Home | Volume 10 | Article number 28

Research

Prediction of low birth weight at term in low resource setting of Gulu city, Uganda: a prospective cohort study

Prediction of low birth weight at term in low resource setting of Gulu city, Uganda: a prospective cohort study

Silvia Awor1,&, Rosemary Byanyima2, Benard Abola3, Paul Kiondo4, Christopher Garimoi-Orach6, Jasper Ogwal-Okeng5, Annettee Nakimuli4, Dan Kaye4

 

1Department of Obstetrics and Gynecology, Faculty of Medicine Gulu University, Gulu, Uganda, 2Department of Radiology, Mulago National referral hospital, and teaching hospital for Makerere University, Kampala, Uganda, 3Department of Mathematics, Faculty of Science, Gulu University, Gulu, Uganda, 4Department of Obstetrics and Gynaecology, School of medicine, Makerere University College of Health sciences, Kampala, Uganda, 5Department of pharmacology, Lira University, Lira, Uganda, 6Department of community health, School of Public health, Makerere University, Kampala, Uganda

 

 

&Corresponding author
Silvia Awor, Department of Obstetrics and Gynecology, Faculty of Medicine Gulu University, Gulu, Uganda

 

 

Abstract

Introduction: despite the widespread poverty in Northern Uganda resulting in undernutrition, not all mothers deliver low birth weight babies. Therefore we developed and validated the risk prediction models for low birth weight at term in Northern Uganda from a prospective cohort study.

 

Methods: one thousand mothers were recruited from 16 - 24 weeks of gestation and followed up until delivery. Six hundred eighty-seven mothers delivered at term. The others were either lost to follow-up or delivered preterm. Used proportions to compute incidence of low birth weight at term, build models for prediction of low birth weight at term in RStudio. Since there were few low birth weight at term, were generated synthetic data using ROSE-package in RStudio by over-sampling low birth weights and under-sampling normal birth weights, and evaluated the model performance against the synthetic data using K (10) - fold cross-validation.

 

Results: mean age was 26.3 years with an average parity of 1.5. Their mean body mass index was 24.7 and 7.1% (49 of 687) had lateral placenta. The incidence of low birth weight was 5.7% (39 of 687). Predictors of low birth weight were gravidity, level of education, serum alanine aminotransferase (ALT), serum gamma-glutamyl transferase (GGT), lymphocyte count, placental location, and end-diastolic notch in the uterine arteries. This predicted low birth weight at term by 81.9% area under the curve (AUC), 76.1% accuracy, 72.9% specificity, and 79.1% sensitivity.

 

Conclusion: a combination of gravidity, level of education, serum ALT, serum GGT, lymphocyte count, placental location, and end-diastolic notch in the uterine arteries can be used for screening for low birth weight in prenatal clinics for screening low birth weight at term.

 

 

Introduction    Down

Low birth weight at term (≥37 weeks of gestation) is diagnosed when the baby weighs less than 2.5Kg at birth [1,2]. It affects 5 - 10% of babies born in the global north [3], and slightly higher number in the global south [4-6]. The causes are multifactorial including prenatal undernutrition, maternal race, and low socioeconomic status [7-9]. There is a known black racial predisposition to low birth weight in multiracial communities [9]. However, inadequate maternal nutrition is the hallmark of sub-Saharan Africa [10]. In northern Uganda, over 60% of the population eats less than three meals a day [11]. This may worsen the pregnancy outcomes associated with nutritional problems [7].

Complications of low birth weight at term include low APGAR score at five minutes of birth, neonatal asphyxia, foetal distress, respiratory distress, neurodevelopmental deficits, impaired renal development, and neonatal death [12-15]. These may require advanced paediatric care available at tertiary level hospitals to manage neonates with low birth weight. Knowledge of the predictors of low birth weight may help prepare better to refer these mothers to tertiary health centers. Despite the widespread poverty and predominantly black population in Northern Uganda, not all mothers deliver low birth weight babies. Therefore we set out to develop and validate the risk prediction models for low birth weight in Northern Uganda.

