Consumer Food Safety Knowledge, Practices, and Demographic Differences: Findings from a Meta-Analysis

Transcription

1 8 Journal of Food Protection, Vol. 68, No. 9, 2005, Pages 8 9 Copyright, International Association for Food Protection Consumer Food Safety Knowledge, Practices, and Demographic Differences: Findings from a Meta-Analysis SUMEET R. PATIL,* SHERYL CATES, AND ROBERTA MORALES RTI International, 00 Cornwallis Road, Research Triangle Park, North Carolina 27709, USA MS 0-570: Received 15 December 200/Accepted 26 March 2005 ABSTRACT Risk communication and consumer education to promote safer handling of food can be the best way of managing the risk of foodborne illness at the consumer end of the food chain. Thus, an understanding of the overall status of food handling knowledge and practices is needed. Although traditional qualitative reviews can be used for combining information from several studies on specific food handling behaviors, a structured approach of meta-analysis can be more advantageous in a holistic assessment. We combined findings from 20 studies using meta-analysis methods to estimate percentages of consumers engaging in risky behaviors, such as consumption of raw food, poor hygiene, and cross-contamination, separated by various demographic categories. We estimated standard errors to reflect sampling error and between-study random variation. Then we evaluated the statistical significance of differences in behaviors across demographic categories and across behavioral measures. There were considerable differences in behaviors across demographic categories, possibly because of socioeconomic and cultural differences. For example, compared with women, men reported greater consumption of raw or undercooked foods, poorer hygiene, poorer practices to prevent cross-contamination, and less safe defrosting practices. Mid-age adults consumed more raw food (except milk) than did young adults and seniors. High-income individuals reported greater consumption of raw foods, less knowledge of hygiene, and poorer cross-contamination practices. The highest raw ground beef and egg consumption and the poorest hygiene and cross-contamination practices were found in the U.S. Mountain region. Meta-analysis was useful for identifying important data gaps and demographic groups with risky behaviors, and this information can be used to prioritize further research. Effective risk communication to inform consumers of the possible health risks of foodborne illnesses and to encourage safer food handling practices in the home probably is the best way to ensure food safety at the consumer end of the food chain (15, 0). Food handling practices differ by gender, ethnicity, age, income, and other demographic characteristics (, 19). Redmond and Griffith (0) found that compared with women, men are less knowledgeable about food safety and have riskier hygiene and cooking practices. For effective risk communication, one must understand the behavioral differences between various subpopulations and target risk communication efforts accordingly (6, 2). Food handling practices for different subpopulations have been evaluated in various studies. However, these studies often target specific behaviors and subpopulations and thus fail to provide a holistic assessment of consumer behaviors based on demographic characteristics. Combining information from several studies is the best way to produce an aggregate measure of consumer behavior. This aggregate measure can be used to evaluate how food handling practices differ by subpopulation. Using meta-analysis methods, researchers can synthesize and compare results from multiple studies and conduct statistical analyses. * Author for correspondence. Tel: ; Fax: ; spatil@rti.org. Present address: Animal Health and Plant Inspection Service, U.S. Department of Agriculture, 1017 Main Campus Drive, Raleigh, NC 27606, USA. Meta-analysis refers to the statistical analysis of a large collection of results from individual studies, such as opinion surveys, experimental studies, and causal models, to integrate the findings (1). Meta-analysis has been used widely in education, psychology, and clinical trial research (8, 27, ). Despite its advantages, its use has been limited in food safety consumer research. The primary advantage of meta-analytical methods over traditional qualitative review methods is the precision introduced into the analysis with respect to the (i) specific purpose of the analysis, (ii) selection of specific studies for the analysis, (iii) similarity of units of analysis and subject matter, and (iv) distribution of study values (10). Other advantages, such as an emphasis on studies with smaller sample sizes, computational efficiency, and ease of various analyses, have been discussed by Patil et al. (28). Unlike meta-analysis, traditional qualitative methods are sensitive to issues such as the social context of the study and theoretical factors. Possible limitations of the application of meta-analysis in food safety consumer research include low statistical power if only a limited number of studies are available, lack of accuracy by defining study variables too broadly or an insufficient number of studies in the case of narrowly defined variables, and insensitivity to a study s social context or theoretical influences, which could be considered in a qualitative review (28). To draw additional inferences, meta-analysis and qualitative review methods can be conducted on the same body

