Critical Appraisal of Child Stunting

One of the core competencies that nurses and medical practitioners require to implement evidence into practice is the critical appraisal of relevant published literature to inform decision-making. Scholars believe that evidence-based medicine can help improve the quality of healthcare delivered. Thus, the medical fraternity should be in a position to identify, critically appraise, and synthesise research to assess the usefulness and validity of research findings (Chang et al., 2013; Smith & Noble, 2018). In line with this, the current report seeks to appraise an article from Rabaoarisoa et al. (2017) on “the importance of public health, poverty reduction programs and women’s empowerment in the reduction of child stunting in rural areas of Moramanga and Morondava, Madagascar.” The research is a case-control epidemiological study, which makes the selection of the Critical Appraisal Skills Programme (CASP, 2018) checklist tool for case-control studies ideal for reviewing the literature. The checklist comprises an eleven-point scale that evaluates research validity, reliability of the results, and feasibility of the recommendations within the population tested. The reason for selecting CASP tools is that they are more succinct and effectively cover the areas needed for critical appraisal, according to Nadelson & Nadelson (2014).

Research Validity

The first step into critically appraising literature is to establish a clear issue of focus, which the researcher may have identified using a research question as well as their study scope (Ferreira & Patino, 2018). The authors do not initiate a clear research question. However, they do concisely state the research’s objective, which is to identify the main determinants of stunting among children aged 6 to 59.9 months in the districts of Moramanga and Morondava. From the objectives, it is evident that the child stunting is the research’s issue of focus. Rabaoarisoa et al. (2017) define child stunting as a common manifestation of chronic malnutrition and refers to children who are short for their age. The rationale for developing the research is contributing knowledge to help combat chronic malnutrition, which causes mortality in children under the age of five. The focus on Madagascar is important as a large percentage of child stunting statistics point mainly to Asia and African continents (Raiten & Bremer, 2020).

The research scope involves a nutritional perspective into child stunting with a socio-economic underpinning of the factors determining its prevalence. Many scholars, including the World Health Organization (2018), note that socio-economically disadvantaged individuals have a higher likelihood of having stunted children. Hence, it is a logical approach as validated by past research. Some of the critical considerations the research makes include the socio-economic conditions of its sample population, macro- and micronutrient deficiencies, food insecurity, inadequate access to water, sanitation, hygiene, and health services. Failure to deploy the study could have short and long-term effects on the educability, future work capacity, income-earning ability, and susceptibility to disease (Rabaoarisoa et al., 2017). The potential harms identified by the authors are valid as they corroborate recent findings regarding the adverse impacts that child stunting can have on children and in their later life. Titaley et al. (2019) discussed cognitive, motor, and socio-economic development challenges, obstetric complications for women with a smaller pelvis, greater vulnerability to non-communicable diseases in adulthood, and perpetuated cycle of malnutrition. 

The researchers fail to provide a rationale for their use of a quantitative methodology. Regardless, it is appropriate to their research objective. First of all, it requires comparisons, meaning that the researchers needed the use of statistical information from their sample participants. An ideal method to use in such circumstances is a case-control approach, which is what the research utilised. Case-control studies are the simplest analytical study design that compares one group of patients with a particular health condition/disease (cases) with a similar comparison group that lacks the trait/disease to assess (control) (Omair, 2016). The design is appropriate for the research by Rabaoarisoa et al. because they need to establish a baseline from which they can determine the relationship that exists between the study variables and child stunting in characteristics specific to Moramanga and Morondava. Although susceptible to bias during recollection about exposure and reverse causality, the case-control design managed to address the study’s objective, which is to compare relationships in the two areas. Further, the use of regression modelling to establish the relationships adds to the credibility of the results. Regression models evaluate relationships between different variables or characteristics of interest in the community (Suárez et al., 2017). 

