Richard F. MacLehose

Estimating predicted probabilities from logistic regression: different methods correspond to different target populations

BACKGROUND We review three common methods to estimate predicted probabilities following confounder-adjusted logistic regression: marginal standardization (predicted probabilities summed to a weighted average reflecting the confounder distribution in the target population); prediction at the modes (conditional predicted probabilities calculated by setting each confounder to its modal value); and prediction at the means (predicted probabilities calculated by setting each confounder to its mean value). That each method corresponds to a different target population is underappreciated in practice. Specifically, prediction at the means is often incorrectly interpreted as estimating average probabilities for the overall study population, and furthermore yields nonsensical estimates in the presence of dichotomous confounders. Default commands in popular statistical software packages often lead to inadvertent misapplication of prediction at the means. METHODS Using an applied example, we demonstrate discrepancies in predicted probabilities across these methods, discuss implications for interpretation and provide syntax for SAS and Stata. RESULTS Marginal standardization allows inference to the total population from which data are drawn. Prediction at the modes or means allows inference only to the relevant stratum of observations. With dichotomous confounders, prediction at the means corresponds to a stratum that does not include any real-life observations. CONCLUSIONS Marginal standardization is the appropriate method when making inference to the overall population. Other methods should be used with caution, and prediction at the means should not be used with binary confounders. Stata, but not SAS, incorporates simple methods for marginal standardization.

Prevalence of symptoms consistent with a diagnosis of vulvodynia: population-based estimates from 2 geographic regions

OBJECTIVE We used validated sensitive and specific questions associated with clinically confirmed diagnoses of unexplained vulvar pain (vulvodynia) to compare the cumulative incidence of vulvar pain and prevalence of care-seeking behavior in Boston metropolitan area (BMA) and in Minneapolis/Saint Paul metropolitan area (MSP) from 2001 through 2005 using census-based data, and 2010 through 2012, using outpatient community-clinic data, respectively. STUDY DESIGN We received self-administered questionnaires from 5440 women in BMA and 13,681 in MSP, 18-40 years of age, describing their history of vulvar burning or pain on contact that persisted >3 months that limited/prevented intercourse. RESULTS By age 40 years, 7-8% in BMA and MSP reported vulvar pain consistent with vulvodynia. Women of Hispanic origin compared to whites were 1.4 times more likely to develop vulvar pain symptoms (95% confidence interval, 1.1-1.8). Many women in MSP (48%) and BMA (30%) never sought treatment, and >50% who sought care with known health care access received no diagnosis. CONCLUSION Using identical screening methods, we report high prevalence of vulvar pain in 2 geographic regions, and that access to health care does not increase the likelihood of seeking care for chronic vulvar pain.

The Influence of Depression and Anxiety on Risk of Adult Onset Vulvodynia

BACKGROUND Studies have shown that women with vulvodynia are more psychologically distressed than women without vulvodynia. These studies, however, have not effectively established temporal associations between diagnosed psychiatric disorders and vulvodynia. METHODS The Structured Clinical Interview for DSM-IV Axis I Disorders (SCID) was administered to 240 case-control pairs of women with and without vulvodynia. Interviews established age at first onset of diagnosed mood and anxiety disorder. Age information was used to determine whether the first episode of mood and/or anxiety was antecedent or subsequent to the first onset of vulvodynia symptoms. Conditional logistic regressions tested whether antecedent depression or anxiety was more likely among women with or without vulvodynia. Cox proportional hazards modeling was then used to estimate risk of subsequent new or recurrent onset of mood or anxiety disorder. RESULTS After adjusting for education, race, age at menarche, age at first tampon use, and age at first sexual intercourse, odds of vulvodynia were four-times more likely among women with antecedent mood or anxiety compared to women without (95% confidence interval [CI] 2.1-7.5). Vulvodynia was associated with new or recurrent onset of mood or anxiety disorder after adjustment (hazard ratio [HR] 1.7, 95% CI 1.1-2.6) and did not significantly change after including history of mood or anxiety disorder before the onset of vulvodynia or reference age of controls in the models. CONCLUSIONS This is the first community-based epidemiologic study demonstrating that DSM-IV-diagnosed antecedent depression and anxiety disorders influence the risk of vulvodynia and that vulvodynia increases the risk of both new and recurrent onset of psychopathology.

