Click for next page ( 88


The National Academies | 500 Fifth St. N.W. | Washington, D.C. 20001
Copyright © National Academy of Sciences. All rights reserved.
Terms of Use and Privacy Statement



Below are the first 10 and last 10 pages of uncorrected machine-read text (when available) of this chapter, followed by the top 30 algorithmically extracted key phrases from the chapter as a whole.
Intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text on the opening pages of each chapter. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.

Do not use for reproduction, copying, pasting, or reading; exclusively for search engines.

OCR for page 87
APPENDIX BBackground Papers for Workshop on Evaluation of Methods for Obtaining Food Consumption Data 1

OCR for page 87

OCR for page 87
Dietary Methodology CHARLOTTE M. YOUNG It is of interest to find two branches of government suddenly showing inter- est in dietary methodology and willing to invest some money in solution to its problems. Literally millions of dollars have been spent on research in which dietary data have been related to various measures of nutritional status. Often the significance of the dietary methods used has been relatively unknown, and it has been difficult to figure out why the method used was chosen. Thirty some years ago I remember questioning the basis for choos- ing dietary methods used in certain studies and finally concluded it usually seemed to be expediency. These observations led me into some feeble ef- forts at looking at dietary methodology, which we did, with no special funds, as a side issue to another study. At that point I left an active interest in the field in pursuit of seeking answers to other practical questions in human nutrition. STATE OF THE ART For years I have tried to encourage nutritionists to make contributions to this field. The "take" has been very limited. Much of the basic work was done 20 to 40 years ago. Though for 20 years there has been a great need for dietary methodology for epidemiological studies dealing with the relation- ship between dietary intake of an individual and his biochemical or health status, we have not come very far. Many reviews of various depth and value concerning dietary survey methods have been written. I shall not attempt to compete with these; instead, I have prepared a list of many of the reviews 89

OCR for page 87
JO CHARLOTTE M. YOUNG including symposia on the subject which occurred in the 1950's and 1960's. The reader may refer to these for details. Recently I have been impressed that the government has become very interested in dietary methodology. You will find the extensive review of Burk and Pao (1976) as they contemplated the 1975 Household Food Con- sumption Survey, the most recent in a series made about every 10 years by the U.S. Department of Agriculture. It is my understanding that several other sections of the government wished through this study to be able to obtain dietary information on both households and individuals. Hence in about 1975-76 the Consumer and Food Economics Institute of the USDA contracted with Response Analysis of Princeton, N.J., to do an extensive field trial after suggesting a series of methodologies to be tested. As I recall, the task was twofold: ( 1) to recommend procedures for a nationwide study of household food use and individual food intake with justification for such recommendations and (2) to recommend whether a pilot study was needed before proceeding with the nationwide data collection, together with a rationale for the conclusion reached. In this study some nine different methods received field trials. Eight of the methods included an effort to get information on household food usage by either some form of 7-day recall, diary, or interviews requiring different types of preparation or training. Four of the methods included data on individual family members using diary or recall for 1 to 3 days from the homemaker for all individuals in the household or from each individual subject as respondent. If the reports can be made available to you, the Committee and its consultants might well wish to examine them. None of the methods proposed incorporated a direct way of measuring its validity or reliability, since there was no observation of all food used or repeat observations in the same households. The Response Analysis survey is probably the most comprehensive study on dietary methodology of which I am aware. It is interesting that so quickly after these studies the Food and Drug Administration has asked the Food and Nutrition Board to conduct an in-depth study with some of the same directions. Such opportunities were what some of us dreamed of years ago. Today it is fortunate that diverse professions have developed interest in dietary methodology so that your Committee will have access not only to the talents of nutritionists of various backgrounds, but also epidemiologists, an- thropologists, economists, sociologists, physicians, statisticians and com- puter scientists. DIETARY STATUS, NOT NUTRITIONAL STATUS It is imperative to remember that in measuring dietary status one is not measuring nutritional status. In using dietary assessments one should be

