The deleterious effects of air pollution on children’s health are broadly discussed in the literature. Among pollutants, the Particulate Matter with diameter \(<{10}\,\upmu \hbox {m}\) (\(\hbox {PM}_{10}\)) is usually associated with an increase in the number of children’s hospital admissions. In this paper, we have assumed that, at low concentrations, there is no effect of the former on the latter, and from an unknown concentration threshold, the admissions would be positively affected by \(\hbox {PM}_{10}\). As far as we know, this scenario has never been explored before. In this study, the number of children (10 years and younger) daily admitted with respiratory diseases, the daily average concentration of the \(\hbox {PM}_{10}\) pollutant and meteorological variables collected in the period of January 2010–August 2014, and a deterministic factor for workdays and days off are considered in the model. In order to deal with the lack of information regarding the residence of the children, regularized canonical correlation analysis was used to select only admissions from address codes with higher correlations with \(\hbox {PM}_{10}\) collected from the considered monitoring station. To explain the assumed behavior and estimate the \(\hbox {PM}_{10}\) threshold, we used segmented regression analysis for count data. The Negative Binomial segmented model provided a better fit. Despite data limitations, this framework proves to be a promising modeling strategy, since it is able to provide an explicit estimate for a threshold of \(\hbox {PM}_{10}\) exposure. Another important finding is that the inadequate adjustment by non-segmented models would underestimate the number of expected hospital admissions for concentrations far from the threshold, which in a cumulative aspect may lead to an inaccurate hospital management. We also developed a Monte Carlo experiment to investigate the effects of misfitting the segmented behavior by classic count models, which ratified the findings in application.

文献广泛讨论了空气污染对儿童健康的有害影响。在污染物中，直径为\（<{10} \，\ upmu \ hbox {m} \）（\（\ hbox {PM} _ {10} \））的颗粒物通常与污染物数量的增加有关。儿童医院住院。在本文中，我们假设在低浓度下，前者对后者没有影响，并且根据未知的浓度阈值，入学将受到\（\ hbox {PM} _ {10} \ ）。据我们所知，这种情况以前从未被探讨过。在这项研究中，每天接受呼吸道疾病治疗的儿童（10岁及以下）数量，\（\ hbox {PM} _ {10} \）的每日平均浓度该模型考虑了2010年1月至2014年8月期间收集的污染物和气象变量，以及工作日和休息日的确定性因素。为了解决有关儿童居住的信息不足的问题，使用规范化规范相关分析从与所收集的\（\ hbox {PM} _ {10} \）具有较高相关性的地址代码中仅选择接纳监测站。解释假定的行为并估算\（\ hbox {PM} _ {10} \）阈值，我们使用分段回归分析获得计数数据。负二项式分段模型提供了更好的拟合度。尽管存在数据限制，但该框架被证明是一种很有前途的建模策略，因为它能够为\（\ hbox {PM} _ {10} \）暴露阈值提供明确的估计。另一个重要发现是，非分段模型的不充分调整将低估浓度远低于阈值的预期入院人数，这在累积方面可能导致医院管理不准确。我们还开发了一个蒙特卡洛实验，以研究经典计数模型对细分行为的不当影响，该模型批准了应用中的发现。