Rapid and accurate influenza diagnostics can improve patient care.

To summarize and compare accuracy of traditional rapid influenza diagnostic tests (RIDTs), digital immunoassays (DIAs), and rapid nucleic acid amplification tests (NAATs) in children and adults with suspected influenza.

6 databases from their inception through May 2017.

Studies in English, French, or Spanish comparing commercialized rapid tests (that is, providing results in <30 minutes) with reverse transcriptase polymerase chain reaction reference standard for influenza diagnosis.

Data were extracted using a standardized form; quality was assessed using QUADAS-2 (Quality Assessment of Diagnostic Accuracy Studies 2) criteria.

162 studies were included (130 of RIDTs, 19 of DIAs, and 13 of NAATs). Pooled sensitivities for detecting influenza A from Bayesian bivariate random-effects models were 54.4% (95% credible interval [CrI], 48.9% to 59.8%) for RIDTs, 80.0% (CrI, 73.4% to 85.6%) for DIAs, and 91.6% (CrI, 84.9% to 95.9%) for NAATs. Those for detecting influenza B were 53.2% (CrI, 41.7% to 64.4%) for RIDTs, 76.8% (CrI, 65.4% to 85.4%) for DIAs, and 95.4% (CrI, 87.3% to 98.7%) for NAATs. Pooled specificities were uniformly high (>98%). Forty-six influenza A and 24 influenza B studies presented pediatric-specific data; 35 influenza A and 16 influenza B studies presented adult-specific data. Pooled sensitivities were higher in children by 12.1 to 31.8 percentage points, except for influenza A by rapid NAATs (2.7 percentage points). Pooled sensitivities favored industry-sponsored studies by 6.2 to 34.0 percentage points. Incomplete reporting frequently led to unclear risk of bias.

Underreporting of clinical variables limited exploration of heterogeneity. Few NAAT studies reported adult-specific data, and none evaluated point-of-care testing. Many studies had unclear risk of bias.

Novel DIAs and rapid NAATs had markedly higher sensitivities for influenza A and B in both children and adults than did traditional RIDTs, with equally high specificities.

Québec Health Research Fund and BD Diagnostic Systems.

快速准确的流感诊断可以改善患者护理。

总结和比较传统快速流感诊断测试 (RIDT)、数字免疫分析 (DIA) 和快速核酸扩增测试 (NAAT) 在疑似流感儿童和成人中的准确性。

从成立到 2017 年 5 月的 6 个数据库。

用英语、法语或西班牙语比较商业化快速检测（即在 30 分钟内提供结果）与逆转录酶聚合酶链反应参考标准进行流感诊断的研究。

使用标准化表格提取数据；使用 QUADAS-2（诊断准确性研究的质量评估 2）标准评估质量。

纳入了 162 项研究（130 项 RIDT、19 项 DIA 和 13 项 NAAT）。从贝叶斯双变量随机效应模型中检测甲型流感的汇总敏感性为 RIDT 为 54.4%（95% 可信区间 [CrI]，48.9% 至 59.8%），DIA 为 80.0%（CrI，73.4% 至 85.6%）和 91.6 NAAT 的百分比（CrI，84.9% 至 95.9%）。检测 B 型流感的 RIDT 为 53.2%（CrI，41.7% 至 64.4%），DIA 为 76.8%（CrI，65.4% 至 85.4%），NAAT 为 95.4%（CrI，87.3% 至 98.7%）。合并的特异性均较高（>98%）。46 项甲型流感和 24 项乙型流感研究提供了儿科特异性数据；35 项甲型流感和 16 项乙型流感研究提供了针对成人的数据。除了快速 NAAT 的 A 型流感（2.7 个百分点）外，儿童的汇总敏感性高 12.1 至 31.8 个百分点。汇总的敏感性有利于行业赞助的研究 6.2 至 34.0 个百分点。不完整的报告经常导致不明确的偏倚风险。

临床变量的少报限制了对异质性的探索。很少有 NAAT 研究报告了针对成人的数据，也没有一项评估床旁检测。许多研究存在不明确的偏倚风险。

新型 DIA 和快速 NAAT 在儿童和成人中对甲型和乙型流感的敏感性明显高于传统 RIDT，具有同样高的特异性。

魁北克健康研究基金和 BD 诊断系统。