2 J. Food Prot., Vol. 68, No. 9 META-ANALYSIS OF CONSUMER BEHAVIORS 85 TABLE 1. Example of a study index table, study index 6 Sample size Consumer behavioral measures Demographic category No. of respondents Study Meta-analysis Effect size (% respondents) Caucasian Hispanic African American Asian Young adult Mid-age adult Senior adult of research (5). However, in this study we focused on meta-analysis only and did not address qualitative review methods. The purpose of this study was to use meta-analysis to (i) identify data gaps where additional research is needed to fully evaluate consumers risky food handling and consumption practices; (ii) evaluate differences in knowledge versus reported practices for various consumer behaviors and demographic subpopulation categories (i.e., gender, ethnicity, age, income, education, geographical region, and metropolitan status); and (iii) evaluate differences in reported risky behaviors among demographic subpopulations. MATERIALS AND METHODS Rosenthal (1), Cook et al. (9), Lipsey and Wilson (21), and Hedges et al. () have provided a detailed explanation of the general meta-analysis methodology. Patil et al. (28) discussed the meta-analysis methodology as applied to food safety consumer research, with its advantages and limitations. The procedure for meta-analysis employed in this study is briefly described here. Problem specification. The first step is to clearly define the problem to guide the search and selection of studies for metaanalysis. The problem statements for this study were as follows. What are the demographic characteristics and proportions of U.S. consumers that engage in risky food handling and consumption practices? Are there any differences in what people know and their reported practices? Are there any differences among demographic groups in risky behaviors? The next step is to develop criteria to guide the inclusion or exclusion of various studies in the meta-analysis. This meta-analysis included studies that evaluated U.S. consumers knowledge and/or reported behaviors in units of mean percentages or proportions. These percentages or proportions are referred to as effect sizes (ESs). To be eligible for inclusion in the meta-analysis, studies had to be published since 1992, provide sample size information, and report the demographic characteristics of the respondents. The following behavioral measures were included in the meta-analysis: (i) consumption of raw or undercooked ground beef, eggs, shellfish, and milk (we selected these foods because the U.S. Department of Agriculture and the Food and Drug Administration have recommended that consumers do not eat these foods raw or undercooked because of the possible risk of foodborne illness, and many surveys have been conducted to collect information on consumption of these foods); (ii) knowledge of good hygiene practices, practices to prevent cross-contamination, proper defrosting methods, apparently safe food sources, and proper cooking and heating practices; and (iii) handling practices for hygiene, prevention of cross-contamination, food holding, cold storage, avoidance of unsafe foods, and cooking and heating. The following demographic characteristics were included in the metaanalysis: (i) gender, (ii) ethnicity (Caucasian, Hispanic, African American, Asian, others), (iii) age (young, to 29 years; midage, 0 to 5; senior, 55), (iv) education (less than high school, high school, more than high school); (v) geographic region (Pacific, Mountain, West North Central, West South Central, East North Central, East South Central, Middle Atlantic, South Atlantic), and (vi) metropolitan (metro, 1,000,000 population) versus nonmetropolitan (nonmetro, 1,000,000 population). Data collection. After determining the eligibility criteria, studies were retrieved from various journals and surveys from government and private organizations. Twenty studies were eligible for inclusion in the meta-analysis. Study indices were identification numbers assigned to each study (or survey) included in the analysis (1, 7,, 1, 16, 19, 20, 22, 26, 29,, 7, 9 1). Two studies (, ) each included two different surveys. For each study, data on sample size and ES separated by the demographic characteristics of interest were collected and stored in tables according to individual study indices as shown as an example in Table 1 for study index 6. Using these tables, data from all studies on a specific behavior were combined as shown as an example in Table 2 for raw or undercooked ground beef consumption. Data analysis. After collecting, indexing, and arranging the data in the proper format, the ES, standard error (SE), and the inverse variance weight ( ) for each study were calculated for each risky behavior and demographic category combination (28). The inverse variance weight for each study was calculated as 1 inverse variance weight ( ) (1) SE 2 The mean ES then was calculated by averaging individual ESs weighted by inverse variance weights. Similarly, the SE of the mean ES was calculated by averaging SEs of individual effects weighted by inverse variance weights. Patil et al. (28) discussed the advantages of using inverse variance weights in averaging. The significance of the mean ES was confirmed using the z-test. The homogeneity of the ES distribution (whether ESs for all studies are based on a similar population) was evaluated using the Q- test (17, 2). If the Q-value exceeded a critical value based on the