 

 

Methods Up    Down

Study design and Study setting: this was a prospective cohort study at St. Mary´s Hospital Lacor. It is a private, not-for-profit hospital, founded by the Catholic Church. It is located six kilometers west of Gulu city along Juba Road in Gulu district (Longitude 30 - 32 degrees East and Latitude 02 - 04 degrees North). St. Mary´s Hospital Lacor is one of the teaching hospitals of Gulu University with a bed capacity of 482. It is staffed by specialists, medical officers, midwives, nurses, laboratory and radiology staff, as well as support and administrative staff. The hospital receives over nine thousand antenatal mothers and conducts about seven thousand deliveries per year [16]

Study population: we recruited 1,000 pregnant mothers 16 - 24 weeks from April 2019 to March 2020, gave them unique identifiers (study numbers). Excluded all participants with lethal congenital anomalies.

Sample size estimation: using Yamane 1967 formula [17] for calculating sample size for cohort studies using finite population size, St. Mary´s hospital Lacor delivers approximately seven thousand mothers per year. Since my study duration is 12 months for recruitment of the mothers, the finite population I can access is about 7,000 mothers. Yamane 1967 formula: Sample size:

Where N is the finite population size= 7,000 mothers; margin of error (e)= 5%; therefore, n = 7,000/1+7,000(0.05)2; n = 379. The required sample size was 379 mothers. We doubled the number (to >758) to cater for loss to follow up, preterm delivery and clients opting out of the study during the follow-up period. We expected over 50% loss to follow-up since rate of hospital delivery in this region is low [18], and some people may deliver from the free government hospitals nearby.

Data collection: a questionnaire was filled with the help of a midwife (research assistant), blood samples taken for complete blood count, liver and renal function tests, and uterine artery Doppler sonography done. The laboratory and sonographers only used the study numbers to identify the participants. The mothers were followed up until September 2020 when the last one was delivered. The delivery team had no access to the questionnaire, laboratory and ultrasound results.

Outcomes: birth weight of the baby <2.5Kg at ≥37 weeks was taken as low birth weight.

Definitions: from maternal history, grouping was based on known classification e.g. level of education in Uganda is grouped as primary, secondary and tertiary education levels. Gravidity was grouped as prime gravida for first pregnancies, multigravida for 2-4 and grand multigravida >4. Laboratory tests were grouped according to the reference ranges given by the laboratory while ultrasound pulsatility and resistive indices were grouped using percentiles. End diastolic notch of the uterine arteries was either present or absent; if present it was either unilateral or bilateral. The participants with incomplete results were dropped from the final analysis.

Statistical analysis: six hundred eighty-seven (687) mothers delivered at term. Data were pre-processed using Stata® 15.0 and built models in RStudio (R version 4.1.1 (2021-08-10)). Used proportions to compute incidence of low birth weight at term. Univariable analysis was done, and all variables with p-values ≤0.20 or were known risk factors for low birth weight were included in a logistic regression model. Built models for prediction of low birth weight at term in RStudio. The predictors with p-value <0.05 in the logistic regression model were taken as independent risk factors for low birth weight. Since there were few low birth weight at term, we used all the participants for the development of the models. For the validation cohort, we generated synthetic data using ROSE-package in RStudio by over-sampling low birth weights and under-sampling normal birth weights to balance the data. We obtained 349 (51.2%) and 332 (48.8%) normal and low birth weights respectively. We evaluated (validated) the model performance against the synthetic data using K (10) - fold cross-validation. The original model was put into a confusion matrix against the ROSE-derived data to calculate the accuracy, sensitivity, and specificity of the model in RStudio.

Ethical consideration: the study was approved by Makerere University School of Medicine Research and Ethics Committee (Reference number 2018-105), Uganda National Council for Science and Technology (Reference number HS258ES), and administrative clearance to conduct the research at St. Mary´s Hospital Lacor was also obtained (Reference number LHIREC Adm 009/11/18). The participants were informed about the study during the morning health education by the midwives when they arrived in the hospital. Those who satisfied the inclusion criteria were approached and requested to join the study. Written informed consent was sought from every participant in either English or Acholi language.

