Prevalence and determinants of anemia among school going children in the state of Tamil Nadu, India: applications of two-level logistic regression model

INTRODUCTION

Anemia remains as one of India’s major public health problems. The prevalence of anemia is higher in India compared to other developed countries. As per the studies conducted by world health organization in 2007, about 30-50% of anemia in children and other groups are caused by iron deficiency.¹ The deficiency of iron is the most common cause of anemia worldwide.² Interestingly, the distribution of this disease is not uniform throughout the world. In some regions, the prevalence of anemia among young children is greater than 50% and even approaches 100% in some localities.³ In India, one of the critical causes of childhood anemia is the deficiency of iron, linked to low nutritional iron consumption.⁴ Lack of vitamin B12 and A, infections with malaria parasite and hookworm infections are some other factors that affect childhood anemia.^5,6 In rural India the level of haemoglobin is connected with the status of iron in children.⁷

The DLHS data provide both national and state level data that gives an opportunity to find out the severity and distribution of anemia and helps to assess the trends and socio-economic determinants of the prevalence of anemia.⁸ Therefore, the present study was conducted to explore the prevalence of anemia among school going children by considering the effects of various socio economic and demographic variables taking into account individual and household and contextual level characteristics in the state of Tamil Nadu. The results of the analysis may provide awareness about the necessary steps to be taken to ensure the health and well-being of school going children.⁸

METHODS

Setting and study design

Present study is a cross sectional study and is based on DLHS 4 pertaining to the year 2012-2013 in the state of Tamil Nadu. DLHS is a household survey conducted at the district level by international institute for population sciences (IIPS). DLHS 4 adopted a multi-stage, stratified, probability proportional to size sample with replacement.⁹ More detailed information about sampling procedure can be obtained from the DLHS-4 report. Among the entire participants, those participants who are in between the age group (6-17 years) and currently attending school are considered. In the survey data, current educational status and age of the study participants is mentioned. For all the selected households hemoglobin level was measured for all individuals above 6 months of age. For subjects whose age is in between 6 months and 18 years consents were taken from their parent or guardian. The present study used the HemoCue method, which is considered to be a standard method for hemoglobin measurement by the world health organisation for field studies.⁸ We studied a total of 23,997 children whose hemoglobin level values were available.

Ethical committee approval

The institutional human ethics committee of PSG institute of medical sciences and research, Coimbatore, gave approval for the study.

Outcome variable

The outcome variable considered in this study was anemia. The severity of anemia was assessed on the basis of haemoglobin levels of study subjects. Haemoglobin levels greater than 11 g/dL was considered as normal level. Hemoglobin levels of 10-10.99 g/dL was considered as mild, 7-9.99 g/dL considered as moderate and less than 7 g/dL was considered as severe anemia.⁸

Independent variables

The individual level factors considered in the study was age, sex and body mass index (BMI). Age in years was categorized as 6-10 and 11-17. Sex of the individuals was classified as male and female. BMI of the individuals was classified as below 18.5, 18.5-24.9, 25-29.9 and 30 and above. Household and contextual level factors considered in the study were family members, household ownership, sanitation facilities, drinking water source and type of locality. Standard of living index (SLI) was measured through wealth index and was classified as low (0-14), medium (15-24) and high (25-67) based on the total score.¹⁰ Family members was classified as less than 4 members, 5-7 members, 8 or more members. Household ownership was classified as owned, rented and others. Sanitation facilities and drinking water source were classified as unimproved and Improved. Type of locality was classified as rural and urban.

Statistical analysis

The relationship between chosen covariates and anemia was analysed using both bivariate and multivariate methods. Weighted prevalence of anemia was initially estimated.^11,12 In bivariate analysis, chi-square test and Z test were used to find out the association between anemia and various factors. In the multivariate analysis, we applied a two-level logistic regression model to examine the risk factors of anemia by reducing the district level effect that exists on the data set. Intra-class correlation coefficient (ICC) was calculated prior to the application of the two-level regression model. The ICC was calculated using the formula,

var(V_oj) is variance of random intercept in the two level empty model. Multilevel regression model is applicable if the variance of the random intercept is statistically significant and the value of ICC is greater than zero.¹³

For this, two phases of the model were fitted. Model-1 was the empty model and it contained only district specific random effects in order to model between district variations in anemia.

