Table 5

Summary of findings (GRADE)

Interventions to reduce the GHG emissions of healthcare compared with usual practice
Patient or population: any healthcare population
Setting: any healthcare setting
Intervention: any intervention initiated by clinicians or healthcare services primarily designed to reduce the GHG emissions of healthcare
Comparison: Usual practice, placebo or a different intervention designed to reduce the GHG emissions of healthcare
OutcomesEffects of interventions to reduce the GHG emissions of healthcare (number of studies, participants)Certainty of the evidence (GRADE)
Reduction in GHG emissions directly or indirectly measured (ie, from costs, waste and/or energy consumptionUncertain—the certainty of the evidence is very low (17 uncontrolled befor–after studies, 1 interrupted time series)
Environmental impact was measured in different ways across studies so data could not be pooled.
17 of 18 studies reported effect estimates favouring the intervention. 1 of the 18 studies was judged to be at low risk of bias and it favoured the intervention. The available effect estimates are presented in table 4.
⊕⊝⊝⊝
Very low*†‡
Financial costsUncertain—the certainty of the evidence is very low (14 uncontrolled before–after studies, 1 interrupted time series)
Financial costs were measured differently across studies so data could not be pooled.
13 of the 15 studies reported effect estimates favouring the intervention. None of these studies were judged to be at low risk of bias. The available effect estimates are presented in online supplemental file 7.
⊕⊝⊝⊝
Very low*†§
Effectiveness (as defined in individual studies according to aim, for example, waste reduction, volatile gas consumption, patient tests)Uncertain—the certainty of the evidence was very low (19 uncontrolled before–after studies, 1 interrupted time series).
Effectiveness was measured in different ways across studies so data could not be pooled.
18 of 20 studies reported effect estimates favouring the intervention. 1 of the 20 studies was judged to be at low risk of bias and favoured the intervention. The available effect estimates are presented in online supplemental file 9.
⊕⊝⊝⊝
Very low*‡
HarmsUncertain—the certainty of the evidence is very low (1 uncontrolled before–after study, 5695 participants)
McAlister et al29 found in-hospital mortality did not differ (OR 1.09 (95% CI 0.75 to 1.59)) before or after implementation of a policy to reduce non-urgent pathology testing to 2 days per week.
⊕⊝⊝⊝
Very low*¶
AcceptabilityUncertain—the certainty of the evidence is very low (1 uncontrolled before–after study, 22 staff members within a labour ward in 3 hospitals).
Most staff using nitrous oxide cracking equipment in Pinder et al31 reported the equipment was easy or very easy to set up (16/22, 73%); to explain to patients how to use (19/22, 86%); and to change the masks and filters between patients (19/22, 86%). However, only 54% of staff (12/22) were satisfied or very satisfied with the machine with concerns about the size of the equipment and some reporting problems with the technology (8/22, 36%).
⊕⊝⊝⊝
Very low*¶**
EngagementUncertain—no studies reported this outcome
Patient-relevant outcomesUncertain—no studies reported this outcome
  • *Observational study design starts at low certainty evidence.

  • †Downgraded one level due to serious indirectness; differences in outcome measures.

  • ‡Downgraded one level due to serious risk of bias; confounding adjustment issues, reporting bias (generalisability, inadequate follow-up and outcome).

  • §Downgraded one level due to serious risk of bias; confounding adjustment issues, reporting bias (generalisability and inadequate follow-up).

  • ¶Downgraded two levels due to very serious imprecision; single study.

  • **Downgraded one level due to serious risk of bias; confounding adjustment issues, reporting bias (generalisability, inadequate follow-up and outcome).

  • GHG, greenhouse gas; GRADE, Grades of Recommendation, Assessment, Development and Evaluation.