Project Description

Using the 2017-2018 demographics and reproductive health datasets from the NHANES website, I will answer the research question: what factors influence a woman’s use of birth control in her lifetime?

The reproductive health data was collected as part of a computer-assisted personal self interview. These were completed as part of a visit to the examination center. The demographic data was collected as part of a questionaire. This data was collected in an interview at home. Data was collected from 2017-2018. The original dataset had 3286 observations of 96 variables.

The data and codebooks are available on the NHANES website: https://wwwn.cdc.gov/nchs/nhanes/continuousnhanes/default.aspx?BeginYear=2017

Measurement

To answer my research question, I used a subset of the demographics and reproductive health data that included survey participants’ responses to the following questions:

What is your total family income (reported as a range value in dollars)? What is the highest grade or level of school you have completed or the highest degree receieved? Have you ever taken birth control pills for any reason? How old were you when you had your first menstrual period? Have you had at least one menstrual period in the past 12 months?

The outcome variable, if a woman has ever used birth control, is coded as RHQ420. It is a factor (categorical) variable. It is coded with the following: “1” = Yes, “2” = No, “7” = Refused, “9” = Don’t know.

The predictor variables are age of menarche (first menstrual period), had regular period in last 12 months, highest grade/level of education, and family income level.

One predictor variable, age at menarche, is coded as RHQ010. It is a numeric data type. It is coded as: “6-19” = the numeric age specified by the value, “0” = hasn’t started yet, “20” = 20 years or older, “777” = refused, and “999” = don’t know. For this variable, the value of ‘0’ was removed. Women who have not started their period are unlikely to have used birth control in their lifetime. 32 women were in this category, which was dropped from analysis. Additionally, the value of ‘20’ includes women who started their period at age 20 or older. The exact age values above 20 were not included in this analysis, so this group was dropped from analysis. This group includes 6 women.

Another predictor variable, had regular periods in past 12 months, is coded by RHQ031. It is a factor/categorical data type. It is coded as: “1” = yes, “2” = no, “7” = refused, and “9” = don’t know.

Another predictor variable is the education level for adults age 20+. It is coded by DMDEDUC2. It is a factor/categorical variable. It is coded as: “1” = less than 9th grade, “2” = 9-11th grade, “3” = High school graduate/GED, “4” = Some college/AA degree, “5” = college graduate, “7” = refused, and “9” = don’t know.

The final predictor variable is household income, coded by INDFMIN2. It is also a categorical/factor data type. It is coded as: “1” = ‘$0 to $4,999’, “2” = ‘$5,000 to $9,999’, “3” = ‘$10,000 to $14,999’, “4” = ‘$15,000 to $19,999’, “5” = ‘$20,000 to $24,999’, “6” = ‘$25,000 to $34,999’, “7” = ‘$35,000 to $44,999’, “8” = ‘$45,000 to $54,999’, “9” = ‘$55,000 to $64,999’, “10” = ‘$65,000 to $74,999’, “12” = ‘$20,000 and over’, “13” = ‘Under $20,000’, “14” = ‘$75,000 to $99,999’, “15” = ‘$100,000 and over’, “77” = ‘Refused’, “99” = ‘Don’t know’. For this variable, the values for ‘under $20,000’ and ‘$20,000 and over’ were removed because these were double counted in other levels. Income levels were combined to make less income categories.

For the birth control, regular period, income, and education variables, the values for don’t know/refused were removed.

Results

Descriptive Statistics

Figure 1. The distribution of the age of menarche.

The distribution of the age of menarche variable has a slight right skew, so the median will be used for descriptive statistics. Here is a table of the basic descriptive statistics for the variables.

Birth control use by individual factors (n = 3286; NHANES 2017-2018)
Yes
(N=1736)		 No
(N=1242)		 Overall
(N=3286)
Age of Menarche 12.0, 1.00 12.0, 2.75 12.0, 2.75
Regular Period in Past 12 Months
Yes 781 (45.0%) 768 (61.8%) 1552 (47.2%)
No 954 (55.0%) 473 (38.1%) 1434 (43.6%)
Annual Family Income
$0 to $24,999 420 (24.2%) 311 (25.0%) 813 (24.7%)
$25,000 to $54,999 480 (27.6%) 346 (27.9%) 905 (27.5%)
$55,000 to $74,999 187 (10.8%) 109 (8.8%) 319 (9.7%)
$75,000 to $99,999 157 (9.0%) 94 (7.6%) 269 (8.2%)
$100,000 and over 299 (17.2%) 215 (17.3%) 554 (16.9%)
Highest Level of Education Completed
Less than 9th grade 84 (4.8%) 104 (8.4%) 231 (7.0%)
9th to 11th grade 144 (8.3%) 95 (7.6%) 276 (8.4%)
High school graduate/GED 373 (21.5%) 196 (15.8%) 618 (18.8%)
Some college/AA degree 649 (37.4%) 217 (17.5%) 932 (28.4%)
College graduate 406 (23.4%) 203 (16.3%) 660 (20.1%)

