Table 3.
Risk | Quality of study limitations: ↓ | Indirectness of evidence: ↓ | Inconsistency: ↓ | Imprecision, range CI effect size >2.0: ↓ | Publication bias, yes or unclear: ↓ | Effect estimate >2.0: ↑ >5.0: ↑↑ |
Dose–response effect: ↑ | Residual confounding: ↑ | Overall certainty (high, moderate, low) |
In-hospital mortality | No (−)* | No (−) | No (−)† | No (−) | Unclear ↓ | Yes ↑↑‡ | Yes ↑ | No (−) | High |
Case mortality | No (−)* | No (−) | No (−)§ | No (−) | Unclear ↓ | Yes ↑↑¶ | Yes ↑ | No (−) | High |
Hospitalisation | No (−)* | No (−) | Yes ↓** | No (−) | Unclear ↓ | Yes ↑↑†† | Yes ↑ | No (−) | High |
*High-quality studies also reported a significant association between age and risk of outcome.
†Moderate heterogeneity (I2=64.1%) was observed on studies using age as a continuous variable and on high-quality studies (I2=58.7%).
‡Although per age RR is 1.057, when comparing different adult groups can have an effect size of greater than 5.0 (ie, a 54-year difference, 18 years vs 72 years results in a RR of 20.0).
§Moderate heterogeneity (I2=68.7%) was observed on studies using age as a continuous variable and on high-quality studies heterogeneity was very low (I2=0%).
¶Although per age RR is 1.074, when comparing different adult groups can have an effect size of greater than 5.0 (ie, a 54-year difference, 18 years vs 72 years results in a RR of 47.2).
**High heterogeneity observed (I2 >85%).
††Although per age RR is 1.034, when comparing different adult groups can have an effect size of greater than 5.0 (ie, a 54-year difference, 18 years vs 72 years results in a RR of 6.08).