• Takao Arai, Shoichi Ohta, Junya Tsurukiri, Kenichiro Kumasaka, Katsuhiro Nagata, Taihei Okita, Taishi Oomura, Akira Hoshiai, Masaharu Koyama, and Tetsuo Yukioka.
• Department of Emergency and Critical Care Medicine, Trauma and Emergency Center, Hachioji Medical Center of Tokyo Medical University, Tokyo 193-0998, Japan. Electronic address: qqaraitakao@yahoo.co.jp.
• Am J Emerg Med. 2016 Nov 1; 34 (11): 2150-2153.

BackgroundWe examined whether the values obtained from principal component analysis (PCA) on laboratory tests can be used to predict bacterial infections and identify bacterial strains in blood culture (BC).MethodThis study is a single-center retrospective analysis of 315 patients suspected of having sepsis. We applied PCA on procalcitonin (PCT) and laboratory test biomarkers, namely, platelet (PLT), white blood cell, and C-reactive protein (CRP) as well as BC.ResultsPrincipal component analysis showed that PCT, CRP, and PLT contributions to component 1 were associated with bacterial infection. The number of patients who had BC-negative results, gram-positive cocci (GPC), and gram-negative rods (GNRs) were 124, 28, and 19, respectively. The mean value of component 1 in GNR-positive patients was 1.58±1.41 and was significantly higher than that in GPC-positive patients (0.28±0.87; P<.0001). Furthermore, the mean values of component 1 in both GNR- and GPC-positive patients were significantly higher than that in BC-negative patients (-0.31±0.65; P<.0001 and P<.002, respectively). One certain range showing higher value more than 2.00 for component 1 and -1.00 for component 2 only included GNR-positive patients. There were no BC-positive patients who showed less than -1.00 for component 1.ConclusionThe present results obtained by PCA on laboratory tests involving PCT, PLT, white blood cell, and CRP suggest the potential of PCA-obtained values to not only predict bloodstream infection but also identify bacterial strains. This provides some clinical significance in the management of sepsis in acute care.Copyright © 2016 Elsevier Inc. All rights reserved.

29 keywords...

hide…