The Effect of Privatization of Inpatient Psychiatric Resources on Jail Inmate Population Jangho Yoon, PhD UC Berkeley School of Public Health June 30, 2009 AcademyHealth ARM Funding: NIMH Motivation o Point-in-time prevalence of severe mental disorders o 7-11% in jails or prisons o < 2% in the general population. o Why are jails and prisons over-populated with persons with severe mental disorders? o The effect of privatized inpatient psychiatry has rarely been a subject of scrutiny. Study Aim o Examine whether privatized inpatient psychiatric resources increases the size of jail populations o Measures of privatization: Market concentration oNonprofit vs. for-profit. Jail Population Growth o Nationwide trend in average daily jail inmates per 100,000 persons, 1985-98 220 200 180 Jail inmates/ 160 100K 140 120 100 85 86 87 88 89 90 91 92 93 Year Sources: Census of Jails; Annual Survey of Jails. 94 95 96 97 98 Inpatient Privatization o Nationwide changes in market concentration of public, nonprofit, and for-profit psychiatric beds, 1985-98 100% 90% 80% 70% Public 67% 45% 60% 50% 40% Nonprofit 30% 20% 23% 41% 10% 0% For-profit 10% 85 86 87 88 89 90 91 92 Year 14% 93 94 95 96 97 98 Source: AHA Survey of Hospitals. Link between Privatization and Jail Population Growth o Hospital ownership Different objectives and incentives Treatment of symptoms Jail incarceration. o For-profit: Treat profitable patients o Lead to the growth of jail populations. o Nonprofit: Value patient well-being but may avoid severe patients o Effect is not clear. Methods o Data o State panel data for 45 states and DC, 85-98 (n = 644) o Sources o o o o o o o o o Survey of Jails & Census of Jails American Hospital Association NRI American Medical Association Uniform Crime Reporting Program CJEE Extract US Census Bureau Bureau of Economic Analysis Bureau of Labor Statistics Methods o Empirical model ln(J)st = α ⋅ Rst + β ⋅ ∑ X st + Ss + Tt + εst o s: o t: State Year Methods o Empirical model ln(J)st = α ⋅ Rst + β ⋅ ∑ X st + Ss + Tt + εst o Dependent variable: ln(J) o ln(average daily jail inmates/100,000) o Sources: Annual Survey of Jails & Census of jails Methods o Empirical model ln(J)st = α ⋅ Rst + β ⋅ ∑ X st + Ss + Tt + εst o Main independent variables: R o Nonprofit market share = Nonprofit psychiatric beds Total psychiatric beds o For-profit market share = For-profit psychiatric beds Total psychiatric beds o Source: American Hospital Association Methods o Empirical model ln(J)st = α ⋅ Rst + β ⋅ ∑ X st + Ss + Tt + εst o Covariates (X) o Mental health/substance abuse (MH/SA) resources o Per-capita psychiatric beds o Per-capita community MH/SA expenditures o Per-capita SA treatment beds o Per-capita psychiatrists Methods o Empirical model ln(J)st = α ⋅ Rst + β ⋅ ∑ X st + Ss + Tt + εst o Covariates (X) o Mental health financing & welfare o Mental health insurance mandates (=1 if mandated) o % residents on Medicaid o % residents on TANF/AFDC Methods o Empirical model ln(J)st = α ⋅ Rst + β ⋅ ∑ X st + Ss + Tt + εst o Covariates (X) o Criminal justice policies and practices o Clearance rates (= total arrests/total reported crime) o Tough sentencing (= daily jail inmates/ total arrests) o Crime rates o Per-capita police o Concealed handgun laws Methods o Empirical model ln(J)st = α ⋅ Rst + β ⋅ ∑ X st + Ss + Tt + εst o Covariates (X) o Demographic and socio-economic factors o % male o % African-American o % metropolitan residents o Age categories o Poverty rate o Unemployment rate o Per-capita income Methods o Empirical model ln(J)st = α ⋅ Rst + β ⋅ ∑ X st + Ss + Tt + εst o S: o T: State fixed effects Year fixed effects Methods o Instrumental Variables (IV) o Instrumented: R (market share variable) o Instruments: Determinants of privatization o # of public psychiatric beds (1-year lagged) o % chain hospitals o # medical/surgical beds o Two-stage IV/GMM estimators o Adjusted for heteroskedasticity and AR(1) Results o Marginal effect of the market share of psychiatric beds on jail inmates by ownership status Key Explanatory Variable (Instrumented): Marginal Effect Model 1 Model 2 Nonprofit Share For-profit Share 1.8% 2.5%* Notes. IV/GMM estimators. Control for all covariates as well as fixed effects. * p < 0.05. Results o Robustness: Marginal effect of market share