The functional form of the empty model is,

where logit (π_i,j) = P(Y_ij= 1), α₀ is the (only) fixed effect in the model and V_0j is the level 2 residual.

Model-2 was the random intercept model in which individual level factors, household and contextual level factors and intra level interactions were considered. The mathematical form of the model is,

where, x_ij denotes the observation-related characteristics and X_j denotes cluster-related characteristics. Similarly, β₁₀ is the fixed slope that denotes the average effect of observation related characteristics in the overall sample and β₀₁ is the necessarily fixed slope which is the average effect of cluster level characteristics in the overall sample.¹⁴ In addition, we have also looked into the random coefficient model and appropriateness of these two models were tested using Akaike information criteria (AIC) and Bayesian information criteria (BIC), p<0.05 was considered as statistically significant. The whole analysis was performed using SPSS (version 24).

RESULTS

Table 1 shows the weighted percentage distribution of school going children classified as having anemia by its degree. The overall prevalence of anemia was found to be 41.4% (95% CI: 40.77-42.02%). Among the entire study participants 20.9% had mild anemia, 19% had moderate anemia and severe anemia was reported in 1.7% of participants. The proportion of anemic children was larger (45.8%) in the age group 6-10 years than in the age group 11-17 years (38.7%). The prevalence of anemia was higher in female children than in male children. The prevalence of anemia was higher in those children whose BMI was 30 kg/m² and above and whose BMI was below 18.5 kg/m² compared to whose BMI 18.5 kg/m² to 24.9 kg/m². The prevalence of anemia was higher when the number of members in family of study participants were 8 or more. The prevalence of anemia was higher for those participants whose sanitation facilities were still unimproved. There was more prevalence of anemia in the rural areas than in the urban areas and the prevalence was higher among participants whose standard of living was low.

We observed a significant (p<0.001) association between districts and the occurrence of anemia (Table 2). The observed district level variance was 0.1, p<0.001 and the intraclass correlation coefficient was 0.0295. This indicates that 2.95% of variations in the prevalence of anemia can be explained by district level factors (Table 3).¹⁵ The prevalence of anemia was more in coastal districts as compared to non-coastal districts (43.85% vs. 40.18%, p<0.001) (Table 4) and (Table 5). It was also observed that there was not much difference neither in standard of living (p=0.423) nor in BMI (p=0.782) between the children residing in the coastal and non-coastal districts. Distribution of the prevalence of anemia across the districts of Tamil Nadu was presented in the Figure 1.

It was observed that random intercept model has AIC=103603 and BIC=103611 whereas random coefficient model has AIC=103757 and BIC=103830. Accordingly, random intercept model was used in our study. The results obtained from this two-level random intercept logistic regression model are displayed in (Table 6). It was observed that younger age group (6-10 years) compared with older age groups (11-17 years) (AOR=1.343, 95% CI: 1.270-1.420), female children when compared with male children (AOR=1.462, 95% CI: 1.387-1.541), rural residence compared with urban residence (AOR=1.103, 95% CI: 1.038-1.172), more than 8 family members in the house compared with 4 or less family members (AOR=1.120, 95% CI: 1.008-1.245), 5to7 members in the family compared with 4 or less members (AOR=1.093, 95% CI: 1.034-1.155), lower standard of living compared with the high standard of living (AOR=1.203, 95% CI: 1.084-1.334) and the medium standard of living compared with the high standard of living (AOR=1.090, 95% CI: 1.004-1.184) were showed significant association with the prevalence of the anemia.

In our study, we have considered the cut off value for anemia as less than 11 g/dL hemoglobin in blood.^8,16 However, in some recent studies, the cut off value for anemia for the age group 6-11 years was recommended as less than 11.5 g/dL hemoglobin and for the 12-17 years age group as less than 12 g/dL hemoglobin in blood.¹⁷ Accordingly when we re-estimated the weighted prevalence, it was more alarming to note that the prevalence of anemia among school going children of age 6-11 years was 59.7% (95% CI: 56.89-62.501) and for school going children of age 12-17 years was 62.1% (95% CI: 61.23-62.96).

