Measurement and Surveillance of Child Food Insecurity and Hunger
The workshop’s final session before wrap-up examined the measurement and surveillance of child food insecurity and hunger. It examined what is known about the adequacy of the current measurement approach, in both conceptualization and implementation. The moderator was Judith Bartfeld, professor of consumer science, School of Human Ecology, University of Wisconsin–Madison. The session had two speakers, each followed by a discussant. The first speaker was Edward Frongillo, professor and chair of the Department of Health, Promotion, Education and Behavior at the Arnold School of Health, University of South Carolina. Mark Nord, Economic Research Service, served as discussant. The second speaker was Elizabeth Adams, Department of Public Health and Preventative Medicine at Oregon Health and Science University. Maureen Black, Department of Pediatrics and the Department of Epidemiology and Public Health at the University of Maryland School of Medicine, served as the session’s final discussant.
STATEMENT OF EDWARD FRONGILLO1
Frongillo stated that he changed the title of his presentation to “assessment and surveillance” rather than “measurement and surveillance” for reasons that would be clear as he proceeded. He stated that his aim in the presentation was to address what is known about the adequacy of the cur-
________________
1Frongillo and his colleagues, Eliza Fishbein and Maryah Fram, prepared a commissioned paper on this topic; see Frongillo, Fishbein, and Fram (2013).
rent assessment approach: Does it capture key dimensions? Are important populations (i.e., the homeless) missing? Does it adequately describe the experience of everyone in the household? How can assessment be improved? Second, he said he would suggest research topics that might enhance and augment the current surveillance system.
Conceptualization of Food Security
Frongillo began with the conceptualization of food security itself as it has been designed and implemented in the Household Food Security Survey Module (HFSSM) in the Current Population Survey (CPS; see Chapter 2). Frongillo said that from the beginning, this work prioritized mothers’ perspectives, because mothers were thought to be the food decision makers and primary actors in acquiring and managing food, and mothers, especially if single, are also more likely to be food insecure. He said most common knowledge about child food insecurity is based on reports from mothers.
He highlighted two underlying assumptions. First, food insecurity as developed in Radimer et al. (1992) is a household issue that involves managing a process. Second, parents sacrifice and try to buffer their children against the effects of suffering in general and food insecurity in particular. Radimer’s work, in addition to that documented in Hamelin et al. (2002) established the basic foundation for the way experiences of food insecurity are considered.
Frongillo said that traditionally food insecurity is viewed as having four domains: quantitative, qualitative, psychological, and social (Hamelin et al., 2002; Radimer et al., 1992; Wolfe et al., 2003). The quantitative domain ranges in severity from (1) least severe: food depletion, low food stocks but adequate calories; (2) more severe: having to eat less food than usual; and (3) most severe: one or more days without food, actual hunger.
The qualitative domain similarly can be thought about in terms of range of severity. The least severe involves having to buy and eat less preferred foods. This is not considered food insecurity in the United States. The more severe state is having to eat a nutritionally inadequate or poor quality diet. The most severe state is not being able to eat the right foods and meals for health.
The psychological domain focuses on a household’s knowledge and perception of its food situation, and how they feel about that situation. The domain has two components. First, the uncertainty of the food situation and not being able to get the right food for healthy eating leads to feelings of worry and anxiety. Second, the lack of choice and the need to make compromises leads to feelings of deprivation and depression.
The social domain relates to accessing foods in socially unacceptable ways, such as getting food from food pantries, having to ask others for food, borrowing money for food, buying food on credit, or other means. The concept is not based on what people say but on social norms. A second component in this domain is having socially or culturally less normative patterns, like eating peanut butter for dinner daily for weeks in a row.
He said, as has been discussed at this workshop, children in food-insecure households do poorly in many different ways, including behavior, mental health, social, academics, developmental trajectories, hospitalizations, and obesity. Frongillo said he believes that this means that the uncertainty component in the psychological domain is particularly salient for children. Referring to National Research Council (2006) work that showed important consequences of food insecurity, including distress and adverse family and social interactions, Frongillo hypothesized that these particular pathways are the ones that are most detrimental for children.
