Liping Tong

An actor-based model of social network influence on adolescent body size, screen time, and playing sports.

Recent studies suggest that obesity may be “contagious” between individuals in social networks. Social contagion (influence), however, may not be identifiable using traditional statistical approaches because they cannot distinguish contagion from homophily (the propensity for individuals to select friends who are similar to themselves) or from shared environmental influences. In this paper, we apply the stochastic actor-based model (SABM) framework developed by Snijders and colleagues to data on adolescent body mass index (BMI), screen time, and playing active sports. Our primary hypothesis was that social influences on adolescent body size and related behaviors are independent of friend selection. Employing the SABM, we simultaneously modeled network dynamics (friendship selection based on homophily and structural characteristics of the network) and social influence. We focused on the 2 largest schools in the National Longitudinal Study of Adolescent Health (Add Health) and held the school environment constant by examining the 2 school networks separately (N = 624 and 1151). Results show support in both schools for homophily on BMI, but also for social influence on BMI. There was no evidence of homophily on screen time in either school, while only one of the schools showed homophily on playing active sports. There was, however, evidence of social influence on screen time in one of the schools, and playing active sports in both schools. These results suggest that both homophily and social influence are important in understanding patterns of adolescent obesity. Intervention efforts should take into consideration peers’ influence on one another, rather than treating “high risk” adolescents in isolation.

Genetic background of patients from a university medical center in Manhattan: implications for personalized medicine.

Background The rapid progress currently being made in genomic science has created interest in potential clinical applications; however, formal translational research has been limited thus far. Studies of population genetics have demonstrated substantial variation in allele frequencies and haplotype structure at loci of medical relevance and the genetic background of patient cohorts may often be complex. Methods and Findings To describe the heterogeneity in an unselected clinical sample we used the Affymetrix 6.0 gene array chip to genotype self-identified European Americans (N = 326), African Americans (N = 324) and Hispanics (N = 327) from the medical practice of Mount Sinai Medical Center in Manhattan, NY. Additional data from US minority groups and Brazil were used for external comparison. Substantial variation in ancestral origin was observed for both African Americans and Hispanics; data from the latter group overlapped with both Mexican Americans and Brazilians in the external data sets. A pooled analysis of the African Americans and Hispanics from NY demonstrated a broad continuum of ancestral origin making classification by race/ethnicity uninformative. Selected loci harboring variants associated with medical traits and drug response confirmed substantial within- and between-group heterogeneity. Conclusion As a consequence of these complementary levels of heterogeneity group labels offered no guidance at the individual level. These findings demonstrate the complexity involved in clinical translation of the results from genome-wide association studies and suggest that in the genomic era conventional racial/ethnic labels are of little value.

Elevated hypertension risk for African-origin populations in biracial societies: modeling the Epidemiologic Transition Study.

Objectives: Blood pressures in persons of African descent exceed those of other racial/ethnic groups in the United States. Whether this trait is attributable to the genetic factors in African-origin populations, or a result of inadequately measured environmental exposures, such as racial discrimination, is not known. To study this question, we conducted a multisite comparative study of communities in the African diaspora, drawn from metropolitan Chicago, Kingston, Jamaica, rural Ghana, Cape Town, South Africa, and the Seychelles. Methods: At each site, 500 participants between the age of 25 and 49 years, with approximately equal sex balance, were enrolled for a longitudinal study of energy expenditure and weight gain. In this study, we describe the patterns of blood pressure and hypertension observed at baseline among the sites. Results: Mean SBP and DBP were very similar in the United States and South Africa in both men and women, although among women, the prevalence of hypertension was higher in the United States (24 vs. 17%, respectively). After adjustment for multiple covariates, relative to participants in the United States, SBP was significantly higher among the South Africans by 9.7 mmHg (P < 0.05) and significantly lower for each of the other sites: for example, Jamaica: −7.9 mmHg (P = 0.06), Ghana: −12.8 mmHg (P < 0.01) and Seychelles: −11.1 mmHg (P = 0.01). Conclusion: These data are consistent with prior findings of a blood pressure gradient in societies of the African diaspora and confirm that African-origin populations with lower social status in multiracial societies, such as the United States and South Africa, experience more hypertension than anticipated based on anthropometric and measurable socioeconomic risk factors.

