Biostatistics Final Project: Predictors of Infant Birth Weight
2023-12-13
Project Overview
CDC 2022 Birth Data Codebook:
https://ftp.cdc.gov/pub/Health_Statistics/NCHS/Dataset_Documentation/DVS/natality/UserGuide2022.pdf
Research Overview
- The research question:
- How do prenatal maternal behaviors and health characteristics, such as when prenatal care began, the number of cigarettes in the first semester, the mother’s education level, and if the mother was on WIC during pregnancy, impact infant birth weight?
- What is the purpose of the research?
- The purpose of this research is to better understand how different variables impact infant birth weight to help design prenatal interventions to improve infant health.
- What is the outcome variable?
- The outcome variable is the infant’s birth weight in grams.
- What are the predictors?
- The predictor variables are the month into pregnancy when prenatal care began, the number of cigarettes the mother had in their first trimester, the mother’s education level, and if the mother was on WIC during their pregnancy.
Data Set Information
- When was it collected?
- Data from the year 2022
- Data from the year 2022
- How was it collected?
- The data came from birth certificate records from the CDC.
- The data came from birth certificate records from the CDC.
- How many observations and variables are there?
- 3229928 observations of 225 variables.
- 3229928 observations of 225 variables.
- How were the outcome and predictors measured?
- start of prenatal care:
- PRECARE
- continuous
- 1-10 month prenatal care began
- PRECARE
- Cigarettes 1st trimester:
- CIG_1
- Continuous
- 00-97 cigarettes daily
- 98: 98 or more cigarettes daily
- 99: unknown or not states
- CIG_1
- Mother’s education level
- MEDUC
- Categorical
- 1: 8th grade or less
- 2: 9th through 12th with no diploma
- 3: High school graduate or GED completed
- 4 Some college credit, but not a degree
- 5: Associates degree (AA,AS)
- 6 : Bachelor’s degree (BA.AB, BS)
- 7: Master’s degree (MA, MS, MEng, MEd, MSW, MBA)
- 8: Doctorate or Professional degree
- 9: Unknown
- MEDUC
- WIC status:
- WIC
- categorical
- Y: Yes
- N: No
- U: Unknown
- WIC
- Infant Birth weight
- DBWT
- Continuous
- 0227-8165: Number of Grams
- 9999: Not stated birth weight
- DBWT
- start of prenatal care:
Descriptive Statistics
The precare variable is abnormally distributed and skewed to the right. The median (3 months) and IQR will be used to describe the distribution of the month that a mother began to receive prenatal care into pregnancy. The cigarettes variable is also skewed to the right, with a median of 0 and an IQR of 0. The final continuous variable, birth weight, appears to be normally distribute, but may also be slightly skewed. The median is 3290 grams and the mean is 3250 grams.
Table
| Overall (N=3497504) |
|
|---|---|
| precare | |
| Mean (SD) | 2.88 (1.48) |
| Median [Min, Max] | 3.00 [0, 10.0] |
| momEducation | |
| 8th grade or less | 103491 (3.0%) |
| 9th-12th grade no diploma | 279115 (8.0%) |
| HS Graduate or GED | 911725 (26.1%) |
| Some college | 651023 (18.6%) |
| Associates degree | 299481 (8.6%) |
| Bachelors Degree | 775509 (22.2%) |
| Masters degree | 368313 (10.5%) |
| Doctorate or Professional deggree | 108847 (3.1%) |
| birthWeight | |
| Mean (SD) | 3250 (585) |
| Median [Min, Max] | 3290 [227, 8140] |
| cigarettes | |
| Mean (SD) | 0.497 (3.04) |
| Median [Min, Max] | 0 [0, 98.0] |
| wic | |
| Yes | 1041520 (29.8%) |
| No | 2455984 (70.2%) |
Predictor Variables vs. Outcome Variable Overview
Birth Weight vs. Months into Pregnancy Prenatal Care Began Scatterplot
The months into pregnancy that prenatal care began does not seem to have a direct relationship with birth weight. However, prenatal care that began later, specifically after 6 months, tends to have higher minimum birth weights compared to pregnancies where prenatal care began later.
Birth Weight vs. # of Cigarettes in 1st Trimester Scatterplot
The graph shows that cases with more cigarettes in the first trimester have birth weights closer to the median, with very few birth weights higher than 5000 grams. There does not seem to be a strong association between cigarettes and birth weight.