3 86 PATIL ET AL. J. Food Prot., Vol. 68, No. 9 TABLE 2. Data collection for raw or undercooked ground beef consumption Demographic category Sample size Total sample ,15 1,620,97 Gender Ethnicity Caucasian Hispanic African American Asian Age (yr) Young ( 29) Mid-age (0 5) Senior (55 ) Income a Low Middle High Education Region b Pacific Mountain West North Central West South Central East North Central East South Central Middle Atlantic South Atlantic New England Metropolitan status d Metro Nonmetro 616 7, ,781 15,082 1,2 1,526 2 Study index 6 15 Effect size (% respondents) , , ,8 2,8 9,675,92 0 2, , ,722 NA c 2,61,682 NA 2,2 2,00 5,626,5 NA 1,276, a Annual household income: low income, $15,000; middle income, $15,000 $9,999; high income, $50,000. b Regions: Pacific (California, Hawaii, Alaska, Washington, Oregon), Mountain (Montana, Idaho, Wyoming, Nevada, Utah, Colorado, Arizona, New Mexico), West North Central (North Dakota, Minnesota, South Dakota, Nebraska, Iowa, Kansas, Missouri), West South Central (Oklahoma, Arkansas, Texas, Loui- chi-square distribution for a behavioral measure, then the ES distribution for that measure was heterogeneous. For heterogeneous distributions, the total variance of ES ( * i ) was calculated as * i i (2) where is the overall random variance between all included studies and i is the variance associated with subject-level sampling error. Mean ES, the SE of mean ES, the significance of ES, and the Q statistic are then recalculated using * i. For homogeneous ESs, only i was used. Hypotheses testing. To evaluate whether the difference between knowledge and the reported use of a safe handling practice for a demographic category was significant, we used the z-test. The z-value was calculated as ES know ES use z () 2 2 SEES SEES N N know where ES know and ES use are the mean ESs of knowledge and reported use, N is the total sample size, and SE (which is * ES i for heterogeneous ES and i for homogeneous ES) is the SE of the mean ES. When the z-value was greater than 6, then the difference between knowledge and reported use of a safe handling practice was significant at To evaluate whether differences in behaviors among demographic groups, such as gender and ethnicity (Caucasian, Hispanic, African American, Asian), were significant, we used an analysis of variance (ANOVA; PROC ANOVA 1999, SAS Institute, Cary, N.C.) to perform multiple comparisons. When the P-value was less than 0.05 for a comparison, then the difference in the behavioral measure for the demographic category analyzed was significant at RESULTS Using the 20 study index tables (see Table 1 for example for study index 6), tables for each behavioral measure of interest were created (see Table 2 for the example of raw or undercooked ground beef consumption). Each table provided information on sample size and ES for the demographic characteristics of interest. For each combination of behavioral measure and demographic category, the number of eligible studies ranged from 1 to 10. Mean ESs and the corresponding SEs for each behavioral measure are reported in Tables and, respectively. Knowledge versus use. Table 5 lists the difference values between knowledge and reported use of safe food handling practices. For proper defrosting methods and food holding practices, information on consumer knowledge and/ siana), East North Central (Wisconsin, Michigan, Illinois, Indiana, Ohio), East South Central (Kentucky, Tennessee, Mississippi, Alabama), Middle Atlantic (New York, Pennsylvania, New Jersey), South Atlantic (West Virginia, Virginia, North Carolina, South Carolina, Georgia, Florida), and New England (Vermont, Maine, New Hampshire, Massachusetts, Connecticut, Rhode Island). c NA, sample size information was not available; hence, effect size is not reported. d Population of county of residence: metro, 1,000,000; nonmetro, 1,000,000. use

4 J. Food Prot., Vol. 68, No. 9 META-ANALYSIS OF CONSUMER BEHAVIORS 87 or safe handling practices was inadequate to calculate the difference between knowledge and reported use; hence, these behavioral measures are not included in Table 5. Demographic categories of ethnicity and metropolitan status also are not presented because of a lack of data. For all behaviors evaluated in the meta-analysis, consumer knowledge of safe handling practices does not correspond with reported use of the practices, suggesting that knowledge is a poor indicator of actual behavior. For most demographic categories, consumers knowledge of good hygiene practices and practices to prevent cross-contamination exceeded their reported use of the corresponding practice. Knowledge of good hygiene practices exceeded reported use of such practices by 10% for the total sample. The largest difference between knowledge and reported use was observed for individuals with more than a high school education (25.6%), men (2.9%), and mid-age adults (21.8%). For low-income individuals and individuals without a high school education, reported use of good hygiene practices exceeded knowledge of safe practices. Knowledge of practices to prevent cross-contamination exceeded reported use of such practices by.2% for the total sample. The largest difference between knowledge of practices to prevent cross-contamination and reported use was observed for men (1.1%) and young adults (1%). For women, seniors, and individuals without a high school education, reported use of practices to prevent cross-contamination exceeded knowledge of safe practices. For most demographic categories, consumers reported use of practices for proper cold storage, avoiding foods from unsafe sources, and proper cooking and heating exceeded their knowledge of safe practices. The largest difference between knowledge and reported use was observed for avoiding foods from unsafe sources (e.g., not using food from open or damaged packaging and following recommended storage times); reported use of such practices exceeded knowledge of the practices by 2.% for the total sample. Reported use of practices for proper cold storage exceeded knowledge of safe practices by 11.% for the total sample. Reported use of proper practices for cooking and heating exceeded knowledge of safe practices by 1% for the total sample. The largest differences between knowledge and use of practices for proper cooking and heating were observed for education. For individuals without a high school education, reported use exceeded knowledge by %; conversely, for individuals with more than a high school education, knowledge exceeded reported use by %. Differences in reported behaviors and knowledge by demographic categories. An ANOVA was used to evaluate differences in behavior among demographic categories. All differences were significant at the 0.05 level except those for good hygiene practices in a few demographic categories. In the case of reported use of good hygiene practices, only the difference between Caucasians and Hispanics and the difference between African Americans and Asians were not significant. Although differences among various demographic categories for a specific behavioral measure are not reported explicitly, they can be calculated using the respective columns in Table. For example, consumption of raw or undercooked ground beef ranged from 10% in the East South Central region to 29% in the Mountain region. Women reportedly consumed 5.5% less raw or undercooked ground beef than did men (26.7 versus 21.2%). African Americans reported the lowest consumption, and Caucasians reported the highest consumption. Mid-age adults reported the highest consumption (2.7%) compared with other age groups. The high-income group reported the highest consumption (29%) compared with the low-income group (%). Individuals with less than a high school education reported lower consumption than did individuals with a high school education or more. Individuals residing in metropolitan areas reported higher consumption than did those in nonmetropolitan areas. Therefore, men, Caucasians, mid-age adults, individuals with a high school education or higher, individuals in the Mountain region, and individuals residing in metropolitan areas or cities reported the highest consumption of raw or undercooked ground beef. Because the demographic categories are not mutually exclusive, it is not possible to generalize about demographic groups engaging in unsafe behaviors (e.g., we cannot generalize that men living in the Mountain region consume the most raw ground beef). For example, results for the male category represent men from all geographical regions. Likewise, the results for geographical regions represent both men and women. Information on the proportion of men for each geographical region was not available; therefore, we could not provide results for men by geographical region, and we refrain from identifying any combinations of demographic characteristics engaging in unsafe behaviors. More people consumed raw or undercooked eggs (7%) than consumed raw or undercooked ground beef (21%), shellfish (%), and raw milk (%). Consumption of raw or undercooked food varied by gender, ethnicity, age, income, education level, and region. Consumers were more knowledgeable about good hygiene practices (88%), practices to prevent cross-contamination (86%), and safe food holding practices (8%) than they were about the other practices included in the analysis. Reported use of good hygiene practices (78%) and proper cooking and heating practices (77%) were more widespread compared with other practices included in the analysis. Based on the available data, we can assess differences in knowledge and reported use of good hygiene practices, practices to prevent cross-contamination, and proper cooking and heating practices by gender, age, and education level. Data are available for some practices included in the analysis to assess differences in the reported use of the practice by various demographic groups. Table 6 summarizes differences in reported knowledge and behavior by demographic category. For each combination of behavioral measure and demographic category, the demographic groups with the highest prevalence () and lowest prevalence () are noted.