Comparison of predictive modeling approaches for 30-day all-cause non-elective readmission risk

BackgroundThis paper explores the importance of electronic medical records (EMR) for predicting 30-day all-cause non-elective readmission risk of patients and presents a comparison of prediction performance of commonly used methods.MethodsThe data are extracted from eight Advocate Health Care hospitals. Index admissions are excluded from the cohort if they are observation, inpatient admissions for psychiatry, skilled nursing, hospice, rehabilitation, maternal and newborn visits, or if the patient expires during the index admission. Data are randomly and repeatedly divided into fitting and validating sets for cross validations. Approaches including LACE, STEPWISE logistic, LASSO logistic, and AdaBoost, are compared with sample sizes varying from 2,500 to 80,000.ResultsOur results confirm that LACE has moderate discrimination power with the area under receiver operating characteristic curve (AUC) around 0.65-0.66, which can be improved to 0.73-0.74 when additional variables from EMR are considered. These variables include Inpatient in the last six months, Number of emergency room visits or inpatients in the last year, Braden score, Polypharmacy, Employment status, Discharge disposition, Albumin level, and medical condition variables such as Leukemia, Malignancy, Renal failure with hemodialysis, History of alcohol substance abuse, Dementia and Trauma. When sample size is small (≤5000), LASSO is the best; when sample size is large (≥20,000), the predictive performance is similar. The STEPWISE method has a slightly lower AUC (0.734) comparing to LASSO (0.737) and AdaBoost (0.737). More than one half of the selected predictors can be false positives when using a single method and a single division of fitting/validating data.ConclusionsTrue predictors can be identified by repeatedly dividing data into fitting/validating subsets and referring the final model based on summarizing results. LASSO is a better alternative to the STEPWISE logistic regression, especially when sample size is not large. The evidence for adequate sample size can be explored by fitting models on gradually reduced samples. Our model comparison strategy is not only good for 30-day all-cause non-elective readmission risk predictions, but also applicable to other types of predictive models in clinical studies.

Depression in the US population during the time periods surrounding the great recession.

OBJECTIVE To determine whether the time periods surrounding the 2008 US economic downturn were accompanied by an increase in prevalence of depression in the US adult population. METHOD We used data from the 24,182 adults aged ≥ 18 years who participated in the National Health and Nutrition Examination Survey during 2005-2012. A cross-sectional analysis was performed at each time period to determine prevalence of major and other depression as assessed by standardized questionnaires based on 9 criteria for major depressive episodes defined by DSM-IV. RESULTS The demographic characteristics of the US population were similar across time periods except for the percentage of adults living in poverty, which increased from 26.43% during 2005-2006 to 33.46% during 2011-2012. The prevalence of major depression increased from 2.33% (95% CI, 1.64%-3.01%) during 2005-2006 to 3.49% (95% CI, 2.84%-4.03%) in 2009-2010 to 3.79% (95% CI, 3.01%-4.57%) in 2011-2012. Prevalence of other depression increased from 4.10% (95% CI, 3.37%-4.88%) in 2005-2006 to 4.79% (95% CI, 4.10%-5.44%) in the 2009-2010 period but then declined to 3.68% (95% CI, 2.84%-4.48%) in the 2011-2012 time period (P = .4). After adjustment for the distribution of age, sex, race/ethnicity, education, insurance status, and poverty status in the US adult noninstitutionalized population, each 2-year period after the 2005-2006 time period was associated with a 0.4% increase in major depression prevalence (P < .001). No significant differences in other depression prevalence were noted by time period (P = .6). CONCLUSIONS The time periods surrounding the recent economic recession were accompanied by a significant and sustained increase in major depression prevalence in the US population. It is plausible that the recession, given its strong, persistent, and negative effects on employment, job and housing security, and stock investments, contributed to the sustained increase in prevalence of major depression in the US population, but other factors associated with the recession time period could have played a role. The impact of the economic downturn on depression prevalence should be considered when formulating future policies and programs to promote and maintain the health of the US population.