OCR for page 87
Dietary Methodology 91 quite clear as to their limitations, realizing that dietary studies give no direct measurement of nutritional status. They give only presumptive evidence. If food intake is quite poor, nutritional status may be below desirable limits. If nutrient intake appears to meet certain dietary standards and if the individual has no factors that may adversely condition his nutritional needs, he proba- bly will be well nourished. Dietary studies should be referred to as diet studies, not nutritional studies. In interpreting dietary information, one must bear in mind that there are pronounced individual differences in nutritional requirements; that our knowledge of absolute nutritional requirements is relatively meager; that a whole series of factors may condition an individual's nutrient needs by either interfering with the ingestion, absorption, or utilization of a nutrient or by increasing his nutritional needs through increased requirements, excretion, or destruction of the nutrient. These so-called "conditioning fac- tors" may be discovered by medical histories and examination as part of a nutritional assessment but may be completely unknown in a dietary study. Furthermore, we have little information on how slight differences in nutrient intake affect the health of individuals, as well as little knowledge as to what level of failure to meet dietary allowances food intake affects the health of the individual. Additional laboratory and clinical tests are necessary before nutri- tional status is established. Given these limitations, dietary studies do have many uses. In survey work alone, a description of food intake may be valuable in interpreting nutritional findings. The studies help to identify apparent dietary deficien- cies or excesses. They also form a concrete basis for action programs, for in the long run, therapy will need to be interpreted into feeding programs. Also, valid and reliable information on dietary intake of an individual over a sufficient period of time may be important in relation to various biochemical and clinical measurements of his nutritional status. AREAS IN WHICH WORK IS STILL NEEDED Though there is need for work in the entire field of dietary methodology, certain areas stand out as especially critical: 1. Validity, which to my understanding does not include the refinements or subdivisions of Burk and Pao. To me, validity means the degree to which the method is a true measure of what the investigator wishes to describe. What one wishes to measure of course depends upon one's objectives. Burk and Pao describe five aspects of validity: accuracy, concurrent, construct, content, and predictive validities. The aspect most usually considered I sus- pect would come under content validity, i.e., does the method give the

OCR for page 87
92 CHARLOTTE M. YOUNG "usual" picture of the subject's eating at the time in which the investigator is interested. Too little information is available on validity, and probably "true" valid- ity cannot be measured. In the early studies efforts were made to determine if two or more different methods gave comparable results or whether similar population samples in the same frame of time gave similar results. In more recent and sophisticated times, there have been an extremely limited number of studies in which actual intake (observed without the subject's knowledge) has been compared to the reported intake whether by some form of record by the subject or by recall interview. Some have only been of a single meal. The ethical question always arises as to whether the subject should be aware that he is being observed. Factors that interfere with getting the desired "usual" picture include: (a) the tendency to eat differently during a record period either deliberately or unconsciously as a matter of convenience; (b) the desire to please or outwit the surveyor, which can lead either to omit- ting certain items or reporting certain foods the subject believes he should eat; (c) the general attitude to the questioner or to surveys in general; (d) ability or willingness to keep accurate records; (e) the condition of the subject's mem- ory or willingness to use it; (f) food choice if the subject knows he is being observed; and (g) the mere act of record-keeping. Some individuals, such as homemakers, certain kinds of students, or certain food faddists, may be much more aware of what they eat than others. 2. Reliability, which to me means repeatability or true reproducibility or the error variation in collecting and processing dietary data. Some people confuse reliability as I am using it with validity. Most reliability studies have related to such matters as how accurately actual intake has been re- ported, errors due to the use of food tables, the differences in nutritional value between calculated and chemically analyzed diets, whether reported studies on the same population in a limited period of time gave similar results, and errors in sampling as induced by failure to get random samples when the objective would indicate its desirability but practical cir- cumstances preclude random sampling. 3. Intraindividual variability in food or nutrient intake, which is espe- cially important in determining the length of time that an individual needs to be studied for a true picture of intake be it current or past. There are more of these studies than those related to validity, but they still are quite limited. 4. Interindividual variability, which is of particular interest in population studies or group averages to determine the size of sample needed for the purpose. More has been done here than in the case of intraindividual varia- bility. In recent years the role of a statistician in the planning stages of a study as well as the later analytical stages has been appreciated by nutritionists.

