Data Resource Profile: Self-reported data in the Million Veteran Program: survey development and insights from the first 850 736 participants

The Department of Veterans Affairs (VA) Million Veteran Program (MVP) is a national voluntary research programme designed to examine genetic influences on health with the goal of improving health for Veterans. Enrollment began in 2011 and, by 20 September 2021, over 850 000 participants have enrolled. Data collection is ongoing with the goal of at least 1 million Veteran participants. MVP Baseline and MVP Lifestyle Surveys are two self-reported surveys that collect participant-provided information to complement genetic and VA electronic health record (EHR) information. MVP Baseline Survey collects information on participant demographics, physical features, uniformed services experience, activities and habits, health status, medical history and health care usage, and family history. MVP Lifestyle Survey assesses personal information, personal habits, activities, psychosocial attributes and well-being, military and environmental experiences, men’s and women’s health, medication usage, vitamin intake, and dietary habits. Among the first 850 736 enrollees (469 487 males and 43 362 females), 513 306 have completed the MVP Baseline Survey and 389 287 have completed the MVP Lifestyle Survey. In this paper we describe the MVP survey development and report overall survey characteristics. Data are currently available to VA and VA-affiliated investigators through approved MVP research projects. Interested future research collaborators can contact MVP at [www.research.va.gov/mvp]. Health surveys have been used for many years as a tool for determining risk factors and disease occurrence in populations.1 Self-reported investigations enable researchers to collect information on a population’s medical, social, economic and lifestyle history which often affect their health. Large-scale studies that use surveys such as the National Health and Nutrition Examination Survey, the National Institutes of Health All of Us and UK Biobank have markedly contributed to biomedical knowledge through survey-based capture of exposures and outcomes. As the field of precision medicine continues to expand, survey data augmented with electronic health record (EHR) and genetic data can also provide invaluable information on social and lifestyle determinants of health which influence the variability of an individual’s genes and risk for disease, treatment and prevention.2 The Department of Veterans Affairs (VA) has a long-established history of using data from the VA’s EHR system for research, but studies have often centred on a specific exposure, disease outcome or service era. Additionally, studies using survey data have often been limited in sample size or are not representative of the entire Veteran population. To expand the scope of research to improve Veteran health, the VA’s Office of Research and Development (ORD) launched the Million Veteran Program (MVP) in 2011 to create a VA-based infrastructure resource for genomic data collection among Veterans.3 As part of the programme’s efforts, surveys were designed to collect individual participant information to enhance the use of MVP genomic data for clinical research. Currently there is no source of large-scale, self-reported health and lifestyle data which exist within VA data sources such as the Corporate Data Warehouse (Health Data Repository), the National Patient Care Database, VA-Medicare merge, national Laboratory and Pharmacy extracts and the VA Vital Status File. MVP methodology has been previously described.3 Briefly, enrollment of users of the Veterans Health Administration (VHA) began in 2011 with the goal of enrolling at least 1 million Veterans. The MVP Baseline Survey is included as part of initial mailed invitational materials; therefore participants may complete the MVP Baseline Survey prior to or following enrollment in MVP. Participation in MVP involves: (i) collection of self-reported data; (ii) provision of blood specimens for genetic research; (iii) access to health records on an ongoing basis; and (iv) agreement for additional research opportunities. Once a Veteran enrols in MVP, the MVP Lifestyle Survey is available for completion and is mailed typically within 2–4 weeks following enrollment. To date, participants have been primarily recruited through mailings and in-person enrollment at VA MVP facilities or events. In 2019, MVP launched MVP Online, an electronic portal whereby Veterans can enroll in MVP, complete surveys and schedule specimen collection activities. Development of the MVP surveys involved extensive review of existing large-scale