Frongillo reported that, in Fram et al. (2011), he and his co-authors saw in their qualitative interviews many examples of responses to food insecurity within families, some of which were disturbing. Parents try to provide for the quality and quantity of food, and they try to provide emotional support around eating, but, as he found in his research, they are not fully successful. The researchers also found protective actions were attempted not just from parents to children, but also from parent to parent and that children also try to protect their parents and siblings. Mothers and children tried to protect other children in the family, particularly younger children, and also poorer children in other families.
There is considerable evidence, he said, that children end up living adult roles in food-insecure households. They prematurely act like adults, doing activities that take away from the activities that they should engage in for their own development. In one study, Frongillo and his coauthors intentionally talked to children as well as their mothers and fathers. They found that mothers talked about their social role as being a manager and their job to protect children. Fathers talked about being the provider, trying to protect the wife and children. But children talked about how they actively contribute and how they act to protect other children and their parents. Children even talked about the fact that they know that it is important to their parents to feel like they are protecting the children. The children said they make sure that they hide what they are doing so their parents will still feel good about themselves.
He noted better understanding about what he termed the myth that parents are protecting their children from food insecurity would be invaluable. The available evidence shows that they are trying but are not fully successful at that protection.
Assessment and Measurement
Frongillo turned next to assessment and measurement, and how the concepts fit together, noting his comments reflect ideas particularly from the psychology and clinical chemistry literatures on assessment and measurement. He noted that in psychometrics, item response theory may be used, and generally it is believed that a scale comprised of multiple items has greater reliability than a single item, as in college entrance examinations like the SAT. If a unidimensional scale is assumed, as is often done, there is one underlying construct. Typically it is assumed that the frequency of affirmation is a function of the severity. Again thinking about the SAT, the items that students get wrong more often are the ones that are considered to be difficult (more severe). That idea has been brought over to food insecurity assessment.
There are a number of options for constructing indicators from measures (Frongillo, Nanama, and Wolfe, 2004). The simplest would be to create a scale and report the average. Second, one could create a scale and construct ordinal categories by making cut-points based on the distribution. A third option would be to create a scale and construct ordinal categories by making cut-points on the scale based on specific meaning of items. Finally, one could construct nominal categories based on the specific meaning of items and not on the scale.
Frongillo provided an example of an assessment and different ways for constructing indicators using a study about food insecurity affected by seasonality in northern Burkina Faso (Frongillo, Nanama, and Wolfe, 2004). The average scale was used to document that pre-harvest household food insecurity was very high (scale scores of 10.7, 7.5, and 6.2), in post-harvest it got much better (scale scores of 4.9 and 4.5), and then it cycled again. He then constructed indicators in several ways. He used this example to point out that whenever one creates a scale and establishes cut-points (the third option above), the process will generally underestimate severity because of the way the psychometrics works.
Current Assessment Method
Frongillo explained that when the predecessor of the HFSSM was deployed in 1995 in the CPS, it was intended to estimate prevalence, overall and for certain groups of households, to answer the question “how many are affected.”
It is now used for other purposes. At the group level, it is used as a potential early warning (to determine when action is needed); to target that action; to monitor whether the situation is changing; and to evaluate impact. At the separate household or individual level, the system has
been used for even more purposes, such as for screening, diagnosis, and monitoring.
He said the current assessment method makes a number of assumptions. First, it focuses on households, but there is now an additional focus on children as a subset of the household. The focus is on access to food that is constrained by money, not by the other causes discussed in this workshop. The assessment collects a mixture of items that refer to households, adults, and children. The method has many items about the quantitative domain, few items about the qualitative domain, and one item about the psychological domain on worry and anxiety, with nothing collected about deprivation. The child measure is the same except that it has no psychological measure. Thus, he said, the tool is only getting at part of the story.
In addition, the scale is unidimensional, so it assumes that frequency equals severity and it only contains items that fit a unidimensional framework. He noted the cut-points were based on specific meaning involving the concept of hunger but that meaning is now suppressed in the reported categories.
Current Assessment System
Frongillo noted a few concerns about the current system. First, he said, there may be inadequate coverage of certain subpopulations, such as people who are institutionalized, in the military, indigenous, homeless, marginally housed, and have chronic disease. He said that it is known that food insecurity is important for those with HIV and chronic diseases like diabetes. He asked whether or not food insecurity is being captured adequately among the mentally ill, certain ethnic groups, and immigrants, particularly undocumented immigrants. Second, he said that the sample sizes for children in households with very low food security are too small to support detailed analysis. He referred to Coleman-Jensen’s earlier point (see Chapter 3) that, even in the largest survey, researchers can end up with a really small sample size.
