Modeling Regional Determinants of Early Motherhood in Colombia and its impact during COVID-19 pandemic

doi:10.21203/rs.3.rs-5327481/v1

Download PDF

Research Article

Modeling Regional Determinants of Early Motherhood in Colombia and its impact during COVID-19 pandemic

https://doi.org/10.21203/rs.3.rs-5327481/v1

This work is licensed under a CC BY 4.0 License

Version 1

posted

You are reading this latest preprint version

Introduction: Early motherhood, defined as occurring in women under 19 years of age, poses significant social challenges with immediate and long-term consequences. This study focuses on early fertility rates (early motherhood) in Colombia and examines their relationship with social and demographic factors.

Methods: Departmental-level data from Colombia's Social Protection Information System (RUAF) and the National Administrative Department of Statistics (DANE) were cleaned and aggregated from 2018-2023. This allowed the estimation of the overall fertility rate, infant fertility rate, and adolescent fertility rate by department and year. A model was constructed to estimate the influence of recognized risk factors such as low educational attainment (% of undereducated population), poverty level (% of population under subsidized health schemes, current GDP), employment rates (% of workforce participation), violence against women (% of female deaths, both general and violent), and infant mortality. The model was refined to improve performance by removing variables with low contributions to early motherhood.

Results: Fertility indicators focusing on early fertility in Colombia were described. Regions that should be prioritized for early fertility prevention strategies were identified because of their increase in the early pregnancy rate during the pandemic. Significant risk factors included the percentage of the population under subsidized health schemes, GDP per capita, percentage of violent deaths, percentage of female deaths, and percentage of deaths among those <18 years of age.

Conclusions: We identified regions that should be prioritized for early fertility prevention strategies, highlighting areas particularly vulnerable to increased early fertility rates during the pandemic. These findings reveal the effects of relaxed public health measures alongside the social and demographic factors needed for targeted interventions. Further research is necessary to quantify the impact of specific interventions on early fertility rates.

Biostatistics

Population Biology

Public Administration

early motherhood

multivariate analysis

public policy

fertility rate

public health

healthcare system

generalized linear regression

Early motherhood (10--19 years old) imposes risks associated with obstetric and mental health complications and social risks due to increased school dropout rates and, consequently, limited access to future employment opportunities, contributing to intergenerational cycles of poverty and the deterioration of human capital in subsequent generations. (1–6) Factors such as educational level, accessibility to health care, violence, family planning policies, and cultural values can influence this measure. (7,8)

Public policies for the prevention of sexual and reproductive health events in childhood and adolescence have focused their efforts on sex education and the promotion of healthy lifestyles, as well as on the restitution of rights in cases of sexual abuse and commercial exploitation, without a clear recognition of the impact of these interventions. (9–11)

During 2021, increases in infant fertility (10-- to 14-year-old females) and adolescent fertility rates (15-- to 19-year-old females) were reported in some regions in Colombia, revealing regional factor inequality during 2020, which demands the formulation of high-quality immediate public policy interventions. (12)

This study analyzes the early motherhood landscape on the basis of public data from Colombia and provides recommendations for policymakers on the basis of a statistical model.

Data sources:

The data were obtained from Colombia's Social Protection Information System and integrated via Power Query. All aggregation and harmonization were performed at the departmental level from 2018--2023 (see Table 1):

Birth data: The birth information on births was extracted from the vital statistics accessible at the Single Registry of Affiliates (RUAF), including maternal age and year of registry.

Mortality data: Information was extracted from the RUAF on deaths due to cause of mortality, age, sex, year of registry, and department of occurrence. A subset of data covering violent deaths, such as transport accidents, falls, firearm injuries, drowning, suffocation, electric shock, exposure to smoke or fire, poisoning, self-inflicted injuries, crimes, homicides, and other external causes (WHO 667), was extracted.

Insurance data: The Unified Database of Affiliates (BDUA) contains data on the number of affiliates with public insurance (with more than 97% of the Colombian people covered) and is classified by affiliation regime, age, sex, and department of residence.