5 88 PATIL ET AL. J. Food Prot., Vol. 68, No. 9 TABLE. Meta-analysis matrix for mean effect sizes for behavioral measures a Demographic category Raw or undercooked food consumption Ground Shellfish beef i Eggs j k Milk Good hygiene b Prevent crosscontamination c defrosting d Safe food holding e cold storage f Avoid unsafe food source g cooking, heating h Total sample Effect size 21.1 No. of studies l 6 Gender Ethnicity Caucasian Hispanic African American Asian Other Age (yr) Young ( 29) Mid-age (0 5) Senior (55 ) Income m Low Middle High Education Region m Pacific Mountain West North Central West South Central East North Central East South Central Middle Atlantic South Atlantic New England Metropolitan status m Metro Nonmetro

6 J. Food Prot., Vol. 68, No. 9 META-ANALYSIS OF CONSUMER BEHAVIORS 89 TABLE. Continued a Values are percentage of respondents. b Includes washing hands before and after handling raw meat or poultry, before and after food preparation, and after bathroom use; colonized persons not handling foods; and washing raw vegetables, fruit, and prepackaged salads before eating. c Includes washing cutting boards before reuse; washing counters with soap and water; not using same plate for raw and cooked meat; not commingling raw foods; and refrigerating raw meat, poultry, and other foods separately. d Includes thawing food in a refrigerator or microwave. e Includes holding cooked food at room temperature no longer than recommended time. f Includes storing foods at proper cold temperatures, cooling cooked foods rapidly in a refrigerator, and refrigerating leftovers in small shallow containers. g Includes not using food from damaged or open packaging and not using improperly frozen food, foods aged beyond the expiration date, food that has been improperly held or stored, or food refrigerated beyond recommended time. h Includes cooking meat and poultry and reheating leftovers to recommended internal temperature measured appropriately. i Includes raw or rare ground beef, hamburgers served rare, meat loaf served rare, and steak tartare. j Includes raw or undercooked eggs eaten whole or as ingredients in other unheated preparations (e.g., cookie dough, egg nog, custards, homemade ice cream). k Includes raw or undercooked oysters, clams, and other molluscan shellfish. l The total number of studies used to estimate the total sample ES can be higher or lower than the number of studies used for specific demographic categories. Studies used for estimating the total sample ES may or may not include studies used for other demographic categories. Therefore, the ES for the total sample can be higher or lower than the average ES across other demographic categories. m For complete explanation, see Table 2 footnotes. There were differences in knowledge and behavior with regard to gender. Compared with women, men reported greater consumption of raw or undercooked foods, less knowledge and reported use of good hygiene practices, less knowledge and reported use of practices to prevent cross-contamination, lower reported use of proper defrosting methods, and lower reported use of practices to identify safe food sources. Men were more knowledgeable than women regarding proper cooking and heating practices; however, their reported use of such practices was lower than that of women. Differences in behaviors according to ethnicity were most apparent for consumption of raw or undercooked foods. For example, African Americans reported the lowest consumption of raw or undercooked foods, and Hispanics reported the highest consumption of raw milk. Analysis of age groups provided mixed results. Midage adults consumed more raw or undercooked foods (except milk) than did young adults and seniors. With the exception of proper cooking and heating, young adults had the poorest safe handling practices and were the least knowledgeable about such practices. Consumers in the high-income category had more widespread consumption of raw or undercooked ground beef and shellfish compared with low- and middle-income groups. Individuals with high incomes also were the least knowledgeable about good hygiene practices and practices to prevent cross-contamination. Individuals without a high school education reported the lowest consumption of raw or undercooked ground beef and shellfish compared with individuals with a high school education or more. They also reported safer handling practices regarding hygiene, practices to prevent cross-contamination, and proper cooking and heating. However, in contrast, individuals without a high school education had the least amount of knowledge of the various safe handling practices. With respect to geographical region, the Mountain region had the highest reported raw or undercooked ground beef and egg consumption and the lowest reported use of good hygiene practices and practices to prevent cross-contamination. Raw or undercooked shellfish consumption was highest in the South Atlantic, and raw milk consumption was highest in the West South Central compared with other regions. Individuals living in metropolitan areas reported the highest consumption of raw or undercooked ground beef and shellfish compared with those in nonmetropolitan areas. DISCUSSION One of the goals of this meta-analysis was to identify areas that need further research to completely describe consumer behaviors. As Table indicates, data on reported consumption of raw or undercooked foods, good hygiene practices, and practices to prevent cross-contamination are more readily available than data on other measures such as proper defrosting, practices to identify safe food, and proper cooking and heating methods. More information also is available on the reported use of safe practices than on

7 90 PATIL ET AL. J. Food Prot., Vol. 68, No. 9 TABLE. Meta-analysis matrix for standard errors for mean effect sizes a Demographic category Ground beef Eggs Raw or undercooked food consumption Shellfish Milk Good hygiene Prevent crosscontamination defrosting Safe food holding cold storage Avoid unsafe food source cooking, heating Total sample SE for mean effect size No. of studies Gender Ethnicity Caucasian Hispanic African American Asian Other Age (yr) Young ( 29) Mid-age (0 5) Senior (55 ) Income Low Middle High Education Region Pacific Mountain West North Central West South Central East North Central East South Central Middle Atlantic South Atlantic New England Metropolitan status Metro Nonmetro a For complete explanations of all categories and measures, see Table footnotes.