Third, he observed that now the HFSSM has been included in several national surveys, but there are unexplained discrepancies in the result of food insecurity assessment across these different surveys.
Direct Assessment of Children
Frongillo turned to direct assessment of children. The questions used in the HFSSM of the CPS and other surveys are typically answered by a single respondent, almost always an adult, for all people in the household. Although children and adults may have similar experiences regard-
ing food security, they need different language to report those experiences (Connell et al., 2004; Nord and Hopwood, 2007). Multiple studies now show a poor agreement between adult and adolescent reports. Nord and Hanson (2012) found that adolescents self-report food insecurity more often than adults report for them. These adult reports are only weakly associated with the self-reports of the adolescents. The finding that parent and child reports do not give the same information has been replicated in the United States by Fram et al. (2013), as well as in Ethiopia (Hadley et al., 2008) and in Venezuela (Bernal, 2011).
Frongillo said that these studies demonstrate that parents are not fully knowledgeable of their children’s food insecurity experiences. Researchers think this is due to lack of communication and because of efforts to protect each other. People in the household are hiding things and they are not talking openly to each other about these issues.
In a qualitative study of 16 families (Escobar-Alegría et al., 2012), all 16 targeted children had cognitive awareness of food insecurity, but only seven parents were fully aware. Four parents were aware of part of what their child was experiencing, and five others were not aware at all. Fifteen children, but only eight parents, had emotional awareness. For two children experiencing physical awareness of food security, no parent was aware. Eight children initiated responses to food insecurity and one child generated resources. None of their parents was knowledgeable about these actions.
Frongillo described his mixed-methods study (Fram et al., 2013) funded by the Research Innovation and Development Grants in Economics Program. In-depth interviews with 100 children were conducted with questions developed based on earlier qualitative work. These questions were used to develop a definitive measure, a very accurate classification of what the children were experiencing, based on these in-depth interviews. The children were classified in terms of six domains that the authors thought were accurate. The parents were asked the HFSSM questions, and their answers were compared with the child reports. For cognitive, emotional, and physical awareness, the child report was accurate. For participation and resource generation, the child report was not accurate. On the other hand, parent reports under cognitive and physical awareness (the only two domains that could be tested) were not accurate. In other words, he said, relying on the child’s report is much more accurate than the parent’s report. Particularly disturbing is that parents missed more than half of the incidences of hunger because the actual prevalence was 33 and parents only picked up only 15 of those. This is further evidence that parents do not know about the experiences that their children are having and, therefore, if parents’ reports are the only gauge for a child’s experience, results will not be accurate.
Improving Assessment of Child Food Insecurity and Hunger
Frongillo said household food insecurity is a powerful stressor and a marker for other stressors, as Laraia posited earlier (see Chapter 8). Further, he observed, it is plausible that non-nutritional pathways (as opposed to nutritional pathways) deliver the most harm to children. Children are accurate reporters of their own experiences, but parents are inaccurate reporters of their children’s experiences. Therefore, he said, the current parent-report system very likely underestimates the prevalence of food insecurity and hunger among children.
The current system is valuable for the purposes for which it was developed—to estimate the prevalence of household food insecurity. It has been a very powerful tool for that purpose and to monitor how food insecurity has changed over time. However, he suggested, there is more to understand, such as ways in which children experience food insecurity, how many children have those experiences, which children have those experiences, and which actions will ameliorate those experiences.
He questioned the most salient causes of child food insecurity, noting lack of money is not the entire story. Parental physical and mental health is crucial, as are transportation barriers to accessing food or food assistance, parental work demands and schedules, stigma, and other social issues. He noted the most salient domains are awareness (cognitive, emotional, physical) and responsibility (participation, initiation, and resource generation).
To respond to child food insecurity, he said an assessment system that builds on existing systems would be important. Frongillo and his co-authors, funded by the Nord Family Foundation, have been working in schools in the last 18 months, testing to see to what extent schools can become a part of the system to address child food insecurity. Schools respond formally to child food insecurity, through the National School Lunch, School Breakfast, and Afterschool Snack Programs, as well as informally. More than one-half of teachers nationally purchased extra food to give to students whom they thought were food insecure and did other things to help children cope with food insecurity (Share Our Strength, 2012b). Holiday food baskets, food backpacks, and other approaches have been used. Schools do many things, he said, and the process is somewhat haphazard.