OCR for page 87
Dietary Methodology 93 Statisticians are extremely valuable in making the investigator clearly define his objectives and then in assisting in the many decisions that must be made such as sampling methods, sample size, appropriate quality control, setting up appropriate analyses, and interpretations of the results once available. 5. Changes in inter- and intraindividual variability over time and par- ticularly in relation to seasonal variation. 6. Epidemiological methods suitable for large-scale surveys of the re- lationship of certain dietary or nutrient intakes or habits to disease condi- tions, two of which are of great current interest: coronary heart disease and cancer. Since such studies are often based on relatively large numbers of subjects, special techniques need to be developed or others adapted to the purpose. Five or six groups have attempted to develop various forms of short questionnaires and to check their validity and reliability. Hankin and her colleagues in a period of over 10 years (first in California and later in Hawaii) have developed and tested such a questionnaire in relation to car- diovascular disease, and more recently, to cancer. Previous techniques have been expensive and tedious and often could not be applied to all kinds of population groups. Also, often they have~concentrated on only one facet of dietary pattern, i.e., the level of nutrient intake, with little information on other parameters of food intake such as specific food usage, changes in food patterns, spacing or apportionment of food over the day, environment, speed of eating, etc. 7. Need for more studies of response rates and factors that affect them. 8. Need for studies of actual time and costs involved in the use of differ- ent methods. 9. Development of methods for use by relatively untrained, inexperi- enced interviewers, or nonnutritionists. 10. The role of different disciplines in the study of dietary habits. The role of the statistician was mentioned earlier, but what about the others? We must recognize our need for them at appropriate times as well as their need in certain studies for the advice of an expert nutritionist. It is again part of the increasing recognition of the interdisciplinary nature of much research concerned with very personal aspects of human beings. WHAT WE HAVE LEARNED; DECISIONS TO BE MADE First and foremost, it is clear that the objectives of the dietary study must be clearly defined, for the objective determines the appropriate methods to be used in collecting, processing and interpreting the dietary data. Past studies would indicate that this point has not always been appreciated. Secondly, it is clear on the basis of past dietary studies that there is a need

OCR for page 87
94 CHARLOTTE M. YOUNG for a clear-cut distinction between methods to be used for the average intake of a group for group comparisons and those suitable for defining the dietary intake of an individual. Dietary methodology studies have demonstrated the much smaller cost in time, money, and subject cooperation to obtain the mean intake of a group. There has been increasing realization of the time, cost, subject cooperation, and degree of precision necessary to obtain a valid and reliable measure of the intake of an individual. Even after the objective is clearly defined, many decisions must be made with regard to methodology before a dietary study is undertaken, particu- larly studies concerned with the intake of individuals. Decisions with regard to the collection of data center on sampling, the schedule or form to be used and/or the interviewer, the instructions or supervision to be given, the time period to be covered, the timing of the recording process or recall period, the methods to be used in determining the amounts of food used, and whether food intake only or food habits as well are of interest. Questions of sample size and type are fundamental. Does the sample need to be random so that one can describe a larger universe from which the sample was drawn? What limitations will be imposed by variation in the willingness of selected subjects to cooperate? Or can the sample be nonran- dom if one is concerned with data on only the sample individuals and the characteristics of the larger universe bearing upon food consumption? The size of the sample varies with the objective and the method chosen but also on the classifications to be made in analyzing the data. For group averages a larger sample may be needed to include interindividual variability; the lower the variability in food consumption, the smaller the number needed for stable averages. The method of sampling should be clearly stated as well as information relative to those unwilling to or unable to cooperate. Too often this point has been neglected in reports. It is well to remember that sampling is a field of statistics in itself. What time period should be covered? Are we interested in current intake, immediately past intake for a fixed period, or the usual or characteristic intake over an extended period in the present or past? In what kind of dietary information are we interested? In nutrients, which ones? Foods? If so, which ones? Are we concerned with quantitative aspects or only the presence or absence of certain foods in the diet? Is it a food pattern in which we are interested? Are we concerned with food habits such as time, place, and circumstances of eating; with whom, regularity, food patterns, seasonings, waste, peculiarities of preparation, etc.? How will amounts be reported? By weight, by household measures, by estimations of size of portions, or merely by the frequency with which some item is eaten? From whom shall such information be obtained, i.e., who will be the respondent? The individual subject, the person responsible for the

OCR for page 87
Dietary Methodology 95 subject's food, or both, or an observer? If the latter, what kind of training is needed? The major determinant of these decisions is the objective of the study, but also to be considered are the resources available in terms of time, money, trained personnel, and the availability of a sufficient supply of suitable subjects (in Burk and Pao's terms, the respondent burden, field survey costs, and data-processing costs). NECESSITY FOR PRETESTING METHODOLOGY All evidence points to the necessity of pretesting the method on the particu- lar population before undertaking a large study. As indicated earlier, possi- ble methods vary with age, sex, culture, and education of the subjects as well as the complexity of the food resources available to the participant. Age is a factor not only for the very young, who cannot write or report, but also for the elderly, where memory may be affected. Education and culture may affect the ability to write and how much interviewers must be involved. Inter- and intraindividual variability should be determined as an indica- tion of the number of subjects needed as well as the time period that must be covered. It is also an opportunity to determine if there is a seasonal variation in intake. All too often the results are disastrous when pretesting has not been done. Quality control for data collection with training of interviewers and careful standardization must be planned. In advance of the study, decisions also must be made with regard to methods of analysis, such as whether one is interested in diet patterns, frequency of intake of particular foods, or specific nutrients. If individual nutrients are to be calculated, what methods and what tables are suitable? If chemically analyzed, what methods? Interpretation or evaluation of ade- quacy then follows. For precision, even when food tables are used, there is increasing emphasis on chemical analyses of certain local foods to check values included in nationally used tables. Some investigators believe that before a major study is begun, the form of anticipated tables of results should be set up to be sure proper data and needed numbers of subjects are being used to meet the objective. TYPES OF METHODS AVAILABLE In general there are two types of dietary studies and within each type a variety of methods may be used: those concerned with food usage of families or institutions sharing common food supplies and those concerned with food usage of individuals. Information is obtained either by food rec- ords or by recall of what has been eaten.

