| Objective: This article attempts to statistically analyze the significant differences in the C-reactive protein(CRP)of premature children infected and non-infected,and determine its critical shear value to diagnose infection or non-infection. Based on these results,we can evaluate the value of C-reactive protein (CRP) used in early diagnosis of infection and non-infection in premature children, and explore the effective use of the C-reactive protein (CRP) with other clinical diagnosis. Working through the above, this article can achieve goals that to reduce the incidence of severe infection and death rates in preterm children, and improve quality of life in premature children.Methods: In this article, the 173 premature infants all were hospitalize in our hospital from July 2009 to February 2010. At the first time, they were takeblood from the heel, and completed CRP immunofluorescence. We measured and recorded their blood and CRP values. In experiments, testing sensitivity of C-reactive protein assay kit is in 0.5~200mg/L, therefore we use a CRP average 0.30mg/L in ordinary people instead of that value less than 0.5mg/L. Combinating of blood and other clinical findings, all premature infants were divided into infection group including 79 premature infants and non-infection group (control group) including 94.Using statistical methods, firstly, we analyze the distribution of CRP values in infection group and non-infection group. Secondly, the significance test of numerical difference in CRP values on two groups is completed, by the rank sum test. Thirdly, We determine CRP optimal operating point (intersection) of the infected group and non-infected group, using ROC curve analysis. At last, analysing the reasons of positive showed in non-infection group, and research the main factors of non-infected group increasing false positive rate for CRP. Results: 1) Infection and non-infected group are all right-skewed distribution. And CRP values of 79 premature children with infectious diseases is in 0.30~83.00mg/L, their mean is 10.40mg/L be much larger than the mode 0.50mg/L, showing that values concentrate in small areas. Their skewness is 2.59, and standard deviation is 15.45, relative to the mean of 10.40mg/L, showing a larger span values. CRP frequency distribution histogram normal distribution curve is very different corresponding to the normal distribution, with an obvious right-skewed distribution. Similarly, non-infected group also has a similar right-skewed distribution. 2) Rank sum test showed that in the Mann-Whitney U test, the significant parameters P=0.000, less than 0.01, the difference is extremely significant. While, in the Kolmogorov-Smirnov Z test, the significant parameters P = 0.000, the same small at 0.01, the difference is extremely significant too. By confirming between the infection group and non-infection group, CRP difference is extremely significant. 3) ROC curves analysis results: When the best operating point obtaines CRP = 2.60mg/L, the positive likelihood ratio reached the maximum 49.45%, corresponding to a sensitivity of 0.544, and a error rate of 0.011, this time to maximize specific degree 0.989. When the best operating point obtaines CRP = 0.40m/L, the Youden Index reached the maximum of 0.637, corresponding to a sensitivity of 0.924, and a error rate of 0.287, this time to maximize sensitivity of 0.924. Obviously, the optimal operating point from two ways is very different, and non-infection group has great disturbance for CRP diagnosis of infection. Combining Clinical needs and characteristics of this study, we come to 0.5mg/L as the optimal working point, this time to the sensitivity of 0.83, and the specificity of 0.73, with a good diagnosis of infections and to exclude non-infection. 4) The false positive rate of C-reactive protein (CRP) in non-infection mainly is due to asphyxia, jaundice, and intracranial bleeding. And these diseases all increase CRP. At this time, application of CRP diagnosis need to increase the optimal operating point to 2.0mg/L, in order to improve the sensitivity and reduce false-positive rate. For omphalitis, flu and so on, according to that clinical diagnosis is more convenient, CRP diagnosis is not recommended.Conclusion: The conclusion of this study are determined by experiment and statistical analysis. 1) the difference C-reactive protein (CRP) blood levels is extremely significant, in premature children infected and non-infected, having potential as early diagnosis of infectious diseases. 2) the optimal operating point of C-reactive protein (CRP) is at 0.5mg/L, this time to the sensitivity of 0.83, and the specificity of 0.73, with a good diagnosis of infections and to exclude non-infection. 3) The false positive rate of C-reactive protein (CRP) in non-infection mainly is due to asphyxia, jaundice, and intracranial bleeding. And these diseases all increase CRP, need to be exclused with other clinical diagnosis. 4) In addition, this study also found that C-reactive protein (CRP) had a lower true positive rate in some infectious diseases, such as respiratory tract infection, need further filter on these infection types having low true positive rate, to improve the diagnostic sensitivity CRP. |
PDF Full Text Request