Multiple Imputation in Practice

Factors affecting increasing radiation dose for mammography in North Carolina from 1997 through 2001: an analysis of Food and Drug Administration annual surveys

RATIONALE AND OBJECTIVES

To determine which factors affected the increase in average glandular dose recorded at the annual US Food and Drug Administration Mammography Quality Standards Act inspections of mammography equipment in North Carolina from 1997 to 2001.

MATERIALS AND METHODS

Average glandular dose, HVL, kVp, ambient light, luminance, equipment age, processing speed, and system speed for every mammography unit at all facilities in the state were collected by state inspectors. A mixed-effect model was used to assess the average changes of glandular dose over time and to identify the factors associated with these changes.

RESULTS

There was a statistically significant increase in the average glandular dose in North Carolina in 1999, 2000, and 2001 when compared with the baseline year of 1997. Factors that were statistically significantly linked to this effect were changes in kVp, processing speed, and system speed.

CONCLUSION

Average glandular dose for mammography has recently increased in North Carolina. This change is likely caused by changes in screen-film products and processing techniques.

The effect of weight change on nursing care facility admission in the NHANES I Epidemiologic Followup Survey

Data from the first National Health and Nutrition Examination Survey Epidemiologic Followup Survey were used to examine whether weight change was associated with an increased relative risk of nursing care facility admission. Hazard ratios were calculated with Cox proportional hazards models and stratified by overweight status at baseline. Moderate and large weight loss was associated with an increased risk of nursing care facility admission in overweight and non-overweight subjects. Large weight gain was associated with an increased relative risk in only overweight subjects. In the process of functional decline that results in nursing care facility admission, weight loss may be a sign of acute illness, starvation, or aging. Preventing weight loss may help delay this process of decline. In overweight subjects, preventing weight gain may also be important in delaying this process of decline.

Iron supplementation during pregnancy, anemia, and birth weight: a randomized controlled trial

BACKGROUND

The need for prophylactic iron during pregnancy is uncertain.

OBJECTIVE

We tested the hypothesis that administration of a daily iron supplement from enrollment to 28 wk of gestation to initially iron-replete, nonanemic pregnant women would reduce the prevalence of anemia at 28 wk and increase birth weight.

DESIGN

Between June 1995 and September 1998, 513 low-income pregnant women in Cleveland were enrolled in the study before 20 wk of gestation. Of these, 275 had a hemoglobin concentration >/= 110 g/L and a ferritin concentration >/= 20 micro g/L and were randomly assigned to receive a monthly supply of capsules containing either 30 mg Fe as ferrous sulfate or placebo until 28 wk of gestation. At 28 and 38 wk of gestation, women with a ferritin concentration of 12 to < 20 micro g/L or < 12 micro g/L received 30 and 60 mg Fe/d, respectively, regardless of initial assignment. Almost all the women received some supplemental iron during pregnancy. We obtained infant birth weight and gestational age at delivery for 117 and 96 of the 146 and 129 women randomly assigned to receive iron and placebo, respectively.

RESULTS

Compared with placebo, iron supplementation from enrollment to 28 wk of gestation did not significantly affect the overall prevalence of anemia or the incidence of preterm births but led to a significantly higher mean (+/- SD) birth weight (206 +/- 565 g; P = 0.010), a significantly lower incidence of low-birth-weight infants (4% compared with 17%; P = 0.003), and a significantly lower incidence of preterm low-birth-weight infants (3% compared with 10%; P = 0.017).

CONCLUSION

Prenatal prophylactic iron supplementation deserves further examination as a measure to improve birth weight and potentially reduce health care costs.

Modern statistical methods for handling missing repeated measurements in obesity trial data: beyond LOCF

This paper brings together some modern statistical methods to address the problem of missing data in obesity trials with repeated measurements. Such missing data occur when subjects miss one or more follow-up visits, or drop out early from an obesity trial. A common approach to dealing with missing data because of dropout is 'last observation carried forward' (LOCF). This method, although intuitively appealing, requires restrictive assumptions to produce valid statistical conclusions. We review the need for obesity trials, the assumptions that must be made regarding missing data in such trials, and some modern statistical methods for analysing data containing missing repeated measurements. These modern methods have fewer limitations and less restrictive assumptions than required for LOCF. Moreover, their recent introduction into current releases of statistical software and textbooks makes them more readily available to the applied data analyses.

Missing... presumed at random: cost-analysis of incomplete data

When collecting patient-level resource use data for statistical analysis, for some patients and in some categories of resource use, the required count will not be observed. Although this problem must arise in most reported economic evaluations containing patient-level data, it is rare for authors to detail how the problem was overcome. Statistical packages may default to handling missing data through a so-called 'complete case analysis', while some recent cost-analyses have appeared to favour an 'available case' approach. Both of these methods are problematic: complete case analysis is inefficient and is likely to be biased; available case analysis, by employing different numbers of observations for each resource use item, generates severe problems for standard statistical inference. Instead we explore imputation methods for generating 'replacement' values for missing data that will permit complete case analysis using the whole data set and we illustrate these methods using two data sets that had incomplete resource use information.

Obesity affects nursing-care facility admission among whites but not blacks

OBJECTIVE

This study examines whether obese individuals have a greater rate of nursing care facility admission than normal weight individuals.

RESEARCH METHODS AND PROCEDURES

Data from the National Health and Nutrition Examination Survey Epidemiological Follow-up Survey were analyzed. Cox proportional hazards models were used to examine the relationship between baseline weight status and subsequent time to first nursing home admission while adjusting for sex, age, race, marital status, height, presence of children, smoking status, education, region, urban residence, income, and physical activity.

RESULTS

Of 5960 adults 45 to 74 years old, 989 individuals were admitted to a nursing care facility over the subsequent 20 years. Body mass index (BMI) was studied using five categories: < 18.5, 18.5 to <25, 25.0 to <30, 30.0 to <35, > or =35 kg/m(2). The effects of BMI differed by race: compared with those with a BMI of 18.5 to < 25 kg/m(2), adults with a BMI > or =30 kg/m(2) or a BMI <18.5 kg/m(2) had a greater rate of nursing home admission in whites, whereas no relationship was found in blacks. The inclusion of time to death with nursing home admission as a joint outcome yielded similar results.

DISCUSSION

The large increase in the prevalence of obesity coupled with the rapid expansion of the number of older Americans will likely increase the demand for nursing facility use. More research is needed to understand differences in factors related to nursing home admission among ethnic groups.

Revisiting field experimentation: field notes for the future

Field experiments in the social sciences were increasingly used in the 20th century. This article briefly reviews some important lessons in design, analysis, and theory of field experiments emerging from that experience. Topics include the importance of ensuring that selection into experiments and assignment to conditions occurs properly, how to prevent and analyze attrition, the need to attend to power and effect size, how to measure and take partial treatment implementation into account in analyses, modern analyses of quasi-experimental and multilevel data, Rubin's model, and the role of internal and external validity. The article ends with observations on the computer revolution in methodology and statistics, convergences in theory and methods across disciplines, the need for an empirical program of methodological research, the key problem of selection bias, and the inevitability of increased specialization in field experimentation in the years to come.