For the age of menarche, the median is the same for the group of women that used birth control and those that have not (m = 12.0). However, the IQR is different for those that use birth control (IQR = 1.0) and those that do not (IQR = 2.75). Though the two categories have different spreads, it seems unlikely that there will be an association between these variables, as the median is similar for both categories. For those who had a regular period in the past 12 months, women who have a regular period are less likely to use birth control (61.8% do not use birth control). Women who do not have a regular period are more likely to have used birth control (55% do use birth control). It seems that there will likely be an association between these variables. For annual family income, the frequencies in each income category are similar between those that have used birth control and those that haven’t. It seems unlikely that there will be an association here. Finally, for the highest level of education variable, the frequency of birth control use increases as education level increases. Thus, there is likely an association between these variables.

Age of Menarche and Birth Control Use

Figure 2. Birth control use by age of menarche according to 3286 participants in a 2017-2018 survey.

The data for those who do not use birth control seems to be normally distributed, as the median is in the middle of the box. However, the data for those who do use birth control is skewed because the median is towards the bottom of the box, so there may be some larger values on the top of the distribution. The IQR for those who do not use birth control is much larger, and this category has fewer outliers. Because the medians of the spreads are similar, it seems unlikely that the age of menarche and birth control use are related.

An independent samples t-test will be used to assess the relationship between the age of menarche and the use of birth control.

H0: There is not an association between the age of menarche and the use of birth control in a woman’s lifetime.

HA: There is an association between the age of menarche and the use of birth control in a woman’s lifetime.

Checking the assumptions:

  • The observations are independent. –> MET
  • There is 1 binary variable and 1 continuous variable. –> MET
  • There is a normal distribution of age within birth control groups.

Figure 3. The distribution of birth control use by age of menarche according to 3286 participants in a 2017-2018 survey.

The distribution is normal for women that have used birth control in their lifetime. However, the distribution is slightly right skewed for women that have not used birth control in their lifetime. Thus, this assumption fails.

  • There is an equal variance for each group. –> MET
## Levene's Test for Homogeneity of Variance (center = median)
##         Df F value Pr(>F)
## group    1  0.0041 0.9491
##       2941

This Levene’s test has a p-value of 0.9, which is not enough to reject the null hypothesis. Thus, this assumption is met.

Because the normality assumption failed, I will conduct a Mann-Whitney U test.

## 
##  Wilcoxon rank sum test with continuity correction
## 
## data:  nhanes2018.clean$period.start by nhanes2018.clean$birth.control
## W = 1059521, p-value = 0.72
## alternative hypothesis: true location shift is not equal to 0

A Mann-Whitney U test comparing the age of first menstrual period for those who use birth control (m = 12.0) and those who did not use birth control (m = 12.0) did not find a statistically significant difference between the two groups (p = 0.72). This indicates that there is no difference in the age of menarche between those who use birth control and those who do not use birth control.

Regular Period and Birth Control Use

Figure 4. Birth control use by period regularity in the last 12 months according to 3286 participants in a 2017-2018 survey.

For those who had a regular period, the frequency of those who used birth control and those who did not was very similar. However, for those who did not have a regular period, there were many more people who used birth control than those who did not. Based on the visualization of this data, it seems likely that period regularity and birth control use are related.

A chi-squared test will be used to assess the relationship between period regularity in the last 12 months and birth control use.

H0: There is no association between having a regular period in the last 12 months and birth control use.

HA: There is an association between having a regular period in the last 12 months and birth control use.

Checking the assumptions:

  • The variables are nominal. –> MET
  • The observations are independent. –> MET
  • At least 80% of the groups have 5 or more observations. –> MET
##    Cell Contents 
## |-------------------------|
## |                       N | 
## |              Expected N | 
## |           N / Row Total | 
## |            Std Residual | 
## |-------------------------|
## 
## =========================================================
##                                   nhanes2018.clean$reg.period
## nhanes2018.clean$birth.control       Yes       No   Total
## ---------------------------------------------------------
## Yes                                  781      954    1735
##                                    903.1    831.9        
##                                    0.450    0.550   0.583
##                                   -4.062    4.232        
## ---------------------------------------------------------
## No                                   768      473    1241
##                                    645.9    595.1        
##                                    0.619    0.381   0.417
##                                    4.803   -5.004        
## ---------------------------------------------------------
## Total                               1549     1427    2976
## =========================================================
## 
## Statistics for All Table Factors
## 
## Pearson's Chi-squared test 
## ------------------------------------------------------------
## Chi^2 = 82.51237      d.f. = 1      p <2e-16 
## 
## Pearson's Chi-squared test with Yates' continuity correction 
## ------------------------------------------------------------
## Chi^2 = 81.83777      d.f. = 1      p <2e-16
## [1] 0.165829

I used the chi-squared test to test the null hypothesis that there was no relationship between having a regular period in the last 12 months and the use of birth control. I rejected the null hypothesis and concluded that there was a statistically significant association between having a regular period in the last 12 months and the use of birth control [\(\chi^2\)(1) = 81.83, p < .001]. There were significantly more people without a regular period that used birth control (n = 954, std res = 4.23), and there were significantly more people with a regular period that do not use birth control (n = 768, std res = 4.80). Although statistically significant, the relationship was weak to moderate (V = 0.17).

