Table 5.
Association Between Neighborhood Characteristics at Birth and Psychotic Experiences at Age 18 years, Based on Full Sample With Complete Exposure and Imputed Outcome and Covariates
| Exposures | Total Na N = 3972 |
N with exposure and outcome n (%) |
P (χ2) [P, χ2 for trend] | Psychotic Experiences at Age 18 (N = 11 879) | ||
|---|---|---|---|---|---|---|
| Crude Model OR (95% CI)b |
Adjusted Model 1 OR (95% CI)c |
Adjusted Model 2 OR (95% CI)d |
||||
| Population densitye | ||||||
| 1 (least densely populated) | 1397 | 85 (6.1) | .001 [.0001] | Ref | Ref | Ref |
| 2 | 1238 | 98 (7.9) | 1.31 (0.97–1.78) | 1.24 (0.91–1.70) | 1.25 (0.91–1.72) | |
| 3 (most densely populated) | 1337 | 134 (10.0) | 1.77 (1.33–2.35)*** | 1.56 (1.16–2.09)** | 1.57 (1.14–2.17)** | |
| Neighborhood deprivation | ||||||
| 1 (least deprived) | 1493 | 97 (6.5) | <.0001 [.0001] | Ref | Ref | Ref |
| 2 | 1413 | 105 (7.4) | 1.13 (0.85–1.51) | 1.01 (0.75–1.35) | 0.89 (0.65–1.21) | |
| 3 (most deprived) | 1066 | 115 (10.8) | 1.79 (1.33–2.39)*** | 1.20 (0.87–1.67) | 0.98 (0.68–1.42) | |
| Inequality | ||||||
| 1 (least inequality) | 1370 | 110 (8.0) | .91 [.78] | Ref | Ref | Ref |
| 2 | 1297 | 106 (8.2) | 1.01 (0.79–1.29) | 1.03 (0.80–1.32) | 1.06 (0.82–1.36) | |
| 3 (most inequality) | 1305 | 101 (7.7) | 0.97 (0.75–1.26) | 1.03 (0.79–1.36) | 1.07 (0.80–1.43) | |
| Social fragmentatione | ||||||
| 1 (least fragmented) | 1357 | 85 (6.3) | .01 [.01] | Ref | Ref | Ref |
| 2 | 1341 | 118 (8.8) | 1.45 (1.07–1.97)* | 1.31 (0.96–1.79) | 1.32 (0.95–1.83) | |
| 3 (most fragmented) | 1274 | 114 (9.0) | 1.47 (1.11–1.96)** | 1.26 (0.92–1.72) | 1.19 (0.85–1.67) |
Note: aN Refers to participants with complete exposure who also have outcome data.
bCrude model.
cAdjusted for: child’s ethnicity; maternal age, education, marital status, social class, and depression.
dAdjusted for all variables in adjusted model 1 + all exposures (population density, deprivation, inequality, and social fragmentation) adjusted for each other.
eIn model 2, there was evidence that population density provided a better fit to the data when modeled as a continuous categorical variable (OR per tertile: 1.30, 95% CI: 1.10–1.55, P = .003). To test this, we compared this model to a more complex model fitted with the categorical term, via Likelihood Ratio Test (P = .60) in complete case analyses because LRT cannot be computed in MI models with cluster robust standard error.
*P ≤ .05, **P ≤ .01, ***P ≤ .0001.