The goal of the Population Health Metrics Research Consortium (PHMRC) Mexico Study was to develop better instruments and methods for measuring population health, particularly in resource-poor settings. The project was part of the Grand Challenges in Global Health initiative funded by the Bill & Melinda Gates Foundation. It addressed Grand Challenge #13, to develop technologies that permit quantitative assessment of population health status.
The study collected mortality and disease prevalence information to help improve strategies for population health measurement and produce instruments that are science-based, standardized, and widely applicable. It also sought to enable policymakers and researchers to address persistent inequities in health outcomes in both the developed and the developing worlds.
The PHMRC Mexico Study focused on two key areas of population health: adult and child causes of death in populations where cause of death coding is incomplete or inadequate and the prevalence and incidence of major diseases.
Collaborators
The National Institute of Public Health of Mexico was responsible for identifying cases and implementing the verbal autopsy and symptomatic diagnosis instruments in the field.
38 Mexican hospitals and clinics participated in the study, including federal government hospitals and hospitals of the secretaries of health in the federal district of Mexico, Morelos and Hidalgo states, as well as a clinic of the Institute for Social Security and Services for State Workers.
Key Activities
- Develop new ways to collect cause of death information through verbal autopsy
In countries with an incomplete or nonexistent vital registration system, researchers may conduct verbal autopsies (VAs) by asking a deceased person’s relatives about signs or symptoms the deceased person may have experienced prior to death. IHME led a VA study in Mexico City and the states of Morelos and Hidalgo with the aim of finding new ways to collect accurate cause of death information in low-resource settings. This project built on the success of verbal autopsy research in other
PHMRC field sites. The validation of the verbal autopsy instrument in Spanish was an important accomplishment. Successfully testing the standardized PHMRC instrument in Spanish provided a demonstration case to suggest that the instrument would work in other Spanish-speaking countries with local adaptations.
Overall, trained study physicians reviewed 8,573 medical records to select the subset of 3,019 cases that qualified for entry into the study based on the quality of available diagnostic evidence. Interviewers carried out 2,049 verbal autopsies between June 2009 and November 2010 in Mexico City and the state of Morelos on selected gold standard cases. These cases contributed to the single largest collection of gold standard cases of verbal autopsies known to exist – more than 12,000 deaths from the PHMRC’s six field sites.
VAs were also administered to a community sample without gold standards in the state of Hidalgo. Interviewers conducted 1,201 verbal autopsies in 13 rural and urban municipalities between August 2010 and December 2010.
Multiple methods have been developed to determine causes of death from verbal autopsy responses, including Bayesian symptom pattern techniques, machine learning decision tree algorithms, data-driven symptom-cause tariff computations, and the InterVA verbal autopsy tool, as well as the current standard in the field, which is physician review of VAs. We evaluated each approach’s ability to accurately determine both individual causes of death and cause-specific mortality fractions.
- Test a new method to measure disease prevalence
IHME has also attempted to better measure the prevalence of major diseases in resource-poor areas. Study members created and tested a new survey tool in Spanish called symptomatic diagnosis (SD) in a validated environment in Mexico City.
Symptomatic diagnosis aims to help measure disease prevalence in resource-poor areas, a difficult undertaking for many chronic diseases of public health importance. Conventional methods of measuring prevalence include biometrics, such as blood antibody tests. These are easy and relatively inexpensive to implement in instances where there is a specific biomarker in the blood that appears if the individual is ill (e.g., glucose in diabetic patients). However, for conditions that involve more than one symptom that is not easy to measure, such as asthma or chronic obstructive pulmonary disease, prevalence measurements are more difficult and costly to obtain. Focusing on 10 specific diseases and conditions of public health importance in Mexico, the SD method utilized a survey about self-reported signs, symptoms, and indicators of chronic diseases.
Study physicians worked in 12 hospitals, including specialized hospitals such as the National Institute of Psychiatry and the National Institute of Respiratory Diseases, to find cases that complied with the study’s gold standards. Interviewers conducted 1,090 successful SD interviews with individuals who suffer from the conditions studied, (asthma, angina pectoris, cataracts, cirrhosis, COPD, depression, hearing loss, osteoarthritis, rheumatoid arthritis, and vision loss). Controls were obtained at the Automated Detection and Diagnosis Clinic (CLIDDA), a preventive health clinic where government workers between 30 and 55 years of age can get a free yearly checkup. Two hundred healthy individuals were interviewed for this study.
Methods used to analyze the gold standard verbal autopsy database are currently being adapted to analyze symptomatic diagnosis data. Ultimately, the study will produce baseline population prevalence for each of the conditions separately, as well as estimates of the probability that each individual suffers from each of the conditions studied.
Impact
Development and validation of the VA and SD instruments in Spanish provides a standardized method of collecting data on causes of death and morbidity in Spanish-speaking countries. It is our hope that these instruments will be used to support evidence-based policy decisions.
Reviewing, recoding, and comparing death certificates with our study’s gold standard definitions provides important information regarding the quality of death certification in Mexico and provides a benchmark for the validation of verbal autopsies.
This study’s verbal autopsy work contributed to the single largest collection of gold standard cases of verbal autopsy known to exist – more than 12,000 deaths from India, the Philippines, Mexico, and Tanzania. As part of this collaborative effort, our methods in instrument design, implementation, and data analysis will be made publicly available.