Economic and educational data: Microdata from the National Administrative Department of Statistics (DANE) were used to obtain the departmental gross domestic product (GDP) per capita at current prices (mean yearly income by person), the number of students enrolled by department, the employment rate (occupation rate: the ratio between the labor force and the working-age population) and demographic information on the indigenous population by department.

Data Cleansing and Preparation

Once extracted and integrated, the data were standardized at the departmental and annual levels in Python Colab, and the integrated dataset and code used for the model are accessible as supplementary material. (13,14)

Since the data used in the study were obtained from anonymous databases, this study did not require the approval of an ethics committee for its execution, as the data used did not involve the direct participation of human subjects.

Descriptive analysis

The indicators used were as follows:

Total Fertility Rate (TFR): This rate was determined by aggregating birth rates across various age groups and multiplying by 5 to represent a standard five-year interval

['TFR] = (

(['Maternity 15 to 19 years old']/['Affiliates 15 to 19 years old']) +

(['Maternity 20 to 24 years old']/['Affiliates 20 to 24 years old']) +

(['Maternity 25 to 29 years old']/['Affiliates 25 to 29 years old']) +

(['Maternity 30 to 34 years old']/['Affiliates 30 to 34 years old']) +

(['Maternity 35 to 39 years old']/['Affiliates 35 to 39 years old']) +

(['Maternity 40 to 44 years old']/['Affiliates 40 to 44 years old']) +

(['Maternity 45--49 years old']/['Affiliates 45--49 years old])

) * 5

Infant Fertility Rate (IFR) (10--14 years): calculated as the number of births classified as infant births (to mothers aged 10--14) divided by the total number of females aged 10--14, again multiplied by 1,000.

[IFR] =

(['Infant maternity']/['Members 10--14 years old']) * 1000

Adolescent Fertility Rate (AFR) (15--19 years): This rate is derived from the number of births to mothers aged 15--19 divided by the total number of females in that age group, multiplied by 1,000.

['AFR'] =

(['Maternity 15 to 19 years']/['Affiliates 15 to 19 years']) * 1000

Early Fertility Rate (EFR) (10--19 years): estimated as the sum of births to mothers aged 10--14 and 15--19 years, divided by the total number of females aged 10--14 and 15--19 years, and then multiplied by 1,000.

['EFR'] = (

(['Infant motherhood'] + ['Motherhood 15 to 19 years old'])/

(['Members 10--14 years old'] + ['Members 15--19 years old']) * 1000

)

The results are presented in heatmaps by department and year, as well as average aggregates for the total period (2018--2023).

Data Modeling

We computed social-demographic determinants at the year and department levels:

Percentage of the population enrolled in school relative to the total number of school-aged individuals
Percentage of violent deaths out of total deaths
The female mortality metrics included the following:
1. Percentage of female deaths out of total deaths
2. Percentage of violent female deaths out of the total number of female deaths
Percentage of infant deaths (<18 years old deaths) out of total deaths
% of the labor force participation rate out of the total number of people of labor age
% percentage of the subsidized population out of total insured people
GDP per capita
Percentage of the indigenous population.

We subsequently constructed a new data frame incorporating these calculated variables with departmental and annual identifiers. To ensure data integrity, we employed a Z score method for outlier detection, identifying and removing observations with Z scores exceeding a threshold of 3. The resulting cleaned dataset provides a robust basis for further analysis, encapsulated in the final data frame, which includes only the relevant demographic rates and excludes outlier observations. This rigorous data processing approach enhances the reliability of subsequent analyses to understand the intricate relationships between demographic factors and health outcomes.

The aggregated and purified data are presented in scatter plots and trend lines showing the correlation between the early fertility rate (EFR) and its potential explanatory variables.

Generalized linear model (GLM) analysis

A generalized linear model (GLM) with a Gaussian family and an identity link function was used to analyze the relationships between the EFR and the selected variables.

The initial GLM included all relevant variables; second, to improve the performance of the model, all variables with high p values (indicating low statistical significance) were iteratively eliminated.