Panel A: Remove outliers • 5 most populous states • 5 states with greatest increases in jail population • 5 states with greatest increases in market share by ownership Panel B: Functional form • Unlogged dependent variable Notes. IV/GMM estimators. * p < 0.05; ** p < 0.01 Model 1 Nonprofit Model 2 For-profit 2.0%* 1.8%* 1.4% 2.7%* 7.6% 3.0%** 1.8 2.6* Conclusion o A greater for-profit market concentration leads to jail population growth o Magnitude is large: From 1985 to 1998, a 4 percentage-point increase in for-profit market share accounts for a 16% increase in the jail inmate population. o A greater nonprofit market share does not appear to increase jail inmate population. Policy Implications o The ownership mix of inpatient psychiatric resources is an important agenda item for mental health policy. o The collaboration between the mental health and jail systems should be continued and expanded. o Jail incarceration of mentally-ill persons as a indicator for mental health system performance. Thank You! E-mail: yoonjangho@gmail.com Appendix: Jail Population Growth o Potential determinants o Changes in demographics and socio-economics o Criminal justice practices o Policing effectiveness; tough sentencing; crime o Changes to the mental health system o Managed care1 o Insurance coverage (e.g., Medicaid)2 o Insufficient public mental health resources3 o Privatized inpatient psychiatry: NOT tested yet! 1 Domino et al. 2004; 2 Morrissey et al. 2007; 3 Palermo et al. 1991; Yoon 2007; Raphael and Stoll 2009. Appendix: Determinants of Privatization o Market niche for inpatient psychiatric services due to public inpatient reductions o Awareness of unmet need o Expansion of insurance coverage Demand o Reduced stigmatization of psychiatric treatment o Mental health practitioner growth o Financing: Medicare prospective payment and bed conversion o State regulations o Managed care, LRA, lawsuits (1990’s) References: Dowart and Schlesinger (1988); Geller (2006). Appendix: IV Tests (1) o Prospective instruments Tested instruments Valid? o CON regulations (1/0) No o % hospitals with formal contract with MCO No o # public psychiatric beds (1-year lagged) Yes o % chain hospitals Yes o # medical/surgical beds Yes Appendix: IV Tests (2) o Two conditions for valid IV o Condition 1: Non-zero correlation between the instrumented and instrumental variables (Strength of IVs) o Condition 2: Exclusion restrictions o Econometric tests employed o Condition 1 o Individual t test, individual LR test, joint F-test, & joint LR test o Test statistics should be significant. o Condition 2 o J test (jointly) & C test (individually) o Test statistics must be insignificant. Appendix: IV Tests (3) o First-stage results and IV tests for Cond. 1 Model 1 Model 2 Instrumented Nonprofit For-profit Public bedt-1 -.31** -.35*** Public bed2t-1 .0097** .0081** Medical/surgical bed n/a .0019* Chain hospital (%) .24* n/a R2 .90 .87 Note. Regressions include all covariates as well as fixed effects. * p < 0.05; ** p < 0.01; *** p < 0.001. Appendix: IV Tests (4) o IV tests of Conditions 1 & 2 Model 1 Model 2 Nonprofit For-profit F statistic 5.4*** 9.4*** Anderson LR statistic 15.8*** 42.8*** 1.1 1.7 Instrumented Joint significance (Cond. 1) Joint exclusion restrictions (Cond. 2) Hansen J statistic * p < 0.05; ** p < 0.01; *** p < 0.001. Appendix: IV Tests (5) o IV tests of Condition 2 (Exclusion Restrictions) Instrumented Model 1 Model 2 Nonprofit For-profit Individual exclusion restrictions ( Sargan C statistic) Public bedt-1 1.0 0.1 Public bed2t-1 0.5 0.2 Medical/surgical bed n/a 1.2 Chain hospital (%) 1.1 n/a * p < 0.05; ** p < 0.01; *** p < 0.001 Appendix: Main Results o Selected variables Model 1 Model 2 -0.005* -0.009*** Mandate -0.300*** -0.259*** Clearance rate 0.035*** 0.031*** Tough sentencing 0.259*** 0.239*** Handgun -0.192** -0.133* Male 0.284** 0.247** Metropolitan 0.032*** 0.031*** Age 20-24 0.112*** 0.148*** Psychiatric Bed * p < 0.05; ** p < 0.01; *** p < 0.001 Appendix: Empirical Model ln(J)st = α ⋅ Rst + β ⋅ ∑ X st + Ss + Tt + εst o o o o s: t: ln(J): R: o X: o S: o Y: State Year Natural logarithm of average daily jail inmates Market composition of nonprofit or for-profit psychiatric beds (instrumented) Covariates State fixed effects Year fixed effects