日本新生児成育医学会雑誌　第29巻　第2号　99～ 108頁（2017年）

日本新生児成育医学会雑誌　29(2)：373-382；2017　

受付日：平成28．10．20

受理日：平成29．01．16

新生児科医による心エコー静止画上の計測の検者間信頼性

Interrater Reliability of Echocardiographic Measurements of Still Images Among Neonatologists

＊1国立成育医療研究センター　新生児科，＊2マクマスター大学（カナダ）　臨床疫学統計学部，＊3大阪市立総合医療センター　新生児科，＊4東京女子医科大学八千代医療センター　新生児科，＊5長浜赤十字病院　小児科，
＊6大阪母子医療センター　新生児科，＊7京都第一赤十字病院　新生児科，＊8埼玉医科大学総合医療センター　小児科，＊9岐阜県総合医療センター　新生児内科，＊10名古屋第二赤十字病院　小児科，＊11埼玉医科大学総合医療センター　小児循環器部門，＊12神奈川県立こども医療センター　新生児科

＊1 Division of Neonatology, National Center for Child Health and Development, ＊2 Department of Clinical Epidemiology & Biostatistics, McMaster University, ＊3 Department of Neonatology, Osaka City General Hospital, ＊4 Department of Neonatology, Tokyo Women’s Medical University, Yachiyo Medical Center, ＊5 Department of Pediatrics, Nagahama Red Cross Hospital, ＊6 Department of Neonatal Medicine, Osaka Women’s and Children’s Hospital, ＊7 Department of Neonatology, Japanese Red Cross Kyoto Daiichi Hospital, ＊8 Department of Pediatrics, Saitama Medical Center, Saitama Medical University, ＊9 Department of Neonatology, Gifu Prefectural General Medical Center, ＊10 Department of Pediatrics, Nagoya Daini Red Cross Hospital, ＊11 Pediatric Cardiology, Saitama Medical Center, Saitama Medical University, ＊12 Division of Neonatology, Kanagawa Children’s Medical Center

甘利昭一郎＊1・諫山哲哉＊2・岩見裕子＊3・佐藤雅彦＊4・山本正仁＊5・野崎昌俊＊6・小谷　牧＊7・川崎秀徳＊8・長澤宏幸＊9・横山岳彦＊10・増谷　聡＊11・豊島勝昭＊12

Shoichiro AMARI ＊ 1，Tetsuya ISAYAMA ＊ 2，Hiroko IWAMI ＊ 3，Masahiko SATO ＊ 4，Masahito YAMAMOTO ＊ 5，Masatoshi NOZAKI ＊ 6，Maki KOTANI ＊ 7，Hidenori KAWASAKI ＊ 8，Hiroyuki NAGASAWA ＊ 9，Takehiko YOKOYAMA ＊ 10，Satoshi MASUTANI ＊ 11，Katsuaki TOYOSHIMA ＊ 12

Key Words：neonatologist，newborn infant，preterm infant，echocardiography，interrater reliability

目的：新生児科医の心エコー計測の検者間信頼性と精度向上の可能性とを検討した。
方法：全国の新生児科医が共通の静止画上で左房径/ 大動脈径比（LA/Ao），左房容積（LAV），動脈管径（DA）を計測した。計測は前期（各2 画像）と後期（各1 画像）に分け，前後期の間に精度向上を目的とした介入を行った。計測誤差（（計測値－真値）/ 真値）と，計測値の変動係数を検討した。
結果：全検者220 名のうち，計測誤差が±10％以内に収まった割合はLA/Ao：71 ～ 84％で，LAV：37 ～ 45％やDA：37 ～ 55％より大きかった。前期2 画像，後期1 画像での変動係数の95％信頼区間は，LA/Ao で0.11 ～ 0.16，0.10 ～ 0.13，0.074 ～ 0.11，LAV で0.20 ～ 0.26，0.24 ～ 0.41，0.21 ～ 0.60，DA で0.17 ～ 0.33，0.20 ～ 0.30，0.13 ～0.20 であった。LA/Ao はLAV，DA より変動係数が小さく，LA/Ao とDA は後期で変動係数の改善を認めた。
結論：新生児科医の心エコー計測にはばらつきがあるが，精度向上に向けた介入が有効な可能性がある。

Objective：To investigate the interrater reliability of neonatologists’ echocardiographic measurements of still images and to identify methods to improve reliability.
Methods：Study participants were neonatologists in Japan who measured three indices, including left atrium-to-aorta ratios（LA/Ao）, left atrial volumes（ LAV） and ductus arteriosus diameters（ DA） on echocardiographic still images in this study. Each index was measured in three different still images, resulting in nine measurements. The measurements were performed in two stages： the first and second images of each index were measured at the initial stage（ six measurements）, and the third image was measured at the next stage （three measurements）. Between the two stages, the participants received feedback on how to improve the reliability of their measurements. A measurement error, defined as the measured-value minus the mean-value divided by the mean-value for each participant, and a coefficient of variance of measured values for all participants, were evaluated for each measurement.
Results：This study included 220 neonatologists who completed the measurements of all images. The proportions of participants with measurement errors within plus or minus 10％ were 71 ～ 84％ for LA/Ao, 37 ～ 45％ for LAV and 37 ～ 55 ％ for DA. The 95％ confidence intervals of the coefficients of variance for measurements of the first, second, and third images were 0.11 ～ 0.16, 0.10 ～ 0.13, and 0.074 ～ 0.11 for LA/Ao, 0.20 ～ 0.26, 0.24 ～ 0.41, 0.21 ～ 0.60 for LAV, and 0.17 ～ 0.33, 0.20 ～ 0.30, and 0.13 ～ 0.20 for DA, respectively. The coefficients of variance for LA/Ao were smaller than those for LAV and DA, and improved in the second stage for LA/Ao and DA.
Conclusion：Serious measurement errors were found in neonatologists’ echocardiographic measurements of still images；however, the intervention for reducing measurement errors could be effective.