Our systematic review revealed that use of empirical therapy according to guidelines, de-escalation of therapy, switch from intravenous to oral therapy, therapeutic drug monitoring, use of a list of restricted antibiotics, and bedside consultation (especially for S aureus bacteraemia) can lead to significant benefits for clinical outcomes, adverse events, and costs, although the quality of evidence is generally low. Treatment according to guidelines and de-escalation of therapy had significant effects on mortality, although heterogeneity between studies was substantial. Most studies that assessed prescribing empirical therapy according to guidelines involved patients with community-acquired pneumonia, which makes it difficult to extrapolate the results to other infectious diseases. We assume, however, that effects would be similar where validated guidelines are available. Reduced mortality was also associated with switching from intravenous to oral therapy, therapeutic drug monitoring, use of a list of restricted antibiotics, and bedside consultation, but these effects were not significant. When patients with S aureus bacteraemia received bedside consultations, mortality was lower and diagnosis of complicated disease were better than those in patients who did not. A study on the effects of infectious disease consultations, published after our literature search was completed, confirms these results.
ガイドラインに則るエンピリック治療、de-escalation、IVから経口へのスイッチ、TDM、抗菌薬制限リスト、ベッドサイドのコンサルテーション（とくにS. aureus菌血症）が臨床アウトカムにとくに役に立つ。が、エビデンスの質は全体的に低かった。例えば、ガイドライン治療やde-escalationは死亡率を減らしたけど論文の異質性が大きかった。エンピリック治療の評価はだいたい市中肺炎（community acquired pneumonia)に関してで、他の感染症に適用できるか分からない。けど、まあ効果は似たようなものだろう。他にも死亡率を減らす介入はいくつかあった。ってなことが書いてあります」
The study also has limitations. We noted substantial heterogeneity between studies in relation to settings, methods, and reported outcomes, and the quality of evidence was generally low. These features make synthesis and interpretation of results difficult. Nevertheless, sensitivity analyses of the pooled mortality rates did not alter the findings. We did not do a grey literature search and we restricted our searches to Embase, Ovid MEDLINE, and PubMed, which introduces an inherent degree of publication bias. Also, as in any review, we might have missed some relevant studies. We chose to report only aggregated data when available. We did so to keep some overview of the results without being overwhelmed by data. We only report data on adults and inpatients because many objectives, such as de-escalation or switching from intravenous to oral therapy, are not applicable in the outpatient setting. We found only three studies that reported effects on rates of infection with Clostridium difficile, one each in three different searches. The final limitation is that we report data on studies published between 1979 and 2014. Treatment (eg, antibiotics used) and the environment (eg resistance rates) have changed substantially over this period and, therefore, not all the results of the included studies will be applicable to the current situation.
まずheterogeneityが問題。エビデンスの質がいまひとつ。grey literature searchをしてないって書いてありますね。これは伝統的でないアカデミックな論文って感じの意味だそうです（https://en.wikipedia.org/wiki/Grey_literature）。政府の文書とか報告書の類。こういうのはWikipediaが役に立ちますね。あとは論文見逃しのリスクなど、メタ分析での定型的なlimitaitionsなどが書いてありますね。ま、そのへんは飛ばしてもよい」