Covariates. Age at Pregnancy was calculated using the date which a pregnancy-related ICD-9-CM code first showed up in the Medicaid claims data and mother’s date of birth; in cases with multiple dates of birth, the most recent observation was used. Age at pregnancy was then transformed into a dichotomous variable: 18.999 years old or younger was code as ‘1’ and 19 years old or older was coded as ‘2’, to account for the change in Medicaid eligibility criteria at that age. Race was a categorical variable including White, Black, Asian, American Indian/Alaskan native, native Hawaiian/Pacific Islander. White was coded as ‘1’, Black was coded as ‘2’, American Indian/Alaskan native was coded as ‘3’, Asian was coded as ‘4’, and native Hawaiian/Pacific Islander was coded as ‘5’. However, given the small sample of individuals who identified as Asian, American Indian/Alaskan Native, Native Hawaiian/Pacific Islander, only those who identified as black or white were included in the final analysis. Severity of CHD was initially coded into five categories based on a modified version of Marelli’s CHD hierarchy: Severe = ‘1’, Shunt = ‘2’, Shunt and Valve = ‘3’, Valve = ‘4’, and Other = ‘5’. This variable was then recoded, collapsing the Marelli-based five hierarchical categories into three groups: ‘1’= Severe, ‘2’ =mild/moderate, and ‘3’ = those CHD patients with an isolated 745.5 ICD-9-CM code (a code that is often used for non-CHD diagnoses). Urban-Rural Residence was computed using the Federal Information Processing Standard (FIPS) county codes to create a dichotomous variable with counties categorized as either urban coded as ‘1’ or rural coded as ‘2’. FIPS county codes were based on the 2010 Census of Population and Housing (48).
Covariates. Individual level covariates also included in the model have each been shown to be associated with anemia, maternal health service utilization, or IFA receipt. These were maternal age at index birth29,30, age of marriage30,31, maternal education32,33, gender composition of living children34, birth order of index pregnancy29,32,35, caste19, religion19, and household wealth quintile19 (See Table 2.2).
Covariates. The covariates available for this analysis included age and sex. We also used data on health insurance status, history of hypertension, history of diabetes, and smoking, abstracted from medical records to describe the initial characteristics of the population with a recurrent AMI.
Covariates. We included the following socioeconomic and demographic variables to adjust for confounding of the relationship between the schooling variables and prior year IPV: residence (urban versus rural), age measured continuously in years, relationship status (married, living together, previously partnered), work status (worked within the past year versus not working), number of children living at home measured continuously, witnessing father beat mother (yes, no), and a continuous score for household wealth. The DHS household wealth index is the standardized score (mean 0, standard deviation of 1) derived from the first principal component of a principal components analysis of recoded items measuring whether or not the household had a specified set of assets and amenities (Xxxxxxxx, 2004). Age of partner was not asked of all women in the previously partnered category and therefore was not included in the analyses. Statistical Analysis All descriptive and inferential analyses were conducted using the PASW version 18.0 statistical package for PC (PASW Statistics 18, 2009) and multivariate logistic regression was conducted using SAS version 9.2 (SAS Institute, 2009). Univariate analyses were conducted for all covariates, outcomes, and variables from which analytic covariates were derived to assess their completeness, distributions, and relative frequencies. Bivariate associations of all covariates were estimated to assess potential co- linearity among these variables. Variables statistically associated with our outcome in the bivariate analysis were included in the multivariate model. Because the outcome variable, prior year IPV (y), for each individual is binary, multivariate logistic regression analysis was used with strata at the regional level and clustering at the primary sampling unit (PSU) to estimate the adjusted associations of explanatory variables on the log odds of having experienced IPV in the prior year. The SAS survey logistic procedure fits linear logistic regression models for survey data using the maximum likelihood method (Binder, 1983; XxXxxxxxx, 1989). The procedure incorporates complex survey sample designs, including designs with stratification, clustering, and unequal weighting for statistical inferences (SAS Institute I, 2004). The model used to estimate the adjusted effects of covariates on the adjusted log odds of having experienced IPV in the prior year was:
Covariates. In addition to diarrheal prevalence, assessing the general health status of all children surveyed was important for this study. Therefore, three important anthropometric indicators were assessed—stunting (low height-for-age), wasting (low weight-for-height), and anemia status (mean hemoglobin)—in order to gauge nutritional status of each study participant. Anthropometric measures including height, weight and age were used to calculate z-scores (using 2005 WHO growth standards) to determine if a child was stunted3, wasted4 or underweight5 and are presented in Table 2. Hemoglobin measurements obtained from fingerpricks were used to determine a child’s anemia status6. Blood smears were obtained from study participants to test for the presence of malaria.