School is a place where food insecurity in children is observed and where children get food. But it is also a place where food insecurity is secret and stigma is perceived. If the government is going to potentiate schools as a system for addressing food insecurity, he said, education and training of school personnel, systematic attention to the problem responses, and meaningful assessment of holistic responses will be important.
Schools only reach school-aged children, he observed, and they can-
not operate in a vacuum. As others have discussed, food augmentation may not be the best response, and it can be overtly harmful. Therefore, community-wide systems that augment the schools will be important, which holistic community assessment and response can help. He provided an example from the implementation of the U.S. Triple P System Population Trial in South Carolina. Triple P is the Positive Parenting Program, described by Sanders (2008). This randomized study showed that training the existing workforce in positive parenting could have large effects on reducing the number of substantiated maltreatment, out-of-home placement of children, and child maltreatment injuries.
Finally, Frongillo stated that—to end child hunger—new thinking is needed based on systems, public health, resource realism, and a holistic approach. Assessment instruments and systems to directly and accurately identify child food insecurity are fundamental. They include questionnaires that cover all domains of child food insecurity, and making use of observation is key. The development of resources and protocols for the actions that can help when children’s food security is identified is beginning. Those efforts involve training school personnel, nurses, pediatricians, clergy, and other professionals who come in contact with children so they can assess, identify, target, act, and then monitor what happens.
STATEMENT OF MARK NORD
Nord said that he appreciated Frongillo’s comments and agreed that the current measure of food insecurity is essentially economic access, although it is likely that other aspects or adjuncts of food insecurity are important. He said he does not believe that this is a weakness of the current measure nor does he support adding other components onto the current measure. Instead, he said, it may be useful to add additional measures addressing other dimensions and adjuncts.
Nord suggested research to address what he termed missing pieces. He said that the greatest incidence of food insecurity in children is in older children. However, it is thought that the greatest impacts of food insecurity, at least through nutritional channels, and possibly through psychological channels, are on younger children. He pointed to evidence presented at the workshop that shows the importance of early childhood. He said that may be good news, because those children appear to be shielded more successfully from food insecurity. But he questioned if that is really the case, since the assessment is based on parent reports. He noted the assessment of the accuracy of parent-reported insecurity for older children, but not for younger children who cannot speak for themselves. He suggested this is an area where further research is indicated. If parents are able to report the food insecurity of younger children
with reasonable accuracy, then the overall monitoring of food insecurity among children is not too bad. Although older children are not being monitored perfectly, they are impacted less. When one combines the older and younger children, he said the overall monitoring may be acceptable.
Monitoring has specific purposes: to retain and improve an effective, efficient program and policy framework in order to prevent, or minimize, the extent and severity of childhood hunger and food insecurity. This purpose includes the ability to maintain public and policy-official awareness and understanding of the extent and severity of children’s food insecurity. The purpose also includes the ability to document trends over time and the distribution by relevant household characteristics and geography. He said that this purpose of the monitoring leads to the following measurement and statistical tool requirements: credible, consistent, timely measurement; understandable to policy officials and the public; at the appropriate geographic level; publicly accessible; regularly interjected into policy and program consideration; prevalent, with some measurement error acceptable if random relative to reported categories; and implementable at a large scale.
He said the current measure meets most of these criteria. Nord said some level of measurement error in a monitoring tool can be tolerated as long as that error is not systematic across the key categories. Random error can matter more in certain research applications. He noted that Frongillo made the point that the purpose of the measure and its use may be different.
He pointed to a major finding (Nord and Hanson, 2012) that youth-reported personal food insecurity did not compare well to what their parents reported. Nord added that there was almost no statistically significant association between their reports. If both the parent and the youth said that the youth was food secure, then the average mean Healthy Eating Index for the youth was higher. If they both thought the youth was food insecure, the Healthy Eating Index was lower. If they were in that in-between category where they disagreed one way or the other—one thought the youth was food insecure, the other did not—their Healthy Eating Index was in the middle, but there was not clear priority of one over the other. Unfortunately, he said, the study did not provide insight into which type of report was most accurate.
He referred to a study (Nord, 2013a) that could shed more light on this issue. It incorporates confidential data from the National Center for Health Statistics to see to what extent adults appear to be shielding youth from food insecurity. Based on adult and youth self-reports in the same household, it shows considerably less food insecurity among the youths than among the adults. So a protection mechanism seems to be working to a considerable extent, though not as completely and strongly as would be ideal.
When the measurement is intended for screening, this different purpose requires that the measurement have high sensitivity, reasonable specificity, low burden, and suitability for the context. He noted screening should identify those individuals who need help, so high sensitivity is important. Reasonable specificity will avoid people who would not qualify requesting assistance. Low burden is important. Nord noted that Frongillo and others have talked about having only one or two items for screening. Finally, the context is important because it does not make sense to ask certain types of questions in certain screening situations.
Collection of food insecurity data to support research presents another set of purposes for the measures, as Frongillo laid out. The purpose of data for research is to identify the causes and consequences of food insecurity, gauge effectiveness of programs and policies, and assess quality and characteristics of the measure. Nord noted that measurement error, particularly systematic measurement error, matters greatly in these situations. With a high level of error, associations in general are underestimated and statistical significance may be elusive as a result. The measure must also be collectable from the population of interest.
He noted that researchers who have collected these data know the difficulty, for example, of obtaining institutional review board approval for surveys that ask children about their food security. He said the idea of using a school-based assessment is innovative, and maybe such an experiment would be useful.
Nord talked about current issues in measurement, some of which were discussed earlier at the workshop. First, how well do adults report food insecurity of younger children? Second, how can frequent or persistent food insecurity within a survey year be addressed? Nord stated that there has been an inconsistency in what is meant by “frequency,” saying that he uses the term to mean “how often it happened”—for example, almost every month, some months, often, sometimes, etc. If frequency of occurrence is perfectly collinear with maximum severity during the year, then that proposed measure does not provide additional information, although it could shed light on what the level of severity means in terms of frequency. According to his research (Nord, 2013b), they are not quite perfectly collinear, but the patterns of lack of agreement seem to make some sense. A third current issue in measurement is improving the household-level measure in households with children. He identified several problems with the current measure. It overstates food insecurity and understates very low food security in households with children relative to those without children. The biases vary depending on the ages of the children. The U.S. Department of Agriculture is considering changing the methodology based on the work done following the National Research Council report (2006). The new procedures would use the child scale and
the adult scale, and, if either shows food insecurity, then the household would be categorized as food insecure. If either one of them shows very low food security then the household would be categorized as having very low food security.
Nord suggested that another useful research issue is the development of a better understanding of the causes of the differences between youth and adult-proxy reports of youth’s food security.
STATEMENT OF ELIZABETH ADAMS
Adams described the Childhood Hunger Coalition2 (CHC) to inform the context of her presentation about a pilot screening project for food insecurity. She concluded with general remarks.
Screening for Food Insecurity
The CHC is an interdisciplinary collaborative in Oregon that includes health care professionals, public health professionals, educators, and antihunger advocates. The coalition focuses on hunger as a public health concern, trying to eliminate problems that may come about because of food insecurity.
In 2008, the CHC conducted a survey of health care providers in Oregon to assess what is known about food insecurity. The survey revealed that providers understood a lot about food insecurity and its impact on children, but identified barriers that limited their ability to screen for it. Nearly 90 percent of health care providers said they would be willing to screen if they had guidance about what questions to ask and how to approach a family about such a sensitive issue. They also said that if they were to screen, they wanted a way to provide assistance to families.
Based on this study, CHC developed tools that health care providers could use and a model of how they could work with families to address food insecurity. These tools included a two-question screen (discussed by Black later in this session), as well as an intervention algorithm, an online continuing medical education course (http://www.ecampus.oregonstatedu/hunger), educational toolkits mailed to providers, a quarterly digest that reports on the current news and findings, and the CHC website. The health care providers were given a set of recommendations or actions that they could take to assess the impacts of food insecurity, as well as information about food assistance programs and other resources for food-insecure families.