Birth Weight vs. WIC Status Boxplot Graph
The box plots comparing WIC status and infant birth weight seem to be very similar, with a median birth weight slightly higher for those not on WIC.
Birth Weight vs. Mother’s Education Boxplot Graph
The box plots of infant birth weight for each education level of the mother seem to be very similar; there is slight variation in the median, with the most variation in the maximum birth weight for each group. There does not seem to be a strong association between mother’s education level and infant birth weight.
Bivariate Tests
Birth Weight vs. Month Prenatal Care Began
H0: There is no correlation between the month prenatal care began and infant birth weight.
HA: There is a correlation between the month prenatal care began and infant birth weight. Visual of the Relationship
There does not appear to be a relationship between birth weight and when prenatal care began.
Assumptions Testing for Pearson’s r correlation analysis
- Observations are independent.
- Both variables are continuous.
- Both variables are normally distributed. X
- The relationship between the two variables is linear. X
- The variance is constant with the points distributed equally around the line. X
The normality distribution assumption was failed because the precare variable is not normally distributed; it is skewed to the right.
The linearity assumption was failed. The left tail of the data is lower than the line, and the right tail of the data is higher than the line.
##
## studentized Breusch-Pagan test
##
## data: birthdata.clean$precare ~ birthdata.clean$birthWeight
## BP = 1861.1, df = 1, p-value < 2.2e-16The Homoscedasticity assumption was not met because there was a statistically significant relationship in the difference in spread across the data (p<.05).
Assumptions Testing Conclusion
The normality distribution, linearity, and homoscedasticity assumptions failed for the Pearson’s R test. The Spearman’s Rho Test will be used as an alternative test.
Assumption Testing Spearman’s Rho Test
- The variables must be at least ordinal or even closer to continuous.
- The relationship between the two variables must be monotonic.
- It is unclear whether this assumption is met from the graphs but for the sake of the analysis, the test will be completed. The results should be analyzed with caution based on this assumption.
##
## Spearman's rank correlation rho
##
## data: birthdata.clean$birthWeight and birthdata.clean$precare
## S = 7.0875e+18, p-value < 2.2e-16
## alternative hypothesis: true rho is not equal to 0
## sample estimates:
## rho
## 0.006033835There was a statistically significant positive correlation between birth weight and the month into pregnancy prenatal care began (rs = .006; p<.001). As the number of months increases, so does the infant birth weight.
Birth Weight vs. # of Cigarettes
H0: There is no correlation between the number of cigarettes the mother had in their first trimester and infant birth weight.
HA: There is a correlation between the number of cigarettes the mother had in their first trimester and infant birth weight. Visual of the Relationship
There appears to be a negative relationship between birth weight and the number of cigarettes consumed by the mother in their first trimester.
Assumptions Testing for Pearson’s r Correlation Analysis
- Observations are independent.
- Both variables are continuous.
- Both variables are normally distributed. X
- The relationship between the two variables is linear. X
- The variance is constant with the points distributed equally around the line. X
The normality distribution assumption was failed because the
cigarettes variable is not normally distributed.
The linearity assumption was failed. The curve dips below the line on the left and then goes above the line for the remainder of the graph.
##
## studentized Breusch-Pagan test
##
## data: birthdata.clean$birthWeight ~ birthdata.clean$cigarettes
## BP = 867.06, df = 1, p-value < 2.2e-16The Homoscedasticity assumption was not because the variance in the variables differs a significant amounts (p< .05).
Assumptions Testing Conclusion
The normal distribution, linearity, and homoscedasticity assumptions were failed for the Pearson’s R test for birth weight versus cigarettes. The Spearman’s Rho test will be used as an alternative.
Assumption Testing Spearman’s Rho
- The variables must be at least ordinal or even closer to continuous.
- The relationship between the two variables must be monotonic.
- It is unclear whether this assumption is met from the graphs but for the sake of the analysis, the test will be completed. The results should be analyzed with caution based on this assumption.
##
## Spearman's rank correlation rho
##
## data: birthdata.clean$cigarettes and birthdata.clean$birthWeight
## S = 7.6957e+18, p-value < 2.2e-16
## alternative hypothesis: true rho is not equal to 0
## sample estimates:
## rho
## -0.07925826There was a statistically significant negative correlation between birth weight and the number of cigarettes the mother had in her first trimester. (rs = -0.08; p<.001). As the number of cigarettes increases, birth weight decreases.
Birth Weight vs. Mother’s Education
H0: Infant birth weight is equal across mother's education level.
