Table A5: GRADE Quality of Evidence*.
| Number of Studies |
Design | Study Quality |
Consistency | Directness | Imprecision | Other Modifying Factors | Overall Quality of Evidence |
|---|---|---|---|---|---|---|---|
| Outcome: Mortality | |||||||
| 6 | RCT | Very serious limitations† | Serious limitations‡ | No serious limitations | No serious limitations | n/a | Very Low |
| Outcome: Hospital Readmissions | |||||||
| 6 | RCT | Very serious limitations† | No serious limitations | No serious limitations | No serious limitations | n/a | Low |
| Outcome: Lung Function | |||||||
| 3 | RCT | Very serious limitations§ | No serious limitations | No serious limitations | Sparse data║ | n/a | Very Low |
| Outcome: HRQOL | |||||||
| 5 | RCT | Very serious limitations¶ | No serious limitations# | Serious limitations** | Sparse data†† | n/a | Very Low |
| Outcome: Mean Length of Stay | |||||||
| 5 | RCT | Very serious limitations‡‡ | Serious limitations§§ | No serious limitations | No serious limitations | n/a | Very Low |
| Outcome: Patient Satisfaction | |||||||
| 3 | RCT | Very serious limitations∥∥ | No serious limitations | No serious limitations | Sparse data¶¶ | n/a | Very Low |
| Outcome: Caregiver Satisfaction | |||||||
| 1 | RCT | Very serious limitation## | n/a | No serious limitations | Sparse data*** | n/a | Very Low |
| Outcome: Patient Preference | |||||||
| 1 | RCT | Very serious limitations## | n/a | Serious limitations††† | Sparse data*** | n/a | Very Low |
| Outcome: Caregiver Preference | |||||||
| 1 | RCT | Very serious limitations## | n/a | Serious limitations††† | Sparse data*** | n/a | Very Low |
Abbreviations: HRQOL, health-related quality of life; ITT, intention-to-treat; n/a, not applicable; RCT, randomized controlled trial.
Study quality was downgraded for the mortality and hospital readmission outcomes because of very serious limitations in many of the studies, including unknown or inadequate allocation concealment (3 of 6 studies); unclear randomization process based on published trials (2 of 6 studies); unclear whether assessor was blinded (single blind) (5 of 6 studies); lack of a priori power calculations (4 of 6 studies) and inadequately powered studies based on post-hoc sample size calculations (mortality: 6 of 6 studies; readmissions: 5 of 6 studies), withdrawals/dropouts > 20% (1 of 6 studies) or unknown (1 of 6 studies), and ITT analysis not used (unknown for 2 studies) or withdrawals/dropouts not considered in ITT analysis (3 of 4 studies).
Downgraded due to lack of consistency between the point estimates.
Study quality was downgraded for lung function outcomes because of very serious limitations in the studies including: unknown or inadequate allocation concealment (3 of 3 studies); unclear randomization process based on published information (2 of 3 studies); unknown whether assessor was blinded (single blind) based on published information (3 of 3 studies); lack of a priori power calculations (2 of 3 studies) and likely underpowered studies but not possible to calculate post-hoc sample size calculations based on information provided (3 of 3 studies), withdrawals/dropouts > 20% (1 of 3 studies), and ITT analysis not used (unknown for 2 studies) or withdrawals/dropouts not considered in ITT analysis (1 of 3 studies).
Downgraded due to sparse data, as the studies each reported different measures of lung function, so there was only 1 study per lung function outcome.
Study quality was downgraded for HRQOL because of very serious limitations in the studies including unknown or inadequate allocation concealment (3 of 5 studies); unclear randomization process based on published information (2 of 5 studies); unknown whether assessor was blinded (single blind) based on published information (4 of 5 studies); lack of a priori power calculations (3 of 5 studies) and likely underpowered studies but not possible to calculate post-hoc sample size calculations based on information provided (5 of 5 studies), withdrawals/dropouts > 20% (1 of 5 studies), and ITT analysis not used (unknown for 2 studies) or withdrawals/dropouts not considered in ITT analysis (3 of 5 studies).
While the GRADE for HRQOL was not downgraded due to inconsistent results because the majority of the studies (4 of 5) showed nonsignificant differences between the hospital-at-home and inpatient hospital groups, there was some inconsistency in the results as 1 study did find a significant difference between the groups for some HRQOL scales.
Downgraded because HRQOL was measured at 2- to 6-months follow-up rather than during the exacerbation itself, and this time point may be too late to observe a change in HRQOL associated with the intervention.
Downgraded due to sparse data because 4 of the 5 studies reported different measures of HRQOL, so there was only 1 study per HRQOL outcome.
Study quality was downgraded for length of stay because of very serious limitations in the studies, including unknown or inadequate allocation concealment (2 of 5 studies); unclear randomization process based on published information (2 of 5 studies); unknown whether assessor was blinded (single blind) based on published information (4 of 5 studies); lack of a priori power calculations (2 of 5 studies) and likely underpowered studies but not possible to calculate post-hoc sample size calculations based on information provided (5 of 5 studies), withdrawals/dropouts > 20% (1 of 5 studies), and ITT analysis not used (unknown for 2 studies) or withdrawals/dropouts not considered in ITT analysis (3 of 5 studies).
Downgraded due to lack of consistency across studies: 1 study reported similar length of stay between groups, 1 study a shorter length of stay in the hospital-at-home group, and 3 studies a longer length of stay in the hospital-at-home group. Some results were significantly different and some were not.
Study quality was downgraded for patient satisfaction with care because of very serious limitations in the studies including unknown or inadequate allocation concealment (2 of 3 studies); unclear randomization process based on published information (1 of 3 studies); unknown whether assessor was blinded (single blind) based on published information (2 of 3 studies); lack of a priori power calculations for this outcome and likely underpowered studies but not possible to calculate post-hoc sample size calculations based on information provided (3 of 3 studies), withdrawals/dropouts > 20% (1 of 3 studies), and ITT analysis not used (unknown for 2 studies) or withdrawals/dropouts not considered in ITT analysis (1 of 3 studies).
Downgraded due to sparse data as none of the studies use the same outcomes to measure satisfaction with care, so there was only 1 study for each outcome.
Study quality was downgraded for caregiver satisfaction and for patient and caregiver preference because of very serious limitations in the study including: unknown allocation concealment (1 of 1 study); unclear randomization process based on published information (1 of 1 study); unknown whether assessor was blinded (single blind) based on published information (1 of 1 study); lack of a priori power calculations for this outcome and likely underpowered but not possible to calculate post-hoc sample size calculations based on information provided (1 of 1 study), withdrawals/dropouts > 20% (1 of 1 study), and unknown whether intention-to-treat (ITT) analysis was used (1 of 1 study).
Downgraded due to sparse data as there was only 1 study that reported this outcome.
Patient and caregiver preference for hospital-at-home care was measured during the study after patients had begun their treatment either in hospital or in hospital-at-home care. Thus, patients and caregivers in the hospital-at-home group had experience with the program, whereas patients and caregivers in the inpatient hospital group did not. While this provides some information that suggests patients and caregivers become more comfortable and accepting of hospital-at-home care after they have experienced it (which may have policy implications), a comparison of patient/caregiver preferences for hospital-at-home care between the groups at baseline is needed and would be less biased.