8 J. Food Prot., Vol. 68, No. 9 META-ANALYSIS OF CONSUMER BEHAVIORS 91 TABLE 5. Values for differences between knowledge and reported use of safe food handling practices a Demographic category Good hygiene Prevent crosscontamination cold storage Avoid unsafe food source cooking, heating Total sample Gender Age (yr) Young ( 29) Mid-age (0 5) Senior (55 ) Income Low Middle High Education Region Pacific Mountain West North Central West South Central East North Central East South Central Middle Atlantic South Atlantic New England a For complete explanation of all categories and measures, see Table footnotes. The difference is calculated as the percentage of respondents who reported knowledge of the safe practice minus the percentage of respondents who reported using the safe practice. All differences are significant at 0.05 using the z-test. knowledge of safe practices. The information in Table and insights regarding the importance of different consumer practices with respect to risk mitigation can be helpful in prioritizing consumer research (25). For example, cooking meat and poultry to the proper internal temperature can be very effective in controlling or eliminating pathogens. However, few studies of proper cooking and heating practices have been conducted (Table ). Therefore, further surveys on consumers use of proper cooking and heating practices (e.g., use of a food thermometer) should take priority over additional data collection on, for example, raw ground beef consumption, which is well documented. Similar priorities can be set based on the information in Table in conjunction with knowledge of the relative risk of consumers food handling practices. Knowledge versus use. A positive correlation between knowledge and actual behavior can be logically expected. If the difference between knowledge and use of the safe handling practices is positive, then we can infer that a negligent attitude or lack of concern, an underestimation of risk, or voluntarily acceptance of the risk are the reasons for not following safe practices despite adequate knowledge. We also can infer that the greater the difference, the higher the consumer s indifference toward food safety. Cases in which knowledge was higher than the reported use of the practices suggest optimistic bias wherein consumers perceived that negative effects are relatively unlikely (6, 0). When the difference between knowledge and use of the practice was negative, we inferred that consumers were unwittingly following safe food handling practices. Additional research is needed to identify reasons other than knowledge that drive the safe behavior of consumers, which can be helpful in developing risk communication programs and materials. Although the relationships between socioeconomic factors, such as poverty, attitude, and culture, and the differences in knowledge and use of food handling practices were not included in this meta-analysis, the existence of such relationships was identified and could be the focus of additional consumer research. In studies that rely on selfreporting of behaviors (e.g., questionnaires or interviews), researchers have found evidence of a tendency for participants to overreport behaviors perceived to be good (11, 2, 6). Therefore, reported behavior may exceed knowledge. Generally, differences between knowledge and re-

9 92 PATIL ET AL. J. Food Prot., Vol. 68, No. 9 TABLE 6. Summary of differences in reported knowledge and behavior by demographic categories a Demographic category Raw or undercooked food consumption Ground beef Eggs Shellfish Milk Good hygiene Prevent crosscontamination defrosting Safe food holding cold storage Avoid unsafe food source cooking, heating Total sample Effect size Gender Ethnicity Caucasian Hispanic African American Asian Other Age (yr) Young ( 29) Mid-age (0 5) Senior (55 ) Income Low Middle High Education Region Pacific Mountain West North Central West South Central East North Central East South Central Middle Atlantic South Atlantic New England Metropolitan status Metro Nonmetro b b b b a For complete explanation of categories and measures, see Table footnotes. indicates highest prevalence for demographic category; indicates lowest prevalence for demographic category. b Differences between Caucasians and Hispanics and between African Americans and Asians were not significant.

