Exploring efficiency and quality of care among hospitals of the US Veteran Health

Administration and Germany

Jonas Schreyögg, PhD

Commonwealth Harkness Fellow, Center for Health Policy, Stanford University, USA

Harkness Fellowships, Final Reporting Seminar, Orlando, June 5-6

$419

$549

$793$862

$1,013$1,015$1,069

$2,337

$0

$500

$1,000

$1,500

$2,000

$2,500

United

States

France Australia Italy Canada OECD

Median

Germany J apan

Hospital Spending per Inpatient Acute Care Day in 2004 adjusted for PPP

Background

Source: OECD Health Data 2006

-> we want to know more!

Research approach

• Measuring efficiency and quality between hospitals with individual level data -> costs and mortality per patient

• To adequately control for case-mix we select three major episodes of care: Acute Myocardial Infarction (AMI), hip replacement, appendectomy

• Advantages of AMI over other care episodes:• Requires immediate medical attention -> no patient selection• High incidence and leading cause of death in the elderly• Quality of care provided by hospitals can substantially avoid mortality• Reflects level of technology utilization

Objective

1) to explain variation in costs as a measure of efficiency and

2) to explain variation in hospital mortality as a measure of clinical quality

between hospitals of the Veteran Health Administration and Germany

0.000

0.020

0.040

0.060

0.080

0.100

0.120

0.01

0.04

0.07

0.10

0.13

0.16

0.19

0.22

0.25

0.28

0.31

0.34

0.37

0.40

0.43

0.46

0.49

0.53

0.56

RG (cost weight) per day

Rela

tive fre

qu

en

cy

Relative frequeny 04 Relative frequeny 05 Relative frequency 06

Compression effect: Most German DRGs pay the same amount

per day -> very similar in Medicare

Why can‘t we just take reimbursement rates?

Schreyögg J, Tiemann O, Busse R (2006) Cost accounting to determine prices: how well do prices reflect costs in the German DRG-system? Health Care Management Science 9 (3): 269-280

Data

Data for year 2005 from two settings were chosen which have a very similardata structure and follow a similar cost accounting approach

1) Hospitals of the US Veteran Health Administration (VHA)– standardized accounting system – subject to extensive periodic audits– maintained and continuously improved since the mid 90ties– we used data from administrative files/EHR – data from 135 hospitals

2) German hospitals – obtained as part of the national cost data study (263 hospitals) maintained by

InEK (Institute for the Calculation of Hospitals reimbursement) since 2004– participation of hospitals requires adherence to cost accounting standards– to increase reliability of cost data we only included hospitals which

participated 2004 and 2005– InEK databases is not accessible for researchers -> each hospital was

approached– 60 hospitals met our sample definition criteria -> 19 delivered us their data

Methodology

1. Multilevel regression is performed-> What determines costs and mortality in both systems?

2. Imposing VHA prediction functions on German sample and imposing German prediction functions on the VHA sample-> How would VHA hospitals perform with German patients and vica versa?

3. Applying a propensity matching score approach to match VA and German patients-> How do hospitals from both systems perform with the same patient co-morbidities?

Modeling strategy

Co-morbidities

Hospital characteristics

Treatment strategies (Use of Technology)

Costs per case/ Mortality

German sample

Definiton of common set of variables

Lower level (Unit: cases)

Estimation of seperate cost/mortality

multilevel functions for German and VHA hospitals

Co-morbidities

Hospital characteristics

Treatment strategies (Use of Technology)

Costs per case/ Mortality

VHA sample

Higher level (Unit: hospitals)

Lower level (Unit: cases)

Higher level (Unit: hospitals)

Sample characteristics (1)

Characteristics VA Gemany

No. of cases in the sample 6598 585No. of hospital in the sample 130 18

Patient characteristicsAge (mean) 65.7 62.3Acute renal failure 3% 4%Cancer 4% 1%Chronic renal failure 8% 5%Congestive heart failure 9% 18%Diabetes 24% 16%Diabetes complications 4% 2%Peripheral vascular disease 4% 2%Pulmonary disease 15% 7%Severe liver disease <1% 0%Angina pectoris 1% 3%Cerebrovascular disease 2% 1%Cardiac dysrhythmias 9% 13%Chronic ischemic heart disease 66% 80%Hypertensive heart disease <1% <1%Shock <1% <1%Hospital mortality 6.3% 1.5%