cohort data collection instruments, evaluation of available data within the VHA EHR and identification of key topics specific to Veteran health and wellbeing. The selected key health risk factors and measures are listed in Table 1. Pilot work was conducted in advance of MVP’s launch, to determine survey length tolerability, optimal method of administration (in-person or self-administration) and timing of completion (pre-/post-enrollment). Various survey versions were created and administered through the mail or in person. By using multiple data capture methods (self-administered vs in-person interview or a combination), the pilot work also helped determine the best method of data capture to maximize enrollment and minimize participant burden. Ultimately, the efficiency of the survey data was determined to be best collected through participant completion in their own time. This was especially important for questions related to family pedigree and health, to allow for consultation with family members and/or records as needed. Summary of the Million Veteran Program (MVP) Baseline and Lifestyle Survey questionnaire items and source PTSD, posttraumatic stress disorder. Summary of the Million Veteran Program (MVP) Baseline and Lifestyle Survey questionnaire items and source PTSD, posttraumatic stress disorder. Based on the pilot work, two surveys were developed as part of enrollment in MVP (Table 1). The MVP Baseline Survey focuses on participant demographics, medical history, family history, military experience and selected lifestyle factors. The MVP Lifestyle Survey focuses on daily/lifestyle habits (such as physical activity, smoking, diet), environmental exposures, gender-specific health and topics specific to Veteran health including combat experience, environmental exposures during military service and post-traumatic stress disorder. The final survey domains and sources are presented in Table 1. Completion time for the MVP Baseline Survey is approximately 15–20 min and 45 min for the MVP Lifestyle Survey. Completed MVP surveys are scanned and digitally archived. Responses to survey questions are processed through a third party vendor who captures and validates data electronically. Indeterminate and write-in values are flagged for later resolution. Given the large number of surveys, data review and cleaning by the vendor are limited to resolving invalid multiple responses by choosing the most conservative response and restricting free response values within a plausible range. Any changes to the original data are flagged to indicate that cleaning was completed on a variable. Ongoing cohort maintenance is managed routinely by the MVP Data Analytics Team who review electronic survey data from newly enrolled MVP participants and apply cleaning macros so that new data are consistently formatted and processed. As of 20 September 2021, 6 365 863 participants have been invited to participate in MVP and 850 736 participants have enrolled, with a response rate of 13.36%. Among enrollees, 593 141 (69.72%) enrolled by ‘walk-ins’ at a study site, 246 293 (28.95%) enrolled at a scheduled appointment and 11 302 (1.33%) enrolled online. Among the enrolled participants, 513 306 have completed the MVP Baseline Survey (91.54% male and 8.46% female). In total, 389 287 participants have completed the MVP Lifestyle Survey (91.92% male and 8.08% female), of whom 345 312 have also completed the MVP Baseline Survey (Figure 1). Flow chart for participant enrollment and survey completion in the Million Veteran Program (MVP). Enrollment is still ongoing, with a target enrollment of 1 million participants. The numbers listed in the figure were updated as of 20 September 2021.VA, Veterans Affairs Briefly, the MVP Baseline Survey includes eight sections with 54 items. As shown in Table 2, the majority of participants enrolled between 2012 and 2017. The enrollment rate decreased in 2020 and 2021 as a result of the COVID-19 pandemic; 79% of males were aged 50–80 and 91% of females were aged younger than 70 years. Black participants accounted for 12.48% of males and 21.08% of females; educational level of 63.01% females and 42.2% males were college or above. Table 2 also briefly presents the number of responders and average levels of physical features, uniformed services experience, activities, habits, health status, lifetime health conditions, health care usage and two samples family history (family histories of 19 diseases were included in the survey). Besides these items presented in Table 2, the Baseline Survey also collected detailed information on 75 health conditions including ‘the year of diagnosis’ and ‘whether taking medications’ for each