Motivation of Study

Maternal health is a cornerstone of public health, influencing the well-being of both mothers and their children. Despite advances in healthcare, complications during pregnancy and childbirth remain a significant concern worldwide, contributing to maternal morbidity and mortality. Identifying and understanding the complex interplay between key physiological indicators, such as blood pressure, heart rate, blood sugar levels, and body temperature, can provide critical insights into the underlying factors that contribute to varying levels of maternal health risks.

This study is motivated by the need to improve risk assessment and preventive care for expectant mothers. By leveraging multivariate analysis techniques, we aim to uncover patterns and relationships among health indicators that differentiate low, medium, and high-risk individuals. Our findings have the potential to guide clinical interventions, inform personalized care strategies, and contribute to evidence-based policymaking to enhance maternal health outcomes globally. Through this analysis, we strive to address gaps in understanding and provide actionable insights to mitigate maternal health risks effectively.

Dataset Description:

This dataset was collected from hospitals, community clinics, and maternal healthcare centers in rural areas of Bangladesh using an IoT-based risk monitoring system. It contains 1,013 observations, with each row representing an individual case, and six features (variables) describing maternal health indicators and associated risk levels. The dataset supports multivariate analysis and is designed for classification tasks.

Variables:

  1. Age: The individual’s age in years (continuous variable).

  2. SystolicBP: Systolic blood pressure measured in mmHg, indicating blood vessel pressure during heartbeats.

  3. DiastolicBP: Diastolic blood pressure measured in mmHg, indicating blood vessel pressure between heartbeats.

  4. BS (Blood Glucose Levels): Blood glucose concentration in mg/dL, a critical factor for detecting diabetes or gestational complications.

  5. BodyTemp: Body temperature in degrees Fahrenheit, reflecting metabolic and physical health conditions.

  6. HeartRate: Heart rate measured in beats per minute (bpm), a vital sign for cardiovascular health.

  7. RiskLevel: A categorical outcome variable indicating maternal health risk level, categorized as:

    • Low Risk

    • Mid Risk

    • High Risk

The dataset aims to identify the relationships between health indicators and maternal health risks to support clinical decision-making and improve maternal care.

Research Question 1:

What are the key factors influencing maternal health risk levels (low, medium, high)?

Maternal health risk levels are critical for identifying at-risk individuals and improving outcomes. Understanding the relationship between key health indicators—blood glucose, blood pressure, and heart rate—and risk levels helps prioritize clinical focus areas. We chose the boxplot for blood glucose levels to reveal differences in distributions across risk categories, as it effectively highlights variations and potential outliers. Stacked bar charts for blood pressure and heart rate were selected to compare proportions across risk levels, showing trends in categorical variables like hypertension and heart rate classifications. Multinomial logistic regression was chosen to quantify the combined effects of these indicators on risk levels, allowing us to identify the most significant predictors with precision.

## # weights:  18 (10 variable)
## initial  value 1113.992861 
## iter  10 value 902.020681
## iter  20 value 837.473153
## iter  20 value 837.473152
## iter  20 value 837.473152
## final  value 837.473152 
## converged
## Call:
## multinom(formula = RiskLevel ~ BS + SystolicBP + DiastolicBP + 
##     HeartRate, data = maternal_data)
## 
## Coefficients:
##          (Intercept)         BS   SystolicBP  DiastolicBP  HeartRate
## low risk   16.990185 -0.7343843 -0.040706681 -0.006788434 -0.0711463
## mid risk    9.479964 -0.3428134  0.009240925 -0.047731800 -0.0444918
## 
## Std. Errors:
##          (Intercept)         BS  SystolicBP DiastolicBP  HeartRate
## low risk    1.549808 0.07871523 0.009884971  0.01251921 0.01434230
## mid risk    1.359308 0.03684657 0.009266688  0.01188558 0.01327421
## 
## Residual Deviance: 1674.946 
## AIC: 1694.946
##          (Intercept)         BS   SystolicBP  DiastolicBP  HeartRate
## low risk   16.990185 -0.7343843 -0.040706681 -0.006788434 -0.0711463
## mid risk    9.479964 -0.3428134  0.009240925 -0.047731800 -0.0444918
##           (Intercept) BS   SystolicBP  DiastolicBP    HeartRate
## low risk 0.000000e+00  0 3.821126e-05 5.876523e-01 7.027819e-07
## mid risk 3.078204e-12  0 3.186578e-01 5.920914e-05 8.030354e-04

Graph 1: Blood Glucose Levels by Maternal Health Risk Levels

Interpretation:
The boxplot demonstrates that blood glucose levels (BS) are substantially higher in the “high risk” category compared to “low risk” and “mid risk.” The median blood glucose level for “high risk” is notably above those of the other categories, with a larger interquartile range indicating more variability. Conversely, the “low risk” category has a narrow range and consistently lower values. This suggests that elevated blood glucose is a strong predictor of higher maternal health risk.