Cross-validation was performed via Sklearn's linear regression algorithm to evaluate the model's performance in predicting the dependent variable. A cross-validation scheme with 5 partitions (KFold) was used, randomly dividing each partition to minimize bias and using a fixed seed to guarantee the reproducibility of the results.

As a measure of the model predictive capacity, the value of the R² coefficient was determined, and the difference between the actual and predicted values was analyzed.

Between 2018 and 2023, the general fertility rate (GFR) decreased from 1.86 to 1.40 births per woman (see Figure 1).

Between 2018 and 2023, the infant fertility rate (IFR) decreased from 3.48 to 2.81 births per 1,000 women between the ages of 10 and 14 (see Figure 2).

Between 2018 and 2023, the AFR (adolescent fertility rate) decreased from 68.66 to 45.99 births per 1,000 women between the ages of 15 and 19 (see Figure 3).

All indicators (GFR, IFR, and AFR) peaked in 2021 for the departments of Vichada, Guainía, La Guajira, Vaupés, Magdalena, Cesar, and Chocó, with improvements below the initial levels in 2023.

The heatmap for the period of 2018-2023 shows that the departments of Vichada, Guainía, La Guajira, Vaupés, and Magdalena are the five departments with the highest rates of early fertility (see Figure 4).

When the scatter plots of the variables potentially associated with early fertility are examined, a positive correlation is observed for the variables of % of the subsidized population, % Indigenous population, % violent deaths and subgroup violent deaths in women, and % infant deaths; there seems to be a negative correlation for GDP per capita (the higher the income, the lower the early fertility), % Female deaths of the total deaths (the higher the female mortality, the lower the early fertility); there is no correlation with the variables % enrolled population (schooled) and labor active population (% occupation) (see Figure 5).

The first equation for the GLM confirms the low significance (p<0.05) of the variables % study population (p=0.682), % employment (p=0.110), and % indigenous population (p=0.073) (see Figure 6).

To improve the model's overall performance, the variables of low significance were eliminated (see Figure 7). Cross-validation via Sklearn's linear regression algorithm demonstrated an average R² value of 0.5188, suggesting that the model has reasonable performance, but considerable variation in the dependent variable that is not explained by the independent variables included in the model remains.

The visualization of the residuals in the scatter plot reveals a good distribution of values around the red line (values without error, with residuals equal to zero). The histogram of residuals showed a different distribution than normal (with asymmetry skewed to the right) (see Figure 8). The above, as expected, occurs in the context of a complex problem (such as early fertility).

The present study developed an exploratory exercise and an explanatory analysis of the early fertility rate in Colombia.

During the descriptive stage, a progressive decrease in general fertility and age-specific fertility rates was observed for childhood (10--14 years) and adolescence (15--19 years), which is similar to what was reported by the National Department of Statistics (DANE) in 2024 (15). The discrepancies in the estimates between our findings and those reported by the DANE underscore the complexity of estimating fertility rates among young adolescents; both reports (the present study and the DANE) are used as sources of births for numerators (vital statistics), but a different source for denominators, where the DANE used population projections from the 2018 census(16). Our analysis used the BDUA database, which provides real-time data on people with health coverage in Colombia. (17)

The descriptive findings highlight the top five vulnerable departments for early fertility; these departments show special fragility during the pandemic (births in 2021 correspond to pregnancies in 2020), demonstrating how the relaxation of measures to promote and prevent early pregnancy during the pandemic has affected early fertility rates (18). Measures of social isolation and reduced mobility may have resulted in reduced access to reproductive health services, including contraceptives and abortion, which could have increased the rate of sexual abuse and unplanned pregnancies. (19)

This finding aligns with the UN's report on disruptions to family planning services in less developed regions, including parts of Latin America. For example, the 2014--2016 Ebola outbreak led to a "post-Ebola baby boom" due to increased unintended pregnancies in Liberia. Similarly, during the COVID-19 pandemic, it was estimated that 12 million women across 115 low- and middle-income countries experienced contraceptive interruptions, leading to 1.4 million unintended pregnancies in 2020. International and national efforts have helped mitigate these disruptions, ensuring continued access to essential reproductive health services. (20)