condition. Distribution proportion of characteristicsa collected from the Million Veteran Program (MVP) Baseline Survey Unless otherwise indicated, data are expressed as proportion. ‘Unknown or missing’ is not counted in the distribution proportion. Distribution proportion was calculated as column percentage except sex, the latter is the row percentage. N: sample without missing for the individual characteristic. GED: General Educational Development which is a high school equivalency. Ever smoking: lifetime smoked a total of at least 100 cigarettes, cigars or pipes and/or ever smoked daily or almost every day for at least 1 year. Always: included respondents of ‘all of the time’, ‘most of the time’ and ‘good bit of the time’. In total, 75 health conditions were collected in Baseline Survey. VA, Veterans Affairs. In total, family histories of 19 diseases were collected in Baseline Survey; family history of a certain condition was defined as any biological family members (mother, father, siblings, maternal grandparents, paternal grandparents) who have been diagnosed with the condition. Distribution proportion of characteristicsa collected from the Million Veteran Program (MVP) Baseline Survey Unless otherwise indicated, data are expressed as proportion. ‘Unknown or missing’ is not counted in the distribution proportion. Distribution proportion was calculated as column percentage except sex, the latter is the row percentage. N: sample without missing for the individual characteristic. GED: General Educational Development which is a high school equivalency. Ever smoking: lifetime smoked a total of at least 100 cigarettes, cigars or pipes and/or ever smoked daily or almost every day for at least 1 year. Always: included respondents of ‘all of the time’, ‘most of the time’ and ‘good bit of the time’. In total, 75 health conditions were collected in Baseline Survey. VA, Veterans Affairs. In total, family histories of 19 diseases were collected in Baseline Survey; family history of a certain condition was defined as any biological family members (mother, father, siblings, maternal grandparents, paternal grandparents) who have been diagnosed with the condition. The MVP Lifestyle Survey includes 10 sections consisting of 70 items. Several items present a series of questions that are based on standardized questionnaires (Table 3) including insomnia disorders,4,5 physical activity pattern,6,7 Mental Disorders Screening Scores of Generalized Anxiety Disorder 2-item (GAD-2) and the 2-item Patient Health Questionnaire (PHQ_2)8,–9 for anxiety and depression, Medical Outcomes Survey (MOS) for Cognitive Functioning,10 MOS Social Supports,11 posttraumatic stress disorder (PTSD) Checklist—Civilian Version,12 Combat Experiences Scale from the Deployment Risk and Resilience Inventory,13,14 and adapted Structured Neurotoxicant Assessment Checklist.15 Besides these primary results, the data collected by the Lifestyle Survey would be further developed and adapted to use in other ways. For example, the frequency of physical activities at different intensities could be converted into average metabolic equivalent for task (MET) values of the overall activity level16 and the frequency of food intakes could be converted to dietary energy and nutrient intakes in combination with the data of food composition table. Levels of major lifestyle factorsa collected from the Million Veteran Program (MVP) Lifestyle Survey MOS, Medical Outcomes Survey; PTSD, Posttraumatic Stress Disorder. Unless otherwise indicated, data are expressed as proportion. ‘Unknown or missing’ is not counted in the distribution proportion. Distribution proportion was calculated as column percentage except sex, the latter is the row percentage. N: sample without missing for the individual characteristic. Mental Disorders Screening using for GAD-2 and PHQ-2, with a score of 3 points as the preferred cut-off for identifying possible cases for generalized anxiety disorder (GAD-2 ≥3)8 and major depressive disorder (PHQ-2 ≥3).9 MOS-CF: MOS cognitive function score, with six items (each ranged 0–6) which assess the frequency during the previous month that one had experienced problems in six categories. MOS-CF36 (ranged 6–36) was calculated as sum of the six questions. Up to one item allowed to be missing. Missing value imputed with average of other responses. MOS-CF100 was on a 0–100 scale, calculated by 100 x (observed score-minimum possible score)/(maximum possible score-minimum possible score). Up to one item allowed to be missing. Missing value imputed with average of other responses. Response categories ‘a good bit of the time’, ‘most of the time’, and ‘all of the time’ were combined as ‘Yes’. MOS-CF scale score: raw scores were converted linearly to a 0 to 100 scale, and higher scores represented higher levels of neurocognitive functioning. PCL-C: PTSD Checklist—Civilian Version. PCL-C (Civilian) is the PTSD checklist, a self-reported measure on current PTSD symptoms. Completed on the Lifestyle questionnaire. A higher score corresponds to more severe PTSD symptoms. The 17 questions were broken down to three clusters. Clusters include: B re-experiencing, questions 1–5; C avoidance/numbing, questions 6–12; D hyperarousal, questions 13–17. Up to one item per cluster allowed to be missing. If one item in a cluster was missing, the value was imputed with the average response of the other items in the cluster. If all 17 questions were answered after imputation by cluster, the sum of the 17 items was taken as the PCL-C score. MOS social supports: MOS (Medical Outcomes Survey) Social Support survey including 18 items categorized into Social Support subscale (each ranged 0 to 5): Emotional/informational support (average response of items 1–8); Tangible support (average response of items 9–12); Affectionate support (average response of items 13–15); and Positive social interaction (average response of items 16–18). MOS Social Support Index was overall average of items. Up to three items allowed to be missing. If three or less missing, responses of missing questions were imputed with the average score of questions answered. MOS Social Support Score was MOS Social Support Index on a 0–100 scale, calculated by 100 x (observed score-minimum possible score)/(maximum possible score-minimum possible score). Deployment Risk and Resilience Inventory (DRRI). Allowed up to three items missing. If three or less missing, response was imputed with the average score of questions answered. DRRI score is sum of the 15 items. Using Combat Experiences Scale from the Deployment Risk and Resilience Inventory; among the responders who answered ‘Yes’ to both ‘deployed’ and ‘served in a combat/war zone’ and who answered all 15 items of the Combat Experiences Scale (score Levels of major lifestyle factorsa collected from the Million Veteran Program (MVP) Lifestyle Survey MOS, Medical Outcomes Survey; PTSD, Posttraumatic Stress Disorder. Unless otherwise indicated, data are expressed as proportion. ‘Unknown or missing’ is not counted in the distribution proportion. Distribution proportion was calculated as column percentage except sex, the latter is the row percentage. N: sample without missing for the individual characteristic. Mental Disorders Screening using for GAD-2 and PHQ-2, with a score of 3 points as the preferred cut-off for identifying possible cases for generalized anxiety disorder (GAD-2 ≥3)8 and major depressive disorder (PHQ-2 ≥3).9 MOS-CF: MOS cognitive function score, with six items (each ranged 0–6) which assess the frequency during the previous month that one had experienced problems in six categories. MOS-CF36 (ranged 6–36) was calculated as sum of the six questions. Up to one item allowed to be missing. Missing value imputed with average of other responses. MOS-CF100 was on a 0–100 scale, calculated by 100 x (observed score-minimum possible score)/(maximum possible score-minimum possible score). Up to one item allowed to be missing. Missing value imputed with average of other responses. Response categories ‘a good bit of the time’, ‘most of the time’, and ‘all of the time’ were combined as ‘Yes’. MOS-CF scale score: raw scores were converted linearly to a 0 to 100 scale, and higher scores represented higher levels of neurocognitive functioning. PCL-C: PTSD Checklist—Civilian Version. PCL-C (Civilian) is the PTSD checklist, a self-reported measure on current PTSD symptoms. Completed on the Lifestyle questionnaire. A higher score corresponds to more severe PTSD symptoms. The 17 questions were broken down to three clusters. Clusters include: B re-experiencing, questions 1–5; C avoidance/numbing, questions 6–12; D hyperarousal, questions 13–17. Up to one item per cluster allowed to be missing. If one item in a cluster was missing, the value was imputed with the average response of the other items in the cluster. If all 17 questions were answered after imputation by cluster, the sum of the 17 items was taken as the PCL-C score. MOS social supports: MOS (Medical Outcomes Survey) Social Support survey including 18 items categorized into Social Support subscale (each ranged 0 to 5): Emotional/informational support (average response of items 1–8); Tangible support (average response of items 9–12); Affectionate support (average response of items 13–15); and Positive social interaction (average response of items 16–18). MOS Social