10 J. Food Prot., Vol. 68, No. 9 META-ANALYSIS OF CONSUMER BEHAVIORS 9 ported use of safe handling practices were greatest for men, young and mid-age adults, and individuals with more than a high school education. These individuals do not necessarily have a higher incidence of foodborne illnesses. Unsafe behavior can increase the exposure to foodborne pathogens, but the risk of illness depends on pathogen dose and the pharmacokinetic response of individuals. Although these consumers were knowledgeable about safe practices (e.g., the need to use a food thermometer to check for a safe internal temperature), they consciously or unconsciously chose to not follow the safe practice. Differences in reported behavior by demographic categories. How behaviors differ according to gender, level of education, or other demographic characteristics can be helpful in targeting risk communication efforts. The reasons for these differences or similarities can be cultural, social, and/or economical and require further research. For example, raw milk consumption by Hispanics and people living in the West South Central region stands out in comparison with other subpopulation groups. Because the West South Central region is near Mexico and raw milk consumption (particularly in the form of fresh cheese) can be a cultural aspect of the Hispanic population (5), high consumption of raw milk in the West South Central region may be attributable to the higher number of Hispanics in that region (8). Such association, although identified in this meta-analysis, cannot be proven with the available data. Further research is needed on the socioeconomic factors and other population characteristics that could explain the differences in safe handling practices and risky food consumption habits by demographic categories identified in this analysis. Specific demographic categories that report use of various unsafe handling practices or risky food consumption habits can be targeted for risk communication to improve their overall food handling practices. Potential behaviors to target include consumption of raw or undercooked eggs and proper cold storage and defrosting methods. Although we have made generalized observations, such as that men have poorer overall hygiene practices than do women, a metaanalysis could be conducted for specific behaviors such as washing hands before cooking or after handling raw meat if sufficient studies were available. Such information could be used to develop targeted risk communication. Based on the application described here, meta-analysis can be used as a structured quantitative analysis tool in food safety consumer research. The meta-analysis findings can be used to identify important data gaps, to prioritize further research, to conduct statistical analyses, to identify demographic categories with risky consumption and handling practices, and to help target risk communication and risk management efforts. ACKNOWLEDGMENTS This work was supported by funding from the Cooperative State Research, Education, and Extension Service and the U.S. Department of Agriculture (grant no ). We thank Dr. Peter Cowen and Dr. Lee-Ann Jaykus (North Carolina State University, Raleigh) for their guidance. REFERENCES 1. Albrecht, J. A Food safety knowledge and practices of consumers in the USA. J. Consum. Stud. Home Econ. 19: Altekruse, S. F., D. A. Street, S. B. Fein, and A. S. Levy Consumer knowledge of foodborne microbial hazards and food-handling practices. J. Food Prot. 59: Altekruse, S. F., S. Yang, B. B. Timbo, and F. J. Angulo A multistate survey of consumer food-handling and food-consumption practices. Am. J. Prev. Med. 16: Audits International home food safety study report. Audits International, Northbrook, Ill. 5. Bell, R. A., V. N. Hillers, and T. A. Thomas The Abuela Project: safe cheese workshops to reduce the incidence of Salmonella Typhimurium from consumption of raw-milk fresh cheese. Am. J. Public Health 89:1. 6. Brealwell, G. M Risk communication: factors affecting impact. Br. Med. Bull. 56: Bruhn, C., and H. G. Schutz Consumer food safety knowledge and practices. J. Food Saf. 19: Chalmers, T. C., J. Berrier, H. S. Sack, H. Levin, D. Reitman, and R. Nagalingam Meta-analysis of clinical trials as a scientific discipline. Stat. Med. 6: Cook, T. D., H. Cooper, D. S. Cordray, H. Hartmann, L. V. Hedges, R. J. Light, T. A. Louis, and F. Mosteller Meta-analysis for explanation: a casebook. Russell Sage Foundation, New York. 10. Cooper, H Synthesizing research: a guide for literature review, rd ed. Sage Publications, Thousand Oaks, Calif. 