GRADE assessment of certainty of estimate of effect size and summary of findings
| Outcome | GRADE assessment of certainty* | Summary of findings | Absolute effect estimates | ||||||||||
| Risk of bias | Inconsistency | Indirectness | Imprecision | Publication bias | Certainty | Number of studies | HF/TAU participants | Effect estimate (95% CI) | Outcomes in control† | Outcomes in HF | Effect (95% CI) | ||
| Mental health | Self-rated mental health | Serious (−1) | None‡ | None | None | Unlikely | +++, moderate | 2 | 1359/1194 | SMD=0.07 (−0.19 to 0.33) | 14.4—gain in score (out of 100) across 24 months | 15.53—gain in score | +1.13 difference in mean change (−3.06 to +5.31) |
| Quality of life | Self-rated physical health | Serious (−1) | None | None | None | Unlikely | +++, moderate | 2 | 1359/1194 | SMD=0.00 (–-0.09 to 0.09) | 6.5—gain in score (out of 100) across 24 months | 6.5—gain in score (out of 100) across 24 months | no difference in mean change (−1.14 to +1.14) |
| Substance use | Problematic substance use | Serious (−1) | None | None | None | Unlikely | +++, moderate | 2 | 465/456 | Several outcomes analysed showing no meaningful difference in changes in problematic substance use between groups across 24 months | |||
| Health service use | Hospitalisations | Serious (−1) | None | None | None | Unlikely | +++, moderate | 2 | 516/519 | IRR=0.76 (0.70 to 0.83) | 1480 hospitalisations per 1000 person-years | 1125 hospitalisations per 1000 person-years | 355 fewer hospitalisations (252 to 444) |
| Time spent hospitalised | Serious (−1) | Serious (−1) | None | None | Unlikely | ++, low | 3 | 898/839 | SMD=−0.14 (−0.41 to 0.14) | 6379 days spent hospitalised per 1000 person-years | 6372 days spent hospitalised per 1000 person-years | 7 fewer days (7 more days to 20 fewer) | |
| Emergency department visits | Serious (−1) | None§ | None | Serious (−1)¶ | Unlikely | ++, low** | 2 | 1359/1194 | IRR=0.63 (0.48 to 0.82) | 2056 emergency department visits per 1000 person-years | 1295 emergency department visits per 1000 person-years | 761 fewer emergency department visits (370 to 1069) | |
| Housing stability | Attaining stable housing | Serious (−1) | None | None | Serious (−1)¶ | Unlikely | ++, low | 3 | 985/1000 | RR=2.46 (1.58 to 3.84) | 37 persons per 100 stably housed at 24 months | 75 persons per 100 stably housed at 24 months†† | 38 more persons stably housed (31 to 46)†† |
| Time spent stably housed | Serious (−1) | None | None | Serious (−1)¶ | Unlikely | ++, low | 2 | 1257/1116 | SMD=1.24 (0.86 to 1.62) | 235 days per person spent stably housed in 24 months | 497 days per person spent stably housed in 24 months | 261 more days stably housed (181 to 432) | |
*For all health outcomes, certainty was assessed as either certainty of direction of effect or certainty of null effect. For housing stability outcomes, assessment was made around effect size,42
†Control means calculated from all useable control data.
‡High heterogeneity was deemed to be explained by differences in effect by age group, therefore not downgraded.
§All effects in same direction with high confidence of non-null effect; rated ‘none’ despite high I2.
¶Score reduction of ‘−1’ here on the basis of relatively small concerns across both inconsistency and imprecision.
**Rated ‘low’ for certainty of effect direction; ‘very low’ for certainty of effect size.
↵††Absolute effect size recalculated from proportion of each group stably housed.
HF, Housing First; IRR, incidence rate ratio; RR, risk ratio; SMD, standardised mean difference; TAU, treatment as usual.