 

 

Results Up    Down

One thousand pregnant mothers were recruited. Six hundred eighty-seven (687) mothers delivered at term. Three hundred thirteen (313) mothers were lost to follow-up or delivered preterm, dropped and not used in the data analysis. The prevalence of birth weight < 2.5Kg at term was 5.7% (39 out of 687) (Figure 1).

General characteristics of the study population: mean maternal age was about 26 years, with majority being informally employed. Only one in five of the mothers had a tertiary level of education (Table 1). Average body mass index was 24.7Kg2 while prevalence of prenatal hypertension of 0.7%. About 7.1% had lateral placental location and 10.2% had bilateral end diastolic notch [19] (Table 2). Average maternal haemoglobin level was 10.8g/dL with a haematocrit of 32.5%. Mean serum alkaline phosphatase (ALP), GGT, aspartate aminotransferase (AST) and ALT were 154.3IU, 21.7IU, 20.2IU and 30.7IU respectively (Table 2). The mothers were followed up for an average of 19.1 weeks over one and half years (April 3rd 2019 to September 30th 2020)

Unadjusted estimates of the variables against low birth weight: all continuous variables were categorized using the laboratory reference ranges and interquartile ranges for the participants. Both ranges were analyzed and the best-fitted range for the models was chosen by the researcher. All variables with unadjusted p-value of ≤0.20 at univariable analysis were taken for multivariable level to build the models (Table 3, Table 4). A model was chosen based on one with fewer variables producing higher accuracy, sensitivity, specificity and AUC.

Risk prediction models for low birth weight: six risk prediction models were built from maternal history and physical examination, obstetric ultrasound parameters and uterine artery Doppler indices, maternal laboratory lab tests and the combinations of maternal history with either laboratory tests or ultrasound parameters or all the variables (Table 5, Table 6). In model 6, (combination of all the variables) (details in Table 6), the predictors of low birth weight were gravidity, level of education, serum ALT, serum GGT, lymphocyte count, placental location and end-diastolic notch in the uterine arteries. Being a prime gravida (aOR = 5.89, 95% CI 1.42 -41.94, p=0.032), having a laterally (one sided) located placenta (aOR = 3.42, 95% CI 1.23 - 9.45, p=0.018) and presence of end-diastolic notch (aOR = 2.59, 95% CI 1.07 - 6.28, p=0.035) were independent risk factors, while having a tertiary level of education (aOR = 0.16, 95% CI 0.04 - 0.69, p=0.013), normal lymphocyte count (aOR = 0.30, 95% CI 0.10 - 0.91, p=0.033) and serum ALT (aOR = 0.22, 95% CI 0.09 - 0.56, p=0.001) were protective against low birth weight.

Evaluation of the performance of the models 1-6 for the prediction of low birth weight: the models´ performance were evaluated using K (10) - fold cross validation against a synthetic data derived from the ROSE package in RStudio, and listed in (Table 2). Model accuracies ranged from 59.3% - 76.1%, while AUC from 62.6% - 81.9%. In the absence of ultrasound scan and laboratory tests, model 1 can predict low birth weight with 62.3% accuracy, 37.3% sensitivity, 88.3% specificity and 65.3% AUC.

