Table 4.
Sub-hazard ratio estimates for unnatural-cause mortality in people with severe mental illness
| Total sample (n=18 201) | Unnatural-cause deaths (n=192) |
Model 1 |
Model 2 |
||||
|---|---|---|---|---|---|---|---|
| Sub-hazard ratios (95% CI) | p value | Sub-hazard ratios (95% CI) | p value | ||||
| Area level indicators | |||||||
| Index of Multiple Deprivation (per increase in fifths; from less to more deprived) | .. | .. | 0·88 (0·80–0·98) | 0·02* | 0·94 (0·83–1·06) | 0·32* | |
| Urbanicity (per increase in fifths; from less to more urban) | .. | .. | 0·87 (0·79–0·97) | 0·01* | 0·94 (0·83–1·07) | 0·33* | |
| Social fragmentation (per unit increase; from less to more fragmented) | .. | .. | 0·98 (0·87–1·10) | 0·70* | 1·02 (0·90–1·15) | 0·80* | |
| Individual-level indicators | |||||||
| Gender | |||||||
| Male | 9610 (52·8%) | 132 (68·8%) | Ref | Ref | |||
| Female | 8591 (47·2%) | 60 (31·3%) | 0·54 (0·40–0·73) | <0·0001 | 0·61 (0·44–0·84) | 0·0026 | |
| Diagnosis | |||||||
| Non-affective | 13 160 (72·3%) | 141 (73·4%) | Ref | Ref | |||
| Affective | 5041 (27·7%) | 51 (26·6%) | 1·05 (0·76–1·45) | 0·75 | 1·01 (0·73–1·41) | 0·95 | |
| Marital status | |||||||
| Married or cohabiting | 2781 (15·3%) | 22 (11·5%) | Ref | Ref | |||
| Divorced, separated, widowed, or single | 15 420 (84·7%) | 170 (88·5%) | 1·29 (0·83–2·00) | 0·26 | 1·08 (0·69–1·68) | 0·75 | |
| Substance use disorder | |||||||
| No substance use disorder | 15 046 (82·7%) | 128 (66·7%) | Ref | Ref | |||
| Life-time substance use disorder | 3155 (17·3%) | 64 (33·3%) | 2·36 (1·74–3·20) | <0·0001 | 2·07 (1·52–2·81) | <0·0001 | |
| Interaction of ethnicity with areal ethnic density | |||||||
| Lowest ethnic density area (0% ethnic minorities) | |||||||
| White British | .. | .. | Ref | Ref | |||
| Ethnic minorities | .. | .. | 0·57 (0·24–1·39) | 0·22 | 0·52 (0·21–1·26) | 0·15 | |
| Highest ethnic density area (95% ethnic minorities) | |||||||
| White British | .. | .. | Ref | Ref | |||
| Ethnic minorities | .. | .. | 0·47 (0·17–1·30) | 0·15 | 0·51 (0·19–1·40) | 0·19 | |
| p value between ethnicity and ethnic density interaction | .. | .. | .. | 0·84 | .. | 0·99 | |
Data are n (%), unless otherwise specified. Competing risks regression models with robust standard errors to adjust for clustering at Lower Super Output Area level were used to estimate sub-hazard ratios. Model 1 is crude estimates. Model 2 is adjusted for age, an interaction between area level own ethnic density and ethnicity, and all other displayed variables. p values are from Wald tests.
*
p value for linear trend.