Evaluation of Multiple Imputation with Large Proportions of Missing Data: How Much Is Too Much?

BACKGROUND: Multiple Imputation (MI) is known as an effective method for handling missing data in public health research. However, it is not clear that the method will be effective when the data contain a high percentage of missing observations on a variable. METHODS: Using data from "Predictive Study of Coronary Heart Disease" study, this study examined the effectiveness of multiple imputation in data with 20% missing to 80% missing observations using absolute bias (|bias|) and Root Mean Square Error (RMSE) of MI measured under Missing Completely at Random (MCAR), Missing at Random (MAR), and Not Missing at Random (NMAR) assumptions. RESULTS: The |bias| and RMSE of MI was much smaller than of the results of CCA under all missing mechanisms, especially with a high percentage of missing. In addition, the |bias| and RMSE of MI were consistent regardless of increasing imputation numbers from M=10 to M=50. Moreover, when comparing imputation mechanisms, MCMC method had universally smaller |bias| and RMSE than those of Regression method and Predictive Mean Matching method under all missing mechanisms. CONCLUSION: As missing percentages become higher, using MI is recommended, because MI produced less biased estimates under all missing mechanisms. However, when large proportions of data are missing, other things need to be considered such as the number of imputations, imputation mechanisms, and missing data mechanisms for proper imputation.

Using direct observations on multiple occasions to measure household food availability among low-income Mexicano residents in Texas colonias.

BACKGROUND: It has been recognized that the availability of foods in the home are important to nutritional health, and may influence the dietary behavior of children, adolescents, and adults. It is therefore important to understand food choices in the context of the household setting. Considering their importance, the measurement of household food resources becomes critical.Because most studies use a single point of data collection to determine the types of foods that are present in the home, which can miss the change in availability within a month and when resources are not available, the primary objective of this pilot study was to examine the feasibility and value of conducting weekly in-home assessments of household food resources over the course of one month among low-income Mexicano families in Texas colonias. METHODS: We conducted five in-home household food inventories over a thirty-day period in a small convenience sample; determined the frequency that food items were present in the participating households; and compared a one-time measurement with multiple measurements.After the development and pre-testing of the 252-item culturally and linguistically- appropriate household food inventory instrument that used direct observation to determine the presence and amount of food and beverage items in the home (refrigerator, freezer, pantry, elsewhere), two trained promotoras recruited a convenience sample of 6 households; administered a baseline questionnaire (personal info, shopping habits, and food security); conducted 5 in-home assessments (7-day interval) over a 30-day period; and documented grocery shopping and other food-related activities within the previous week of each in-home assessment. All data were collected in Spanish. Descriptive statistics were calculated for mean and frequency of sample characteristics, food-related activities, food security, and the presence of individual food items. Due to the small sample size of the pilot data, the Friedman Test and Kendall's W were used to assess the consistency of household food supplies across multiple observations. RESULTS: Complete data were collected from all 6 Mexicano women (33.2y +/- 3.3; 6.5 +/- 1.5 adults/children in household (HH); 5 HH received weekly income; and all were food insecure. All households purchased groceries within a week of at least four of the five assessments. The weekly presence and amounts of fresh and processed fruits and vegetables, dairy, meats, breads, cereals, beverages, and oils and fats varied. Further, the results revealed the inadequacy of a one-time measurement of household food resources, compared with multiple measures. The first household food inventory as a one-time measure would have mistakenly identified at least one-half of the participant households without fresh fruit, canned vegetables, dairy, protein foods, grains, chips, and sugar-sweetened beverages. CONCLUSIONS: This study highlights the value of documenting weekly household food supplies, especially in households where income resources may be more volatile. Clearly, the data show that a single HFI may miss the changes in availability--presence and amount--that occur among low-income Mexicano households who face challenges that require frequent purchase of foods and beverages. Use of multiple household food inventories can inform the development and implementation of nutrition-related policies and culturally sensitive nutrition education programs.

The social construction of "evidence-based'' drug prevention programs: a reanalysis of data from the Drug Abuse Resistance Education (DARE) program.