OCR for page 87
96 Household or Institution Methods Three methods are available: CHARLOTTE M. YOUNG 1. Food accounts- a simple running description of food purchased, re- ceived as gifts, or produced for household use over a given period of time. These are not precise and rarely used today. 2. Food records a weighted inventory of foods on hand at the begin- ning and close of the study together with a day-to-day record of food brought into the home or institution over the period of the study with or without an accounting of kitchen and plate waste or food fed to pets. This is the method the USDA used in its early household consumption studies. 3. Food lists a method by which an interviewer obtains from the person responsible for the food an estimate of the quantities of food used over a given period of time. This is the method used by the USDA in the more recent household food consumption studies. The methods vary widely in cost of collecting and processing the data and to some extent the uses to which the data may be put. It is well to remember that there is no merit in using a more elaborate or expensive method than is necessary to obtain the data needed to meet the defined objective of the study. Individual Methods Dietary data on individuals are collected either (1) to obtain average nutrient intake, food intake, or food habits of groups for comparison with other groups or (2) to obtain nutrient intake of a given individual for correlation with clinical or biochemical measurements obtained on that individual. Studies vary from a qualitative type of food habit inquiry to those of a very much more precise quantitative nature. Each type of study has its use; the important thing is to consider carefully what kind of facts are needed for a given purpose and which method provides these facts at the lowest cost. Methods used with individuals include the following: 1. An estimation by recall, in which the subject, or in the case of young children perhaps the parent, recalls the food intake over the previous 24 hours or longer, with dependence on memory. 2. Records of food eaten by an individual kept by weights, household measurements, or by estimated quantities over a stated period of time. 3. Dietary history, in which by recall or repeated food records or both the interviewer aims to discover the usual eating patterns over a relatively long

OCR for page 87
Dietary Methodology 97 period of time. It is a time-consuming process requiring professional per- sonnel. 4. More recently foodfrequency questionnaires, either self-administered or interviewer-administered, have been studied as a means of acquiring information on general dietary intake or specific foods or nutrients over a longer period of time at less expense and with less personnel resources for epidemiological studies in which people may be grouped in extremes of intake. Unfortunately, a number of these initial efforts have not been done with typical American cultures. 5. Weighted intake for precise measurements during which all food eaten is carefully weighted and nutritive value either calculated from food tables or determined by laboratory analysis of duplicate samples; usually this method is used only in research groups with special facilities for collection and analysis. FACTORS INVOLVED IN CHOICE OF METHODS OR STATE OF THE ART WITH REGARD TO EACH METHOD Three statements of Keys (1968, 1979) may well be borne in mind as we consider the factors involved in choice of method. The first is that good surveys are difficult, costly, and require professional expertise in planning, operation, and interpretation. The second is that little is known of the long- time intraindividual variability of diet or that diet at any one time is rep- resentative over longer periods of time; and third, that even the best dietary surveys cannot provide answers about biochemical relationships. Many comments may be made about the various dietary methods based on heterogenous dietary methodology studies that have been made over the last 40 years. More than most reviewers, Burk and Pao (1976) have brought them together in one place in relation to each type of method and is the latest review now in print. Hence the following discussion draws heavily on their review. Greater detail may be found in the review itself. Food Accounts This method is not widely used in this country. Great Britain and Israel are two countries with considerable experience with the method. In general, the food account gives a record of food entering the house or institution in a given period but not whether it is actually consumed in that period. It is less accurate than other household or institution methods but is felt by reviewers to give a good picture of the general diet. There is usually no indication of food discarded for spoilage, plate waste, other waste or fed to animals. Sampling can be large since less is demanded of respondents. However,