11. Curtis, V., S. Cousens, T. Mertens, E. Traore, B. Kanki, and I. Diallo Structured observations of hygiene behaviors in Burkina- Faso validity, variability, and utility. Bull. W.H.O. 71:2 2.. Food Marketing Institute Trends in the United States: consumer attitudes and the supermarket. Food Marketing Institute, Washington, D.C. 1. Glass, G. V Primary, secondary, and meta-analysis of research. Educ. Res. 5: Gravani, R., D. Williamson, and D. Blumental What do consumers know about food safety? FSIS Food Saf. Rev. 2: Griffith, C., D. Worsfold, and R. Mitchell Food preparation, risk communication and the consumer. Food Control 9: Headrick, M. L., B. Timbo, K. C. Klontz, and S. B. Werner Profile of raw milk consumers in California. Public Health Rep. 1: Hedges, L. V Estimating effect size from a series of independent experiments. Psychol. Bull. 92: Hedges, L. V., J. A. Shymansky, and G. Woodworth A practical guide to modern methods of meta-analysis. National Science Teachers Association, Washington, D.C. 19. Klontz, K. C., B. Timbo, S. Fein, and A. Levy Prevalence of selected food consumption and preparation behaviors associated with increased risk of food-borne diseases. J. Food Prot. 58: Knabel, S. J Scientific status summary foodborne illness: role of home food handling practices. Food Technol. 9: Lipsey, M. W., and D. B. Wilson Practical meta-analysis. Applied Social Research Methods Series 9. Sage Publications, Thousand Oaks, Calif. 22. Maciorowski, K. G., S. C. Ricke, and S. G. Birkhold Consumer poultry meat handling and safety education in three Texas cities. Poult. Sci. 78: Manun-Ebo, M., S. Cousens, P. Haggerty, M. Kalengaie, A. Ashworth, and B. Kirkwood Measuring hygiene practices: a comparison of questionnaires with direct observations in rural Zaire. Trop. Med. Int. Health 2: McIntosh, W. A., L. B. Christensen, and G. R. Acuff Perceptions of risks of eating undercooked meat and willingness to change cooking practices. Appetite 22: Medeiros, L. C., V. N. Hiller, P. A. Kendall, and A. Mason Food safety education: what should we be teaching to consumers? J. Nutr. Educ. : Meer, R. R., and S. L. Misner Food safety knowledge and

11 9 PATIL ET AL. J. Food Prot., Vol. 68, No. 9 behavior of expanded food and nutrition education program participants in Arizona. J. Food Prot. 6: Olkin, I Meta-analysis: methods for combining independent studies. Stat. Sci. 7: Patil, S. R., R. Morales, S. Cates, D. Anderson, and D. Kendall An application of meta-analysis in food safety consumer research to evaluate consumer behaviors and practices. J. Food Prot. 67: Raab, C. A., and M. J. Woodburn Changing risk perception and food-handling practices of Oregon household food preparers. J. Consum. Stud. Home Econ. 21: Redmond, E. C., and C. J. Griffith Consumer food handling in the home: a review of food safety studies. J. Food Prot. 66: Rosenthal, R Meta-analytic procedures for social research. Applied social research methods series, vol. 6. Sage Publications, Thousand Oaks, Calif. 2. Rosenthal, R., and D. B. Rubin Comparing effect sizes of independent studies. Psychol. Bull. 92: Sacks, H. S., J. Berrier, D. Reitman, V. A. Axcona-Berk, and T. C. Chalmers Meta-analysis of randomized controlled trials. N. Engl. J. Med. 16: Shiferaw, B., S. Yang, P. Cieslak, D. Vigia, R. Marcus, J. Koehler, V. Deneen, F. Angulo, and the FoodNet Working Group Prevalence of high-risk food consumption and food-handling practices among adults: a multistate survey, 1996 to J. Food Prot. 6: Slavin, R. E Best evidence synthesis: an intelligent alternative to meta-analysis. J. Clin. Epidemiol. 8:9. 6. Stanton, B. F., J. D. Clemens, K. M. A. Aziz, and M. Rahman hour recall, knowledge attitude practice questionnaires, and direct observations of sanitary practices a comparative-study. Bull. W.H.O. 65: Unklesbay, N., J. Sneed, and R. Toma College students attitudes, practices and knowledge of food safety. J. Food Prot. 61: U.S. Census Bureau Current population survey, March Ethnic and Hispanic Statistics Branch, Population Division. Table.1. Available at: hispanic/pp1-165tab-l.xls. Accessed 1 March Williamson, D. M., R. B. Gravani, and H. T. Lawless Correlating food safety knowledge with home food-preparation practices. Food Technol. 6: Woodburn, M. J., and C. A. Raab Household food preparers food-safety knowledge and practices following widely publicized outbreaks of foodborne illness. J. Food Prot. 60: Zhang, P., K. Penner, and J. Johnston Prevalence and selected unsafe food-consumption practices and their associated factors in Kansas. J. Food Saf. 19: