Table 2 Impact of modelled interventions on proportion of population experiencing violence.

Baseline

–

77.3% (66.5–83.3%)

–

Remove homelessness entirely

β_I = 0

54.0% (36.3–67.5%)

29.8% (14.5–50.2%)

Cease all police displacement

η_I = 0

43.7% (16.3–61.8%)

42.7% (23.2–77.4%)

Remove homelessness and cease all police displacement

β_I = 0; η_I = 0

25.2% (13.2–37.7%)

67.3% (53.1–81.1%)

Additional housing rate as seen in housing first²⁴

α_I = α_B + 1.8

73.5% (59.9–80.5%)

5.1% (2.14–11.4%)

Reduce police displacement rate by 39%

η_I = 0.61η_B

74.2% (63.0–80.9%)

3.9% (2.35–6.92%)

Additional housing and reduce policing together

α_I = α_B + 1.8; η_I = 0.61η_B

69.6% (55.5–77.1%)

10.2% (5.91–19.6%)

Remove difference in policing between homeless and not homeless

θ_I = 1

63.1% (43.3–74.6%)

17.9% (8.0–38.5%)

Being housed allows 5% to transition to indoor work and reduces violence

See supplementary methods; \(\pi\) and \(\kappa\)

75.2% (65.1–81.1%)

2.7% (2.03–3.33%)

Being housed allows 50% to transition to indoor work and reduces violence

See supplementary methods; \(\pi\) and \(\kappa\)

56.9% (51.5–59.9%)

26.5% (22.6–28.2%)

Being housed allows 75% to transition to indoor work and reduces violence

See supplementary methods; \(\pi\) and \(\kappa\)

46.5% (43.9–48.1%)

39.8% (34.2–42.4%)

Increase rate of housing as seen during COVID-19 lockdown

α_I = 12

66.8% (51.2–75.8%)

13.5% (6.91–25.2%)

Increase rate of policing as seen in COVID-19 lockdown

η_I = 3.2η_B

81.9% (71.6–87.4%)

Increase by 5.7% (3.55–9.94%)

Both COVID-19 changes together

α_I = 12; η_I = 3.2η_B

76.8% (63.6–83.8%)

0.7% decrease (8.7% decrease–4.1% increase )