Adams said that the Oregon Food Bank and Oregon Health and Sci-
________________
2See http://www.childhoodhunger.org [August 28, 2013].
ence University are now carrying out a pilot study to assess how well these screening and intervention procedures are working. The pilot is being conducted in two pediatric clinic sites. It will quantify prevalence of food insecurity, as well as the use of available resources by food-insecure households before and after screening. The pilot will also assess the impact on families and on providers through qualitative studies to see how well they think the project is working. These quantitative and qualitative results will be used to develop best practices and “lessons learned.” The CHC intends to share information with health care providers in other clinical settings and then with schools and Head Start programs, which have expressed interest in tested strategies for screening and intervention.
The screening and intervention protocol for the pilot project is as follows: Physicians screen as part of routine care, provide the parent of a child identified at risk with a link to resources, and follow up with the child per clinical judgment. Interventions are not standardized, but may be based on the CHC algorithm.
Families of children who were under 18 and who screened positive for the food insecurity screen between December 2012 and March 2013 of this year were eligible and recruited for the follow-on study. The study will follow the screened child for six months. In recruitment, the researcher reviewed the protocol and met with the parent to answer questions and obtain informed consent. Following the initial visit, the family was given an electronic medical record summary that provides information about all the resources for food insecurity available in the family’s community. The summary provided to families included information about resources, such as the OregonHelps website; the Supplemental Nutrition Assistance Program; the Special Supplemental Nutrition Program for Women, Infants, and Children; child nutrition programs; emergency foods; gardening; gleaning; and farmers’ markets.
Families in the study are asked to complete a baseline questionnaire that covers issues of health, household situation, and food security over the last six months. Then in six months, they complete a follow-up survey focusing on the first child that was seen. The follow-up survey includes changes in the child’s health, changes in demographic characteristics, and changes in employment status over the last six months.
To date, 1,130 patients have been screened in the first four months of the study, and 143 have screened positive for food insecurity. Providers have supported the project and are surprised by the number of families screening positive. Families have appreciated the issue being raised.
Response to Workshop Presentations
Adams shifted into the role of discussant. Referring to Frongillo’s overview of the child food insecurity screening and assessment tools, she said it is important to work together toward a better system for assessing and monitoring children’s experiences of food insecurity. She sees a parent-child discrepancy in food insecurity reporting in a clinical setting, in which one may hear different things from the child than from the parent. She agreed with Frongillo’s suggestion about coordinating the efforts of public health, medical, and other disciplines that focus on children.
Adams noted the importance of better understanding the experiences of food insecurity, and about triggers and outcomes in the food-insecure population and what they mean for assessment and monitoring. As mentioned earlier in the workshop, understanding the root causes of hunger among children is important, and information about the intergenerational patterns of food insecurity and what happens over the life course will help. These determinants and experiences can have implications for policy and program planning.
Finally, Adams said that community-based resource systems are very important to link the families to the programs that are available in their communities. Food stamps, for example, are an important part of the safety net for many reasons. But she thinks there may be even more resources that individual communities would be able to use. There are models for community-based organizing systems, such as the F.E.A.S.T. model (Food, Education, Agriculture Solutions Together) that comes out of Oregon. This approach is starting to be applied in other states and may be a good model to pursue.
Adams said many potential sources for data exist to address the gaps that have been identified. Existing surveillance systems collect data about food insecurity and health outcomes of parents and children, and these systems can be used or built upon to collect information that will inform strategies to address childhood hunger. Regional clinical health information networks also could be a source of data. There are many research opportunities to collaborate with existing programs like schools and Head Start. Adams suggested it is also important to integrate food security research with the emerging regional and national transformations currently under way in health care and education.
STATEMENT OF MAUREEN BLACK
Black first discussed the global measure of food insecurity, observing that more than 90 percent of the world’s children live in low- and middle-income countries where food insecurity is an enormous concern. She then discussed the food insecurity screener and parental assessment
of body size, observing that Frongillo had talked about parents’ ability to recognize food insecurity among their children, and closed by providing a few thoughts about future research topics.
Global Food Insecurity
Looking at household access to food and at children’s growth from a global perspective, food insecurity is often grouped in three categories: availability of food, access to food, and utilization. There is evidence that more than 50 percent of child deaths worldwide are associated with malnutrition. Viewed from a chronic perspective, stunting occurs when a child’s linear growth slows down, especially during the first two years of life. Wasting is acute undernutrition. Black said food insecurity in childhood relates to stunting and wasting. She described data from the MAL-ED3 Study (Psaki et al., 2012), for which she is a consultant. The study is fielded in eight countries—Bangladesh, Brazil, India, Nepal, Pakistan, Peru, South Africa, and Tanzania—and includes 800 families, 100 per site, with children between 24 and 60 months of age.