Takeaway:
Blood glucose levels are a significant indicator of maternal health risks, with higher levels strongly associated with the “high risk” group.

Graph 2: Blood Pressure Categories by Maternal Health Risk Levels

Interpretation:
The stacked bar chart reveals a clear pattern in blood pressure distribution across risk levels:

  • The “high risk” category is predominantly composed of individuals classified as “Hypertensive.”

  • In contrast, the “low risk” group has a substantial proportion of individuals with “Normal” blood pressure, with only a small percentage falling into the “Hypertensive” category.

  • The “mid risk” group represents a transitional mix, where “Hypertensive” cases are prevalent but with a larger “Normal” population compared to “high risk.”

Takeaway:
Hypertension is a critical factor in maternal health risk, with a sharp contrast between “high risk” and “low risk” categories.

Graph 3: Heart Rate Categories by Maternal Health Risk Levels

Interpretation:
The stacked bar chart for heart rate categories shows that nearly all individuals, regardless of risk level, fall into the “Normal” heart rate range. Only a small number of “Bradycardia” cases are observed, primarily in the “high risk” category.

Takeaway:
Heart rate, while mostly within the normal range across all risk levels, might not be as strongly associated with maternal health risk as blood glucose or blood pressure.

Statistical Analysis: Multinomial Logistic Regression

Interpretation:
The regression model assesses the predictive power of blood glucose (BS), systolic blood pressure (SystolicBP), diastolic blood pressure (DiastolicBP), and heart rate (HeartRate) on maternal health risk levels. Key findings include:

  1. Blood Glucose (BS): The negative coefficient for BS in “low risk” and “mid risk” categories relative to “high risk” suggests that higher blood glucose levels increase the likelihood of falling into the “high risk” category.

  2. Systolic and Diastolic Blood Pressure: Both blood pressure metrics show small but significant coefficients, further supporting their importance as predictors of maternal health risk.

  3. Heart Rate: The coefficients for heart rate are smaller and less significant, indicating that it is a weaker predictor compared to BS and blood pressure.

Takeaway:
Blood glucose and blood pressure (both systolic and diastolic) are significant predictors of maternal health risk levels, with blood glucose being the most influential. Heart rate appears to have a less pronounced effect.

Conclusion

Blood glucose and blood pressure are key factors in determining maternal health risk, with elevated glucose and hypertension being strong indicators of higher risk levels. Heart rate, while primarily normal across groups, is less predictive. This analysis highlights the need for targeted interventions focusing on glucose and blood pressure management to mitigate maternal health risks effectively.

Research Question 2:

How does maternal health risk level distribution differ by age group and blood pressure category?

Maternal health risk levels are critical for identifying at-risk individuals and improving outcomes. Exploring how risk levels vary across age groups and blood pressure categories provides a nuanced understanding of how demographic and physiological factors influence health risks. We chose the bar chart of risk level proportions by age group to highlight age-based trends in health risks. A heatmap of systolic vs. diastolic blood pressure, stratified by risk levels, reveals specific blood pressure ranges associated with different risk categories. Additionally, multinomial logistic regression was chosen to quantify the interaction effects between age, blood pressure, and risk levels, enabling the identification of high-risk subpopulations with precision.

## # weights:  15 (8 variable)
## initial  value 1113.992861 
## iter  10 value 982.540462
## final  value 981.914581 
## converged
## Call:
## multinom(formula = RiskLevel ~ Age + SystolicBP + DiastolicBP, 
##     data = maternal_data)
## 
## Coefficients:
##          (Intercept)         Age   SystolicBP DiastolicBP
## low risk    8.362300 -0.02572630 -0.047439155 -0.02094696
## mid risk    5.986604 -0.02611426 -0.003994701 -0.05578983
## 
## Std. Errors:
##          (Intercept)         Age  SystolicBP DiastolicBP
## low risk   0.7046960 0.006761704 0.008351776  0.01034795
## mid risk   0.7084185 0.006895472 0.008134047  0.01014494
## 
## Residual Deviance: 1963.829 
## AIC: 1979.829
##          (Intercept)         Age   SystolicBP DiastolicBP
## low risk    8.362300 -0.02572630 -0.047439155 -0.02094696
## mid risk    5.986604 -0.02611426 -0.003994701 -0.05578983
##          (Intercept)          Age   SystolicBP  DiastolicBP
## low risk           0 0.0001419724 1.345941e-08 4.294330e-02
## mid risk           0 0.0001523788 6.233496e-01 3.813554e-08

Graph 1 Interpretation:

Risk Level Proportions by Age Group
The bar chart reveals distinct trends in maternal health risk levels across age groups:

  • Women in the <20 age group are predominantly in the low risk category, with minimal proportions in the mid risk and high risk categories.

