PDF(Pubmed)
Abstract：
In the past two decades, there has been a growing recognition of the need to establish indigenous standards or reference growth charts, particularly following the WHO multicenter growth study in 2006. The availability of accurate and reliable growth charts is crucial for monitoring child health. The choice of an appropriate model for constructing growth charts depends on various data characteristics, including the distribution\'s tails and peak. While Pakistan has reported some reference growth charts, there is a notable absence of indigenous charts for children under two years of age, especially for infants aged 0-6 months who are exclusively breastfed. Additionally, acquiring data poses a significant challenge, particularly for low-income countries, as it demands substantial resources such as finances, time, and expertise. The Multiple Indicator Cluster Survey (MICS) constitutes a large-scale national survey conducted periodically in low-income countries under the auspices of UNICEF. In this study, we propose methods for generating selection variables utilizing the \"Novel Case Selection Method,\" as previously published. Further our approach enables to select and fit appropriate model to the MICS data, selected, and to develop the standard growth charts.
Out of the 11,478 children under 6 months of age included in MICS-6 (Pakistan), 3,655 children (1,831 males and 1,824 females) met the specified criteria and were selected using the \"Novel Case Selection Method\". The sample was distributed across provinces as follows: 841 (23.0%) from KPK, 1,464 (40.1%) from Punjab, 819 (22.4%) from Sindh, and 531 (14.5%) from Balochistan. This sample encompassed both rural (76.4%) and urban (23.6%) populations. Following data cleaning and outlier removal, a total of 3,540 records for weight (1,768 males and 1,772 females) and 3,515 records for height (1,759 males and 1,756 females) were ultimately available for the development of standard charts. The Bayesian Information Criterion (BIC) was employed to determine the optimal degrees of freedom for L, M, and S using RefCurv_0.4.2. Three families within the gamlss class-namely, Box Cox Cole and Green (BCCG), Box Cox T (BCT), and Box Cox Power Exponential (BCPE)-were applied, each with three smoothing techniques: penalized splines (ps), cubic splines (cs), and polynomial splines (poly). The best-fitted model was selected from these nine combinations based on the Akaike Information Criteria.
The Novel Case Selection Method yielded 3655 cases as per criteria. After cleaning the data, this method lead to selection of 3540 children for \"weight for age\" (W/A) and 3515 children for \"height for age\" (H/A). The \"BCPE\" family and \"ps\" as smoothing method proved to be best on AIC for all four curves, i.e. the W/A male, W/A female, H/A male, and H/A female. The optimum selected degrees of freedom for the curve \"W/A\", for both genders were (M = 1, L = 0, S = 0). The optimum degrees of freedom for H/A male were again (M = 1, L = 0, S = 0), but for females the selected degrees of freedom were (M = 1, L = 1, S = 1). The indigenous fitted standard curves for Pakistan were on lower trajectory in comparison to WHO standards.
This study uses the Novel Case Selection Method with introduced algorithms to construct tailored growth charts for lower and middle-income countries. Leveraging extensive MICS data, the methodology ensures representative national samples. The resulting charts hold practical value and await validation from established data sources, offering valuable tools for policy makers and clinicians in diverse global contexts.
Out of the 11,478 children under 6 months of age included in MICS-6 (Pakistan), 3,655 children (1,831 males and 1,824 females) met the specified criteria and were selected using the \"Novel Case Selection Method\". The sample was distributed across provinces as follows: 841 (23.0%) from KPK, 1,464 (40.1%) from Punjab, 819 (22.4%) from Sindh, and 531 (14.5%) from Balochistan. This sample encompassed both rural (76.4%) and urban (23.6%) populations. Following data cleaning and outlier removal, a total of 3,540 records for weight (1,768 males and 1,772 females) and 3,515 records for height (1,759 males and 1,756 females) were ultimately available for the development of standard charts. The Bayesian Information Criterion (BIC) was employed to determine the optimal degrees of freedom for L, M, and S using RefCurv_0.4.2. Three families within the gamlss class-namely, Box Cox Cole and Green (BCCG), Box Cox T (BCT), and Box Cox Power Exponential (BCPE)-were applied, each with three smoothing techniques: penalized splines (ps), cubic splines (cs), and polynomial splines (poly). The best-fitted model was selected from these nine combinations based on the Akaike Information Criteria.
The Novel Case Selection Method yielded 3655 cases as per criteria. After cleaning the data, this method lead to selection of 3540 children for \"weight for age\" (W/A) and 3515 children for \"height for age\" (H/A). The \"BCPE\" family and \"ps\" as smoothing method proved to be best on AIC for all four curves, i.e. the W/A male, W/A female, H/A male, and H/A female. The optimum selected degrees of freedom for the curve \"W/A\", for both genders were (M = 1, L = 0, S = 0). The optimum degrees of freedom for H/A male were again (M = 1, L = 0, S = 0), but for females the selected degrees of freedom were (M = 1, L = 1, S = 1). The indigenous fitted standard curves for Pakistan were on lower trajectory in comparison to WHO standards.
This study uses the Novel Case Selection Method with introduced algorithms to construct tailored growth charts for lower and middle-income countries. Leveraging extensive MICS data, the methodology ensures representative national samples. The resulting charts hold practical value and await validation from established data sources, offering valuable tools for policy makers and clinicians in diverse global contexts.