Annual Family Income and Birth Control Use

Figure 5. Birth control use by annual family income according to 3286 participants in a 2017-2018 survey.

For all income levels, more people used birth control than didn’t use birth control. There does not seem to be any clear pattern between birth control use and income level overall; most income categories have a similar breakdown of those who used birth control and those who did not. Based on the frequencies, it seems unlikely that income and birth control use are related.

A chi-squared test will be used to assess the relationship between annual family income and birth control use.

H0: There is no association between annual family income and birth control use.

HA: There is an association between annual family income and birth control use.

Checking the assumptions:

  • The variables are nominal. –> MET
  • The observations are independent. –> MET
  • At least 80% of the groups have 5 or more observations. –> MET
##    Cell Contents 
## |-------------------------|
## |                       N | 
## |              Expected N | 
## |           N / Row Total | 
## |            Std Residual | 
## |-------------------------|
## 
## ===============================================================================
##              nhanes2018.clean$income
## nh2018.$.    $0 t $24,   $25,000 t   $55,000 t   $75,000 t   $100,000     Total
## -------------------------------------------------------------------------------
## Yes                420         480         187         157         299     1543
##                  430.8       486.8       174.5       147.9       302.9         
##                  0.272       0.311       0.121       0.102       0.194    0.589
##                 -0.522      -0.310       0.950       0.745      -0.226         
## -------------------------------------------------------------------------------
## No                 311         346         109          94         215     1075
##                  300.2       339.2       121.5       103.1       211.1         
##                  0.289       0.322       0.101       0.087       0.200    0.411
##                  0.626       0.371      -1.138      -0.893       0.271         
## -------------------------------------------------------------------------------
## Total              731         826         296         251         514     2618
## ===============================================================================
## 
## Statistics for All Table Factors
## 
## Pearson's Chi-squared test 
## ------------------------------------------------------------
## Chi^2 = 4.571276      d.f. = 4      p = 0.334

I used the chi-squared test to test the null hypothesis that there was no relationship between annual family and the use of birth control. I did not reject the null hypothesis and concluded that there was not a statistically significant association between annual family income and the use of birth control [\(\chi^2\)(4) = 4.58, p = 0.334].

Highest Education Level and Birth Control Use

Figure 6. Birth control use by highest level of education according to 3286 participants in a 2017-2018 survey.

For all education groups except less than 9th grade, there is a greater percentage of women who use birth control than those who do not. Additionally, the frequency of birth control use seems to increase as education level increases. Thus, it seems likely that education and birth control use are associated.

A chi-squared test will be used to assess the relationship between highest education level and birth control use.

H0: There is no association between education level and birth control use.

HA: There is an association between education level and birth control use.

Checking the assumptions:

  • The variables are nominal. –> MET
  • The observations are independent. –> MET
  • At least 80% of the groups have 5 or more observations. –> MET
##    Cell Contents 
## |-------------------------|
## |                       N | 
## |              Expected N | 
## |           N / Row Total | 
## |            Std Residual | 
## |-------------------------|
## 
## ===============================================================================
##              nhanes2018.clean$education
## nh2018.$.    Lss t 9 g   9t t 11 g   H s g/GED   Sm c/AA d   Cllg grdt    Total
## -------------------------------------------------------------------------------
## Yes                 84         144         373         649         406     1656
##                  126.0       160.2       381.3       580.4       408.1         
##                  0.051       0.087       0.225       0.392       0.245    0.670
##                 -3.741      -1.278      -0.427       2.849      -0.106         
## -------------------------------------------------------------------------------
## No                 104          95         196         217         203      815
##                   62.0        78.8       187.7       285.6       200.9         
##                  0.128       0.117       0.240       0.266       0.249    0.330
##                  5.333       1.821       0.608      -4.061       0.151         
## -------------------------------------------------------------------------------
## Total              188         239         569         866         609     2471
## ===============================================================================
## 
## Statistics for All Table Factors
## 
## Pearson's Chi-squared test 
## ------------------------------------------------------------
## Chi^2 = 72.57547      d.f. = 4      p = 6.49e-15
## [1] 0.1713794

I used the chi-squared test to test the null hypothesis that there was no relationship between the highest level of education achieved and the use of birth control. I rejected the null hypothesis and concluded that there was a statistically significant association between the highest level of education received and the use of birth control [\(\chi^2\)(4) = 72.57, p < .001]. There were significantly more people with less than 9th grade education that did not use birth control (n = 104, std res = 5.33), and there were significantly more people with some college education that do use birth control (n = 649, std res = 2.85). Although statistically significant, the relationship was weak to moderate (V = 0.17).