Support Index was overall average of items. Up to three items allowed to be missing. If three or less missing, responses of missing questions were imputed with the average score of questions answered. MOS Social Support Score was MOS Social Support Index on a 0–100 scale, calculated by 100 x (observed score-minimum possible score)/(maximum possible score-minimum possible score). Deployment Risk and Resilience Inventory (DRRI). Allowed up to three items missing. If three or less missing, response was imputed with the average score of questions answered. DRRI score is sum of the 15 items. Using Combat Experiences Scale from the Deployment Risk and Resilience Inventory; among the responders who answered ‘Yes’ to both ‘deployed’ and ‘served in a combat/war zone’ and who answered all 15 items of the Combat Experiences Scale (score A of MVP the for ongoing data capture from As MVP continues to of enrollees is to expand to multiple such as and Additionally, with the of MVP Online, to over 20 000 Baseline and Lifestyle Surveys have been completed electronically. survey completion for data collection a result of and for and also survey so that participants are presented with the items. The self-reported data collected from MVP surveys are an invaluable resource for Veteran research. MVP Baseline and Lifestyle Surveys were designed to capture participant self-reported information on demographics, family history of disease, health status, and lifestyle factors that can influence or be by genetic which is for and the of personal factors on Given the large sample MVP Baseline and Lifestyle Survey data can be used to examine Veteran health and or in with electronic health records (EHR) and genomic data to improve of Veteran health. EHR provide data for researchers and health care to the disease of a over but EHR not provide information for health is by and environmental factors. Data from the MVP Baseline and Lifestyle Surveys this by detailed and environmental information that is consistently a large sample of Veterans. For is that the MVP Baseline and MVP Lifestyle Surveys be in future MVP to enhance the of Veteran health and wellbeing. MVP participants a of and including and MVP data have been used to a of health conditions Research topics health, disease, PTSD, use disease, and Research that the of MVP is to the VHA but with the overall National Health and Nutrition Examination Survey In the of and is higher among in MVP research has also in programme health and of health conditions for Veterans in the MVP the for to of the Veteran care through the Lifestyle factors an important in development and of disease a of lifestyle factors were in Veterans using MVP such as overall dietary score of to and physical Data collected from MVP Baseline and Lifestyle Surveys also provide important information to support MVP genetic research including the genetic study of PTSD, use and of PTSD on use disorder and and genetic to was a risk for disease, disease and Data collected from MVP Baseline and Lifestyle Surveys were also used in the study of which that COVID-19 was with an higher and history of or metabolic data from the surveys and information in the VA EHR provide a to and a medical using self-reported information. For example, not all MVP enrollees of their care from the so data can provide information to create a more Additionally, as MVP to Veterans who may not use the VA for health care the of data as MVP the infrastructure to access health A key is that not all MVP participants complete the MVP Baseline and/or MVP Lifestyle to the If data from EHR are used to of missing data but not provide the from the currently collection of the MVP surveys is are to a for data As in in MVP and MVP access to MVP data and/or samples is by the scope of MVP and VA and review by VA review Data are currently available to VA investigators and other approved with for to investigators in the can be to at For the a is collaborators from and other with VA investigators are This research was conducted to the of the of and approved by the Department of Veteran Affairs approved in has been from the participants in with all VA and the of the VA Data above. data are available at online. Analytics and data MVP and and review of for and All have and to the of the This research is based on data from the VA Million Veteran Program by from the Department of Veterans Affairs. This not the of the Department of Veterans Affairs or the The have no of the VA Million Veteran Program is in the available as data at online. The the members of the Million Veteran Program are in the available as data at who have contributed to the Million Veteran and the Veteran participants for their