HA: Infant birth weight is not equal across mother's education level. Visual of the Relationship
Assumptions Testing for ANOVA test
- Continuous variable and independent groups.
- Independent observations.
- Normal distribution within groups.
- Equal variances within groups. X
Each group appears to be normally distributed, with a slight skew to the right.
## Levene's Test for Homogeneity of Variance (center = mean)
## Df F value Pr(>F)
## group 7 409.87 < 2.2e-16 ***
## 3497496
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1The assumption of homogeneity of variances is not met. The difference in variance between groups is statistically significant (p<.05).
Assumptions Testing Conclusion
The assumptions were not met for the ANOVA test because the equal variances within groups assumption failed. The Welch’s test will be used as an alternative.
Welch’s Test
##
## One-way analysis of means (not assuming equal variances)
##
## data: birthWeight and momEducation
## F = 4056, num df = 7, denom df = 749058, p-value < 2.2e-16The results show a statistically significant difference in the mean of the birth weight variable by mother’s education level [Fw(7, 749058) = 4056; p<.05].
Birth Weight vs. WIC Status
H0: The distribution of birth weight in infants is the same for mother's who are on WIC during pregnancy and those who are not on WIC during pregnancy.
HA: The distribution of birth weight in infants is different for mother's who are on WIC during pregnancy and those who are not on WIC during pregnancy.Visual of the Relationship
### Assumptions Testing for Independent-samples T-Test - Continuous
variable and independent groups. - Independent observations. - Normal
distribution within groups. - Homogeneity of variances. X
The graphs appear to be mostly normally distributed.
## Levene's Test for Homogeneity of Variance (center = median)
## Df F value Pr(>F)
## group 1 22.273 2.365e-06 ***
## 3497502
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1The assumption of homogeneity of variances is not met (p<.05). The difference in variance between the yes WIC group and the no WIC group is statistically significant.
Assumptions Testing Conclusion
The assumptions for the independent-samples t-test were not met because the homogeneity of variances assumption failed. The Kolmogorov-Smirnov test will be used as an alternative.
Kolmogorov-Smirnov Test
##
## Asymptotic two-sample Kolmogorov-Smirnov test
##
## data: wic.no and wic.yes
## D = 0.062111, p-value < 2.2e-16
## alternative hypothesis: two-sided
While it is a small difference between between birth weight for those on WIC, and those not on WIC during pregnancy, the difference of .062 is statistically significant (D=.062, p<.05).
Discussion
Based on the 2022 birth data from the CDC, the month prenatal care began, the number of cigarettes consumed by the mother in the first trimester, WIC status, and education level are good predictor variables of infant birth weight. While the relationship between the predictor variables and the outcome variable is not obvious from the graphs, the statistical analyses revealed a significant relationship between the variables. Other studies have examined the variables that impact birth weight, such as socioeconomic status, vitamins consumed during pregnancy, and previous low birth weight pregnancies (Khan et al. 2014). Another study looked at maternal lifestyle factors, such as smoking, gestational weight gain, and physical activity during pregnancy to examine their impact on birth weight (Xi et al. 2020). They found that smoking, improper weight gain, and gestational hypertension were associated with low birth weight (Xi et al. 2020). Other studies have looked at the impact of the Special Supplemental Nutrition Program for Women, Infants and Children (WIC). One study found the important role WIC has in increasing birth weight, length of gestation, NICU admission, and other pregnancy outcomes (Sonchak 2016). By better understanding what contributes to low birth weight and other outcome variables of pregnancy, interventions, like WIC, can be designed to improve infant and mother health.
Low birth weight is an important factor to study as it is a risk factor for infant mortality and can possibly impact diseases developed later in life. A retrospective cohort study in Brazil found that babies with extremely low birth weights (<1000 g), had a higher risk for infant mortality (Vilanova et al. 2019). A study from 2020 examined the relationship between low birth weight and the development of hypertension and chronic kidney disease later in life. They found that women who were born with a low birth weight were at a greater risk for hypertensive disorders in pregnancy, resulting in an increased risk for cardiovascular disease and end-stage renal disease (Kanda et al. 2020). Birth weight is an extremely important public health concept to study, as it has many contributing factors and many outcomes related to it. Further research needs to be done to better understand how interventions can prevent low birth weight, especially for people with risk factors for low birth weight. By increasing the understanding of birth weight and how to prevent low birth weight, many other outcomes such as infant mortality and the development of disease can be addressed.