OCR for page 87
194 D. D. SCHNAKENBERG et. al. At sea, there is insufficient refrigerated storage space to keep fresh milk available for periods of more than =7 days. This problem is described in Figure 5a, where the ship ran out of fresh milk on July 15, was resupplied TABLE 12 Food-Type Consumption of Subjects with Low, Marginal, or Adequate Vitamin A Intakes, uss Saratoga, July-August 1977 Quantity (g/day) Consumed by Subjects Whose Daily Vitamin A Intakes Were: Lowa Marginalb AdequateC No. of subjects 41 63 99 Milk 137 236 178 Cheeses and ice cream 19 29 31 Tomatoes and tomato products 9 25 38 Carrots, raw and cooked 0.2 1.5 8 Sweet potatoes 0.3 1.8 2.0 Liver 0 0 1 .3 Leafy and green vegetables 6 14 31 Melons, peaches, plums 11 12 28 Eggs 24 41 42 Potatoes, french fries 8 11 11 a Nutrient ratio <0.7. b Nutrient ratio 0.7 to < 1 . 0. c Nutrient ratio ~1 . 0. DAIRY PRODUCTS AND BEVERAGE CONSUMPTION DAILY TOTALS 2000- TRANSIT | PORT , 1 500 - 6 CD ~ 1 000 - - z C' 500- | F Ll G HT OPS DAIRY PRODUCTS OTHER BEVERAGES O- ~ / / ~ 3 5 7 2 14 6 18 20 22 24 26 28 30 ~ 1977 AUGUST USS SARATOGA FIGURE Sa

OCR for page 87
Diary-Interview Technique with filled milk while in port, and ran out again on July 31. The consump- tion of carbonated and noncarbonated beverages markedly increased when milk was not available. The percentage of the population with low and marginal intakes of calcium (Figure Sb) and riboflavin (Figure 5c) substan- tially increased on days when milk was not available. At our urging, a low-fat (25-30 percent fat calories) vitamin A-fortified (30 percent RDA per 195 CALCI U M DAILY TOTALS TRANSIT | PORT | FLIGHT OPS Rn ~ Z O O ~ ,_ 60 Z ~ LLI J UJ o ~ to 40 o ~ 1 2 1 4 1 6 1 8 20 22 24 26 28 30 1 JULY 1977 USS SARATOGA AUGUST OADEaUATE MARGINAL LOW FIGURE Sb RIBOFLAVIN DAILY TOTALS 100 - 80 - ~ Z O O 60- Z ~ 111 A O ~ to ~ O Q ~ 40 - 20 12 14 16 18 20 22 24 26 28 30 1 JULY 1977 USS SARATOGA . TRANSIT | PORT | FLIGHT OPS 3 6 7 AUGUST DADEQUATE ~ MARGINAL ~ LOW FIGURE Sc

OCR for page 87
196 D. D. SCHNAKENBERG ET. AL. serving) milk shake, which is reconstituted aboard ship from a dry shelf stable base, is being tested on the uss Saratoga. Based upon the food-type selection patterns previously shown in Table 12, we have projected that this product will effectively reduce the incidence of low and marginal intakes of calcium, riboflavin, and vitamin A as shown in Table 13. One of the pri- mary purposes of our current follow-up study on the uss Saratoga is to assess the validity of our projections. As a result of our recent experiences from seven dietary assessment studies in a variety of military settings and our attempts to validate our methodology, we have made some modifications to our dietary-interview technique. Most significantly, we have reduced the time the participants are asked to maintain their food diaries from 14-17 days to no more than 7-8 consecutive days. We believe that 7-8 days is long enough to obtain a reliable sample of an individual's eating patterns and food selection habits, but short enough to maintain a cooperative attitude in most of the partici- pants. It is still essential though that a diary review session be scheduled 3~ days after the initial interview. We are incorporating some sort of validation procedures in each of our studies, most of which focus on obtaining an estimate of the energy expenditure level of each subject. However, we are keenly aware that the magnitude of the error in estimation of energy expen- TABLE 13 Projected Nutritional Impact of Offering Vitamin A-Enriched Milk Shakes,a uss Saratoga Percentage of Population Nutrient Calcium intake Jul.-Aug. 1977 Projectedb Riboflavin intake Jul.-Aug. 1977 Projectedb Vitamin A intake Jul.-Aug. 1977 Projectedb c Low Marginal Adequate 2 21 77 2 10 88 1 24 75 1 11 88 20 3 1 49 5 13 82 a Shake mix enriched to provide 1,800 flu vitamin A per 12.5 fl. oz. serving. b projections based upon Jul.-Aug. 1977 data selecting only the days when milk was available. c Based upon presumption that individuals who consumed at least one glass of milk per day when available will consume one enriched milk shake per day. NOTE: These projections assumed that milk shakes will be available at both forward and aft galleys.