Chi-square=556.374 and p<0.001. CI*: Confidence Interval.

DISCUSSION

The present study comprehensively indicates that half of the school going children are anemic in the state. Prevalence of anemia among school going children showed a district level variation and was more significantly associated with districts in the coastal areas. Besides, we were not able to observe much differences neither in BMI nor in standard of living between those residing in coastal districts and non-coastal districts.

Many studies reported that hookworm infection is among major causes of anemia particularly among people residing in coastal areas.^18,19 Also reported that hookworm infection increases risk of anemia and causes negative effects on growth, iron status and cognition in children with high intensity infections.²⁰ However, there are no sufficient data to quantify benefits of deworming as prevention of anemia and further studies are warranted.

Children of age 6-10 years more affected with anemia compared to children of age 11-17 years and supports findings of Zelalem et al. However, our findings do not support evidence that absorbed iron requirements increases with age, compared to other energy requirements in children.¹⁶

In our study, prevalence of anemia was higher among female children compared to male children. One of the reasons may be menstrual blood loss. In some studies, it was reported that menstrual losses range from 10-250 mL (4-100 mg of iron) per period.²¹ These iron losses in female children double their need to absorb iron in comparison to male children.

In our study it was observed that the prevalence of anemia was higher among children residing in family with eight or more members compared to four or less than it. This could be due to the competition for food, nutritional deficiencies and other related factors. This finding is supported by similar studies.²²

In our study the prevalence of anemia was higher in rural areas and those having lower standard of living. This could be due to the fact that those who have high standard of living are more likely to provide balanced macro and micro nutrients to their children and children from these families have more chance of accessing health care services. Besides, poor sanitation and worm infestation can also contribute to their higher prevalence. Our results, with the support of other studies confirm that children from lower standard of living are vulnerable to various nutritional disorders including anemia.²³

CONCLUSION

The potential strength of our study is the use of all available data (DLHS-4) for the state of Tamil Nadu which enabled us to have a larger sample size. In addition, the use of two-level logistic regression analysis helped us to observe the district level variation of the data and to have a better insight to locate the high and low hotspot areas of anemia in the state. Limitations of the study include the cross-sectional nature of the study and therefore the temporal relationship between anemia and explanatory factors can not be established. Other confounders like presence of fever and diarrhoea were not included in the analysis. Nevertheless, these biases are non-differential as they are independent of the characteristics of children. In conclusion, we observed that half of the school going children were anemic in the state. The prevalence of anemia was significantly higher among coastal districts. It was also observed that it was more prevalent in the younger ages and those residing in lower standard of living. Therefore, the interventions of government and regional health centres should consider these factors also to avert the consequences of anemia in school going children.

ACKNOWLEDGEMENTS

Author would like to thanks to Dr. Tadury Madhukar Subbarao, Principal, PSG Institute of medical sciences and research, for permitting us to do this study. We express our gratitude to Dr. Sudha Ramalingam, director of research and innovation, HOD, department of community medicine, PSG institute of medical sciences and research, for permitting and providing the necessary support and facilities for doing this study. We also express our gratitude to Dr. M Sivamani, professor community medicine for his valuable suggestions in improving the quality of the paper. We are also indebted to Dr. S. L. Ravishankar, professor, department of community medicine, PSG institute of medical sciences and research, for his valuable guidance.

