventional and intensive glycaemic control were about$Int12 500 and $Int13 600 per DALY averted, respectively, implying that glycaemic control is potentially cost effective at less than two times per capita GDP. Considering the incremental cost effectiveness of intensive compared with conventional glycaemic control, the more intensive strategy had an incremental ratio of around$Int16 900. Lipid control for patients with diabetes would produce a much smaller overall population benefit, but at an average cost effectiveness ratio similar to that for glycaemic control. Blood pressure control would produce an overall benefit between that for lipid control and that for glycaemic control. The cost effectiveness ratio for blood pressure control, however, was considerably lower than those for the other two interventions. At around $Int8500 per DALY averted, blood pressure control would be classified as highly cost effective according to standard international benchmarks. ### Sensitivity of results to choice of health metric In our base case analysis, we have focused on measuring population health effects of interventions using DALYs that reflect differential age weights, such that health outcomes experienced during young adulthood are weighted more heavily than those during childhood or older adult years. We examined the sensitivity of results to the inclusion of age weights by recalculating all health effects and cost effectiveness ratios with equal weights at all ages. The appendix figure on bmj.com compares the yearly population health effects for all 101 intervention strategies under scenarios with and without age weights, indicating a high degree of consistency. Examining the average cost effectiveness ratios in the scenario without age weighting (shown for all interventions in the appendix table on bmj.com), we found that only two interventions moved from one broad category (that is, highly cost effective, potentially cost effective, not cost effective) to another. These were nicotine replacement therapy, which shifted from not cost effective to potentially cost effective with the removal of age weights, and angiotensin converting enzyme (ACE) inhibitors for acute myocardial infarction, which shifted from potentially cost effective to highly cost effective; both of these shifts were based on relatively small changes in the ratios. Overall, this analysis indicated that the results in general were highly robust to the choice of age weighted DALYs averted as the primary indicator of population health benefits. ## Discussion In this study we estimated the population health impacts and costs for various interventions targeting non-communicable diseases in Mexico. As in the companion papers in this series presenting analyses for two world regions,22 23 24 25 26 27 28 we found wide variation across interventions in both costs and effects; yet we also observed that, in every major disease cluster we examined, at least some of the interventions provided excellent value for money. At the high value end of the spectrum, several interventions cost <$Int250 for each year of life they added to the population—including interventions aimed at heavy alcohol use, tobacco use, cataracts, and cardiovascular disease risks. At a small fraction of the per capita income in Mexico, such gains are considered to be exceptionally good value, based on standard benchmarks for cost effectiveness analysis.

Economic evaluations in middle income countries like Mexico often require extrapolation of evidence and assumptions from other countries or from regional databases.29 In our analysis, the availability of local epidemiological and economic information, and explicit consideration of local norms for clinical and public health practice, helped to reveal limitations in the transferability of conclusions from one setting to another and to develop national estimates of cost effectiveness for selected interventions. For example, distinct features of the screening norm in force for breast cancer in Mexico have important implications for analysis of the economic efficiency of different intervention strategies. We found that the aggressive screening policy in place in Mexico for early detection of breast cancer had an estimated cost effectiveness ratio between one and three times GDP per capita, despite guidelines in other countries such as the US that discourage testing before age 50.30 As further liberalisation of Mexican norms for early breast cancer screening since July 2011 begins to shape practice, information on costs and expected health effects of wider screening coverage is essential in preparing for the expected increase in demand for screening services. Another example of the importance of incorporating local information in economic evaluation relates to variation in costs of interventions. For interventions with intensive use of drugs, differences across settings in purchasing practices can have an important effect on the costs of these interventions.