  • The 20-30 and 31-40 age groups show a balanced distribution among low risk, mid risk, and high risk levels, with a slight increase in high risk cases for 31-40.

  • The >40 age group has a noticeably higher proportion in the high risk category, highlighting advanced maternal age as a key risk factor.

Takeaway:
As age increases, the proportion of women categorized as high risk grows significantly, particularly for those over 40. Conversely, younger women under 20 are less likely to fall into the high risk category.

Graph 2 Interpretation :

Scatter Plot of Systolic vs. Diastolic Blood Pressure by Risk Level

The scatter plot illustrates the distribution of systolic and diastolic blood pressure values across different maternal health risk levels:

  • Low risk cases are tightly clustered in the lower-left region (systolic < 120 mmHg and diastolic < 80 mmHg), corresponding to normal blood pressure values.

  • Mid risk cases are more dispersed, spanning intermediate systolic and diastolic blood pressure ranges, reflecting mild deviations from normal values.

  • High risk cases occupy the upper-right region, where both systolic and diastolic blood pressure values are significantly elevated, indicating hypertension or severe cardiovascular stress.

Notable Patterns:
- High-risk individuals show a clear tendency toward higher systolic (≥ 130 mmHg) and diastolic (≥ 80 mmHg) pressures.
- Mid-risk individuals often overlap with both low- and high-risk groups, underscoring the transitional nature of this category.

Takeaway:
Elevated systolic and diastolic blood pressure levels are strongly associated with high maternal health risk, while normal values are predominantly linked to low risk. This pattern highlights hypertension as a key factor in maternal health risks and emphasizes the importance of monitoring blood pressure for early detection and management.

Statistical Analysis: Multinomial Logistic Regression

The model assesses the effects of Age, SystolicBP, and DiastolicBP on maternal health risk levels, with high risk as the reference category.

Key Findings:

  1. Age:
    • Negative coefficients for both low risk (-0.0257) and mid risk (-0.0261) indicate that increasing age raises the likelihood of being in the high risk category.
    • Highly significant (p < 0.001).
  2. SystolicBP:
    • Higher systolic blood pressure reduces the likelihood of being low risk (-0.0474, p < 0.001), but its effect on mid risk is not significant (-0.00399, p = 0.623).
  3. DiastolicBP:
    • Higher diastolic blood pressure lowers the odds of being low risk (-0.0209, p = 0.043) and mid risk (-0.0558, p < 0.001).

Takeaway:

Age and blood pressure are significant predictors of maternal health risk, with older age and higher blood pressure strongly linked to high risk. These findings highlight the importance of targeted interventions for these factors.

Conclusion

The analysis demonstrates clear patterns in maternal health risk levels based on age and blood pressure:
1. Age: Advanced maternal age (particularly >40) is strongly linked to higher risk levels, emphasizing the importance of targeted monitoring for older pregnant women.
2. Blood Pressure: Elevated blood pressure, particularly cases classified as hypertensive, is a significant determinant of high maternal health risk. Normal blood pressure is a protective factor for lower risk levels.
These findings highlight the need for age-specific and blood pressure-focused interventions to mitigate maternal health risks, particularly in high-risk subpopulations.

Research Question 3:

How do heart rate and blood sugar levels interact and vary across different maternal health risk levels?

Understanding the interaction between heart rate and blood sugar levels is crucial in assessing maternal health risks, as these physiological indicators are closely tied to cardiovascular and metabolic health. Variations in these factors can signal potential complications during pregnancy, such as gestational diabetes, preeclampsia, or other conditions that may endanger maternal and fetal outcomes. By investigating how these variables differ across low, medium, and high maternal health risk levels, we can uncover critical patterns that help identify at-risk populations earlier. This analysis focuses on the relationship between blood sugar levels (measured in mg/dL), heart rate (beats per minute), and the maternal health risk categories, providing actionable insights into preventive care and personalized health interventions.