Black showed a world map with global estimates of stunting in children under five years old, mostly in South Asia, sub-Saharan Africa, and Guatemala. With stunting, not only is a child’s growth hampered, but also his/her development, academic performance, and work productivity, and it is a lifespan intergenerational concern. The study used a household food insecurity and access scale developed by Food and Nutrition Technical Assistance Program4 of the U.S. Agency for International Development. The scale is similar to the household food security scale used in the United States, except that it reflects the past four weeks. There are nine “occurrence” questions; if the household endorses any one of those questions, then there is a frequency question. There is one question on anxiety, three questions on quality, and five questions on food intake and consequences. Sometimes the last three questions are used as a household hunger set. In terms of access to food, 37 percent of households reported no problem.
How does food insecurity relate to children’s growth? She said that stunting was seen in all eight countries, with the overall rate of stunting in this sample about 43 percent, an enormous problem. There is not as large a proportion of children wasted as stunted. However, it is still a major problem in several countries.
Food insecurity is related to both stunting and wasting. In particular, there is a clear relationship with height for age. After adjusting for
________________
3See http://www.fnih.org/work/key-initiatives/mal-ed [August 13, 2013].
4See http://www.fantaproject.org/ [August 13, 2013].
socioeconomic status (including water source, maternal education, and household density), she explained that food insecurity is associated with a definitive negative shift in the distribution of height for age (as indication of stunting). This is a consistent relationship across countries, even beyond the eight countries in this study. The study found no relationship with weight for height, the indicator of acute status of wasting. The study also found no relationship with hunger. From the study, some cultural aspects in responding to the questionnaire could be seen in the data from Nepal and Tanzania. In spite of that, food insecurity does relate to poor growth and height beyond socioeconomic status. Multiple factors relate to child growth, not only food insecurity.
Developing a Screener
Black discussed the process of developing the two-item screener for food insecurity (see Hagar et al., 2010), of which she is a coauthor, noting that this is the screener used by Adams and others in Oregon. The objective was to identify those questions that are most often endorsed by food-insecure families. She and her colleagues started with a screener used by Children’s HealthWatch and chose the first two questions addressing anxiety and food intake. These questions are slight rewordings of the first two questions in the HFSSM.
If either one or both of the statements is endorsed, then the household is said to be at risk for food insecurity. They validated the results using data from more than 30,000 Children’s HealthWatch participants (children under three years old from seven medical centers across the country). They found sensitivity of 97 percent, meaning the screener captured 97 percent of food-insecure families. The specificity was 83, meaning that 17 percent of food-secure families were incorrectly classified as insecure. However, these households were certainly still at risk for food insecurity.
Black compared the two-question screener with the full 18-item HFSSM in terms of odds for a number of negative outcomes: fair/poor child health, child hospitalizations, developmental risk, fair/poor caregiver health, and caregiver depressive symptoms. These odds were adjusted for site, race/ethnicity, U.S.-born versus immigrant mother, marital status, education, child gender, caregiver employment, breastfeeding, and low birthweight.
The screener odds ratios were slightly attenuated from using the entire scale, but were still significantly associated with children’s health, hospitalizations, developmental risk, caregiver health, and caregiver depressive symptoms. As noted by Adams, this screener is being used in Oregon, and it enables health care providers to assess risk for food insecurity and immediately provide support to the families.
Black stated that in the United States, there is not the association with the children’s height or weight that was obvious in the global perspective. Food insecurity is often invisible, and a way to identify it would be invaluable. More universal interventions might ensure that children have the nutrition and the food patterns that will help them grow, she commented.
Parental Assessment
An open question has been whether parents can/do give an accurate assessment of child food insecurity. Black provided an example of parental assessment in terms of reporting a child’s size. Using pictures of four toddlers of different body sizes, she said parents express a preference for a large body size and concern about a smaller body size. Parents participate in the assessment with a sense of how they would like their children to look. This is not just true for parents of toddlers. If one asks adolescents and parents to judge both how the adolescent’s body size is and how they would like it to be, the parent and adolescent agree if the adolescent is within the normal range. But if the adolescent is overweight, the parent thinks it is fine while the adolescent does not.
