Skip to main content
. 2021 Mar 26;6(5):e309–e323. doi: 10.1016/S2468-2667(21)00013-X

Table 2.

Sensitivity analysis of the effect of recent homelessness or unstable housing on risk of HIV or HCV acquisition in PWID compared with PWID who are not homeless or are stably housed

Number of estimates Effect size (95% CI) p value I2 pheterogeneity
Effect on risk of HIV acquisition
Unadjusted effect estimates 17 1·55 (1·23–1·95) 0·0002 62·7 0·0003
Adjusted effect estimates 9 1·39 (1·06–1·84) 0·019 65·5 0·0031
Published unadjusted estimates 5 1·65 (1·11–2·44) 0·012 77·3 0·0015
Unpublished unadjusted estimates* 12 1·46 (1·12–1·91) 0·0048 42·7 0·058
Only longitudinal studies 15 1·62 (1·27–2·07) 0·0001 63·4 0·0005
Only studies with hazard ratios 10 1·56 (1·08–2·25) 0·018 73·5 <0·0001
Only studies with at least 90% of participants injecting recently 11 1·44 (0·97–2·13) 0·068 75·4 0·0001
Only studies at low-to-moderate risk of bias 10 1·56 (1·14–2·12) 0·0052 72·6 0·0001
Estimate for homelessness 12 1·44 (1·13–1·83) 0·0036 52·7 0·016
Estimate for unstable housing 5 1·82 (1·13–2·95) 0·014 67·8 0·014
Effect on risk of HCV acquisition
Unadjusted effect estimates 28 1·65 (1·44–1·89) <0·0001 44·8 0·0060
Adjusted effect estimates 14 1·64 (1·43–1·89) <0·0001 9·6 0·35
Published unadjusted estimates 11 1·61 (1·18–2·19) 0·0029 66·9 0·0008
Unpublished unadjusted estimates* 17 1·69 (1·49–1·92) <0·0001 14·2 0·29
Only longitudinal studies 25 1·61 (1·40–1·86) <0·0001 38·6 0·027
Only studies with hazard ratios 18 1·59 (1·36–1·86) <0·0001 45·6 0·019
Only studies with at least 90% of participants injecting recently 20 1·54 (1·32–1·80) <0·0001 39·0 0·039
Only studies at low-to-moderate risk of bias 19 1·67 (1·37–2·03) <0·0001 52·4 0·0042
Estimate for homelessness 19 1·66 (1·37–2·00) <0·0001 55·3 0·0018
Estimate for unstable housing 10 1·72 (1·48–1·99) <0·0001 0·0 0·47

HCV=hepatitis C virus. PWID=people who inject drugs.

*

Unpublished estimates are those that have been calculated from raw data for the current study and have not been presented in previously published papers.

One study (unpublished estimate, provided by Wijnand and colleagues) provided estimates for both unstable housing and homelessness. For all sensitivity analyses, the effect of unstable housing from this study was used, except when pooling estimates specifically for homelessness or unstable housing. Hence, the number of studies for these two sensitivity analyses add up to 29.