Another key consideration in a critical appraisal is whether the research presents selection bias, which can undermine the overall research validity. Selection bias is a systematic error that infers a non-representative selection of controls (Aigner et al., 2018). It is not clear whether the research accounts for all bias, but then again, selection bias is unavoidable. Researchers should try as much to minimise systematic errors as possible (Tripepi et al., 2010). Rabaoarisoa et al. try to minimise selection bias by carefully defining their cases, ensuring that their participants are representative of the defined population, use an established reliable system for selection, and use random sampling to arrive at the case and controls to use for the study. According to Malone et al. (2014), it is possible to minimise selection bias when selecting a sample from a population, at the same time increasing the generalizability of the result findings. 

Rabaoarisoa et al. identify their case and control studies from a sufficient number of cases from the population, which one can argue are representative of their respective geographical regions. There was a total of 894 children (431 cases and 463 controls) from a total of 8,627 participants in Moramanga. The researchers selected 932 children (420 cases and 512 controls) from a population of 2,379 in Morondava. The authors offer no specific distribution on the cases and controls, but instead highlight collective information regarding them. For example, the sample population from Morondava was younger than that from Moramanga, averaging at 30.4 months compared to Moramanga’s 33.2 months. Also, there was a larger number of females from Moramanga (471, 52.6%) compared to 423 (47.3%) of males. The sampling strategy relied on WHO reference populations, and the universal standard from which most child stunting literature appears to base their analysis on – children under the age of five years. Past researchers have used five years as the threshold from which they determine global child stunting statistics (Titaley et al., 2019; Raiten & Bremer, 2020). The criteria for selection excluded any children with disabilities, participants exhibiting both wasting and stunting, and those not accompanied by their mothers or caregivers. 

The time frame for the study was relevant to the exposure – nutritional composition – with the researchers assigning each area six months to experiment. The observations conform to the idea that the research was more interested in the cyclic trends associated with child stunting. Cyclic trends (weather or health-related) refer to increases or decreases in the frequency of a disease or other phenomena over several years or within a year (Friis, 2017). A simple random sampling of malnourished and stunted children helped create the case and control groups. There was nothing special about the controls, except for the fact that they did not exhibit stunting or wasting. The above conditions render the selection criteria relatively acceptable for both the case and control studies. However, there is the issue of non-response bias. The research has no record of its non-responders, meaning there was no way to evaluate whether the response rates were high or low in either situation. When response rates are low, critics might question the validity of the experiment due to non-response bias risk, which occurs when survey estimates based on respondent outcomes differ from those of the total sample selected (Studer et al., 2013). They acknowledge this in their limitations section, which can help during replication of the case-control study. They did, however, ensure rigorous training and competence in data collection and entry procedure to ensure the collection of high-quality data. 

The analysis points towards the minimal potential for selection bias, given the rigorous process undertaken to ensure that the experiment had a sufficient sample size and participants to inform the study. However, it does indicate a potential for non-response bias, which they mitigated by ensuring that most of the respondents provided high-quality data. Another issue that could have an impact on the validity of the results is recall bias, as a limitation to the case-control study design (Omair, 2016). According to Rabaoarisoa et al., the research does suffer from recall bias, occurring due to differences in accuracy or completeness of recollections from study participants (Kesmodel, 2018). The authors note the incompleteness of some information provided by the parents to some of the participants. Several households had missing information regarding birth spacing in Morondava (22%), thus requiring the experiments to carry on with missing observations. However, it does not undermine the validity of the findings as they manage to replicate another statistically significant finding from a previous model that integrated missing variables in its analysis. The researchers managed to measure the discrepancy to minimise the bias by adding the missing variables in a category of their own for evaluation.

Based on the evaluation above, the research presents minimal bias, largely because the researchers make a point of mitigating against such potential and minimising the impact it would have on validity. The issue is that they do not explicitly state the reason for employing a particular method, which is critical for replication studies. The literature byCoiera et al. (2018) notes that the ability to describe context explicitly helps with task replication by quantifying study replicability, and in translating research evidence into real-world application. Other than that, the researchers treat all groups equally and use a relatively common statistical method in their analysis – backward stepwise logistic regression analysis. The backwards stepwise method is common in epidemiological research. It involves the elimination of all possible explanatory variables that have the least significance until the remaining variable constitutes a statistically significant model (Smith, 2018). It is ideal in settings where the full model can fit without problems, no inordinate number of potential confounders to consider, and a lack of clear and strong heterogeneity (Greenland et al., 2016). The only design flaw in the research might have been the lack of sensitivity in identifying opportunistic parasites may have underestimated their prevalence among the population.


