Socio-economic Status and Clustering of Child Deaths in Rural Punjab
Socio-economic Status and Clustering of Child Deaths in Rural Punjab
复制标题
旁遮普邦农村地区的社会经济状况和儿童死亡的聚集性
DOI：
--
复制
发表时间：
1997
期刊：
影响因子：
--
通讯作者：
M. Gupta
中科院分区：
文献类型：
--
作者：
M. Gupta
近年来，对儿童死亡率的几项研究都集中在儿童死亡率的死亡聚集或家庭间异质性问题上（Das Gupta 1990; Curtis、Diamond和McDonald 1993; Guo 1993; Ronsmans 1993; Zaba和大卫1994）。早先对世界生育率调查数据的分析（Meegama，1980年; Hobcraft、McDonald和Rutstein，1985年）以及在各种情况下进行的其他研究都表明了这一点。“异性恋可能对生殖健康和儿童生存方案产生相当大的影响。在印度，同许多其他国家一样，保健服务主要是根据需求提供的。如果儿童死亡主要集中在某些家庭，这就意味着，通过采取更具成本效益的办法，将保健资源专门集中用于儿童死亡风险高的家庭分组，可以大大降低儿童死亡率。聚类需要仔细检查，因为有几个原因可以解释为什么观察到的风险异质性可能是数据分析方法的人为因素。因此，当我们分析从不同地点汇总的数据时，我们可能会期望聚类，就像通常对国家一级数据所做的那样。不同地区居民的生活条件可能有很大差异，并可能影响死亡率（例如，通过获得保健服务、营养和疾病风险方面的差异）。同样的考虑也适用于死亡率随时间变化的地方研究;在这种情况下，汇总不同年龄组妇女的经验将产生与汇总不同区域样本类似的效果。在人口中聚集的另一个原因是，不同社会经济地位群体的家庭死亡率可能不同。因此，Guo（1993年）发现，危地马拉家庭中的大多数聚集现象可以用家庭的经济地位和母亲的教育程度来解释。正如他所指出的，由于这些众所周知的因素而导致的聚集性的发现并没有增加我们对儿童死亡率的理解（考德威尔，1979; Ware，1984; Cleland and and货车Ginneken，1988）。在控制了所有与儿童死亡率有关的因素之后，任何家庭群体的聚集现象都会消失;只有当它指出了迄今为止尚未分析的因素时，它才是有意义的。在本文中，我们探索了其他模型来测试聚类，并将其应用于按社会经济地位、母亲教育程度和儿童死亡年龄分层的旁遮普群体的数据。儿童的性别也被包括在内，因为众所周知，在旁遮普，有一个以上女儿的家庭中的女孩受到选择性歧视，死亡率过高（Das Gupta，1987年），这可能是造成一些集群现象的原因。只有在社会经济和教育水平最低的群体中才有明显的集群迹象。我们还探讨了家庭风险和家庭建设因素之间的关系，如短生育间隔和高产次分娩，因为众所周知，这些因素与较高的儿童死亡率有关。这种关联似乎与其说是造成儿童死亡集中的原因，不如说是一种结果。
In recent years several studies of child mortality have been focused on the problem of death clustering or inter-family heterogeneity in child mortality (Das Gupta 1990; Curtis, Diamond and McDonald 1993; Guo 1993; Ronsmans 1993; Zaba and David 1994). This was indicated by earlier analyses of World Fertility Survey data (Meegama 1980; Hobcraft, McDonald and Rutstein 1985), as well as in other studies in various settings.' Heterogeneity can have considerable implications for reproductive health and child survival programmes. In India, as in many other countries, health services are made available largely in response to demand. If child deaths are heavily concentrated in some families, this would suggest that substantial improvements in child mortality could be achieved by adopting the more costeffective techniques of focusing health-care resources specifically on the sub-group of families with a high risk of child death. Clustering needs to be examined carefully because there are several reasons why observed heterogeneity in risks could be an artifact of the method of data analysis. Thus, we might expect clustering when we analyse data aggregated from different locations, as is often done with countrylevel data. Living conditions of residents in different areas may be very different, and could affect mortality (e.g. through differences in access to health services, nutrition, and exposure to disease). The same considerations apply to local studies in a setting in which mortality has changed over time; in this case aggregation of the experience of different age groups of women will have an effect similar to that of aggregating samples from different regions. Another reason for clustering within a population is that mortality may differ in families in different socio-economic status groups. Thus, Guo (1993) found that most of the clustering in Guatemalan families could be explained by the household's economic status and the mother's education. As he points out, the discovery of clustering owing to such well-known factors does not add much to our understanding of child mortality (Caldwell 1979; Ware 1984; Cleland and van Ginneken 1988). Clustering in any group of families will disappear after all the relevant factors in child mortality have been controlled; it is of interest only if it points to factors that have hitherto not been analysed. In this paper we explore aternative models to test for clustering and apply them to data for Punjab in groups stratified by socio-economic status, mother's education, and child's age at death. The child's sex is also included because it is known that in Punjab, girls in families with more than one daughter suffer selective discrimination and excess mortality (Das Gupta 1987), and this could account for some clustering. Significant evidence of clustering is found only in the lowest socioeconomic and education groups. We also explore the relationship between familial risk and familybuilding factors such as short birth intervals and high-parity births, as these are well known to be associated with higher child mortality. It appears that this association may be more an effect than a cause of clustering of child deaths.