The departments identified in the top five are recognized disparities in the distribution of resources, inequality, poverty, and corruption, with limited access to sexual and reproductive education and the greatest tolerance for sexual violence and child abuse, as highlighted in other reports. (12,21,22)

In addition, school closures and decreased economic activity may have increased the time children and adolescents spend at home, which has led to increased exposure to situations of sexual abuse and domestic violence, which are important risk factors for childhood and adolescent pregnancies. The lack of demand (sociocultural factor) or encouragement for postpartum contraception in childhood and adolescence has been reported in other studies as a cause of intergenerational cycles of early pregnancy. (23)

The integration of data and the use of generalized linear models (GLMs) for explanatory purposes have been widely reported in the scientific literature for their ability to adjust to situations where the outcome variables have a different distribution than normal. (24)

The results obtained in this study allowed the identification of deterministic variables of the early fertility rate event, providing inputs for prioritizing public health interventions to reduce this phenomenon.

As a limitation of GLM, we highlight the inherent limitations of the explanatory power of cross-sectional data, given that correlation results do not necessarily indicate a causal relationship between the variables.(8,20,25–27)

The use of more robust models (e.g., system dynamics) is needed, as is the inclusion of new variables, to propose explanatory and prescriptive hypotheses that allow the magnitude of the effect on the intervention of the factors described to be projected.(24,25,28,29)

As recommended by the results presented, public policies should prioritize interventions focused on education, access to health, access to modern contraceptives, and violence prevention, particularly in regions with more subsidized populations, to mitigate these disparities and improve overall reproductive health outcomes. (30–33)

This study provides a detailed analysis of fertility indicators in Colombia, with a focus on early fertility. The findings identified some regions that faced significant challenges in controlling early fertility rates during the pandemic, exacerbating existing social and demographic vulnerabilities, such as poverty, access to education, and healthcare availability.

The pandemic's impact on these regions highlights the urgency of bolstering resilience in health and education systems to mitigate the long-term consequences of early pregnancies.

Further research is essential to estimate the specific impact of various interventions. This additional analysis could help refine public policies and improve the effectiveness of programs for reducing early fertility and supporting the well-being of young women across Colombia.

Ethics approval and consent to participate: Not applicable; since the data used in the study were obtained from anonymized databases, this study did not require the approval of an ethics committee for its execution, as the data used did not involve the direct participation of human subjects.

Consent for publication: All authors give their permission to publish this manuscript.

Availability of data and material: The supporting data and Python code information are available and can be downloaded as supplementary material.

Competing interests: The authors declare that they have no conflicts of interest.

Funding: This research received no external funding.

Authors' contributions: F.D. and F.C.; methodology, F.D.; software, F.D.; validation, F.D. and F.C.; formal analysis, F.D.; investigation, F.D.; resources, F.D.; data curation, F.D.; writing—original draft preparation, F.D.; writing—review and editing, F.D. and F.C.; visualization, F.D.; supervision, F.C. All authors have read and agreed to the published version of the manuscript.

Acknowledgments: Not applicable.