The researchers conducted the same experiment in the areas of Moramanga and Morondava. The stepwise regression arrived at the following results. For Moramanga, the significant determinants of child stunting, from a nutritional element perspective were age group, infection by Trichuris trichiura, and household characteristic wealth. In Morondava, the significant determinants of child stunting were age group, birth weight, birth spacing, being the firstborn, mother’s working status, and the occurrence of respiratory symptoms within the previous two weeks. The analysis is appropriate to the case-control design as it helps illustrate the differences that exist between the two regions. Understanding the difference in factor determinants can help foster development initiatives that are specific to a given area.

The association between exposure and outcomes are as follows. The highest association between malnutrition and child stunting in the age group category was children aged between 12-23 months for both adjusted and non-adjusted measures. In Moramanga, this age category was 4.0 times more likely to be stunted (p-value = 0.000) when adjusted for confounding variables and 3.8 times higher when unadjusted (p-value = 0.000). The difference is not significant, meaning that confounding can still explain the association. Morondava had similar observations, only that unadjusted model reported exposed children to be 2.7 times more likely to stunt (p-value = 0.000) compared to the adjusted model, showing exposed victims 2.5 times higher chance of stunting (p-value = 0.002). Those infected with Trichuris trichiura were 2.4 more times likely to be stunted (adjusted, p-value = 0.002), and 2.7 times more likely (unadjusted, p-value = 0.01). The significance may be less than when adjusted for confounding factors, but the observation and significance to the model maintains. Finally, the poorer and poorest households, whose odds ratio was constant at 2.3 times higher than the socio-economically advantaged groups (p-value = 0.000). 

In Morondava, participants with low birth weights were 1.6 more times likely to stunt, with the significance being higher when unadjusted for confounding variables (p-value = 0.001 vs p-value = 0.004). Firstborns were 0.6 more times likely to stunt with confounding variables increasing significance (p-value = 0.02) compared to the adjusted model (p-value = 0.04). Individuals born within 24-48 months of the earlier birth were both 0.5 more times likely to stunt during development in both categories, with confounding factors having lesser influence (p-value = 0.005 compared to adjusted p-value = 0.003). Working mothers contributed 1.7 more times to their children stunting (p-value = 0.001), with the significance reducing when confounding variables are present (p-value = 0.01). Finally, participants with an occurrence of respiratory symptoms had similar occurrence outcomes, with confounding factors have lesser significance (p-value = 0.009 vs 0.005). 

Based on the results, the associations between occurrence and outcomes, and size of the effects, it is highly likely that the results are reliable. The size of the significance values is hard to ignore. The fact that there is minimal difference between the significance of the adjusted and unadjusted effect ratios means that confounding has little to do with the significance of the variables measured. Furthermore, it is clear from the earlier discussion that the design and methods used in the analysis are not sufficiently flawed to make the results unreliable. Using the Bradford Hills criteria for causality, it is apparent that the research results are valid (Fedak et al., 2015). First of all, the fact that the researchers use experimentation is a strong basis for results validity. The results evidence strength of association, are consistent with other research findings, and also evidence some temporality. The findings that variables, such as economic status and age, determine child stunting, corroborate the secular variation component of the health condition (Friis, 2017; Prendergast & Humphrey, 2014). Also, the results are plausible and evidence some level of coherence owing to the available body of literature, and it is possible to make analogies from the results.

