Consumer Directed Health Plans:
New evidence on cost and utilization
iHEA Conference, Barcelona, SpainJuly, 2005
Roger Feldman, Stephen T. Parente, and Jon B. Christianson
Sponsored by the Robert Wood Johnson Foundation’sHealth Care Financing & Organization Initiative (HCFO)
Presentation Overview
Past Results New questions Methods and results from new
analysis Conclusions
2004 Study Design Reported in August 2004 Health Services
Research* A large employer that offered CDHP in 2001,
alongside existing POS and PPO plans Employees who worked for firm from 2000-02 3 cohorts:
Joined CDHP in 2001 and stayed in 2002 Always in POS 2000-02 Always in PPO 2000-02
Control for several factors to adjust cost & use estimates
*ST Parente, R Feldman, JB Christianson. “Evaluation of the Effect of a Consumer Driven Health Plan on Medical Care Expenditures and Utilization,” Health Services Research, 39: 4, Part II, pp. 1189-1209, August 2004.
What was the adjusted impact on total expenditure?
NOTES: (1) These are results from a restricted continuously enrolled sample of 50% to 60% of the total employee population and are not a reflection of the plans’ full PMPM expenditures. (2) Patient expenditures from the Personal Care Account (PCA) are included in the employer payment category. ( 3) Employee payment reflects deductibles, copayments, and coinsurance expenses.
Health Plan Cohorts Mean Mean Mean
CDHP Cohort N=531Total Expenditure 4,396.22$ 6,154.36$ 8,149.26$
POS Cohort N=1,551Total Expenditure 5,284.53$ 6,773.62$ 7,197.50$
PPO Cohort N=1,554Total Expenditure 5,228.42$ 7,050.59$ 8,377.78$
Regressions adjusted by annual trends, health plan choice, age, gender, income, illness burden, health shock, number of covered lives in contract, and use of healthcare flexible spending account. The estimates are based on a two-part model.
2000 2001 2002
Did the adjusted impact on total expenditure differ by type of service?
NOTE: These are results from a restricted continuously enrolled sample of 50% to 60% of the total employee population and are not a reflection of the plans’ full PMPM expenditures.
Health Plan Cohorts MeanDeviationMean Mean
CDHP Cohort N=531Hospital Expenditure 1,369.97$ 1,999.25$ 3,468.53$ Physician Expenditure 2,093.70$ 2,935.84$ 3,510.83$ Pharmacy Expenditure 935.29$ 1,103.72$ 1,341.78$
POS Cohort N=1,551Hospital Expenditure 1,842.80$ 1,796.37$ 1,956.83$ Physician Expenditure 2,381.08$ 2,959.90$ 3,088.22$ Pharmacy Expenditure 1,107.64$ 1,498.54$ 1,640.25$
PPO Cohort N=1,554Hospital Expenditure 1,779.06$ 2,049.76$ 2,367.17$ Physician Expenditure 2,245.22$ 2,834.32$ 3,294.47$ Pharmacy Expenditure 1,007.95$ 1,484.91$ 1,789.26$
Year 2000 Year 2001 Year 2002
What was the adjusted impact on use of different services?
NOTE: These are results from a restricted continuously enrolled sample of 50% to 60% of the total employee population and are not a reflection of the plans’ full admissions and prescription drug experience.
Health Plan Cohorts Mean Mean Mean
CDHP Cohort N=531Hospital Admission Rate 0.05 0.10 0.16Physician Visits 5.74 7.49 7.15Prescriptions Filled 18.89 22.23 25.25
POS Cohort N=1,551Hospital Admission Rate 0.07 0.06 0.09Physician Visits 6.75 7.56 7.29Prescriptions Filled 22.23 22.59 30.89
PPO Cohort N=1,554Hospital Admission Rate 0.07 0.07 0.11Physician Visits 5.78 6.54 6.95Prescriptions Filled 20.63 23.79 24.50
2000 2001 2002
Summary of 2001-2002 CDHP Results
CDHP cohort began with favorable selection but experienced highest rate of growth in total $
CDHP cost and use trends were highest for hospital care: hospital expenditures and admission rate were dramatically higher by 2002
No dramatic differences in use physician services Relatively better control of Rx spending in CDHP
Cohort Design for New Study
Experimental Cohort #1 (“long” cohort): Worked for firm from 2000-03 Joined CDHP in 2001 and stayed through 2003 Control groups: worked for firm from 2000-03
and stayed in PPO or POS the entire time Experimental Cohorts #2 &3 (“short”
cohorts): Another plan in 2001, CDHP in 2002 Another plan in 2002, CDHP in 2003 Control groups for short cohorts: enrolled in
PPO or POS for matching two-year period
Research questions
Will another year of data show that the CDHP can control trends in total cost and use?