Graph 1

The density plots compare the distributions of blood sugar and heart rate across different maternal health risk levels (high, mid, and low risk). These distributions reveal how these variables are associated with different risk levels:

  • For blood sugar, low-risk individuals have a tightly packed distribution around lower blood sugar levels (~7 mg/dL), while mid- and high-risk groups display more variability, with high-risk individuals skewing toward higher values.

  • For heart rate, mid-risk individuals exhibit a wide range of values, while high-risk individuals are clustered toward higher heart rate levels, suggesting potential physiological stress or anomalies in this group.

Key Takeaway: Blood sugar and heart rate exhibit distinct patterns for each risk level. Blood sugar is a clear differentiator between high-risk and low-risk groups, and higher heart rates are more indicative of high-risk individuals.

Graph 2: PCA plot

This PCA plot visualizes the relationships among key variables (blood sugar, heart rate, systolic and diastolic blood pressure) across risk levels. The first two principal components capture the most variance in the data:

  • Blood sugar (BS) and heart rate contribute significantly to the variance along the first component, aligning with their importance in distinguishing risk levels.

  • Clusters of low-, mid-, and high-risk individuals are moderately separable, with high-risk individuals occupying the extremes of the first principal component, suggesting greater variance and abnormality in their health indicators.

Key Takeaway: The PCA plot underscores the multivariate nature of maternal health risks, with blood sugar and heart rate being the dominant factors differentiating risk groups. This plot highlights patterns that align with the density plot insights while providing a broader context for how all health indicators interplay.

Together, these graphs provide a robust understanding of the interplay between blood sugar, heart rate, and maternal health risk levels. The density plots showcase variable-specific differences, while the PCA plot integrates these insights into a multivariate framework, revealing broader patterns and relationships. These findings emphasize the importance of monitoring blood sugar and heart rate as critical indicators for maternal health risk stratification.

Research Question 4:

Are there correlations between different maternal health indicators?

Maternal health is influenced by a variety of physiological indicators that often interact in complex ways. Understanding the correlations between these indicators, such as systolic blood pressure, diastolic blood pressure, heart rate, and blood glucose levels, is critical for identifying patterns that could signal potential health risks.

## Call:psych::corr.test(x = numerical_vars, method = "pearson")
## Correlation matrix 
##               Age SystolicBP DiastolicBP    BS BodyTemp HeartRate
## Age          1.00       0.42        0.40  0.47    -0.26      0.08
## SystolicBP   0.42       1.00        0.79  0.43    -0.29     -0.02
## DiastolicBP  0.40       0.79        1.00  0.42    -0.26     -0.05
## BS           0.47       0.43        0.42  1.00    -0.10      0.14
## BodyTemp    -0.26      -0.29       -0.26 -0.10     1.00      0.10
## HeartRate    0.08      -0.02       -0.05  0.14     0.10      1.00
## Sample Size 
## [1] 1014
## Probability values (Entries above the diagonal are adjusted for multiple tests.) 
##              Age SystolicBP DiastolicBP BS BodyTemp HeartRate
## Age         0.00       0.00        0.00  0        0      0.03
## SystolicBP  0.00       0.00        0.00  0        0      0.46
## DiastolicBP 0.00       0.00        0.00  0        0      0.28
## BS          0.00       0.00        0.00  0        0      0.00
## BodyTemp    0.00       0.00        0.00  0        0      0.01
## HeartRate   0.01       0.46        0.14  0        0      0.00
## 
##  To see confidence intervals of the correlations, print with the short=FALSE option

Graph 1: Heatmap of Correlations Between Health Indicators

Interpretation: The heatmap shows significant positive correlations between systolic and diastolic blood pressure, indicating a strong interdependence between these measures of cardiovascular health. Age shows moderate positive correlations with both systolic and diastolic blood pressure, suggesting that blood pressure tends to increase with age. Blood glucose levels (BS) exhibit weak correlations with blood pressure, hinting at a potential but limited relationship. In contrast, heart rate and body temperature display little to no significant correlation with other variables, indicating their independence within this dataset. Overall, the heatmap highlights key relationships among blood pressure indicators and age while suggesting that variables like heart rate and body temperature may require separate analysis or additional contextual factors to explain variability.

Takeaway: The heatmap reveals clear relationships between blood pressure indicators and age, while other variables like heart rate and body temperature appear less interdependent. This provides insights into which indicators may be jointly analyzed in future studies, such as focusing on blood pressure and age as related health factors.