Application of Results and Recommendations

The recommendations made for improving children’s growth in Madagascar include poverty reduction, women empowerment, public health programs focusing on WASH and increasing acceptability, and increased coverage and quality of child/maternal health services. The recommendations are feasible from a general viewpoint, seeing that other researchers have recommended the same approaches in combating child stunting. The researchers offer no way to distinguish the participants in terms of ethnic or cultural distribution, so there is no way of telling whether the results are generalisable. However, it is possible to forecast the feasibility of the results within the two broad areas from which the researchers derived their study population. The primary benefit accrued from the recommendations proposed to the local population within these towns, and perhaps those surrounding them is the development of better nutritional programs. Empowering women and girls in the community enhance household appreciation for better and high-quality nutrition for their households, including themselves, to maintain nutritional behaviours, which will eventually contribute to declining child stunting cases. They do this by influencing the health of their children directly through intra-household resource allocation and childcare practices, and indirectly through their health and nutritional status (Siddhanta & Chattopadhyay, 2017; Chipili et al., 2018).

Policies touching on poverty recommendation are perhaps the most cited recommendation for reducing child stunting. The World Health Organization (2018) argue that failure to alleviate the people living in poor conditions will help maintain, if not increase, the prevalence rates of child stunting at the individual and population level. Social protection is one of the proposed methods to help improve the impoverished conditions that children with stunted growth live in at present. Social protection reduces chronic poverty by enabling poor households to meet their basic consumption needs an achieve better health, nutrition, and education that can help with creating more awareness on the social issue (World Health Organization, 2018). Countries may also engage communities with programs or measures that help improve the country’s overall GDP by leveraging individual productivity and increase national income. However, improving a country’s overall national income should go hand-in-hand with improving the diets of children and initiating nutritional interventions, as observed by McGovern et al. (2017).

It is also important to develop programs or policies that promote education and awareness regarding the positive impact of safe drinking water, sanitation, and hygiene (WASH) on public health and the health of infants and young children in particular. Water quantity interventions relate to mechanisms improving the quantity of water available to a household, while the latter refers to increasing the microbial quality of drinking water (Cumming & Cairncross, 2016). Sanitation refers to technologies and behaviours serving to contain waste and preventing human contact, while hygiene generally refers to washing hands with soap during critical times (Cumming & Cairncross, 2016). Focusing on nutritional, WASH, and well-being interventions can enhance long-term health outcomes in children. However, the lack of education and proper awareness can undermine the progress made by such efforts to reduce stunting. It is the reason why the researchers included the need for support, both from political actors and the community itself to reach stunting reduction goals, as proposed by Kwami et al. (2019).

In conclusion, the research study has an appropriate design that results in meaningful and useful results consistent with the recommendations made in past literature. The researchers live up to their objective to determine the main determinants of child stunting and contribute to the issue by presenting evidence-based solutions exhibiting high validity. The primary challenge is that they are not as explicit in content generation and do not make it apparent why they include some of the practices they do in the methodology. They identify the most significant limitations and account for them during analysis, which reduces the risk to reliability and validity. Overall, the research by Rabaoarisoa et al. (2017) has immense value in developing knowledge regarding determinants of child stunting and adds to the existing body of literature that future researchers can use as a reference point. Also, their deployment of the study design is one that future researchers can learn from and even implement in their respective epidemiology studies.


Aigner, A., Grittner, U., & Becher, H. (2018). Bias due to differential participation in case-control studies and review of available approaches for adjustment. PloS one13(1).

Chang, S. C., Huang, C. Y., Chen, S. Y., Liao, Y. C., Lin, C. H., & Wang, H. H. (2013). Evaluation of a critical appraisal program for clinical nurses: A controlled before-and-after study. The Journal of Continuing Education in Nursing, 44(1), 43-48.

Chipili, G., Msuya, J., Pacific, R., & Majili, S. (2018). Women Empowerment and the Nutrition Status of Children Aged Between 6-59 Months. J. Nutr. Health Sci5, 208.

Coiera, E., Ammenwerth, E., Georgiou, A., & Magrabi, F. (2018). Does health informatics have a replication crisis?. Journal of the American Medical Informatics Association25(8), 963-968.

Critical Appraisal Skills Programme UK. (2018). CASP checklists for Case-Control Studies. Retrieved from!casp-tools-checklists/c18f8. Accessed 28 May 2020. 

Cumming, O., & Cairncross, S. (2016). Can water, sanitation and hygiene help eliminate stunting? Current evidence and policy implications. Maternal & child nutrition12, 91-105.

