Using Willingness to Pay to Evaluate Hospital Mergers: Results from 16 Mergers Presented by Rich Lindrooth Co-authors: David Dranove Mark Satterthwaite Northwestern University This research was funded by Robert Wood Johnson’s HCFO Initiative Using WTP to Evaluate Mergers • Technique was refined in Capps Dranove and Satterthwaite (Rand 2002) • WTP is a measure based on the value health plan enrollees place on inclusion of a hospital in a managed care network. • Direct theoretical link between WTP estimates and hospital prices Key Institutional Details • Health plans assemble networks • Health plans negotiate with local hospitals • Traditional competition models do not apply • Must instead invoke bargaining models • Health plans market networks to local employers • Must provide geographic coverage coincident with where employees live • Services that employee are likely to need • Hospitals that are most attractive to networks bargain for the highest prices Computing WTP • Hypothetical WTP calculation • MCO considering adding hospital X to network • Hospital X is a leader in CABG, but is not conveniently located • What is a typical enrollees WTP to have access to this hospital? Example continued • Typical enrollee considers future medical needs • Most likely (e.g., 90%) they will remain healthy: WTP = 0 • Small chance (e.g. 9.9%) of requiring hospital for routine needs • Local hospital will do just fine • WTP to have access to X if routine problem arises = $1000 • Or for our purposes 9.9%*1000=$99 • Very small chance (e.g. 0.1%) of requiring CABG • Hospital X is best in town • WTP to access X if CABG required = $20,000 • Or for our purposes 0.1%*20000=$200 - Overall WTP to have access to X = $299. - This is the maximum amount that X can “squeeze” out of the negotations (on a per patient basis) - This will take the form of higher overall prices. Three possible health states for person i: Torn Knee ligament, Need a CABG, Healthy Interim Conditions Torn Ligament (A) Interim Choice Probabilities: Healthy Pˆi1 , Pˆi 2 ,..., PˆiJ CABG (B) Pi1 , Pi 2 ,..., PiJ WTPij = 0 WTPi ,Interim (Diagnosis Torn Ligament) k WTPi ,Interim (Diagnosis Need CABG) k EU Interim G 1,..., k ,..., J | Torn Ligament EU Interim G 1,..., k ,..., J | CABG EU Interim G / k | Torn Ligament EU Interim G / k | CABG PR(Torn Ligament | Demographicsi ) Probabilities of Each Condition PR(CABG | Demographicsi ) WTPi ,k EU i ({G}) EU i ({G / k }) D PR(Diagnosis d | Demographicsi ) WTPi ,Interim (Diagnosis d ) k d 1 Aggregate Over The Population WTPk N WTPik f ( X i )dX Mathematically: V (G ) V (G / k ) W G N EY , Z , 1 1 N ln f (Yi , Z i , i ) dYi dZi d i , Y , Z , 1 sk G, Yi , Z i , i EA k Key limitations • Do not observe prices • Do not observe MD information • Omitting MD does not bias the model • MD choice of hospital reflects patient preferences • Analyst could infer that the MD “ran the show” when in fact the MD was deferring to the power of the hospital • Assumes network is formed to fully reflect employee preferences A look at 16 mergers • We use a sample of 16 mergers occurring in 1995-2000: • • • • • • • Bakersfield, CA (2) Buffalo, NY (3) Daytona, FL (4) Denver, CO (2 total; use 1) Jacksonville, FL (2, one de-merger) Rochester, NY (4) Seattle (Use as a control. 1 hospital switched systems in 2000) • Milwaukee (Market too tumultuous over time to model) Example: Daytona Market • Ownership in many of the markets was tumultuous during this period • Many mergers, some de-mergers, and some exits. • The following maps trace merger activity in Daytona from 1995 until 2000. • Each color represents a system. • A white dot is a independent hospital Data • Patient-level data • the Healthcare Cost and Utilization Project State Inpatient Database (HCUP-SID) • Hospital Characteristics • American Hospital Association Annual Survey (AHA) • Hospital Financial Data • CMS Medicare Cost Reports (MCR) Estimating the Effect of the Mergers on Net Inpatient Revenue • Step 1: Calculate WTP • Value a hospital/system brings to a network • Step 2: Estimate effect of change in WTP on Net Inpatient Revenue • Step 3: Measure Percent change in Net revenue: Change in Net Revenue/Pre-merger Net Revenue Calculating WTP • Estimate WTP for independent hospitals and system combinations one year prior to the first merger in the market. • Prior to system consolidation calculate the sum of the independent hospital’s WTP • After system consolidation use the system WTP • The difference reflects the change in WTP is solely a result of the merger WTP Estimates Bakersfield Merger 1 Merger 2 Buffalo Merger 1a Merger 1b Merger 2 Denver Merger 1 Daytona Merger 1a Merger 1b Merger 2 Merger 3 Jacksonville Merger 1 Demerger 1 Merger 2 Rochester Merger 1 Merger 2 Merger 3 Merger 4 Year 1997 2000 # of hospitals Pre-WTP Post-WTP % Change 3+1 4,468 5,127 14.7% 1+1 