Graph 2: Pair Plot of Health Indicators

Interpretation: The pair plot provides an in-depth visualization of the relationships between the health indicators. Systolic and diastolic blood pressures exhibit a strong positive linear relationship (correlation = 0.787), reinforcing their interdependence. Age shows a moderate positive correlation with systolic (0.416) and diastolic blood pressure (0.398), suggesting that blood pressure increases with age. Blood glucose levels (BS) have a moderate positive correlation with both systolic (0.473) and diastolic blood pressure (0.425), indicating a potential association between glucose levels and cardiovascular indicators. Body temperature has weak negative correlations with most variables, particularly systolic blood pressure (-0.287) and age (-0.255), suggesting inverse relationships. Heart rate exhibits very weak correlations with all other variables, highlighting its independence within this dataset.

Takeaway: The pair plot quantifies key interdependencies among health indicators, particularly the strong correlation between systolic and diastolic blood pressure and their moderate association with age and blood glucose levels. These insights suggest that age and glucose levels could play a role in influencing blood pressure, while variables like heart rate and body temperature appear less related to other indicators. This emphasizes the importance of focusing on blood pressure and glucose interactions in further analysis while treating heart rate and body temperature independently.

Statistical Analysis: Pearson Correlation Test

Interpretation: The Pearson correlation test provides a detailed analysis of the relationships between maternal health indicators. Significant positive correlations were observed between systolic and diastolic blood pressure (r = 0.79, p < 0.001), as expected due to their physiological connection. Blood glucose levels also show moderate positive correlations with both systolic blood pressure (r = 0.43, p < 0.001) and diastolic blood pressure (r = 0.42, p < 0.001), indicating a potential link between glucose regulation and blood pressure. Age is moderately positively correlated with systolic blood pressure (r = 0.42, p < 0.001) and diastolic blood pressure (r = 0.40, p < 0.001), reflecting the natural increase in blood pressure with aging. Body temperature exhibits weak negative correlations with blood pressure, suggesting less direct interaction. Lastly, heart rate shows weak correlations with other variables, indicating its relatively independent behavior among these indicators. These results highlight both strong and weaker interactions, offering valuable insights into maternal health dynamics.

Takeaway: The Pearson correlation test suppports strong relationships between systolic and diastolic blood pressure, moderate correlations between blood glucose and blood pressure, and age-related increases in blood pressure, emphasizing their interconnected roles in maternal health. Weak correlations involving heart rate suggest it operates more independently. These findings highlight key interactions among maternal health indicators and suggest potential areas for further exploration, such as causal pathways and confounding factors.

Conclusion

Our analysis confirms that there are significant correlations between different maternal health indicators, with notable relationships identified between systolic and diastolic blood pressure (strong positive correlation), blood sugar levels and blood pressure (moderate positive correlations), and age with blood pressure (moderate positive correlations). These findings highlight the interconnected nature of these variables, suggesting that changes in one indicator could be linked to shifts in others. However, weaker correlations, such as those involving body temperature and heart rate, indicate that some health indicators operate more independently. Overall, the results provide a comprehensive understanding of how maternal health indicators interact, offering a foundation for further investigation and potential predictive modeling in maternal healthcare.

Summary

Our project investigates maternal health indicators and their relationships with maternal risk levels through four research questions, focusing on key factors influencing risk levels, risk distributions by age and blood pressure, interactions between heart rate and blood sugar, and correlations between health indicators. Using visualizations like heatmaps, scatter plots, and PCA graphs, complemented by statistical tests such as logistic regression and correlation analyses, the report provides well-supported and accurate conclusions. It identifies significant factors like systolic blood pressure, glucose, and heart rate, reveals age and blood pressure’s impact on risk levels, and highlights variations and correlations across different indicators. The findings are aligned with the research objectives, delivering meaningful insights into maternal health risks.

Questions for Future Work

While our project addressed several important research questions about maternal health risks, there are still areas left unexplored that could benefit from future investigation. Below, we outline key questions and the reasons they remain unanswered, along with their potential significance.

  1. How do socioeconomic factors (e.g., income, education, and access to healthcare) influence maternal health risk levels?

Why Future Work: Our project focused on physiological indicators, but we recognize that maternal health is deeply influenced by socioeconomic conditions. Incorporating these variables would require additional data collection and integration, which was beyond the scope of our current analysis.

Link to Completed Work: Our findings on the physiological differences across risk levels highlight the need to investigate external factors that might amplify or mitigate these risks.

  1. How does the combination of multiple physiological indicators predict maternal health risks?

Why Future Work: While we analyzed individual metrics like systolic blood pressure and heart rate, understanding their combined effects (interactions) would require more complex statistical or machine learning models. These tools were outside the techniques applied in this project.

Link to Completed Work: Observing significant trends for individual factors (e.g., differences in heart rate trends across risk levels) underscores the value of examining their interactions to improve predictive accuracy.