Skip to main content
. Author manuscript; available in PMC: 2021 Feb 17.
Published in final edited form as: J Psychiatr Pract. 2021 Jan 21;27(1):33–42. doi: 10.1097/PRA.0000000000000524

Table 2.

Aggregated effects of identified discontinuities in psychiatric bed numbers for psychiatric admissions to general hospitals and jail inmate populations in U.S. localities.

HRR Psych Bed Psych Admits (Standard Dev) Psych Admits at Non-Psych Hospitals (Std Dev)
N = 3,030 N = 3,249 N = 3,249
Bed Reduction −90.2 [−121.9, −58.5] ** −0.079 [−0.138, −0.020] ** 0.160 [−0.020, 0.340]
Control Index −42.7 [−113.7, 28.3] 0.953 [0.751, 1.155] ** 0.892 [−0.342, 2.126]
Log(HRR Population) −143.4 [−253.7, −33.2] ** 0.422 [0.034, 0.809] ** −0.069 [−0.752, 0.613]
Bed Addition 46.5 [11.4, 81.7] ** −0.034 [−0.116, 0.047] 0.336 [−0.133, 0.804]
Control Index −44.4 [−124.0, 35.2] 0.954 [0.755, 1.154] ** 0.880 [−0.318, 2.079]
Log(HRR Population) −142.4 [−250.5, −34.4] ** 0.425 [0.034, 0.815] −0.042 [−0.738, 0.653]
Inmate (Standard Dev) Inmate Number
N = 2,955 N = 2,955
Bed Reduction 0.054 [−0.008, 0.117] * 256.2 [3.3, 509.1] **
Control Index 0.099 [−0.029, 0.227] 40.6 [−297.0, 378.2]
Log(HRR Population) 0.282 [−0.032, 0.596] * 2495.7 [942.8, 4048.6] **
Bed Addition −0.024 [−0.097, 0.048] −199.3 [−457.4, 58.8]
Control Index 0.099 [−0.026, 0.225] 45.5 [−293.9, 385.0]
Log(HRR Population) 0.276 [−0.038, 0.590] * 2457 [917.4, 3996.7] **

The table above shows the results of ten regressions summarizing the difference-in-difference effects for 178 HRRs experiencing psychiatric bed additions or reductions from 1988–2011. The outcomes of the regressions are given by the headers with quadrant numbers (1) representing the number of psychiatric beds, (2) the effect psychiatric admissions observed from hospital discharge data given in standard deviations, (3) the effect for inmate number given in standard deviations and (4) the raw effect on inmate number. Mean coefficient effects are given along with 95% confidence intervals given in square brackets.

**

indicates Wald statistic significant at level p<0.05,

*

indicates Wald statistic significant at level p<0.1. All regressions include year and HRR-fixed effects. All standard errors clustered at the HRR level.