Funding: No funding sources

Conflict of interest: None declared

Ethical approval: The study was approved by the Institutional Ethics Committee

REFERENCES

1. 1. World Health Organization. Conclusions and recommendations of the WHO Consultation on prevention and control of iron deficiency in infants and young children in malaria-endemic areas. Food Nutr Bull. 2007;28:621-7.
2. 2. Warner MJ, Kamran MT. Iron Deficiency Anemia. StatPearls Treasure Island (FL). 2021.
3. 3. Lutter CK. Iron deficiency in young children in low-income countries and new approaches for its prevention. J Nutr. 2008;138(12):2523-8.
4. 4. Onyeneho, NG, Ozumba C, Subramanian SV. Determinants of Childhood Anemia in India. Sci Rep. 2019;9(1):16540.
5. 5. Calis JC, Phiri KS, Faragher EB, Brabin BJ, Bates I, Cuevas LE et al. Severe anemia in Malawian children. N Engl J Med. 2008;358(9):888-99.
6. 6. Chaparro CM, Suchdev PS. Anemia epidemiology, pathophysiology, and etiology in low- and middle-income countries. Ann New York Acad Sci. 2019;1450(1):15-31.
7. 7. Pasricha SR, Black J, Muthayya S, Shet A, Bhat V, Nagaraj S et al. Determinants of anemia among young children in rural India. Pediatrics. 2010;126(1):140-9.
8. 8. Madhusudan JV, Jeetendrayadav, Singh KH, Utpal Dhar Das. Prevalence and Determinants of Anemia among School going Children in Meghalaya. Demogr India. 2018;47:38-46.
9. 9. Swain PK, Singh P, Priyadarshini S. Determinants of home deliveries - Findings from India DLHS 4 analysis. J Family Med Prim Care. 2020;9:4723-8.
10. 10. International Institute for Population Sciences. National Family Health Survey (NFHS-2), 1998–99: India. Mumbai: International Institute for Population Sciences (IIPS) and ORC Macro. 2000.
11. 11. Solon G, Haider SJ, Wooldridge JM. What Are We Weighting For? J Hum Resource. 2015;50:301-16.
12. 12. International Institute for Population Sciences. District Level Household and Facility Survey (DLHS-3), 2007-2008; India. Mumbai: IIPS. 2010.
13. 13. Talukder A, Hossain MZ. Prevalence of Diabetes Mellitus and Its Associated Factors in Bangladesh: Application of Two-level Logistic Regression Model. Sci Rep. 2020;10:10237.
14. 14. Sommet N, Morselli D. Keep Calm and Learn Multilevel Logistic Modeling: A Simplified Three-Step Procedure Using Stata, R, Mplus, and SPSS. Int Rev Soc Psycol., 2017;30:203-18.
15. 15. Park S, Lake ET. Multilevel modeling of a clustered continuous outcome: nurses' work hours and burnout. Nurs Res. 2005;54:406-13.
16. 16. Anteneh ZA, Van Geertruyden JP. Spatial variations and determinants of anemia among under-five children in Ethiopia, EDHS 2005-2016. PLoS One. 2021;16:e0249412.
17. 17. World Health Organization. Nutritional anaemias: tools for effective prevention and control. Geneva: World Health Organization. 2017.
18. 18. Smith JL, Brooker S. Impact of hookworm infection and deworming on anaemia in non-pregnant populations: a systematic review. Trop Med Int Health. 2010;15:776-95.
19. 19. Hotez PJ, Brooker S, Bethony JM, Bottazzi ME, Loukas A, Xiao S. Hookworm infection. N Engl J Med. 2004;351:799-807.
20. 20. Kuong K, Fiorentino M, Perignon M, Chamnan C, Berger J, Sinuon M et al. Cognitive Performance and Iron Status are Negatively Associated with Hookworm Infection in Cambodian Schoolchildren. Am J Trop Med Hyg. 2016;95:856-63.
21. 21. Saljoughian M. Iron Deficiency Anemia: A Closer Look. US Pharm. (2007). 32: HS26-HS37.
22. 22. Moschovis PP, Wiens MO, Arlington L, Antsygina O, Hayden D, Dzik W et al. Individual, maternal and household risk factors for anaemia among young children in sub-Saharan Africa: a cross-sectional study. BMJ Open. 2018;8:e019654
23. 23. Kim JY, Shin S, Han K, Lee CK, Kim JH, Choi YS et al. Relationship between socioeconomic status and anemia prevalence in adolescent girls based on the fourth and fifth Korea National Health and Nutrition Examination Surveys. Eur J Clin Nutr. 2014;68:253-8.