Good practices for quantitative bias analysis

Quantitative bias analysis serves several objectives in epidemiological research. First, it provides a quantitative estimate of the direction, magnitude and uncertainty arising from systematic errors. Second, the acts of identifying sources of systematic error, writing down models to quantify them, assigning values to the bias parameters and interpreting the results combat the human tendency towards overconfidence in research results, syntheses and critiques and the inferences that rest upon them. Finally, by suggesting aspects that dominate uncertainty in a particular research result or topic area, bias analysis can guide efficient allocation of sparse research resources. The fundamental methods of bias analyses have been known for decades, and there have been calls for more widespread use for nearly as long. There was a time when some believed that bias analyses were rarely undertaken because the methods were not widely known and because automated computing tools were not readily available to implement the methods. These shortcomings have been largely resolved. We must, therefore, contemplate other barriers to implementation. One possibility is that practitioners avoid the analyses because they lack confidence in the practice of bias analysis. The purpose of this paper is therefore to describe what we view as good practices for applying quantitative bias analysis to epidemiological data, directed towards those familiar with the methods. We focus on answering questions often posed to those of us who advocate incorporation of bias analysis methods into teaching and research. These include the following. When is bias analysis practical and productive? How does one select the biases that ought to be addressed? How does one select a method to model biases? How does one assign values to the parameters of a bias model? How does one present and interpret a bias analysis?. We hope that our guide to good practices for conducting and presenting bias analyses will encourage more widespread use of bias analysis to estimate the potential magnitude and direction of biases, as well as the uncertainty in estimates potentially influenced by the biases.

What's for dinner? Types of food served at family dinner differ across parent and family characteristics.

OBJECTIVE To examine the types of food served at family dinner in the homes of adolescents and correlations with parent and family sociodemographic characteristics, psychosocial factors and meal-specific variables. DESIGN A cross-sectional population-based survey completed by mail or telephone by parents participating in Project F-EAT (Families and Eating and Activity in Teens) in 2009-2010. SETTING Homes of families with adolescents in Minneapolis/St. Paul urban area, MN, USA. SUBJECTS Participants included 1923 parents/guardians (90·8% female; 68·5% from ethnic/racial minorities) of adolescents who participated in EAT 2010. RESULTS Less than a third (28%) of parents reported serving a green salad at family dinner on a regular basis, but 70% reported regularly serving vegetables (other than potatoes). About one-fifth (21%) of families had fast food at family dinners two or more times per week. Variables from within the sociodemographic domain (low educational attainment) psychosocial domain (high work-life stress, depressive symptoms, low family functioning) and meal-specific domain (low value of family meals, low enjoyment of cooking, low meal planning, high food purchasing barriers and fewer hours in food preparation) were associated with lower healthfulness of foods served at family dinners, in analyses adjusted for sociodemographic characteristics. CONCLUSIONS There is a need for interventions to improve the healthfulness of food served at family meals. Interventions need to be suitable for parents with low levels of education; take parent and family psychosocial factors into account; promote more positive attitudes toward family meals; and provide skills to make it easier to plan and prepare healthful family meals.

Food-Related Parenting Practices and Adolescent Weight Status: A Population-Based Study

OBJECTIVE: To examine food-related parenting practices (pressure-to-eat and food restriction) among mothers and fathers of adolescents and associations with adolescent weight status within a large population-based sample of racially/ethnically and socioeconomically diverse parent-adolescent pairs. METHODS: Adolescents (N = 2231; 14.4 years old [SD = 2.0]) and their parents (N = 3431) participated in 2 coordinated population-based studies designed to examine factors associated with weight status and weight-related behaviors in adolescents. Adolescents completed anthropometric measurements and surveys at school. Parents (or other caregivers) completed questionnaires via mail or phone. RESULTS: Findings suggest that the use of controlling food-related parenting practices, including pressure-to-eat and restriction, is common among parents of adolescents. Mean restriction levels were significantly higher among parents of overweight and obese adolescents compared with nonoverweight adolescents. However, levels of pressure-to-eat were significantly higher among nonoverweight adolescents. Results indicate that fathers are more likely than mothers to engage in pressure-to-eat behaviors and boys are more likely than girls to be on the receiving end of parental pressure-to-eat. Parental report of restriction did not differ significantly by parent or adolescent gender. No significant interactions by race/ethnicity or socioeconomic status were seen in the relationship between restriction or pressure-to-eat and adolescent weight status. CONCLUSIONS: Given that there is accumulating evidence for the detrimental effects of controlling feeding practices on children’s ability to self-regulate energy intake, these findings suggest that parents should be educated and empowered through anticipatory guidance to encourage moderation rather than overconsumption and emphasize healthful food choices rather than restrictive eating patterns.

Eat this, not that! Parental demographic correlates of food-related parenting practices

To understand how parents of adolescents attempt to regulate their childrens eating behaviors, the prevalence of specific food-related parenting practices (restriction, pressure-to-eat) by sociodemographic characteristics (parent gender, race/ethnicity, education level, employment status, and household income) were examined within a population-based sample of parents (n=3709) of adolescents. Linear regression models were fit to estimate the association between parent sociodemographic characteristics and parental report of food restriction and pressure-to-eat. Overall, findings suggest that use of controlling food-related parenting practices, such as pressuring children to eat and restricting childrens intake, is common among parents of adolescents, particularly among parents in racial/ethnic minority subgroups, parents with less than a high school education, and parents with a low household income. Results indicate that that social or cultural traditions, as well as parental access to economic resources, may contribute to a parents decision to utilize specific food-related parenting practices. Given that previous research has found that restriction and pressure-to-eat food-related parenting practices can negatively impact childrens current and future dietary intake, differences in use of these practices by sociodemographic characteristics may contribute, in part, to the disparities that exist in the prevalence of overweight and obesity among adolescents by their race/ethnicity and socioeconomic status.