Did CDHP enrollees use more resource-intensive hospital care?
Do we observe inappropriate substitution away from Rx in the CDHP, leading to higher hospital spending, more inpatient admissions, and more ER visits?
Methods: Research Question #1
Prob ($it > 0) = b0 + b1 Cohorti + b2 Time +
b3 Time x Cohorti + b4 Xit + eit
Ln ($it| $it > 0) = f (same variables) $ = annual expenditure per contractCohort = indicators for CDHP and PPO, relative to omitted POSTime = indicators for 2001, 2002 and 2003, relative to 2000X = contract-holder age, sex, and income, contract-level illness burden, health shock, number of covered lives, and use of FSAb3 coefficients indicate difference-in-difference effects
Research Question #1, continued
Estimate same model for short cohorts that joined CDHP in 2002 or 2003
Strongest evidence in favor of pent-up demand would be stable or declining spending for experimental cohort #1 compared with POS, but high spending for short cohorts #2 and #3
We cannot rule out permanent moral hazard if all CDHP cohorts experience high cost & use
Research Question #2
Ln($/ADMit|ADMit > 0) = b0 + b1 Cohorti + b2 Time + b3 Time x Cohorti + b4 Xit +
b5 IBit + eit
IB = illness burden measured annually to control for factors related to hospitalization
Research Question #3
Yit = b0 + b1 Cohorti + b2 Time + b3 Time x Cohorti + b4 Xit + b5 RXit-1 + eit
Y = hospital or emergency room use or $ (in 2 parts)
RX = lagged prescription drug expenditures
Table 1. Impact of CDHP and PPO on Cost
NOTE: These are results from a restricted continuously enrolled sample of 28% of the total employee population and are not a reflection of the plans’ full PMPM expenditures. Bolded numbers are significant at p<.05.
Annual Plan Effects Compared With POS Plan
2000 Model 2001 2002Health Plan Cohorts Mean Stage Plan Effects Plan Effects Plan Effects
CDHP Cohort N=429Total Expenditure 4,037$ Probit -0.111 0.269 0.187
GLM 4.1% 26.0% 22.6%
Employer Expenditure 3,627$ Probit -0.077 0.222 0.222
GLM 11.2% 38.9% 30.1%
Employee Expenditure 410$ Probit -2.083 -2.149 -1.723
GLM 43.0% 36.8% 39.6%
PPO Cohort N=1,025Total Expenditure 4,661$ Probit -0.082 -0.087 -0.161
GLM 8.3% 16.8% 9.5%
Employer Expenditure 4,172$ Probit -0.104 -0.127 -0.183
GLM 11.2% 20.0% 12.4%
Employee Expenditure 490$ Probit -0.041 -0.003 -0.060
GLM -7.0% -3.3% -9.6%
Regressions control for annual trends, health plan choice, age, gender, illness burden, health shock,income, number of covered lives in contract, and use of healthcare flexible spending account
2003
Table 2. Impact of CDHP and PPO on Physician, Hospital, and Pharmacy Cost
NOTE: These are results from a restricted continuously enrolled sample of 28% of the total employee population and are not a reflection of the plans’ full PMPM expenditures. Bolded numbers are significant at p<.05.
Annual Plan Effects Compared With POS Plan
2000 Model 2001 2002Health Plan Cohorts Mean Stage Plan Effects Plan Effects Plan Effects
CDHP Cohort N=429Hospital Expenditure 1,332$ Probit -0.109 0.069 -0.050
GLM 60.8% 119.7% 75.5%
Physician Expenditure 1,891$ Probit -0.089 0.311 -0.007
GLM 10.7% 20.2% 25.1%
Pharmacy Expenditure 814$ Probit -0.086 -0.061 0.256
GLM -14.7% -5.1% -3.9%
PPO Cohort N=1,025Hospital Expenditure 1,669$ Probit 0.109 0.106 -0.091
GLM 23.8% 24.4% 29.0%
Physician Expenditure 1,958$ Probit -0.105 -0.055 -0.174
GLM 5.9% 10.2% 6.7%
Pharmacy Expenditure 1,034$ Probit -0.029 -0.096 0.047
GLM 8.3% 22.7% 9.9%
Regressions control for annual trends, health plan choice, age, gender, illness burden, health shock,income, number of covered lives in contract, and use of healthcare flexible spending account
2003
Table 3. Impact of CDHP and PPO on Visits, Hospital Admissions, and Scripts
NOTE: These are results from a restricted continuously enrolled sample of 28% of the total employee population and are not a reflection of the plans’ total expenditures. Bolded numbers are significant at p<.05.
