肺炎の予測モデル

Construction and verification of a nomogram prediction model of severe adult community-acquired pneumonia
 
OBJECTIVE: To construct and verify the nomogram prediction model based on inflammatory indicators, underlying diseases, etiology and the British Thoracic Society modified pneumonia score (CURB-65 score) in adults with severe community acquired pneumonia (CAP). METHODS: The clinical data of 172 adult inpatients first diagnosed as CAP at Taikang Xianlin Drum Tower Hospital from January 2018 to December 2021 were divided into severe and non-severe diseases groups according to the severity of their conditions. The baseline conditions (including gender, age, past history, comorbidities and family history), clinical data (including chief symptoms, onset time, CURB-65 score), first laboratory results on admission (including whole blood cell count, liver and kidney function, blood biochemistry, coagulation function, microbiological culture results) and whether the antimicrobial therapy was adjusted according to the microbiological culture results were recorded in both groups. Univariate analysis was used to screen for differential indicators between severe and non-severe patients. After covariate analysis, multi-factor Logistic regression analysis was performed based on the Aakaike information criterion (AIC) forward stepwise regression method to rigorously search for risk factors for constructing the model. Based on the results of the multi-factor analysis, a nomogram prediction model was constructed, and the discriminatory degree and calibration degree of the model were assessed using the receiver operator characteristic curve (ROC curve) and calibration curve. RESULTS: A total of 172 adult CAP patients were included, 48 in severe group and 124 in non-severe group. The median age was 74 (57, 83) years old, onset time was 5.0 (3.0, 10.0) days, total number of comorbidities was 3 (2, 5), including 58 cases (33.7%) with hypertension and 17 (9.9%) with heart failure, 113 (65.7%) with CURB-65 score ≤ 1, 34 cases (19.8%) had a CURB-65 score = 2 and 25 cases (14.5%) had a CURB-65 score ≥ 3. Univariate analysis showed that there were statistically significant differences between the two groups in age, smoking history, CURB-65 score, heart rate, onset time, total comorbidity, pathogenic microorganisms, fibrinogen (FIB), D-dimer, C-reactive protein (CRP), procalcitonin (PCT), platelet-to-lymphocyte ratio (PLR), neutrophil-to-lymphocyte ratio (NLR), and alanine aminotransferase (ALT) and aspartate aminotransferase (AST). Multi-factor Logistic regression analysis showed that hypertension [odds ratio (OR) = 3.749, 95% confidence interval (95%CI) 1.411 to 9.962], heart failure (OR = 4.616, 95%CI was 1.116 to 19.093), co-infection (OR = 2.886, 95%CI was 1.073 to 7.760), history of smoking (OR = 8.268, 95%CI was 2.314 to 29.537), moderate to high CURB-65 score (OR = 4.833, 95%CI was 1.892 to 12.346), CRP (OR = 1.012, 95%CI was 1.002 to 1.022), AST (OR = 1.015, 95%CI was 1.001 to 1.030) were risk factors for severe CAP (all P < 0.05). The filtered indicators were included in the nomogram model, and the results showed that the area under the ROC curve (AUC) for the model to identify patients with severe adult CAP was 0.896, 95%CI was 0.840 to 0.937 (P < 0.05), and the calibration curve showed that the predicted probability of severe CAP was in good agreement with the observed probability (Hosmer-Lemeshow test: χ2 = 6.088, P = 0.665). CONCLUSIONS: The nomogram model has a good ability to identify patients with severe adult CAP and can be used as a comprehensive and reliable clinical diagnostic tool to provide a evidence for timely intervention in the treatment of adults with severe CAP.
 
成人市中肺炎重症化ノモグラム予測モデルの構築と検証
 
目的：成人の重症市中肺炎（CAP）における炎症指標、基礎疾患、病因、英国胸部学会修正肺炎スコア（CURB-65スコア）に基づくノモグラム予測モデルを構築し検証する。 方法：2018年1月から2021年12月までに台康仙林鼓楼病院で初めてCAPと診断された成人入院患者172人の臨床データを、重症度に応じて重症群と非重症群に分けた。ベースライン条件（性別、年齢、既往歴、併存疾患、家族歴を含む）、臨床データ（主症状、発症時間、CURB-65スコアを含む）、入院時の最初の検査結果（全血球数、肝機能、腎機能、血液生化学、凝固機能、微生物培養結果を含む）、微生物培養結果に応じて抗菌薬療法を調整したかどうかを両群で記録した。単変量解析を用いて、重症患者と非重症患者との間の差異指標をスクリーニングした。共変量解析の後、モデルを構築するためのリスク因子を厳密に探索するために、Aakaike情報量規準（AIC）前進ステップワイズ回帰法に基づいて多因子ロジスティック回帰分析を行った。多因子解析の結果に基づいてノモグラム予測モデルを構築し、受信者動作特性曲線（ROC曲線）と検量線を用いてモデルの識別度と検量度を評価した。 結果：合計172人の成人CAP患者が対象となり、重症群48人、非重症群124人であった。年齢中央値は74歳（57歳、83歳）、発症日数は5.0日（3.0日、10.0日）、合併症総数は3例（2例、5例）であり、その内訳は高血圧58例（33.7％）、心不全17例（9.9％）、CURB-65スコア≦1が113例（65.7％）、CURB-65スコア＝2が34例（19.8％）、CURB-65スコア≧3が25例（14.5％）であった。単変量解析の結果、年齢、喫煙歴、CURB-65スコア、心拍数、発症時間、合併症の総数、病原微生物、フィブリノゲン（FIB）、CURB-65スコア≧1、CURB-65スコア＝2、CURB-65スコア≧3において、両群間に統計学的有意差が認められた、フィブリノゲン（FIB）、Dダイマー、C反応性蛋白（CRP）、プロカルシトニン（PCT）、血小板対リンパ球比（PLR）、好中球対リンパ球比（NLR）、アラニンアミノトランスフェラーゼ（ALT）およびアスパラギン酸アミノトランスフェラーゼ（AST）。多因子ロジスティック回帰分析では、高血圧［オッズ比（OR）＝3.749、95％信頼区間（95％CI）1.411～9.962］、心不全（OR＝4.616、95％CIは1.116～19.093）、同時感染（OR＝2.886、95％CIは1.073～7.760）、喫煙歴（OR＝8.268、95％CIは2.314～29.537）、CURB-65スコアが中等度～高値（OR＝4.833、95％CIは1.892～12.346）、CRP（OR＝1.012、95％CIは1.002～1.022）、AST（OR＝1.015、95％CIは1.001～1.030）が重症CAPの危険因子であった（すべてP＜0.05）。フィルタリングした指標をノモグラムモデルに含めた結果、重症成人CAP患者を同定するモデルのROC曲線下面積（AUC）は0.896、95％CIは0.840～0.937（P＜0.05）であり、検量線は重症CAPの予測確率が観察確率とよく一致することを示した（Hosmer-Lemeshow検定：χ2＝6.088、P＝0.665）。 結論ノモグラムモデルは、重症成人CAP患者を同定する優れた能力を有し、重症CAPを有する成人の治療において時宜を得た介入を行うためのエビデンスを提供する包括的で信頼性の高い臨床診断ツールとして使用できる。臨床に応用できるといいね。