Covariates. The primary endpoints and selected secondary endpoints will be evaluated in subgroup analysis by subjects with or without prior blinatumomab, and by subjects with or without prior inotuzumab. Such subgroup analyses may not be performed if too few (eg, n < 5) subjects in the mITT set have received prior blinatumomab or prior inotuzumab at the time of the analysis. Additional covariates and subgroup analyses will be outlined in the Statistical Analysis Plan.
Covariates. In addition to the key independent variable, we controlled for several socio- demographic factors (see Table 1) and current smoking status. Drawing from the literature, we believe that smoking has a major impact on the severity of asthma and asthma control, thus we included whether enrollees currently smoke as a measure of their increased risk for emergency department utilization. Table 1: List of Covariates Variable Name Variable Type Description Age Continuous Ranging from 18 to 64 Gender Binary Female yes (1), no (0) for male Smoking Status Binary Yes (1) No (0) Race Categorical 1 Caucasian 2. African-American 3 Other / multiple (including Asian) Marital Status Categorical 1 Married 2 Widowed/Divorced/Separated 3 Never Married Region Categorical 1 Northeast 2 Midwest 3 South 4 West Educational level Categorical 1 High School or less 2 Any college education 3 Any Postgraduate education Health Status Categorical 1 Excellent 2 Very Good 3 Good 4 Fair 5 Poor Income as % FDL Categorical 1 Poor / Negative 2 Near Poor 3 Low Income 4 Middle Income 5 High income Statistical Analysis We conducted the data analysis with a two-part model. Two-part models are often used when dealing with utilization or expenditure data because due to large numbers of non-users of health services, there are excess zeros in the resource use or cost data and the assumption of normality of the error term is not satisfied [34]. These models usually involve outcomes that have two different statistical meanings, first that the outcome is larger than zero and second the outcome, based on the assumption that it was a positive outcome in the first part. [36]. In the two-part model, a binary choice model is estimated for the probability of observing a zero versus positive outcome. Then, conditional on a positive outcome, an appropriate regression model is estimated for the positive outcome observed [37]. Two- part models are shown to perform better than single-equation models in terms of split sample (in our case: ER expenditure versus none) mean-squared forecast error as they accommodate heterogeneity between users and non-users as well as heterogeneity across users based on level of use [38]. It has also been shown that co-linearity problems and violation of the bivariate normality assumption for the error term, likely in health data sets like MEPS, lead to poor performance of selectivity models and we performed tests to check the model fit in STATA, the modifed Xxxxxx-Xxxxxxxx test and the ...
Covariates. All information in the matched multiple birth data set is based on what is documented on the birth certificate or death certificate of the individual infant. Maternal race was categorized as white, black, or other. Maternal age was classified into 5 different categories: <20 years, 20-24 years, 25-29 years, 30-35 years, and >35 years. Maternal education was categorized as (1) less than high school, (2) completed high school, (3) some college, (4) completed college or greater, and (5) unknown. Gestational age at delivery is reported based on either mother’s last menstrual period or clinical ultrasound, depending on the method recorded on the birth certificate. Gestational age was categorized into 4 classes: <28 weeks, 28-32 weeks, 33-36 weeks, or ≥37weeks. The data also reported information on any congenital anomalies that were recorded on the infant birth or death certificate. The data contains specific information about 21 different congenital anomalies, as well as a category for any other unknown or unspecified anomalies. Distributions were calculated for each anomaly separately, and a new variable created to represent the presence of any type of anomaly. Presence of congenital anomaly was further classified as present, not present, or unknown or unspecified. Maternal parity was categorized as either nulliparous or multiparous. Adequacy of prenatal care was classified as either (1) early prenatal care (entry in to prenatal care during the first trimester), (2) late prenatal care (entry into prenatal care after the first trimester), or (3) no prenatal care or unknown. Placental abruption was classified as present, not present, or unknown.
Covariates. Several relevant socio-demographic covariates were analyzed. Two of them (age and number of children) were continuous while the rest were divided into categories. These included age, number of children, race (Black, White, Other), education (<9th grade, 9th grade – high school, beyond high school), household income (<$10,000, >=$10,000), employment status (employed, unemployed), marital status (single/separated/divorced, married, not married but in an intimate relationship) and housing status (currently homeless, not homeless, homeless, but not currently). Statistical analysis All analyses were completed utilizing STATA 13 SE. Chi-square tests of association analyses were performed for bivariate associations between the categorical socio- demographic covariates and the categorical outcome variable (HIV testing in the past year). Student t tests, as well as analyses of variance were performed to determine bivariate associations between the categorical socio-demographic covariates and the continuous outcome variable (Years since last HIV test). T tests and analyses of variance were also used to decipher associations between the two IPV variables, the continuous socio-demographic variables (age, number of children) and the categorical outcome variable (HIV test in the past year). A multivariate linear regression was generated to explore the association between IPV and HIV testing uptake (as measured by years since last HIV test), while accounting for potential socio-demographic covariates. Finally, in order to determine the potential confounders in the association between IPV (as measured by the index of psychological abuse scale and severity of violence against women scale) and HIV testing behavior in the past year, tests of correlation and analyses of variance were performed. All analyses were conducted at the 95% confidence level. CHAPTER 3: RESULTS
Covariates. Participants reported age, gender, marital status (single, married, divorced/separated/widowed), and years of formal schooling. Participants were categorized according to whether they had major responsibilities in their household food economy: that is, I distinguished between dependents living with parents or guardians, versus male and female heads of household and females living with parents but sharing food economy responsibilities (shopping and cooking). Participants reported whether they were members of traditional finance-pooling clubs called їqub. An їqub typically comprises a group of friends or co-workers who pool an amount of money monthly, and one member takes the sum in turn. Participants estimated monthly household incomes at all three rounds. At Rounds 2 and 3, participants reported household composition (i.e. adults and children regularly sleeping and eating in the house). At each round, estimated household incomes were divided by the total number of people in the household to yield monthly household per capita incomes in Ethiopian Birr/mo (converted to USD/mo using a rounded exchange rate of 10 Birr to 1 USD current at the time of the study). Household composition was not reported at Round 1. Since average household composition did not change between Rounds 2 and 3, we assumed that it also had not changed from Round 1 to Round 2. Thus we divided household incomes reported at Round 1 by the total number of people in the household at Round 2. Participants answered three dichotomous questions addressing household economic coping “to fulfill basic needs” in the past three months: 1) starting a new income-generating activity, 2) selling household goods, and 3) keeping students home from school to help in income-generation or food preparation. Participants were categorized according to whether they or anyone in their households engaged in one or more of these coping measures versus none in the three months prior to survey. At each round, participants reported whether they were receiving free food aid from non-governmental organizations, and what kinds of foods they were receiving. Wheat grain or flour was the most common type of food aid reported; in 2007 and the first part of 2008, free wheat was accessed often from NGOs like Xxxxx and Xxxxxx, and was commonly traded for cash by recipients. Participants were categorized based on whether they were receiving free wheat at the time of the survey. At each round, participants reported their total numb...