Efficient Calculation of P‐value and Power for Quadratic Form Statistics in Multilocus Association Testing

We address the asymptotic and approximate distributions of a large class of test statistics with quadratic forms used in association studies. The statistics of interest take the general form D=XTA X, where A is a general similarity matrix which may or may not be positive semi‐definite, and X follows the multivariate normal distribution with mean μ and variance matrix Σ, where Σ may or may not be singular. We show that D can be written as a linear combination of independent χ2 random variables with a shift. Furthermore, its distribution can be approximated by a χ2 or the difference of two χ2 distributions. In the setting of association testing, our methods are especially useful in two situations. First, when the required significance level is much smaller than 0.05 such as in a genome scan, the estimation of p‐values using permutation procedures can be challenging. Second, when an EM algorithm is required to infer haplotype frequencies from un‐phased genotype data, the computation can be intensive for a permutation procedure. In either situation, an efficient and accurate estimation procedure would be useful. Our method can be applied to any quadratic form statistic and therefore should be of general interest.

Accelerometer-measured physical activity is not associated with two-year weight change in African-origin adults from five diverse populations

Background Increasing population-levels of physical activity (PA) is a controversial strategy for managing the obesity epidemic, given the conflicting evidence for weight loss from PA alone per se. We measured PA and weight change in a three-year prospective cohort study in young adults from five countries (Ghana, South Africa, Jamaica, Seychelles and USA). Methods A total of 1,944 men and women had baseline data, and at least 1 follow-up examination including measures of anthropometry (weight/BMI), and objective PA (accelerometer, 7-day) following the three-year study period. PA was explored as 1-minute bouts of moderate and vigorous PA (MVPA) as well as daily sedentary time. Results At baseline; Ghanaian and South African men had the lowest body weights (63.4 ± 9.5, 64.9 ± 11.8 kg, respectively) and men and women from the USA the highest (93.6 ± 25.9, 91.7 ± 23.4 kg, respectively). Prevalence of normal weight ranged from 85% in Ghanaian men to 29% in USA men and 52% in Ghanaian women to 15% in USA women. Over the two-year follow-up period, USA men and Jamaican women experienced the smallest yearly weight change rate (0.1 ± 3.3 kg/yr; −0.03 ± 3.0 kg/yr, respectively), compared to South African men and Ghanaian women greatest yearly change (0.6.0 ± 3.0 kg/yr; 1.22 ± 2.6 kg/yr, respectively). Mean yearly weight gain tended to be larger among normal weight participants at baseline than overweight/obese at baseline. Neither baseline MVPA nor sedentary time were associated with weight gain. Using multiple linear regression, only baseline weight, age and gender were significantly associated with weight gain. Discussion From our study it is not evident that higher volumes of PA alone are protective against future weight gain, and by deduction our data suggest that other environmental factors such as the food environment may have a more critical role.

The social patterning of risk factors for noncommunicable diseases in five countries: evidence from the modeling the epidemiologic transition study (METS)

BackgroundAssociations between socioeconomic status (SES) and risk factors for noncommunicable diseases (NCD-RFs) may differ in populations at different stages of the epidemiological transition. We assessed the social patterning of NCD-RFs in a study including populations with different levels of socioeconomic development.MethodsData on SES, smoking, physical activity, body mass index, blood pressure, cholesterol and glucose were available from the Modeling the Epidemiologic Transition Study (METS), with about 500 participants aged 25–45 in each of five sites (Ghana, South Africa, Jamaica, Seychelles, United States).ResultsThe prevalence of NCD-RFs differed between these populations from five countries (e.g., lower prevalence of smoking, obesity and hypertension in rural Ghana) and by sex (e.g., higher prevalence of smoking and physical activity in men and of obesity in women in most populations). Smoking and physical activity were associated with low SES in most populations. The associations of SES with obesity, hypertension, cholesterol and elevated blood glucose differed by population, sex, and SES indicator. For example, the prevalence of elevated blood glucose tended to be associated with low education, but not with wealth, in Seychelles and USA. The association of SES with obesity and cholesterol was direct in some populations but inverse in others.ConclusionsIn conclusion, the distribution of NCD-RFs was socially patterned in these populations at different stages of the epidemiological transition, but associations between SES and NCD-RFs differed substantially according to risk factor, population, sex, and SES indicator. These findings emphasize the need to assess and integrate the social patterning of NCD-RFs in NCD prevention and control programs in LMICs.

Analytic solutions for D-optimal factorial designs under generalized linear models

We develop two analytic approaches to solve D-optimal approximate designs under generalized linear models. The first approach provides analytic D-optimal allocations for generalized linear models with two factors, which include as a special case the

Association of polymorphisms in the aldosterone-regulated sodium reabsorption pathway with blood pressure among Hispanics

2^2