All significantly different at P<0.001

Characteristics VA Gemany p-value

Treatment strategiesNo procedure 29% 26% 0.034Catheter + NOPCI 17% 14% 0.0046PTCA 14% 1% <0.0001STENTS 1% 1% <0.0001STENTSDRUG 4% 0% <0.0001Catheter + PTCA 8% 35% <0.0001Catheter + STENTS 7% 20% <0.0001Catheter + STENTSDRUG 20% 3% <0.0001

LOS (mean no.) 5.5 8 <0.0001

Hospital characteristicsSelected AMI cases (mean no.) 48 35.6 0.59Nurses in FTE per bed (mean no.) 2.8 0.59 <0.0001

Number of beds (mean no.) 137.8 503.3 <0.0001Member of the council of teaching hospitals 35% 50% 0.63

Located in urban area > 200,000 38% 22% 0.55

Sample characteristics (2)

VHA Germany

Radiology 118 119Laboratory 286 107Therapy + Warda

- Nursing Staff 2,549 727 - Physicians and other staff 874 686 - Drugs 555 128 - Overhead 1,736 848 - Total 5,988 2,579General Overheadb 3,646 797Total costs 11,034 3,650

a This category includes all costs incurred in the catheter lab and all wards costs for pre- and post therapy care. Material costs are included under overhead.b General overhead includes costs for administration, food, laundry, energy, rent etc.

Median costs by cost category

Factor 3.5 (40% due to wages/ 60% due to nursing ratio)

Factor 4.6 (higher admin. per bed ratio 2.0-VA/ 0.12-Ger, higher admin. wages -> more documentation, bureaucracy etc. )

Regressions with dependent "ln_costs per case"

VHA GermanyCoefficients

Intercept 7.8572*** 7.5321***

Co-morbidities included included

Catheter + NOPCI 0.4320*** 0.2398*PTCA 0.000731 0.3605STENTS 0.2893** 0.7371*STENTSDRUG 0.3037*** 0Catheter + PTCA 0.4426*** 0.7204***Catheter + STENTS 0.4646*** 0.7127***Catheter + STENTSDRUG 0.4986*** 0.8850***

Selected AMI cases 0.001332 -0.00244Member of the council of teaching hospitals 0.1061 -0.3088Located in urban area 0.1648** 0.5082*Nurses per bed 0.04701 -0.3144Number of beds 0.001235** 0.000919*

* P<0.05, ** P<0.01, *** P<0.001

Regressions with dependent “hospital mortality"

VHA GermanyCoefficients

Intercept -2.1132*** -4.1002***

Co-morbidities included included

Catheter + NOPCI -0.00537 -0.3913**PTCA received -0.3589*** -0.3153*STENTS received 0.04042 -0.2010

Selected AMI cases -0.00286 -0.00095Member of the council of teaching hospitals -0.07425 -0.3128Located in urban area 0.07042 -0.01648 Nurses per bed 0.09391 0.455Number of beds 0.001401 0.000827

* P<0.05, ** P<0.01, *** P<0.001

How do VHA hospitals perform with German patients and vice versa?

VA Germany VA Germany VA Germany

Based on comorbidities Total costs (in $) 3,675 3,953 14,909 15,565 14,909 3,953 Hospital mortality (in %) 2.5 1.5 6.3 6.4 6.3 1.5

Based on comorbidities and treatment strategies Total costs (in $) 3,498 3,953 14,909 15,488 14,909 3,953 Hospital mortality (in %) 2.0 1.5 6.3 6.0 6.3 1.5

Based on comorbidities, treatment strategies and hospital characteristics Total costs (in $) 3,748 3,953 14,909 15,686 14,909 3,953 Hospital mortality (in %) 2.9 1.5 6.3 6.8 6.3 1.5

Predicted mean values based on German function

Predicted mean values based on VHA function

Observed mean values

How do hospitals from both systems perform with the same patient co-morbidities?