 

 

Discussion Up    Down

We developed and validated risk prediction models for low birth weight at term in Northern Uganda from a prospective cohort study. From maternal history, the predictors of low birth weight were education level and gravidity. This predicted low birth weight at term by 65.3% AUC, 62.3% accuracy, 88.3% specificity, and 37.3% sensitivity. In Ethiopia, similar demographic characteristics were used to predict low birth weight. At a 26% false positive rate, they predicted low birth weight with 83% AUC with 82% specificity and 71% sensitivity [20]. While in India, Singh et al. [21] found the prediction model AUC of 79% with 72% sensitivity and 56% specificity. In the USA, maternal history predicted low birth weight with 75.3% accuracy [22]. Considering the uterine artery Doppler indices, the predictors of low birth weight were placental location and end-diastolic notch in the uterine arteries. This predicted low birth weight at term by 62.6% AUC, 59.3% accuracy, 42.5% specificity, and 75.4% sensitivity. In Denmark, uterine artery pulsatility index predicted low birth weight with 74% AUC [23], while in Saudi Arabia, placental thickness of <2cm and diameter of <18cm predicted low birth weight with 88.6% AUC [24]. This probably outline the differences in population and techniques used in the data analysis. When the maternal history is combined with uterine artery Doppler indices, the predictors of low birth weight were education level, gravidity, placental location, and end-diastolic notch. This predicted low birth weight at term by 71.6% AUC, 62.3% accuracy, 64.8% specificity and 61.8% sensitivity. In India, a combination of uterine artery Doppler indices and maternal history predicted low birth weight with 65.9%AUC, 45.4% sensitivity and 84.6% specificity [25].

While we found the predictors of low birth weight to be serum GGT, serum ALT and lymphocyte count to have predicted low birth weight at term by 66.9% AUC, 59.3% accuracy, 35.8% specificity and 81.7% sensitivity, there is limited data on the prediction of low birth weight using maternal full haemagram, liver and renal function tests. There is no evidence that maternal blood levels of alpha-feto protein (AFP), human chorionic gonadotropin (hCG), or pregnancy-associated plasma protein A (PAPP-A) used as a single predictor are useful to predict low-birth-weight newborns [26]. When the laboratory blood tests were combined with maternal history, the predictors of low birth weight were gravidity, level of education, serum ALT, serum GGT and lymphocyte count. This predicted low birth weight at term by 66.9% AUC, 66.7% accuracy, 59.6% specificity and 73.4% sensitivity. Addition of blood glucose levels to maternal history in Mexico predicted low birth weight with 72% AUC [27]. After combining all the variables from maternal history, laboratory tests, and uterine artery Doppler indices, the predictors of low birth weight were gravidity, level of education, serum ALT, serum GGT, lymphocyte count, placental location, and end-diastolic notch in the uterine arteries. These predicted low birth weight at term by 81.9% AUC, 76.1% accuracy, 72.9% specificity and 79.1% sensitivity. Considering the few predictors, this model can be used for screening low birth weights in prenatal clinics. This makes our model favorably compared to the other models. This is a baseline study in Northern Uganda. We hope it will open doors to a wide range of research in sub-Saharan Africa. There were many loss to follow-up or preterm birth. This could have skewed the models in other ways. Future research should be done in several other locations for external validation of these models to ensure generalizability.

 

 

Conclusion Up    Down

In places with no laboratory tests and ultrasound scan, the predictors of low birth weight from maternal history alone are level of education and number of pregnancies (gravidity). These predicted low birth weight by 65.3% AUC with 62.3% accuracy.

Funding: this study was funded as a PhD project by Makerere University - Sweden bilateral research agreement for junior staff development in academic institutions. The funders had no influence over the topic which a student chooses.

What is known about this topic

  • Black women are more at risk of low birth weight in multiracial communities;
  • Malnutrition is a predictor of low birth weight;
  • Low birth weight is known to be associated with increased risks of early neonatal death.

What this study adds

  • Being a prime gravida, and having laterally located placenta are risk factors for low birth weight while tertiary level of education is protective;
  • The prevalence of low birth weight in Gulu city, Northern Uganda is comparable to those in multiracial communities.

 

 

Competing interests Up    Down

The authors declare no competing interests.

 

 

Authors' contributions Up    Down

Silvia Awor, wrote the proposal, collected data and drafted the manuscript. Benard Abola built the models and cross-validated the models. Rosemary Byanyima, Paul Kiondo and Christopher Garimoi-Orach are doctoral committee members who provided expert opinion and guided to write the manuscript. Jaspar Ogwal-Okeng, Annettee Nakimuli and Dan Kabonge Kaye are doctoral supervisors who guided through the concept and writing of the manuscript. All the authors have read and agreed to the final manuscript.