Analyses of the cost effectiveness of an array of interventions across categories ranging from community services to high intensity clinical services offers insights that in some cases contradict the conventional wisdom on value for money within broad categories of interventions. It is not the case that all community and public health interventions have extremely low cost effectiveness ratios; neither is it true that high complexity interventions are universally expensive in relation to the health gains they provide. Cataract surgery seems to be among the best buys in health interventions in Mexico, less expensive per unit of health benefit than increasing taxation of tobacco products, itself a highly cost effective intervention at a little more than \$Int100 per healthy year gained. Likewise, other clinical services such as antidepressants and some interventions for secondary prevention of cardiovascular disease have cost effectiveness ratios that are well below the threshold of GDP per capita for each DALY averted. Thus, where conventional wisdom on value for money in public health and medicine fails to provide an adequate basis for intervention choice, the development of rigorous information on costs and health benefits from a wide range of intervention strategies offers essential information to planners and policy makers in resource constrained settings.

Information on the cost effectiveness of interventions provides a formal basis for evaluating efficiency of allocated resources, to maximise overall population health gains under resource constraints. As such, cost effectiveness is only one consideration among a range of important criteria in setting priorities for health interventions or designing and evaluating healthcare reform. In addition to attaining high overall gains in population health, decision makers in Mexico, as elsewhere, are also concerned with ensuring that benefits are distributed equitably across the population. Much has been written about concerns for fairness with respect to the Mexican reform, and these concerns were central to the design of the reform and to its subsequent critical evaluation.31 32 33

Another limitation is that the analyses presented here, while covering a broad scope of different interventions, inevitably omit some interventions that are relevant to clinical practice and health policy. For example, although we report on secondary prevention for managing risk factors in diabetes, we have not evaluated the range of available strategies relating to screening and treatment for microvascular complications of diabetes. On the other hand, we included in our analyses a strategy of intensive glycaemic control for diabetes, even as recently revised guidelines have prompted debate over the pros and cons of such a strategy.34 35 As another example, we report here on two types of surgery for cataracts used widely in Mexico—phacoemulsification and conventional extracapsular cataract extraction—but we have not yet evaluated manual small incision cataract surgery. As clinical practice guidelines constitute a moving target, and as new information emerges on health outcomes and resource requirements associated with interventions against chronic diseases and other health challenges, cost effectiveness analyses require periodic updating and revision to incorporate the best currently available evidence and to answer the most urgent policy questions.

We also recognise that this study—in presenting an expansive overview on the range of analyses undertaken across disease clusters—has summarised results relatively parsimoniously, without exhaustive details on uncertainty around each point estimate or sensitivity of each result to particular parameter values and assumptions. Specific choices with regard to technology adoption or detailed practice guidelines demand more precision than what is offered here, along with a more comprehensive characterisation of key uncertainties. The intent in this study, on the other hand, was to offer a broad perspective on the comparative costs and health impacts of a wide array of different interventions and strategies, in view of high level decision making in the context of health reform. To date, the results from this study have provided a methodological foundation for further work on cost effectiveness analysis by interdisciplinary teams within the Mexican Ministry of Health, at the same time that the findings have informed decision making around provision and financial coverage for several interventions and provided an evidence base for other public health policies. Specific examples include design of national programmes such as the breast cancer and cervical cancer screening, diagnosis, and treatment programmes, and national legislation such as the General Law on Tobacco Control. These results have also been referenced in discussions around accelerating or delaying coverage of specific interventions in SSPH.

Over the past 50 years, the epidemiological transition in Mexico has produced a dramatic rise in the importance of non-communicable diseases. Between 1955 and 2005, the proportion of all deaths from non-communicable diseases increased from 23% to 75%.5 This rise has presented new challenges to a health system that traditionally prioritised programmes for communicable disease and reproductive health. As major reforms in Mexico have sought to extend healthcare coverage to the substantial fraction of the population lacking insurance, and to develop new financial mechanisms to protect families from catastrophic health spending, evidence on the costs and health benefits associated with different types of interventions has been—and will continue to be—an essential input to the development of effective, efficient, and fair policies.