OCR for page 87
Diary-lnterview Technique 197 diture and short-term caloric deficit or surfeit of individuals may be as large as the error in estimating energy consumption. Therefore, we are beginning to concentrate more on utilizing a nutrient ratio, nutrient balance, or, if you prefer, nutrient density approach to the analysis and interpretation of the data rather than limiting our interpretations to reported average daily quan- tities of nutrients consumed. We are rapidly expanding our computer pro- gramming capabilities so that we can more fully utilize the information contained in our data bank. Mr. Morris has developed the necessary software to retrieve individual nutrient intake data by demographic and anthropometric characteristics, by date, by time (breakfast, lunch, supper, etc.), by source (home, restaurant, vendor, dining hall, etc.), or by combi- nation of several variables. We can also retrieve, by individual or group, the quantity of any food item (limited to the ~2,500 different food items currently on our Nutrient Factor File), or food type consumed. Dr. Kretsch has recently overseen the addition of cholesterol, animal/plant protein, and animal/plant/marine fat values for nearly all food items on the file and a significant number of vitamin E, vitamin B6, vitamin Bit, folio acid, mag- nesium, copper, zinc, manganese, pantothenic acid, and fatty acids values have been added. These improvements in our data-processing and retrieval system should significantly contribute to the knowledge base of nutrient intake patterns of not only military personnel but our young adult population in general. REFEREN CES Departments of the Army, the Navy, and the Air Force. 1976. Army Regulation 40-25, BUMED Instruction 10110.3E, and Air Force Regulation 160-95. Medical services nutritional stan- dards. Departments of the Army, the Navy, and the Air Force, Washington, D. C., 30 August 1976 (as corrected). Gersovitz, M., J. P. Madden, and H. Smiciklas-Wright. 1978. Validity of the 24-hr. dietary recall and seven-day record for group comparisons. J. Am. Diet. Assoc. 73:48.

OCR for page 87
Possible Alternative Me~ods for Data Collection on Food Consumption and Expenditures ROBERT B. PEARL There are few behavioral patterns in our society that are more extensively documented than the purchase and consumption of food products. Countless practitioners, both inside and outside of government, are engaged in pro- ducing statistics on this subject either on a continuous or intermittent basis. Moreover, there are probably almost as many different methodologies and techniques employed in gathering these data as there are participants in the field. In view of these facts, it would be surprising if almost every possible means of developing these statistics had not already been attempted or at least considered. This likelihood poses a problem for anyone charged with presenting challenging and innovative ideas along these lines. The tempta- tion exists to opt for exotic possibilities, such as using spy satellites to chart the eating behavior of the population. The writer intends to resist this temp- tation in favor of a few notions that, if not entirely novel, at least do not appear to have entered the mainstream of statistical endeavors in this sector. In the main, these will be aimed at some of the principal deficiencies that are believed to exist in nutritional and dietary data. INTERMITTENT PANELS One of the limitations of the methodologies customarily employed in studies of food consumption is the relatively short period of observation. Because of cooperation problems and concern about reporting errors, comprehensive family consumption data have usually been restricted to a period of 1 week or less and individual intake data to a span of a very few days. Even food 198

OCR for page 87
Possible Alternative Methods for Data Collection 199 purchase or expenditure information has rarely been collected for more than a 2- to 4-week period. Assuming samples of sufficient size that are adequately distributed over time, these restrictions should not impair such measures as average family or individual consumption. The limitation arises when an effort is made to provide distributions of families and persons by nutritional status. Since studies have indicated that dietary inadequacies exist in every socioeconomic group, the focus may be more on distributions and individual differences than on group averages. In view of the variations that are likely to occur in food consumption from day to day, week to week, or season to season, it is difficult to see how observation for only a brief period can validly depict the nutritional status of any one family or person. The preferable course would probably be to observe each subject for an extensive, continuous period, perhaps up to a year. Unfortunately, experi- ence has indicated that samples used for continuous expenditure surveys become highly biased because of attrition (Quackenbush and Shaffer, 1960), and there would be even less hope for success in the case of indi- vidual food consumption measurement. One alternative would be to attempt an intermittent panel procedure, whereby the same subjects would be can- vassed at various points of the year, but for a relatively brief period each time. The data from the various observations would be accumulated to provide the measure of nutritional status. Although some further ex- perimentation might be needed, a canvass once each quarter could be adequate for a particular case. Some analyses have indicated that a 3-week period of observation each time for family consumption might be the op- timum in this type of design, balancing the need for stable measurements against that of avoiding undue sample attrition or fatigue (Ferber and Sud- man, 1971; Report and Recommendations, 1976~. For a specific family, the observation period could be randomly scheduled to occur at different times each quarter in order to cover a greater variety of situations. The designation of a 3-week observation period would probably permit the use of a food purchase approach in data collection, in lieu of the more complex food consumption procedure, since the two are likely to balance out for periods of that length (especially where there would be four quarterly observations to aggregate). A food purchase approach, in turn, would likely entail the use of diary keeping, which has been found to be relatively successful for that purpose (Pearl, 1977, 19781. For individuals as opposed to families, the procedure would probably call for maintaining records of consumption for possibly a 3-day period randomized within each 3-week cycle; the family data could then be used as a control (for home consump- tion) over the individual data. The use of a panel always raises problems of securing adequate coopera-