UNFPA. Motherhood in Childhood: The Untold Story [Internet]. 2022. Available from: https://www.unfpa.org/featured-publication/motherhood-childhood-untold-story?cm_ven=ExactTarget&cm_pla=All+Subscribers&cm_ite=recently+released+report&cm_ainfo=&&&&&%25
Jaramillo-Mejía MC, Chernichovsky D. Early adolescent childbearing in Colombia: Time-trends and consequences. Cad Saude Publica. 2019;35(2).
Karra M, Lee M. Human capital consequences of teenage childbearing in South Africa. Poppov Res Netw. 2012;(March).
Fraser B. Adolescent pregnancy in Latin America and the Caribbean. Lancet Child Adolesc Heal. 2020;4(5):356–7.
Ribas CR. Adolescent pregnancy, public policies, and targeted programs in Latin America and the Caribbean: A systematic review. Vol. 45, Revista Panamericana de Salud Publica/Pan American Journal of Public Health. 2021.
Abdool S, Ardon E, Aspilcueta D, Brumana L. Accelerating progress toward the reduction of adolescent pregnancy in Latin America and the Caribbean. World Health Organization. 2016. 56 p.
Demography - Fertility rates - OECD Data [Internet]. [cited 2023 Jun 5]. Available from: https://data.oecd.org/pop/fertility-rates.htm
Sagalova V, Nanama S, Zagre NM, Volmer S. Long-term consequences of early marriage and maternity in West and Central Africa: Wealth, education, and fertility. J Glob Heal [Internet]. 2021;11(13004). Available from: https://pmc.ncbi.nlm.nih.gov/articles/PMC8397277/
Glenton C, Colvin C, Carlsen B, Swartz A, Lewin S, Noyes J, et al. Barriers and facilitators to the implementation of lay health worker programmes to improve access to maternal and child health: Qualitative evidence synthesis. Cochrane Database Syst Rev. 2013 Feb 28;2013(2).
Whitaker R, Hendry M, Booth A, Carter B, Charles J, Craine N, et al. Intervention Now To Eliminate Repeat Unintended Pregnancy in Teenagers (INTERUPT): a systematic review of intervention effectiveness and cost-effectiveness, qualitative and realist synthesis of implementation factors and user engagement. BMJ Open [Internet]. 2014 Apr 1 [cited 2022 Jun 8];4(4):e004733. Available from: https://bmjopen.bmj.com/content/4/4/e004733
Soto-Méndez C, Carrillo-Cruz YA. Anticoncepción en menores de edad: retos para el derecho desde la bioética. Rev Lasallista Investig. 2020;17(1):28–43.
Laboratorio de Economía de la Educación. Embarazo infantil y adolescente en Colombia [Internet]. INFORME ANÁLISIS ESTADISTICO LEE (Laboratorio de la Economía de la Educación). 2022. Available from: https://lee.javeriana.edu.co/documents/5581483/5629089/INF-65-Embarazo-infantil-y-adolescente-en-Colombia-2022.pdf/6f02b26c-4b4b-9762-c919-4a366d39977e?t=1667584449431%0A
Carneiro T, Da Nobrega RVM, Nepomuceno T, Bian G Bin, De Albuquerque VHC, Filho PPR. Performance Analysis of Google Colaboratory as a Tool for Accelerating Deep Learning Applications. IEEE Access. 2018;6:61677–85.
Zhou XH, Eckert GJ, Tierney WM. Multiple imputation in public health research. Stat Med [Internet]. 2001 May 15 [cited 2023 May 17];20(9–10):1541–9. Available from: https://onlinelibrary.wiley.com/doi/full/10.1002/sim.689
Estadísticas Vitales [Internet]. [cited 2013 May 2]. Available from: http://www.dane.gov.co/files/investigaciones/fichas/Estadisticas_vitales.pdf
Proyecciones de población [Internet]. [cited 2013 Jun 4]. Available from: http://www.dane.gov.co/index.php?option=com_content&view=article&id=75&Itemid=72
PAHO/WHO. Colombia - Country Profile. Health in the Americas. Prospects/Measures to achieve universal health coverage. 2022; Available from: https://hia.paho.org/en/countries-2022/colombia-country-profile
Justice for Colombia. Coronavirus: what is the impact in Colombia? Justice for Colombia [Internet]. 2021 [cited 2024 Oct 17]; Available from: https://justiceforcolombia.org/news/coronavirus-what-is-the-impact-in-colombia/