Fedak, K. M., Bernal, A., Capshaw, Z. A., & Gross, S. (2015). Applying the Bradford Hill criteria in the 21st century: how data integration has changed causal inference in molecular epidemiology. Emerging themes in epidemiology,12(1), 14.

Ferreira, J. C., & Patino, C. M. (2018). Critical appraisal of the literature. Why do we care?. Jornal Brasileiro de Pneumologia44(6), 448-448.

Friis, R. H. (2017). Epidemiology 101. Jones & Bartlett Learning.

Greenland, S., Daniel, R., & Pearce, N. (2016). Outcome modelling strategies in epidemiology: traditional methods and basic alternatives. International journal of epidemiology45(2), 565-575.

Kesmodel, U. S. (2018). Information bias in epidemiological studies with a special focus on obstetrics and gynecology. Acta obstetricia et gynecologica Scandinavica97(4), 417-423.

Kwami, C. S., Godfrey, S., Gavilan, H., Lakhanpaul, M., & Parikh, P. (2019). Water, Sanitation, and Hygiene: Linkages with Stunting in Rural Ethiopia. International journal of environmental research and public health16(20), 3793

Malone, H., Nicholl, H., & Tracey, C. (2014). Awareness and minimisation of systematic bias in research. British Journal of Nursing23(5), 279-282.

McGovern, M. E., Krishna, A., Aguayo, V. M., & Subramanian, S. V. (2017). A review of the evidence linking child stunting to economic outcomes. International journal of epidemiology46(4), 1171-1191.

Nadelson, S., & Nadelson, L. S. (2014). Evidence‐based practice article reviews using CASP tools: a method for teaching EBP. Worldviews on Evidence‐Based Nursing11(5), 344-346.

Omair, A. (2016). Selecting the appropriate study design: Case–control and cohort study designs. Journal of Health Specialties4(1), 37.

Prendergast, A. J., & Humphrey, J. H. (2014). The stunting syndrome in developing countries. Paediatrics and international child health34(4), 250-265.

Rabaoarisoa, C. R., Rakotoarison, R., Rakotonirainy, N. H., Mangahasimbola, R. T., Randrianarisoa, A. B., Jambou, R., … & Randremanana, R. V. (2017). The importance of public health, poverty reduction programs and women’s empowerment in the reduction of child stunting in rural areas of Moramanga and Morondava, Madagascar. PloS one12(10).

Raiten, D. J., & Bremer, A. A. (2020). Exploring the Nutritional Ecology of Stunting: New Approaches to an Old Problem. Nutrients12(2), 371.

Siddhanta, A., & Chattopadhyay, A. (2017). Role of women’s empowerment in determining child stunting in Eastern India and Bangladesh. Social Science Spectrum3(1), 38-51.

Smith, G. (2018). Step away from stepwise. Journal of Big Data5(1), 32.

Smith, J., & Noble, H. (2018). Reviewing the literature: Table 1. Evidence Based Nursing, 19(1), 2–3. doi:10.1136/eb-2015-102252

Studer, J., Baggio, S., Mohler-Kuo, M., Dermota, P., Gaume, J., Bertholet, N., … & Gmel, G. (2013). Examining non-response bias in substance use research—Are late respondents proxies for non-respondents?. Drug and alcohol dependence132(1-2), 316-323.

Suárez, E., Pérez, C. M., Rivera, R., & Martínez, M. N. (2017). Applications of Regression Models in Epidemiology. John Wiley & Sons.

Titaley, C. R., Ariawan, I., Hapsari, D., Muasyaroh, A., & Dibley, M. J. (2019). Determinants of the stunting of children under two years old in Indonesia: A multilevel analysis of the 2013 Indonesia Basic Health Survey. Nutrients11(5), 1106.

Tripepi, G., Jager, K. J., Dekker, F. W., & Zoccali, C. (2010). Selection bias and information bias in clinical research. Nephron Clinical Practice115(2), c94-c99.

World Health Organization. (2018). Reducing stunting in children: equity considerations for achieving the global nutrition targets 2025.