#### What is already known on this topic

• The advanced epidemiological transition in Mexico has produced a large and growing burden of non-communicable diseases

• Health reforms since 2003 in Mexico have dramatically expanded insurance coverage in the population, prompting difficult policy choices on which services to provide to beneficiaries of the new social insurance programmes

• An analysis of the cost effectiveness of 101 interventions targeting nine clusters of non-communicable diseases in Mexico showed that within each cluster there was at least one intervention that was considered to provide excellent value for money based on conventional benchmarks

• High value can be found even among highly complex clinical services; conversely, public health strategies are not guaranteed to be highly cost effective

## Notes

Cite this as: BMJ 2012;344:e355

## Footnotes

• doi: 10.1136/bmj.e586
• , doi: 10.1136/bmj.e615
• , doi: 10.1136/bmj.e609
• , doi: 10.1136/bmj.e612
• , doi: 10.1136/bmj.e607
• , doi: 10.1136/bmj.e608
• , doi: 10.1136/bmj.e614
• We thank Julio Frenk, Mariana Barraza-Llorens, Raúl Porras-Condey, Héctor Peña-Baca, Octavio Gómez-Dantés, Jaime Sepúlveda, Héctor Hernández-Llamas, Felicia Knaul, Rafael Lozano, Norman Daniels, Dov Chernichovsky, Colin Mathers, Christopher Murray, Majid Ezzati, Emmanuela Gakidou, Ken Hill, Michael Lisman, Kevin Thomas, Philip Clarke, Gretchen Stevens, Rodrigo Dias, Dennis Feehan, Sandeep Kulkarni, Kristen Loncich, Ben Peterson, Jane Kim, Steven Sweet, Jeremy Barofsky, Chloe Bryson-Cahn, Sue Goldie, Jochen Profit, Jennifer Yeh, Anila Gopalakrishnan, Jeremy Goldhaber-Fiebert, Melanie Bertram, Piali Mukhopadhyay, Simon Barquera, Guilherme Borges, Eric Monterrubio Flores, Jürgen Rehm, Juan Rivera Dommarco, Leonora Rojas Bracho, Jorge Villatoro, Miriam Zuk, and Tessa Tan-Torres for their useful input to this study. We thank and acknowledge Dan Chisholm, Jeremy Lauer, Stephen Lim, and Monica Ortegon for their contributions to the data analyses. We also acknowledge the experts who participated in a series of workshops in Mexico for their inputs on analyses and revisions during this project.

• Contributors: JAS designed the study, led the data collection, analysis, and interpretation, and drafted and revised the manuscript. He is guarantor. NC collected and analysed data and contributed to design of the study, interpretation of results, and writing and revision of the manuscript. CG-D contributed to study design, data collection and analysis, and revision of the manuscript. RO, AM, DRH, DL, YM, LS, and MEM-M contributed to intervention analyses and revision of the manuscript. EGP contributed to study design, analysis and interpretation of results, and revision of the manuscript.

• Funding source: This project was supported by funding from the Ministry of Health, Mexico.

• Role of sponsor: The sponsor defined the scope of analysis for the project, but played no other role in study design; collection, analysis, and interpretation of data; writing of the article; or the decision to submit it for publication.

• Competing interests: All authors have completed the ICMJE uniform disclosure form at www.icmje.org/coi_disclosure.pdf (available on request from the corresponding author) and declare: no support from any organisation for the submitted work; no financial relationships with any organisations that might have an interest in the submitted work in the previous three years; no other relationships or activities that could appear to have influenced the submitted work.

• Ethical approval: Not required.

• Data sharing: No additional data available.

This is an open-access article distributed under the terms of the Creative Commons Attribution Non-commercial License, which permits use, distribution, and reproduction in any medium, provided the original work is properly cited, the use is non commercial and is otherwise in compliance with the license. See: http://creativecommons.org/licenses/by-nc/2.0/ and http://creativecommons.org/licenses/by-nc/2.0/legalcode.

View Abstract