Whole samplesMatched on propensity

scores for comorbidtiesa

Matched on propensity scores for comorbidties

and treatment strategiesb

Ln_cost Coefficient for Germany -1.3793 (0.1293)*** -1.505 (0.1546)*** -1.4085 (0.1418)***

Hospital Mortality Coefficient for Germany -1.8057 (0.2328)*** -1.7901 (0.3609)*** -2.0869 (0.4688)***

**P<0.001 *** P<0.0001S.E. in parenthesisa PSM matched samples for comorbidities include n=2840 for VA and n=568b PSM matched samples for comorbidities and treatment strategies include n=1431 for VA and n=477

How do AMI costs compare to Medicare and other European countries?

$396$1,025

$1,861

$3,953

$5,014 $5,599$5,916

$7,450

$14,908

$11,318

$0

$2,000

$4,000

$6,000

$8,000

$10,000

$12,000

$14,000

$16,000

Hungary Poland Spain Germany England Netherlands France Italy VHA Medicare

w/o physician costs and drugs

Comparison to Medicare hospitals (results from other studies):• costs for Medicare tend to be higher/ increase with complexity of procedure• Medicare hospitals perform more procedures• Mortality in Medicare hospitals is lower

Source: data for other European countries based on findings of EU HealthBasket Project; data for Medicare based on estimates for Medicare part A from MEDPAR Inpatient Hospital National Data for Fiscal Year 2005, Centers for Medicare and Medicaid Services.

Hospital costs per AMI case adjusted for PPP

• The analysis demonstrates the potential of health system comparison with micro-level data

• German hospitals are more efficient and hospital mortality is lower-> VHA: higher nursing ratio, higher wages, use of more complex technologies, higher overheads, higher mortality -> Germany: higher LOS and higher relative marginal costs of technology-> it is not either the prices or the utilization which make the US health care system the most expensive health care system in the world it is a combination of both

Next steps:

• further decomposition of costs into input prices and utilization

• Inclusion of more US subsystems or countries into the analysis

Conclusion

Implications for Germany:• vertical integration can improve public health planning abilities• central purchasing lowers prices for technology and drugs

substantially -> but is the VHA purchasing model transferable?

Implications for the VHA or US in general:• Utilization of more advanced technology does not necessarily lead to

lower mortality• Bureaucracy has to be reduced in order to cut down overheads

(documentation requirements due to liability issues etc.)• Can staffing ratios be reduced in order to increase efficiency?

General Implications:• Financial incentives are obviously not the major driver for utilization• Quality and availability of data in health care should be improved -> Each country should maintain an accessible representative micro data

panel to improve cross-country learning possibilities

Policy Implications

Additional Slides

Generalizability of German results

Study sample InEK sampleNumber of hospitals 18 263

Total costs for AMI treatment according to our $3953 $3852

Treatment strategiesNo procedure 26% 38%Cardiac catheters 17% 17%PCIs 57% 45%

LOS 8.5 8.16Hospital beds 503 278Nursing ratio 0.59 0.77

Definition of variables

Dependent variables:

• Costs per case in PPP-US$: costs incurred during hospitalization for treatment after AMI • Hospital mortality: if a patient admitted with AMI dies during hospitalization

Explanatory variables:

• Hospital characteristics: beds, urban status, teaching status, no. AMI cases, nursing ratio• Co-morbidities:

– Ontario AMI prediction rules and Charlson Co-morbidity Index were used– ICD-9 for VHA and ICD-10 for Germany – crosswalk tables used

Cardiac cath. w/o. PCICardiac cath. w. PTCACardiac cath. w. bare-metal stentCardiac cath. w. drug-eluting stent

PTCA w/o. cardiac cath. Bare-metal stent w/o. cardiac cath.Drug-eluting stent w/o. cardiac cath.

If cardiac catheter was performed during same stay

If cardiac catheter was performed before the stay

• Treatment strategies:– treatment strategies identified for cost function