OCR for page 87
200 ROBERT B. PEARL lion over time. An appropriate incentive system might well be necessary, with a modest reward for completing any one quarterly effort and a sizeable bonus for persisting through the entire cycle. Since panel members could move from quarter to quarter, some follow-ups would be required each time, but these are not likely to exceed 5 percent of the total at any one point. UNI VERS AL PRODUCT CODES Another deficiency in much of the information that has been collected in this field, whether through diaries or interviews, has been inadequate and incon- sistent descriptions of products. This problem is especially acute where conversion to a nutrient basis is the objective. The proliferation of prepared foods, containing a variety of ingredients, obviously adds to the difficulty. One possible way of reducing this problem would be to make use of the Universal Product Codes (uPc) that now appear on nearly all canned and packaged food products and even on some fresh items that are repackaged in stores. Experimentation would be needed to determine whether survey re- spondents could be induced to enter these codes in diaries, in addition to a brief product description, and, moreover, whether they could do so accu- rately. If this could be achieved, and assuming that the necessary data on ingredients could be obtained from the respective manufacturers and stored in computers, conversion to a nutrient basis could become a routine matter. Moreover, consistency and reliability in reporting would be materially ad- vanced and clerical coding and processing costs could be minimized. AUTOMATED CHECKOUT INFORMATION Another technological development of major interest, still limited in scope but offering promise for the future, is the appearance of automated checkout systems in food stores based on the Universal Product Codes. As these become more prevalent, they could serve as the basis for obtaining current food sales data in a level of detail never before contemplated. Although sales data are of only limited value for purposes of measuring family or individual food consumption, they could supply control totals for usually deficient survey data or at least provide a standard for evaluating survey results. Moreover, if all purchases are made by credit card in some imagina- ble future era, the combination of the sales data from mechanized systems and the personal characteristics of credit card holders in the possession of various financial agencies could produce the best of all possible statistical worlds. (This target may sound like " 1984," but is not likely to occur by that date.) Perhaps a more realistic application would be the use for survey purposes of the individual cash register tapes produced by the automated systems,

OCR for page 87
Possible Alternative Methods for Data Collection 201 which generally provide a far more adequate description of the products purchased than has heretofore been available. These tapes could serve as a basic reminder to survey respondents in filling diaries or responding in interviews and, in some cases, could even serve as a substitute for complet- ing parts of diaries. PORTABLE TAPE RECORDERS No matter how carefully surveys of this kind are planned, serious reporting errors are likely to occur because of memory biases, tedium in filling diaries, and other reasons. A possible aid in overcoming these deficiencies would be to provide respondents with small, pocket-size tape recorders to use during the survey period. Respondents would then be asked to record purchases as they are made in the store or are being unpacked at home (possibly including uPc numbers) or, alternatively (if the consumption ap- proach is being used), as various items are being used in food preparation. Moreover, individuals within families might be given such recorders in which to report meals and snacks purchased outside the home. The use of recorders could be especially useful in the case of less literate respondents. DIARY CHECKING PROCEDURE Some recent studies have indicated that there may be considerable dis- parities in the relative accuracy and completeness of reporting various food products in diaries. For example, comparatively costly items, such as meat and poultry, and perishable and frequently used products, such as milk, eggs, and bread, seem to be rather well reported, whereas food staples such as flour, sugar, and shortening tend to be seriously understated (Pearl, 19784. There appears to be a need, therefore, for special checking proce- dures to correct for these imbalances. Some general checking procedures are specified in most diary operations, but what may be needed are specific sets of check questions aimed at those product categories most commonly underreported. In fact, the entire process may become much more of a combination of diaries and interviews than has usually been the case. A companion approach might be to take a brief shelf inventory of these types of staple products, at the time the diaries are being collected, and to inquire about the date of purchase of any that are found in order to determine whether the information was properly recorded. WAS T A GE It is difficult to review the literature without finding a good deal of hand wringing about the inability to measure waste in food use (Pearl, 1977~.