Annual Plan Effects Compared With POS Plan
2000 Model 2001 2002Health Plan Cohorts Mean Stage Plan Effects Plan Effects Plan Effects
CDHP Cohort N=429Physician Visits 5.84 Probit -0.565 -0.389 0.167
GLM -15.3% -17.9% 17.3%
Hospital Admission Rate 0.11 Probit 0.410 0.206 0.457
GLM 8.8% -0.2% 6.2%
Prescriptions Filled 17.19 Probit -0.114 0.051 0.099
GLM 0.7% -18.5% 18.7%
PPO Cohort N=1,025Physician Visits 5.95 Probit -0.087 0.049 0.047
GLM 0.7% 4.1% -6.2%
Hospital Admission Rate 0.16 Probit 0.068 0.066 -0.023
GLM -8.5% -4.3% -8.2%
Prescriptions Filled 21.98 Probit -0.009 -0.009 -0.009
GLM -0.9% -24.9% -0.7%
Regressions control for annual trends, health plan choice,age, gender, illness burden, health shock,
income, number of covered lives in contract, and use of healthcare flexible spending account
2003
Table 4. Short (1 year) Impact of CDHP and PPO on Total Expenditure
NOTE: These are results from two two-year cohorts and are not a reflection of the plans’Full PMPM expenditures. Bolded numbers are significant at p<.05.
Annual Plan Effects Compared With POS PlanHealth Plan Cohorts 2001 Mean 2002 Mean Stage Plan Effects Plan Effects
CDHP Cohort N=187 N=336 Cohort #2 Cohort #3Total Expenditure 3,063$ 6,014$ Probit 1.826 0.269
GLM 2.7% -1.2%
Employer Expenditure 2,740$ 5,314$ Probit 1.839 0.288
GLM 6.7% 1.3%
Employee Expenditure 323$ 700$ Probit -0.315 -0.432
GLM 67.6% 9.1%
PPO Cohort N=994 N=700Total Expenditure 5,131$ 6,308$ Probit 0.251 0.017
GLM 16.3% 12.1%
Employer Expenditure 4,593$ 5,651$ Probit 0.190 0.031
GLM 18.7% 12.6%
Employee Expenditure 538$ 657$ Probit 0.292 -0.137
GLM 8.3% 12.0%
Regressions control for year, plan choice, age, gender, illness burden, health shock,
income, number of covered lives in contract, and use of healthcare flexible spending account
Table 5. Impact of CDHP and PPO on Hospital Expenditure (all costs) per Admission
NOTE: These are results from a restricted continuously enrolled sample of 28% of the total employee population and are not a reflection of the plans’ full PMPM expenditures. Bolded numbers are significant at p<.05.
Annual Plan Effects Compared With POS Plan
2000 Model 2001 2002 2003Health Plan Cohorts Mean Stage Plan Effects Plan Effects Plan Effects
CDHP Cohort N=429Hospital Expenditure (all) 6,828$ GLM -12.7% 70.4% -25.3%
per Admission
PPO Cohort N=1025Hospital Expenditure (all) 6,747$ GLM 22.0% 53.4% 70.8%
per Admission
Regressions control for annual trends, health plan choice, age, gender, illness burden, contemporaneous health shock,
income, number of covered lives in contract, and use of healthcare flexible spending account
Table 6. Impact of Prior-year Pharmacy Spending on Hospital and ER Cost and Use
NOTE: These are results from a restricted continuously enrolled sample of 28% of the total employee population and are not a reflection of the plans’ full PMPM expenditures. Bolded numbers are significant at p<.05.
Effect of Lagged Rx Spending is Assumed Equal Across Plan Cohorts
2001 Model Lagged RxDependent Variable Mean Stage Effect
Hospital Expenditure 1,686$ Probit 0.0050Effect / $100 Lagged Rx GLM 0.79%
Admission Rate 0.22 Probit 0.0047 Effect / $100 Lagged Rx GLM 0.16%
Emergency Room Visits 0.16 Probit 0.0024 Effect / $100 Lagged Rx GLM 0.20%
Regressions control for annual trends, health plan choice, health plan choice x trend,age, gender, income, number of covered lives in contract,and use of healthcare flexible spending account
Conclusions
#1 Addition of 3rd year of CDHP data does not change our previous result: the CDHP is unable to control medical expenditures over time in this large employer
The CDHP had annual gap of only $500 between the personal care account and full insurance coverage for single contracts, and $1,000 gap for family contracts
#2 There is some evidence of pent-up demand, but not enough to deny the existence of a permanent moral hazard problem
Conclusions, continued
#3 We pinpoint the source of the moral hazard problem: the CDHP cohort spent considerably more money on hospital care, and this difference persists over time
#4 The analysis of expenditure per hospital admission produces inconclusive results
#5 Lagged prescription drug spending is associated with higher hospital spending, admissions, and ER use
Summary
The CDHP had too little out-of-pocket cost sharing to be effective
Copies of presentation and draft paper are available from www.ehealthplan.org