摘要：
背景：在过去的二十年中，人们越来越认识到有必要建立土著标准或参考增长图，特别是在2006年世卫组织多中心生长研究之后。准确可靠的生长图的可用性对于监测儿童健康至关重要。构建增长图的适当模型的选择取决于各种数据特征，包括分布的尾部和峰值。虽然巴基斯坦报告了一些参考增长图表，两岁以下儿童明显没有土著海图，特别是0-6个月纯母乳喂养的婴儿。此外,获取数据构成了重大挑战，特别是低收入国家，因为它需要大量的资源，如财政，时间,和专业知识。多指标类集调查（MICS）是在儿童基金会主持下在低收入国家定期进行的大规模国家调查。在这项研究中,我们提出了利用新的案例选择方法来生成选择变量，“如前所述。此外，我们的方法还可以选择合适的模型并将其拟合到MICS数据中，选定,并制定标准增长图。
方法：在MICS-6（巴基斯坦）中包括的11,478名6个月以下儿童中，3,655名儿童（1,831名男性和1,824名女性）符合指定标准，并使用“新型病例选择方法”进行选择。样本分布在各省如下：来自KPK的841(23.0%)，旁遮普1,464（40.1%），819（22.4%）来自信德省，531人（14.5%）来自俾路支省。该样本涵盖了农村（76.4％）和城市（23.6％）人口。在数据清理和异常值删除之后，总共有3,540份体重记录(男性1,768份,女性1,772份)和3,515份身高记录(男性1,759份,女性1,756份),采用贝叶斯信息准则(BIC)来确定L的最佳自由度，M,和S使用RefCurv_0.4.2。gamlss类中的三个家庭-即，BoxCoxColeandGreen(BCCG),BoxCoxT(BCT),并应用了BoxCox幂指数(BCPE)，每个都有三种平滑技术：惩罚样条(ps)，三次样条(cs),和多项式样条（聚）。根据Akaike信息标准从这9种组合中选择最佳拟合模型。
结果：根据标准，新的病例选择方法产生了3655例。清理数据后，这种方法导致选择3540名儿童为“年龄体重”（W/A），选择3515名儿童为“年龄身高”（H/A）。对于所有四条曲线，“BCPE”族和“ps”作为平滑方法被证明在AIC上是最好的，即W/A男性，W/A女性,H/A男性，和H/A女性。曲线“W/A”的最佳选择自由度,男女均为(M=1，L=0，S=0)。H/A男性的最佳自由度再次为（M=1，L=0，S=0），但是对于女性，选择的自由度为（M=1，L=1，S=1）。与世卫组织标准相比，巴基斯坦的土著拟合标准曲线处于较低的轨道。
结论：本研究使用新的案例选择方法和引入的算法，为中低收入国家构建量身定制的增长图。利用广泛的MICS数据，该方法确保具有代表性的国家样本。生成的图表具有实用价值，并等待已建立数据源的验证，在不同的全球背景下，为政策制定者和临床医生提供有价值的工具。
方法：在MICS-6（巴基斯坦）中包括的11,478名6个月以下儿童中，3,655名儿童（1,831名男性和1,824名女性）符合指定标准，并使用“新型病例选择方法”进行选择。样本分布在各省如下：来自KPK的841(23.0%)，旁遮普1,464（40.1%），819（22.4%）来自信德省，531人（14.5%）来自俾路支省。该样本涵盖了农村（76.4％）和城市（23.6％）人口。在数据清理和异常值删除之后，总共有3,540份体重记录(男性1,768份,女性1,772份)和3,515份身高记录(男性1,759份,女性1,756份),采用贝叶斯信息准则(BIC)来确定L的最佳自由度，M,和S使用RefCurv_0.4.2。gamlss类中的三个家庭-即，BoxCoxColeandGreen(BCCG),BoxCoxT(BCT),并应用了BoxCox幂指数(BCPE)，每个都有三种平滑技术：惩罚样条(ps)，三次样条(cs),和多项式样条（聚）。根据Akaike信息标准从这9种组合中选择最佳拟合模型。
结果：根据标准，新的病例选择方法产生了3655例。清理数据后，这种方法导致选择3540名儿童为“年龄体重”（W/A），选择3515名儿童为“年龄身高”（H/A）。对于所有四条曲线，“BCPE”族和“ps”作为平滑方法被证明在AIC上是最好的，即W/A男性，W/A女性,H/A男性，和H/A女性。曲线“W/A”的最佳选择自由度,男女均为(M=1，L=0，S=0)。H/A男性的最佳自由度再次为（M=1，L=0，S=0），但是对于女性，选择的自由度为（M=1，L=1，S=1）。与世卫组织标准相比，巴基斯坦的土著拟合标准曲线处于较低的轨道。
结论：本研究使用新的案例选择方法和引入的算法，为中低收入国家构建量身定制的增长图。利用广泛的MICS数据，该方法确保具有代表性的国家样本。生成的图表具有实用价值，并等待已建立数据源的验证，在不同的全球背景下，为政策制定者和临床医生提供有价值的工具。