 

 

Acknowledgments Up    Down

We thank Mr. Ronald Kivumbi and Mr. Ronald Waiswa, the biostatisticians, for helping me with part of the pre-processing of the data for analysis. The authors would like to acknowledge the support provided for writing this paper by the British Academy Writing Workshop Programme 2022, ’Addressing Epistemic Injustice: Supporting Writing about Inclusive and Life-long Education in Africa’.

 

 

Tables and figure Up    Down

Table 1: social demographic characteristics of mothers who delivered at term

Table 2: clinical and laboratory characteristics of mothers who delivered at term

Table 3: unadjusted risk ratio of social demographic characteristics and low birth weight

Table 4: unadjusted risk ratio of clinical and laboratory characteristics with low birth weight

Table 5: model 1 - 4 for the prediction of low birth weight

Table 6: model 5 - 6 for the prediction of low birth weight

Table 7: model performance evaluation using K-fold cross-validation

Figure 1: participant flow through the study

 

 

References Up    Down

  1. World Health Organization. International statistical classification of diseases and related health problems ICD10. 10th revision, 5th ed. World Health Organization; 2010. Accessed August 30, 2022.

  2. World Health Organization, United Nations Children's Fund. Low birthweight: country, regional and global estimates. Geneva: World Health Organization; 2004. Accessed August 30, 2022.

  3. DeMarco N, Twynstra J, Ospina MB, Darrington M, Whippey C, Seabrook JA. Prevalence of Low Birth Weight, Premature Birth, and Stillbirth Among Pregnant Adolescents in Canada: A Systematic Review and Meta-analysis. J Pediatr Adolesc Gynecol. 2021 Aug;34(4):530-537. PubMed | Google Scholar

  4. Shibre G, Tamire M. Prevalence of and socioeconomic gradient in low birth weight in Ethiopia: further analysis of the 2016 demographic and health survey data. BMC Pregnancy Childbirth. 2020 Oct 8;20(1):608. PubMed | Google Scholar

  5. Alemu A, Abageda M, Assefa B, Melaku G. Low birth weight: prevalence and associated factors among newborns at hospitals in Kambata-Tembaro zone, southern Ethiopia 2018. Pan Afr Med J. 2019 Oct 2;34:68. PubMed | Google Scholar

  6. Louis B, Steven B, Margret N, Ronald N, Emmanuel L, Tadeo N et al. Prevalence and Factors Associated with Low Birth Weight among Teenage Mothers in New Mulago Hospital: A Cross Sectional Study. J Health Sci (El Monte). 2016 Jul-Aug;4:192-199. PubMed | Google Scholar

  7. Kogan MD. Social causes of low birth weight. J R Soc Med. 1995 Nov;88(11):611-5. PubMed | Google Scholar

  8. Negrato CA, Gomes MB. Low birth weight: causes and consequences. Diabetol Metab Syndr. 2013 Sep 2;5:49. PubMed | Google Scholar

  9. Kleinman JC, Kessel SS. Racial Differences in Low Birth Weight. New England Journal of Medicine. 1987;317(12):749-53. PubMed | Google Scholar

  10. Ramakrishnan U. Nutrition and low birth weight: from research to practice. Am J Clin Nutr. 2004 Jan;79(1):17-21. PubMed | Google Scholar

  11. Statistics UBOS. Statistical abstract. In: statistics Bo, editor: Government of Uganda; 2020. p. 339.

  12. Nunes A, Melo F, Silva JE, Costa A, Bispo MA, Palminha JM. [Importance of J. Brazy's neurobiological index. Prediction of the number and severity of complications in very low birth weight infants]. Acta Med Port. 1998 Jul;11(7):615-21. PubMed