Direction of Future Research
Black ended with remarks about measuring food insecurity and how it fits within a larger perspective of children’s early nutritional health and growth. While food insecurity may have an impact on whether children meet dietary guidelines, it is not the whole story. Fixing food insecurity does not necessarily fix helping children meet optimal dietary guidelines. Children do not do well in terms of feeding patterns—how they use food and when they eat—nor do they do well eating breakfast. Parents often use food to manage behavior and teach their children to do so. She stated that a larger perspective would be to determine what can be done to ensure children’s early eating patterns and health, which goes beyond thinking about the problem of food insecurity and not having enough food.
OPEN DISCUSSION
Hilary Seligman (University of California, San Francisco) followed up on a statement by Frongillo concerning the lack of concordance between parent and child’s reports of food insecurity. She said, as was discussed earlier at the workshop, the HFSSM questions can be viewed as a threat to the parent’s identity as a good parent. Because being the parent of
a child who is not being adequately fed causes so much cognitive dissonance, researchers may not be able to accurately get to the bottom of food security experiences of the children using the parents as a proxy even in the context of a qualitative interview. At some level, parents might understand the food insecurity experiences of their children but are not able to communicate those experiences to an interviewer or perhaps even to themselves. Seligman suggested conceptualizing this as “lack of knowledge” because in the context of a really stressful experience—a parent unable to feed their child—it might help drive improvement in the questions.
Nord expressed general agreement but noted the result of comparing what a parent said about children and what the children said about themselves using National Health and Nutrition Examination Survey (NHANES) data was not strongly systematically biased one way or the other. The marginals were somewhat, but not significantly, different. A participant stated that when she was working with the NHANES data, she saw data for families that answered the household questions both in the home and at the Mobile Examination Center during the same time period. She recalled that a large number of families did not answer the same question in the same way within two weeks, suggesting a possible problem in getting people to answer the questions accurately. Nord replied that in the method that he and his coauthor used in their paper, they also compared what adults said about adults in the family, and what the sampled adult said about his or her personal food security in the Mobile Examination Center interview. He noted the agreement was a lot higher for single-adult households and even in multiple-adult households than it was between adult reports for youth and what the youth said.
Frongillo stated that the issue is not just whether parents are knowledgeable but whether, at the moment of the interview, they will express their views. He said this willingness depends on how they are feeling at the moment. He said that is why he tried to emphasize the importance of the context of roles and shared beliefs about those roles.
Pérez-Escamilla stated that Mexico has now included the HFSSM in the Mexican equivalent of NHANES. The data have been released for 2011 and 2012 and show a very strong relationship between household food insecurity and stunting, but not wasting, similar to the results discussed by Black. He said he is struggling with these results about why a measure of food insecurity in the past three months or four months would be causally predictive of stunting and not wasting.
Black responded that it could depend on the age of the child and whether one takes seasonality into the perspective. Food insecurity in developing countries is more severe and more chronic than in the United States, and the children are undoubtedly micronutrient deficient. The
rates of wasting are much lower than the rates of stunting as well. She said a grim possibility is that the most severely wasted children are dead by the time the survey is taken. Another participant agreed, saying that these deaths probably happened early in life and that 50 percent of the deaths of children are associated with malnutrition.
Another participant commented on the issue of measurement for adults versus children, saying that when doing dietary information intake it would be considered ludicrous to expect parents to know what a teenager ate. As a result, some surveys, including NHANES, ask the teenager directly. The speaker asked if something similar could take place for the measurement of food security. The speaker also asked whether work has been done to develop the appropriate wording for requesting the information from teenagers. Frongillo said he finds that school-aged children can answer accurately, with occasional failures. For the most part, children as young as age six or seven can answer the questions, but whether that would work in the context of NHANES is a different challenge. While children are able to report experiences of pain and other things, he said middle school children are most likely to hide information. Older children do not care what they say, and younger children will talk about whatever they are thinking and whatever has happened. Frongillo said it would be possible to do a direct assessment of children, and he and his collaborators have been doing so for about a year and a half in a school district’s elementary and middle schools. They are trying to develop a model from those observations, augmented by reports from teachers and others.