1,591 1,724 8.3% 1996 1998 1997 2+1 3+2 1+1+1+1 13,871 14,573 11,720 14,573 17,623 14,170 5.1% 20.9% 20.9% 1997 1+1 13,805 14,555 5.4% 1999 2000 1998 1996 3+1 4+1 1+1 1+1 8,696 9,885 10,390 1,012 9,885 11,595 11,118 1,053 13.7% 17.3% 7.0% 4.1% 1997 1999 1999 3+1 3 1+1 21,959 27,164 3,849 27,164 21,959 4,093 23.7% -19.2% 6.3% 1998 1996 1999 2000 3+1 1+1 1+1 1+1 35,284 3,014 6,930 1,964 36,153 3,101 7,213 2,463 2.5% 2.9% 4.1% 25.4% HHI (Just for reference) Bakersfield Merger 1 Merger 2 Buffalo Merger 1a Merger 1b Merger 2 Denver Merger 1 Daytona Merger 1a Merger 1b Merger 2 Merger 3 Jacksonville Merger 1 Demerger 1 Merger 2 Rochester Merger 1 Merger 2 Merger 3 Merger 4 Year 1997 2000 # of hospitals in merger 3+1 1+1 1996 1998 1997 2+1 3+2 1+1+1+1 1,488 2,146 1,694 1,694 3,201 2,146 206 1,055 452 1997 1+1 2,047 2,463 416 1999 2000 1998 1996 3+1 4+1 1+1 1+1 3,920 4,230 3,762 3,739 4,230 4,794 3,920 3,762 310 564 158 23 1997 1999 1999 3+1 3 1+1 2,831 4,269 2,831 2,840 1,438 0 9 1998 1996 1999 2000 3+1 1+1 1+1 1+1 2,545 2,530 2,611 2,614 2,611 2,545 2,614 3,250 66 15 3 636 Pre-HHI 4,796 4,853 Post-HHI 4,853 4,853 Change 57 0 Do changes in WTP due to consolidation increased net inpatient revenue? • Regress net inpatient revenue: • • • • • • WTP, WTP*Yrs since merger indicator Hospital fixed effect Hospital payer mix Bedsize, Total admissions, and Control for whether the facilities were combined. Details of regression • Unit of observation: Entity • Independent hospital or system • 6 years of data • Unbalanced, some hospitals don’t report in some years • If one hospital in the system doesn’t report • Drop the system observation for the year System Fixed Effect Results WTP 3909.923** (1937.421) WTP*Second Year of Merger WTP*Third Year of Merger WTP*Forth Year of Merger 6649.468*** (2383.316) 4606.421** (2143.295) 155.476 (341.553) 1110.533* (660.681) -78.326 (933.719) 234.834 (303.028) 324.111 (625.334) -1096.700 (829.123) 340.953 (299.064) 657.558 (499.414) -255.871 (606.881) -4024.480*** (990.239) -376.645 (1269.217) WTP*First Year of Second Merger WTP*Second Year of Second Merger WTP*First Year of De-merger WTP*Second Year of De-merger R-squared Controls 8800.893*** (2654.161) 0.36 None 409.136 (1058.589) 1244.982 (1303.386) 0.37 Year FE -2732.502*** -2642.907*** (905.473) (894.974) 1370.700 403.522 (1141.233) (1009.329) 974.265 (983.494) 1359.085 (1143.716) 0.61 All Controls 0.65 All Controls Results (in words) • WTP coefficient is statistically different than zero with 95-99% confidence • Though not significant if only time trends in the model without WTP*Merger year interactions • Magnitude of the coefficient varies from about $3900 to $8800. • Favored specification indicates an increase of about $6,650 per unit of WTP Effect of mergers on net inpatient revenues MSA Bakersfield Merger 1 Merger 2 Buffalo Merger 1a Merger 1b Combination Merger 2 Denver Merger 1 Daytona Merger 1a Merger 1b Combination Merger 2 Merger 3 Jacksonville Merger 1 Merger 2 Rochester Merger 1 Merger 2 Merger 3 Merger 4 Total Increase in Net Change in WTP inpatient Revenue Premerger net Percent inpatient revenue Increase 658.6 132.8 $4,379,723 $883,080 $178,039,547 $18,509,967 2.46% 4.77% 702.3 3049.6 $353,153,430 $335,892,953 2450.5 $4,670,295 $20,280,040 $24,950,335 $16,295,958 $511,717,061 1.32% 6.04% 7.07% 3.18% 749.9 $4,987,035 $311,633,985 1.60% 1189.0 1709.9 $230,319,971 $228,534,594 728.1 41.7 $7,906,963 $11,371,128 $19,278,091 $4,841,599 $277,145 $208,648,796 $63,230,045 3.43% 4.98% 8.37% 2.32% 0.44% 5204.3 243.8 $34,608,462 $1,621,071 $499,638,496 $306,650,559 6.93% 0.53% 868.6 87.4 283.3 499.1 $5,776,257 $581,163 $1,884,231 $3,318,776 $416,377,613 $172,894,372 $505,253,995 $40,754,389 1.39% 0.34% 0.37% 8.14% 18,599 $123,682,924 $4,381,249,771 2.82% Conclusions • Only four mergers led to an increase in net inpatient revenue greater than 5%. • However, this is a very conservative estimate • The denominator revenue number includes all payers. • If the increase was due to solely private payers then the denominator should include only private payer’s revenue. • For example if 50% of the net revenues were private then the average % increase would be 5.6% Conclusions • Shows promise for prospective merger analyses • R-squared was much lower than what is observed in a single market. • Not surprising given unmeasured variation across markets Caveats • Would prefer to have data on private profits rather net inpatient revenues. • We ignore any efficiencies (or inefficiencies) that result from mergers. • The prospective WTP estimates do not always coincide with the realized post-merger estimates (a problem with any prospective measure). • The second mergers during the period perform worse (may be due to small sample of second mergers). There are other factors at work that need to be addressed in future research.