This study explores the possibility that any drug prevention program might be considered ;;evidence-based'' given the use of data analysis procedures that optimize the chance of producing statistically significant results by reanalyzing data from a Drug Abuse Resistance Education (DARE) program evaluation. The analysis produced a number of statistically significant differences between the DARE and control conditions on alcohol and marijuana use measures. Many of these differences occurred at cutoff points on the assessment scales for which post hoc meaningful labels were created. Our results are compared to those from evaluations of programs that appear on evidence-based drug prevention lists.

Availability of healthier options in traditional and nontraditional rural fast-food outlets.

BACKGROUND: Food prepared away from home has become increasingly popular to U.S. families, and may contribute to obesity. Sales have been dominated by fast food outlets, where meals are purchased for dining away from home or in the home. Although national chain affiliated fast-food outlets are considered the main source for fast food, fast foods are increasingly available in convenience stores and supermarkets/grocery stores. In rural areas, these nontraditional fast-food outlets may provide most of the opportunities for procurement of fast foods. METHODS: Using all traditional and nontraditional fast-food outlets identified in six counties in rural Texas, the type and number of regular and healthier menu options were surveyed using on-site observation in all food venues that were primarily fast food, supermarket/grocery store, and convenience store and compared with 2005 Dietary Guidelines. RESULTS: Traditional fast-food outlets represented 84 (41%) of the 205 opportunities for procurement of fast food; 109 (53.2%) were convenience stores and 12 (5.8%) supermarkets/grocery stores. Although a similar variety of regular breakfast and lunch/dinner entrées were available in traditional fast-food outlets and convenience stores, the variety of healthier breakfast and lunch/dinner entrées was significantly greater in fast food outlets. Compared with convenience stores, supermarkets/grocery stores provided a greater variety of regular and healthier entrées and lunch/dinner side dishes. CONCLUSION: Convenience stores and supermarkets/grocery stores more than double the potential access to fast foods in this rural area than traditional fast-food outlets alone; however, traditional fast food outlets offer greater opportunity for healthier fast food options than convenience stores. A complete picture of fast food environment and the availability of healthier fast food options are essential to understand environmental influences on diet and health outcomes, and identify potential targets for intervention.

Underage driving as an indicator of risky behavior in children and adolescents.

Driving among children and adolescents below the legal driving age in Texas was examined. There were 4170 accidents between 1995 and 2000 (66.5% involved injury/fatality). Drivers were more often male, and underage driving was greatest during the late afternoon/early evening. Risk of severe injury or death was inversely related to speed and nighttime driving, and was greater in rural areas, in accidents involving passengers and among black children and youth compared with white.

Familial aggregation of severity of diabetic retinopathy in Mexican Americans from Starr County, Texas.

OBJECTIVE: Diabetic retinopathy is a major cause of blindness. To determine whether retinopathy itself or only its severity aggregates in families, we examined the occurrence and severity of diabetic retinopathy in Mexican-American siblings with type 2 diabetes. RESEARCH DESIGN AND METHODS: Using stereoscopic fundus photography of seven standard fields, we measured retinopathy in 656 type 2 diabetic patients from 282 Mexican-American families from Starr County, Texas. Retinopathy severity was scored using the Early Treatment of Diabetic Retinopathy Study system and classified as no retinopathy, early nonproliferative diabetic retinopathy (NPDR-E), moderate-to-severe nonproliferative diabetic retinopathy (NPDR-S), or proliferative diabetic retinopathy (PDR). RESULTS: Of 249 siblings of randomly selected probands with retinopathy, 169 (67.9%) had retinopathy, compared with 95 of 125 siblings of unaffected probands (76.0%; P = 0.11). Proband retinopathy class was associated (P = 0.03) with sibling retinopathy class, with significant odds ratios (ORs) for NPDR-E versus no retinopathy (OR 0.57 [95% CI 0.35-0.93]) and PDR versus NPDR-E (2.02 [1.13-3.63]); the contrast of NPDR-S versus NPDR-E approached significance (1.78 [0.99-3.20]). With the more severe classes (PDR and NPDR-S) combined in one group and the less severe ones (none and NPDR-E) in another, more severe proband retinopathy was associated with more severe sibling retinopathy (1.72 [1.03-2.88]). CONCLUSIONS: More severe diabetic retinopathy showed evidence of familial aggregation, but the occurrence of diabetic retinopathy per se did not. The factors involved in the onset of diabetic retinopathy may differ from those involved in its progression to more severe forms.