Data Resource Profile: The Global Health and Population Project on Access to Care for Cardiometabolic Diseases (HPACC)

Though more than four in five deaths due to cardiovascular disease (CVD) occur in low-income and middle-income countries, there have been few data sources that allow for empirical estimation of key relationships relevant to the epidemiology, health behaviour and health services of CVD risk factors at the level of the individual. The Global Health and Population Project on Access to Care for Cardiometabolic Diseases (HPACC) is a novel data resource that fills this gap. The HPACC dataset currently consists of 76 nationally representative, population-based surveys conducted in or after 2005 of adults aged ≥15 years. For 46 countries, surveys are available in multiple years. The dataset currently includes 1 269 542 individual participants. The HPACC dataset includes demographic and socioeconomic variables; anthropometric and biological measures of disease status; data on health service utilization for hypertension, diabetes and hyperlipidemia; and self-reported information on health behaviours relevant for CVD. We are constantly updating and curating the dataset and are expanding it to include data on mental health, cervical cancer, HIV and injuries, where available. The HPACC collaboration includes a global team of physicians, epidemiologists, economists, demographers, and health policymakers. Most surveys contained in the dataset are publicly available; the harmonized, de-identified data and an accompanying dictionary can be requested from the corresponding author (Jennifer Manne-Goehler: [email protected]). Cardiovascular disease (CVD) is the leading cause of morbidity and mortality globally.1 In recent decades, a rising prevalence of major CVD risk factors including diabetes, hypertension and hyperlipidemia has been observed in many low- and middle-income countries (LMICs) where CVD was not previously considered a major health priority. As economic development drives urbanization and changes in lifestyles in many LMICs, this trend is expected to continue.2–4 Although ∼80% of CVD deaths now occur in LMICs,5 there are few data sources that allow empirical estimation of key CVD indicators and relationships relevant to epidemiology, health behaviour, and health services. In this profile, we provide an overview of a novel data resource for the study of CVD risk factors in LMICs, entitled the Global Health and Population Project on Access to Care for Cardiometabolic Diseases (HPACC). HPACC was initiated in 2016 by researchers at the University of Göttingen, Harvard University and Heidelberg University, in collaboration with in-country partners, with the objective to gather and harmonize household survey data on CVD and its risk factors. The HPACC dataset comprises nationally representative, population-based, individual-level surveys conducted in 2005 or later that are collated and then harmonized. The harmonization process is guided by the World Health Organization STEPwise approach to Surveillance (STEPS) survey instrument. This process is complex and includes aligning variable definitions across surveys such as income and asset data, checking data quality and ensuring consistency especially for biomarker measures, documenting variable skip patterns and confirming sample weights. To date, our efforts have resulted in a dataset covering 76 LMICs (Figure 1) and including data from 1 269 542 participants. These 76 countries represent 4.4 billion adults, which is 79% of the LMIC and 65% of the global adult population. The dataset is dynamic and new surveys are added as they become available. Furthermore, in many countries, eligible surveys are conducted periodically, at least once every 5–10 years. For the most part, these repeat surveys consist of a subsequent cross-section of the population but do not follow the same participants longitudinally. The current version of the HPACC dataset includes the most recent data in a cross-sectional harmonization but there is an ongoing effort to expand the dataset to include repeated cross-sections where available. Availability of repeat surveys by country is provided in detail in Table 1. The dataset is not publicly available, but access can be granted upon request. Map of geographic coverage of biomarker data on cardiovascular risk factors in the HPACC data set. Fiji had information on blood glucose but not blood pressure. Countries marked as high-income countries were consistently classified as high-income from 2005 to 2020 by the World Bank. The following countries are represented by dots: Cabo Verde, Comoros, Fiji, Kiribati, Marshall Islands, Samoa, São Tomé and Principe, Seychelles, Solomon Islands, St. Vincent and the Grenadines, Tonga, Vanuatu. BP, blood pressure; BG, blood glucose; HIC, high-income country. The map was generated using the ggplot2 package in R. Survey type and year, sample size, response rate and available measures and indicators by country HH wealth, household wealth; STEPS, STEPwise Approach to Non-Communicable Disease Risk-Factor Surveillance. Median response rate. Albania, Egypt, Fiji, Indonesia, Mexico, Peru, Romania, Uganda and Ukraine only collect data on tobacco use. Repeat data are already pooled for the following countries and years: Bangladesh 2011 and 2018, Benin 2015 and 2018, Bhutan 2014 and 2019, Brazil 2013 and 2019, Ecuador 2012 and 2018, Chile 2009–2010 and 2016–2017, China 2009 and 2015, Fiji 2009 and 2011, Mexico 2009–2012 and 2018–2019, Kyrgyz Republic 2012 and 2013, Lebanon 2008–2009 and 2017, Lesotho 2012 and 2014, Mongolia 2009 and 2013, Mozambique 2005 and 2015, Nepal 2013 and 2019, São Tomé and Principe 2009 and 2019, South Africa 2012 and 2016, India 2015–2016 and 2017–2019. The dataset was collated by an international group of collaborators to meet the need for rigorous, nationally representative evidence on key epidemiological relationships and health system performance relating to CVD risk factors in these understudied country contexts (Table 2). Its structure allows the comparison of population groups across the socio-economic and demographic spectra within and across countries. Furthermore, it yields results that are representative of the entire national population of the countries where the included surveys took place. Key details about the Global Health and Population Project on Access to Care for Cardiometabolic Diseases (HPACC) data resource Includes adults aged 15 years and older Nationally representative surveys Individual microdata available Contains at least one of the following three biological measures: blood pressure, blood glucose/HbA1c, and lipids Response rate ≥50% Conducted in or after 2005 Conducted in a LMIC (according to World Bank) in the year of the survey 1 269 542 individuals from 76 LMICs covering 73% of the LMIC population Regional composition Africa: 26 countries Americas: 10 countries Eastern Mediterranean: 7 countries Europe: 11 countries South East Asia: 8 countries Western Pacific: 14 countries Date range: 2005–2020 STEPS Surveys (57) Non-STEPS surveys (20) Demographic assessment: age, sex, education, marital status, rural/urban residence, household wealth quintile, working status Biological measures: anthropometry, blood pressure, blood glucose/HbA1c, lipids Measures of healthcare utilization: health care, cascade measures (ever screened, diagnosed, treated), lifestyle counselling regarding diet, exercise, smoking cessation, salt reduction Behavioural risk factors: fruit and vegetable intake, salt intake, physical activity, tobacco use,alcohol use Current data frontiers: HIV, oral health, suicide, mental health, injury and violence Prevalence of CVD risk factors Unmet need for healthcare services for CVD risk factors Socio-economic and demographic patterns of CVD risk factors Quantifying associations between CVD risk factors Modelling costs and effects of hypothetical intervention scenarios CVD, cardiovascular disease; HbA1c, glycated haemoglobin; LMIC, low- and middle-income country; STEPS, WHO STEPwise approach to Surveillance. A systematic search and vetting process is used to identify eligible surveys for harmonization and inclusion in the HPACC data set. The inclusion criteria are detailed in Table 2. In brief, the survey needs to have been conducted in a LMIC in or after 2005. Furthermore, individual-level data from surveys representative of a country’s adult population need to be available and include either a blood pressure, blood glucose or lipid measurement. The first systematic search was conducted in 2016. The search was updated in 2021, identifying additional surveys conducted in 2010 or later. We use a four-part search methodology to screen surveys for inclusion in the dataset (Figure 2). First, all countries in which a World Health Organization (WHO) Stepwise Approach to Non-Communicable Disease (NCD) Risk-Factor Surveillance (STEPS) survey has been conducted are identified.6 All eligible STEPS survey data are requested from the WHO STEPS data repository.7 Prior to 2019, when these were not available from a repository, data requests were made directly to the country survey team. Second, a search of three well-regarded survey resources—the Demographic and Health Surveys (DHS),8 the WHO Study on global AGEing and Adult Health (SAGE)9 and the Gateway to Global Aging studies10—is conducted. Third, we search three other commonly used summary resources—the NCD Risk-Factor Collaboration (NCD RisC),11 the Global Health Data Exchange (GHDx)12 and the International Diabetes Federation (IDF) Diabetes Atlas.13 If these yield a potentially eligible survey, we perform a Google search to confirm and request the most recent survey data from the designated point of contact. Fourth, a systematic Google search is conducted using pre-specified search terms (Figure 2) and the first 30 returned results are reviewed. If two surveys are available for a particular country, we prioritize either the most recent survey and/or the one that contains a larger number of variables of interest. To access non-publicly available survey data that we identify at any step during the search process, we contact each survey's lead investigator(s) at least twice to invite them to share data and participate as an HPACC collaborator. Repeat surveys are included based on the same search methodology and eligibility criteria, as new data become available. The data set is updated periodically when new surveys are released. Table 1 indicates the countries for which repeat surveys are already available and the years in which the repeat survey was conducted. Global Health and Population Project on Access to Care for Cardiometabolic Diseases (HPACC) search strategy. DHS, Demographic and Health Survey; GHDx, Global Health Data Exchange; IDF, International Diabetes Federation; NCD RisC, NCD Risk-Factor Collaboration; SAGE, Study on global AGEing and Adult Health; STEPS, STEPwise Approach to Non-Communicable Disease Risk-Factor Surveillance. See the Supplementary Material (available as Supplementary data at IJE online) for a list of countries that were classified as low-income or middle-income for at least 1 year from 2005 to 2021. The HPACC dataset contains 57 STEPS surveys. The STEPS survey instrument is designed to help countries build and strengthen their surveillance capacity by obtaining core data on established chronic disease risk factors. These surveys are the official approach developed by the WHO for monitoring NCD risk factors in adults at the population level and include questionnaires, physical measurements and biochemical measurements, with ‘core’, ‘expanded’ and ‘optional’ modules to allow flexibility by country capacity and interest. The ‘core’ self-reported measures include data on demographic characteristics, tobacco and alcohol use, dietary behaviour (fruit and vegetable intake), physical activity and history of raised blood pressure, diabetes, raised total cholesterol, heart disease and stroke. In addition, there are 20 non-STEPS surveys in the HPACC data set. Among these, the DHS instrument is frequently used. The DHS, in partnership with country governments, are nationally representative household surveys that collect data on a wide range of health-related monitoring and impact evaluation indicators.14 The standard surveys have large sample sizes (usually >5000 participants) and are conducted, on average, once every 5 years. Although they have historically focused on maternal and child health and more recently HIV, later surveys have increasingly incorporated questions about health behaviour, nutrition and tobacco use. These recent DHS have also begun to measure blood pressure and incorporate biomarkers for diabetes, consistently with the HPACC inclusion criteria. Most surveys in the HPACC dataset use multi-stage cluster random sampling of respondents and almost all provide detailed documentation on the applied sampling strategy. In all analyses, sampling weights account for item non-response and are rescaled to allow either population-size weighting or equal weighting for each country, depending on the primary research question of interest. All surveys included in the HPACC dataset contain information on participants’ age and sex, along with a measure of years of schooling or educational attainment. The majority of these surveys also include information about location of residence (rural/urban, 46 surveys), marital status (68 surveys), working status (72 surveys), pregnancy status (70 surveys) and a measure of household wealth (62 surveys). The construction of variables capturing economic status within each survey depends on the available household economic indicators. Surveys using an assets, or wealth, index for economic status assessment have surveyed household ownership of a range of assets, dwelling characteristics and other country-specific wealth indicators. Based on the standard DHS approach, we use principal component analysis to derive an asset index, from which we create household wealth quintiles.15 Countries using a household income-based measurement mainly followed the WHO STEPS template questionnaire. In these cases, respondents were asked about the household earnings over the past year: weekly or monthly average, or year total. In cases in which this question was left unanswered, respondents were asked to place their household in one of the given income brackets. These brackets were usually defined according to a country’s national household income quintile thresholds. Both continuous income and income brackets are used in the creation of the wealth quintiles. We assume that household incomes within a country follow a log-normal and we are to income and In 11 we income brackets as they large with to the continuous income range or not be as the were only asked of respondents that had not a continuous this has resulted in only information The dataset includes and measurements in all surveys and surveys include and/or In terms of the three CVD risk factors of surveys blood pressure and diabetes and lipid and surveys blood pressure and diabetes biomarkers only (Table pressure measurement was the number of measurements that were and survey from one to The most in the assessment of diabetes biomarkers as a of measures were used. surveys blood glucose four glycated and five surveys We have documentation of measurement details relating to each approach to measurement of the For blood pressure measurements, we the number of measurements, details of the and and used. on the diabetes biomarker includes information on type of or measurement were or are and the status of The lipid measurements were also to in with countries either a approach or a measurement. In of the the was most frequently the surveys with lipid measurements, all total and many and of these measurement are available in Supplementary and (available as Supplementary data at IJE A of the HPACC data set is that surveys information on measures of healthcare service utilization for hypertension, diabetes and/or For the majority of this includes a question about a had a for the they have been with this and they are or currently for the surveys include information on a and on they currently use for the of the included surveys asked respondents about history of CVD including heart or stroke. In addition, in the of diabetes, of the surveys with information on diabetes contained questions about use of oral and surveys also included a of questions about a had lifestyle counselling from a healthcare with to physical activity, salt intake, fruit and vegetable and/or smoking The majority of surveys include self-reported information on health behaviours relevant to CVD. This includes information about dietary patterns of fruit and vegetable and salt surveys). In addition, include a physical activity mainly using the Global The measures and activity at and during and to and from This the of a of activity in of which is in a of the working rate to their rate. Furthermore, surveys The has been for use in LMICs previously and is a for physical activity level in Surveys included in the HPACC data set also questions about alcohol surveys) and tobacco use surveys), including the history of use, and The STEPS surveys and of the other surveys included in the HPACC data set also modules on other chronic that are of particular These include about mental health, cervical cancer, and oral health Furthermore, recent STEPS surveys include information on injury and The include additional modules about health behaviours or HIV, and biomarkers To harmonize we a detailed of the survey documentation with to the and from which the data were The harmonization process is based on and guided by the STEPS survey instrument. We defined the core group of variables that were across and established definitions for variables for which response were We perform detailed data quality to harmonization and The current HPACC dataset contains data from 1 269 542 individuals aged ≥15 years were eligible for at least one of the three blood pressure, blood or The age of all individuals in the data set is years and are The of educational of the individuals in the data set is with with at least primary or primary and with at least or all individuals in the had hypertension, had diabetes, and had prevalence are from the biological or physical measurement and self-reported use of the prevalence are for population The HPACC dataset has already provided and novel about the epidemiological relationships and health system performance for CVD risk factors in To date, research using the HPACC dataset have been in These a range of such as an evaluation of health system performance for the of hypertension the in eligibility for hypertension depending on of hypertension individual and the between CVD risk and hypertension Health system performance for the of diabetes has been in three using the HPACC data, which included and LMICs, Furthermore, of the between diabetes and socio-economic as as anthropometric measures, including index and adult have been conducted. In to focused on diabetes and hypertension, the HPACC data have been used to the cascade of for and coverage in Health in this data set has also included a of country indicators for of CVD risk Furthermore, the HPACC data have been used to risk factors for CVD and diabetes, including the patterns of fruit and vegetable as as in dietary study the prevalence of cervical the HPACC data set is used in such as the impact of and of The harmonization of survey data for the HPACC dataset is only to the large number of HPACC collaborators from the that the individual surveys included in These researchers an and in all that have been and that are We have ongoing efforts to contact and collaboration with the survey of publicly available data Furthermore, the of researchers from LMICs is one of and efforts to do are The HPACC dataset has First, the data are at the individual level and population-based, and are to allow of key epidemiological and health including across demographic and socio-economic groups within countries. the data set includes nationally representative survey data from 76 countries population is billion of the adult population in The geographic coverage has at the global and along with Second, all HPACC surveys contain biomarkers and biological measurements, including blood pressure, blood and/or as as that CVD risk factors of are defined based on these measurements, the HPACC data set a as few population-based surveys measurements to 2005. Third, most of these surveys include at least one measure of household income or asset which have been of economic The of this measure has a of relationships between prevalence and for CVD risk factors and an socio-economic The of the HPACC dataset on the of measures across many surveys from geographic and This includes the measures of economic status, which across surveys. In addition, there is in the measurement of biomarkers of diabetes and To we have these including the approach to status, biomarker measurement and about the measurement that can impact the of such as blood glucose We to that surveys self-reported information are within survey and A is that the measures of access to for all CVD risk factors are self-reported by survey participants and to of access be by or health which lead to in the prevalence of health behaviours and to are of the and Third, the included surveys were conducted in or after 2005. The older surveys not the current in the countries. the data set is updated on a and includes the available Fourth, due to the cross-sectional of the data, it is currently not to is to harmonize and include older and conducted in LMICs, which allow to at changes over surveys contained in the HPACC dataset are publicly available. The two most data sources are the WHO data and the DHS additional surveys have been requests of survey data are not already made The harmonized, de-identified HPACC dataset and accompanying data dictionary have been a partnership between Harvard University, University of and Heidelberg University, in collaboration with all survey Access can be requested the corresponding information about HPACC including additional contact information for the collaboration can be at or by at [email The included population-based surveys from the country’s to data All surveys followed such as for participants’ to participate in the The collated HPACC dataset is de-identified and can contact or The for the of that are from an which include information about is by the in such a that the of the be directly or to the the not contact the and the not The HPACC data set was designated as by the Harvard of Health in See Data Supplementary data are available at IJE and the and the and the HPACC and this All in the data and the and are the of the STEPS survey is by the country’s and the DHS are by the for International and the country’s The of the other surveys are by a country’s and international and by The creation of the collated data set has been by the Harvard and the as as from the of Heidelberg and We to each of the survey and study participants made the of this data set We also the and data service by the of and the and the and