Feeding a Family in a Recession: Food Insecurity Among Minnesota Parents

OBJECTIVES We assessed current levels of food insecurity among a large, diverse sample of parents and examined associations between food insecurity and parental weight status, eating patterns, and the home food environment. METHODS Project F-EAT (Families and Eating and Activity Among Teens) examined the home food environments of adolescents. Parents and caregivers (n = 2095) living with adolescents from the Minneapolis-St. Paul, Minnesota school districts completed mailed surveys during a 12-month period in 2009-2010. We performed our assessments using multivariate regressions. RESULTS Almost 39% of the parents and caregivers experienced household food insecurity, whereas 13% experienced very low food security. Food insecurity was significantly associated with poorer nutrition-related variables such as higher rates of parental overweight and obesity, less healthy foods served at meals, and higher rates of binge eating. Food-insecure parents were 2 to 4 times more likely to report barriers to accessing fruits and vegetables. CONCLUSIONS Food insecurity was highly prevalent. Environmental interventions are needed to protect vulnerable families against food insecurity and to improve access to affordable, healthy foods.

Parent conversations about healthful eating and weight associations with adolescent disordered eating behaviors

IMPORTANCE The prevalence of weight-related problems in adolescents is high. Parents of adolescents may wonder whether talking about eating habits and weight is useful or detrimental. OBJECTIVE To examine the associations between parent conversations about healthful eating and weight and adolescent disordered eating behaviors. DESIGN Cross-sectional analysis using data from 2 linked multilevel population-based studies. SETTING Anthropometric assessments and surveys completed at school by adolescents and surveys completed at home by parents in 2009-2010. PARTICIPANTS Socioeconomically and racially/ethnically diverse sample (81% ethnic minority; 60% low income) of adolescents from Eating and Activity in Teens 2010 (EAT 2010) (n = 2793; mean age, 14.4 years) and parents from Project Families and Eating and Activity in Teens (Project F-EAT) (n = 3709; mean age, 42.3 years). EXPOSURE Parent conversations about healthful eating and weight/size. MAIN OUTCOMES AND MEASURES Adolescent dieting, unhealthy weight-control behaviors, and binge eating. RESULTS Mothers and fathers who engaged in weight-related conversations had adolescents who were more likely to diet, use unhealthy weight-control behaviors, and engage in binge eating. Overweight or obese adolescents whose mothers engaged in conversations that were focused only on healthful eating behaviors were less likely to diet and use unhealthy weight-control behaviors. Additionally, subanalyses with adolescents with data from 2 parents showed that when both parents engaged in healthful eating conversations, their overweight or obese adolescent children were less likely to diet and use unhealthy weight-control behaviors. CONCLUSIONS AND RELEVANCE Parent conversations focused on weight/size are associated with increased risk for adolescent disordered eating behaviors, whereas conversations focused on healthful eating are protective against disordered eating behaviors.

Blood lipid levels, lipid-lowering medications, and the incidence of atrial fibrillation: the atherosclerosis risk in communities study.

Background— Several cardiovascular risk factors have been associated with the risk of atrial fibrillation (AF). Limited and inconsistent evidence exists on the association of blood lipid levels and lipid-lowering medication use with AF risk. Methods and Results— We analyzed 13 969 participants (25% African American, 45% men) free of AF at baseline from the Atherosclerosis Risk in Communities study. Fasting high-density lipoprotein cholesterol (HDLc), low-density lipoprotein cholesterol (LDLc), triglycerides, and total cholesterol were measured at baseline (1987–1989) and each of 3 follow-up visits. The incidence of AF was ascertained through 2007. The association of the use of statins and other lipid-lowering medications with AF was estimated in 13 044 Atherosclerosis Risk in Communities participants attending visit 2 (1990–1992), adjusting for covariates from the previous visit. During a median follow-up of 18.7 years, there were 1433 incident AF cases. Multivariable hazard ratios (HRs) and 95% CIs of AF associated with a 1-SD increase in lipid levels were as follows: HDLc, 0.97 (0.91–1.04); LDLc, 0.90 (0.85–0.96); total cholesterol, 0.89 (0.84–0.95); and triglycerides, 1.00 (0.96–1.04). Participants taking lipid-lowering medications had an adjusted HR (95% CI) of AF of 0.96 (0.82–1.13) compared with those not taking medications, whereas those taking statins had an adjusted HR of 0.91 (0.66–1.25) compared with those taking other lipid-lowering medications. Conclusions— Higher levels of LDLc and total cholesterol were associated with a lower incidence of AF. However, HDLc and triglycerides were not independently associated with AF incidence. No association was found between the use of lipid-lowering medications and incident AF.