  13. Njim T, Atashili J, Mbu R, Choukem SP. Low birth weight in a sub-urban area of Cameroon: an analysis of the clinical cut-off, incidence, predictors and complications. BMC Pregnancy Childbirth. 2015 Nov 4;15:288. PubMed | Google Scholar

  14. Kwinta P, Klimek M, Drozdz D, Grudzien A, Jagla M, Zasada M et al. Assessment of long-term renal complications in extremely low birth weight children. Pediatr Nephrol. 2011 Jul;26(7):1095-103. PubMed | Google Scholar

  15. Chen SD, Lin YC, Lu CL, Chen SC. Changes in outcome and complication rates of very-low-birth-weight infants in one tertiary center in southern Taiwan between 2003 and 2010. Pediatr Neonatol. 2014 Aug;55(4):291-6. PubMed | Google Scholar

  16. St. Mary´s Hospital Lacor. St. Mary´s Hospital Lacor Annual Report Financial Year July 2015 - June 2016. Uganda: St. Mary's Hospital Lacor; 2020.

  17. Kasiulevičius V, Šapoka V, Filipavičiūtė R. Sample size calculation in epidemiological studies. Gerontologija. 2006;7(4):225-31. Google Scholar

  18. Ediau M, Wanyenze RK, Machingaidze S, Otim G, Olwedo A, Iriso R et al. Trends in antenatal care attendance and health facility delivery following community and health facility systems strengthening interventions in Northern Uganda. BMC Pregnancy Childbirth. 2013 Oct 18;13:189. PubMed | Google Scholar

  19. Awor S, Byanyima R, Abola B, Nakimuli A, Orach CG, Kiondo Pet al. Incidence of preeclampsia and retention to prenatal care in Northern Uganda. East Afr Med J. 2022 Aug 23;99(6):4885-96. PubMed | Google Scholar

  20. Hassen HY, Gebreyesus SH, Endris BS, Roro MA, Van Geertruyden JP. Development and Validation of a Risk Score to Predict Low Birthweight Using Characteristics of the Mother: Analysis from BUNMAP Cohort in Ethiopia. J Clin Med. 2020 May 23;9(5):1587. PubMed | Google Scholar

  21. Singh A, Arya S, Chellani H, Aggarwal KC, Pandey RM. Prediction model for low birth weight and its validation. Indian J Pediatr. 2014 Jan;81(1):24-8. PubMed | Google Scholar

  22. Gaziano EP, Freeman DW, Allen TE. Antenatal prediction of women at increased risk for infants with low birth weight. A discriminant analysis of maternal characteristics known at 24 weeks. Am J Obstet Gynecol. 1981 May 1;140(1):99-107. PubMed | Google Scholar

  23. Sinding M, Peters DA, Frøkjær JB, Christiansen OB, Petersen A, Uldbjerg N et al. Prediction of low birth weight: Comparison of placental T2* estimated by MRI and uterine artery pulsatility index. Placenta. 2017 Jan;49:48-54. PubMed | Google Scholar

  24. Habib FA. Prediction of low birth weight infants from ultrasound measurement of placental diameter and placental thickness. Ann Saudi Med. 2002 Sep-Nov;22(5-6):312-4. PubMed | Google Scholar

  25. Verma D, Gupta S. Prediction of adverse pregnancy outcomes using uterine artery Doppler imaging at 22-24 weeks of pregnancy: A North Indian experience. Turk J Obstet Gynecol. 2016 Jun;13(2):80-84. PubMed | Google Scholar

  26. Goto E. Association of high maternal blood alpha-fetoprotein level with risk of delivering small for gestational age: a meta-analysis. Pediatr Res. 2021 May;89(7):1742-1750. PubMed | Google Scholar

  27. Hernández-Castro F, Berlanga-Garza A, Cruz-Gutiérrez MD, Soria-López JA, Villagómez-Martínez GE, Dávila-Escamilla IV. Predicción de bajo peso al nacer con hipoglucemia en la prueba de tolerancia a la glucosa. Revista de Saúde Pública. 2021 May 17;55. Google Scholar