Discussion

The research question was answered - both period regularity and education level influence a woman’s use of birth control in her lifetime.

Women who do not have a regular period are more likely to have used birth control in their lifetime. This is consistent with other studies about oral contraceptive use for irregular periods. Women who have abnormal uterine bleeding are likely to respond positively to low-dose oral contraceptives. In a study of women who had abnormal periods for at least 3 months, 80% of women reported improvements after the use of a triphasic birth control pill, compared to only 50% of women who used a placebo. Additionally, birth control pills come with the ability to reduce the number of periods a woman gets in a year, which may seem appealing to women who have irregular periods. Thus, the results of this study make sense in the context of what is known about the effectiveness of oral contraceptives (Kaunitz, 2002).

Women with less than a 9th grade education were less likely to have used birth control in their lifetime, and women with some college education were more likely to have used birth control in their lifetime. These results are consistent with previous studies. A study of mothers in Kinshasa found that higher levels of education are inversely related with the number of children women want. Mothers that are more educated typically want to be able to educate their children, which is costly. This means that they will often desire to have fewer children than less educated mothers. Women who do not want more children typically engage in the use of birth control. In these mothers, the number of women who have ever used birth control methods rises as the education level of the mothers rises. This is, again, consistent with the results shown here. However, most of these women preferred condoms as their birth control method of choice (Shapiro & Tambashe, 1994).

Income level did not seem to be associated with the use of birth control. This is inconsistent with previous studies. A national survey conducted in Canada showed that women with a lower household income were 15% less likely to use birth control pills than those with higher household incomes. The researchers suggested that women from higher income levels were more likely to use birth control pills because they were more likely to have access to birth control, as cost was not a barrier to them. These women were also more likely to have prescription insurance to pay for birth control. Further research on this topic is needed (Nethery et al., 2019).

Age of menarche did not seem to be associated with the use of birth control. There is little research about this association, but some studies did show an association between the two. In a field study of over 13,000 women, women who had a later age of menarche were found to have more profound side effects from the birth control pill, such as heavy withdrawal bleeding, more frequent spotting, weight gain, depression, and headaches. This study reported that women with late menarche were more likely to discontinue the use of birth control pills due to the side effects, showing that there may be an association here. Again, further research on this topic is needed (Wenderlein, 1980).

By studying what factors influence a woman’s use of birth control, public health professionals can understand what interventions may be needed to make birth control more accessible to women who want and need it most. I suggest that further research be done on this topic to determine other factors that may influence a woman’s use of birth control pills in her lifetime. Women use birth control for a variety of reasons other than contraception, though that is the primary reason women use the pill. About 14% of pill users take birth control for other reasons, primarily for health benefits. Cramps, regulation of menstrual period, acne, and endometriosis are the primary reasons women take birth control for health benefits. Most users who use birth control for other reasons are teenagers (Jones, 2011). Additionallly, the Canadian study found that younger age groups were more likely to use oral birth control pills (Nethery et al., 2019). The study of mothers in Kinshasa suggests that employed women also have higher rates of birth control use. Additionally, this study found that women in their first marriage were more likely to use birth control than those in their second or third marriages (Shapiro & Tambashe, 1994). These are additional factors that may be researched in order to improve access to birth control in these groups of women.

References

Jones, R. K. (2011). Beyond Birth Control: The Overlooked Benefits of Oral Contraceptive Pills. Guttmacher Institute.

Kaunitz, A. M. (2002). Noncontraceptive Health Benefits of Oral Contraceptives. Reviews in Endocrine and Metabolic Disorders, 3, 277–283.

Nethery, E., Schummers, L., Maginley, K. S., Dunn, S., & Norman, W. V. (2019). Household income and contraceptive methods among female youth: A cross-sectional study using the Canadian Community Health Survey (2009–2010 and 2013–2014). CMAJ Open, 7(4). https://doi.org/10.9778/cmajo.20190087

Shapiro, D., & Tambashe, B. O. (1994). The impact of Women’s employment and education on contraceptive use and abortion in Kinshasa, Zaire. Studies in Family Planning, 25(2), 96. https://doi.org/10.2307/2138087

Wenderlein, V. J. M. (1980). Menarche of young girls and tolerance of hormonal contraceptives - results of a field study of 33,000 cases. Zentralblatt Für Gynäkologie, 102, 974–980.