OCR for page 87
202 ROBERT B. PEARL This writer has no inspired solutions to offer for this long-standing problem. Instead of the rather futile attempts to measure this element as part of a regular consumption survey, however, it might be more useful to carry out a series of small-scale studies devoted specifically to finding out how much of various food items are ordinarily wasted. Some spot visits to small house- hold samples might be made right after the various meals to obtain quick information of this kind. Respondents might also be asked to record wastage on portable tape recorders, etc. In any event, the objective would be to develop adjustment factors that could be used to discount the usual purchase or consumption data prior to conversion to nutrients. MEALS EATEN OUT Another matter of some discouragement has been the impracticality of ob- taining reliable information from survey respondents on the contents of meals eaten out. Since one out of every three food dollars is now spent in this manner, and the ratio is expected to rise to one in two, this problem is rather central in any effort to measure dietary adequacy. One possibility, in addition to obtaining whatever descriptions can be provided by respondents of meals eaten out, would be to record the names and addresses of the eating places. The objective (and this again might be attempted only for small samples) would be to solicit the recipes and lists of ingredients directly from the eating places (hopefully avoiding trade secrets). From a succession of such investigations, a glossary could possibly be developed thatat least in nutrient form could describe a large proportion of meals eaten out (the preponderance of meals eaten in syndicated fast food establishments could help in this respect). Although obviously not without limitations, this ap- proach could possibly improve what is one of the weakest links in the chain of dietary measurements. INDICATOR FOODS One of the problems in many complex surveys is that so vast an array of information is solicited that respondents experience memory failures or fail- ures of execution, and survey personnel become so lost in the details that they are unable to institute adequate checking procedures. An apparent solution would be to reduce the amount of information to be requested from any one respondent, that is, in our immediate field of interest, to ask only about selected foods instead of the entire gamut. Since we are interested in the overall dietary adequacy of individual families and persons, the question is how to accomplish a reduction in detail without defeating the main objec- tives of the inquiry.

OCR for page 87
Possible Alternative Methods for Data Collection 203 One possibility would be to attempt to develop a set of "indicator" foods, that is, a selection of foods that when translated into nutrient form would approximate the nutrient value of the complete diet of a given family or individual. This approach would represent a form of sampling similar to that used in estimating the overall price index from only a "market basket" of items. The data for the indicator foods would, of course, have to be appro- pnately weighted to arrive at a total nutrient value. In fact, regression equations might be developed to aid the estimation. Various means might be utilized in the development of indicator lists. The knowledge and experience of dietary specialists would represent an impor- tant input. Experiments could be conducted with complete data sets (such as those from the ARS Food Consumption Surveys (see Murray, 1975), whereby venous successive samples of items could be drawn and compared in nutrient value with the total consumption of the family or individual. For every item on an indicator list, there could be two or three options, in case the family did not use a particular product. Moreover, it would probably be necessary to develop different indicator lists for the various geographic regions, socioeconomic groups, and season of the year. REFERENCES Ferber, R., and S. Sudman. 1971. Experiments in obtaining consumer expenditure data by diary methods, J. Am. Stat. Assoc. 66:336, December. Murray, J. 1975. Household food consumption surveys of the U.S. Department of Agriculture: A historical review of data collection methods, 1894-1969. Agricultural Research Service. Pearl, R. B. 1977. Data systems on food demand and consumption: Properties, uses, future prospects. Pages 116-139 in Food demand and consumption behavior. Workshop spon- sored by the S-119 Southern Regional Research Committee, State Agricultural Experiment Stations, and the Farm Foundation, Atlanta, March. Pearl, R. B. 1978. The 1972-73 U.S. Consumer Expenditure Survey: A preliminary evalua- tion. U.S. Bureau of the Census Technical Paper Series. In press. Quackenbush, G. G., and J. D. Shaffer. 1960. Collecting food purchase data by consumer panel. Mich. State Univ. Agric. Exp. Stn. Tech. Bull. 279. East Lansing, Mich. Report and recommendations to the food and nutrition service on a proposed national food stamp program consumer panel survey. April 1976. Unpublished.

OCR for page 87