UNICEF. Key Messages on Gender Based Violence , Teenage Pregnancy and Child Protection in the context of COVID-19. United Nations Children’s Fund. 2020;(June). Available from: https://covidlawlab.org/wp-content/uploads/2020/12/Gender-Based-Violence-Pregnancy-and-Child-Protection-during-COVID.pdf
United Nations Department of Economic and Social Affairs-Population division. World Population Policies 2021: Policies related to fertility [Internet]. 2021. Available from: https://www.un.org/development/desa/pd/sites/www.un.org.development.desa.pd/files/undesa_pd_2021_wpp-fertility_policies.pdf%0A
Heaton TB, Forste R. Education as Policy: The Impact of Education on Marriage, Contraception, and Fertility in Colombia, Peru, and Bolivia. Soc Biol. 1998;45(3–4):194–213.
Sociedad Colombiana de pediatria. Embarazo en Adolescentes Problema de Salud Pública Creciente en Colombia - Sociedad Colombiana de Pediatría | SCP. Soc Colomb Pediatría [Internet]. 2013 [cited 2022 Jun 7];(84):1–8. Available from: https://scp.com.co/editorial/embarazo-en-adolescentes-problema-de-salud-publica-creciente-en-colombia/
Zarocostas J. UNFPA warns of threats to sexual and reproductive health. Lancet (London, England). 2024;403(10436):1528.
Ahmed S, Albalawi O, Shabbir J. A novel approach for estimating fertility rates in finite populations using count regression models. Sci Rep [Internet]. 2024 Jan 22 [cited 2024 Oct 15];14(1):1–24. Available from: https://www.nature.com/articles/s41598-024-51734-z
Ali GG, El-Adaway IH, Dagli CH. A system dynamics approach for study of population growth and the residential housing market in the US. In: Procedia Computer Science. 2020. p. 154–60.
Gurmu E, Mace R. Fertility decline driven by poverty: The case of Addis Ababa, Ethiopia. J Biosoc Sci. 2008 May;40(3):339–58.
Caldwell JC, Caldwell P. Africa: The new family planning frontier. Stud Fam Plann. 2002;33(1):76–86.
Ibeji JU, Zewotir T, North D, Amusa L. Modeling fertility levels in Nigeria using Generalized Poisson regression-based approach. Sci African. 2020 Sep 1;9:e00494.
Kazembe LN. Modeling individual fertility levels in Malawian women: A spatial semiparametric regression model. Stat Methods Appl. 2009 Jul;18(2):237–55.
Upadhya KK, Trent ME, Ellen JM. Impact of individual values on adherence to emergency contraception practice guidelines among pediatric residents: implications for training. Arch Pediatr Adolesc Med [Internet]. 2009 Oct [cited 2013 Jul 25];163(10):944–8. Available from: http://www.ncbi.nlm.nih.gov/pubmed/19805714
Brown KE, Hurst KM, Arden MA. Improving adolescent contraceptive use: evaluation of a theory-driven classroom-based intervention. Psychol Health Med [Internet]. 2011 Mar [cited 2013 Jul 25];16(2):141–55. Available from: http://www.ncbi.nlm.nih.gov/pubmed/21328143
Peach E, Morgan C, Scoullar MJL, Fowkes FJI, Kennedy E, Melepia P, et al. Risk factors and knowledge associated with high unintended pregnancy rates and low family planning use among pregnant women in Papua New Guinea. Sci Rep [Internet]. 2021;11(1):1222. Available from: https://pubmed.ncbi.nlm.nih.gov/33441569/
Office of Population Affairs’ Teen Pregnancy Prevention (TPP). About the Teen Pregnancy Prevention Program. 2023.

The authors declare no competing interests.

SupplementExcelData03092024.xlsx
Exceldatafile
SupplementaryMaterialHTML.html
pythoncode

Download PDF

Version 1

posted

You are reading this latest preprint version

Modeling Regional Determinants of Early Motherhood in Colombia and its impact during COVID-19 pandemic

Status:

Version 1

Abstract

Figures

Introduction

Materials and Methods

Data sources:

Data Cleansing and Preparation

Data Modeling

Generalized linear model (GLM) analysis

Results

Discussion

Conclusions

Declarations

References

Additional Declarations

Supplementary Files

Status:

Version 1