VA Health Economics Course Presentation

1
VA Health Economics Course Presentation
3Costing Methods

• May 2, 2007

2
Costing Methods

• Mark W. Smith, PhD
• Associate Director
• VA Health Economics Resource Center

3
Focusing Question

• What is the cost of a health care intervention?
• Example
• CSP 519 compares separate PTSD and smoking
  cessation therapy to combined therapy.

4
Cost of Health Care

• Outside of health, most items that we purchase
  daily have a readily observable cost
• Not true with health care
• Insurance buffers patient from true cost
• Charges, payments may not equal cost

5
Cost Estimation Approaches

• Two general approaches to costing
• Microcosting
• Average costing (gross costing)

6
Estimating Costs Micro-costing

• Determine each input, find its price, then sum
  (quantityprice) across all inputs
• DSS uses this approach
• Researchers use this approach in some
  circumstances
• Gold standard but resource intensive

7
Estimating Costs Average Costing

• Over a long period, divide total cost by total
  units of care provided
• Less precise than micro-costing

8
Costing Spectrum
micro
average
Pseudo-bill
Cost regression
Reduced list costing
Direct measurement
Average cost per day
Estimate Medicare payment
9
Microcost method 1

• Direct Measurement

10
Direct Measurement

• Used to the find the cost of
• interventions
• care unique to VA (e.g., CSP 519)
• Method
• Measure staff activity
• Find labor cost
• Find cost of supplies, capital, overhead

11
Finding Unit Cost

• Average cost
• Total program cost/number of units
• Assumes homogeneous products
• Relative Values needed for heterogeneous products
• Find Relative Value of each product
• Find cost per relative value unit (RVU)
• Use this to find cost of each product

12
Staff Activity Analysis
Direct Measurement

• Methods of finding staff activities
• Track staff activity in a log
• Estimate activity
• Need not be comprehensive can sample activity
• Estimate labor cost

13
Characterizing Staff Activities
Direct Measurement

• Cost of patient care may include non-patient care
  time
• Activities that produce several products may need
  to be included, depending on perspective
• e.g., time spent on clinical research may be
  regarded as a research cost, or a patient care
  cost, depending on analytical goal

14
Exclude and Include
Direct Measurement

• Exclude development cost
• Exclude research-related costs
• Should measure when program fully implemented
• Should measure at constant returns to scale

15
Direct vs. Indirect vs. Overhead

• Direct costs costs that are tied to a
  particular encounter (e.g., staff time,
  medications)
• Overhead costs that cannot be tied to
  particular procedures (e.g., VA police,
  maintenance, food service)

16
Direct vs. Indirect vs. Overhead

• Indirect
• sometimes means overhead
• sometimes means non-salary benefits
• (e.g., health care, annual leave)
• sometimes means secondary impact of treatment on
  other health care use Example patient
  receives better depression care
• at VA and later has fewer visits for other
  causes

17
Discussion

• Which of these should be included in the cost of
  an intervention?
• Non-salary benefits
• Secondary impact on other health care services
• Overhead costs

18
Other Costs
Direct Measurement

• Survey or actual measure of supply costs
• Alternatives for overhead
• Cost report data
• Standard rates
• Alternatives for capital
• Cost report
• Rental rates

19
Microcost method 2

• Pseudo-Bill

20
Pseudo-bill

• Itemize all services utilized/provided
• Use schedule of cost/reimbursement for each
  service
• Example HERC outpatient costs
• Itemized all CPT codes
• Used relative value weights to assign costs to
  procedures

21
Microcost method 3

• Reduced List Costing

22
Reduced List Costing

• Some utilization items in pseudo-bill explain
  most of variation in cost
• e.g., surgical procedures
• Costing major items may be sufficient
• Schedule of cost/reimbursement must be adjusted
• e.g., new rate for surgical procedures that
  includes cost of laboratory services

23
Microcost method 4

• Cost Regression

24
Cost Regression

• Dependent variable is charges or cost-adjusted
  charge from non-VA data
• Independent variables
• Clinical information
• Diagnosis Related Group
• Diagnosis
• Procedures
• Vital status at discharge
• Length of stay
• Days of ICU care

Anything that predicts cost and is in both
datasets.
25
Transformation of Dependent Variable

• Cost data are frequently skewed
• Skewed errors violates assumptions of Ordinary
  Least Squares
• Error terms not normally distributed with
  identical means and variance
• Transformation
• Typical method log of cost
• Can make OLS assumptions more tenable

26
References - I

• Duan, N. (1983) Smearing estimate a
  nonparametric retransformation method, Journal of
  the American Statistical Association, 78,
  605-610.
• Manning WG, Mullahy J. Estimating log models to
  transform or not to transform? J Health Econ 2001
  Jul20(4)461-94.

27
References - II

• Basu A, Manning WG, Mullahy J. Comparing
  alternative models log vs Cox proportional
  hazard? Health Economics 2004 Aug13(8)749-65.

28
HERC Web Site FAQs

• E1. How do I estimate costs with a clinical cost
  function?
• http//www.herc.research.va.gov/resources/
  faq_e02.asp
• E2. What is retransformation bias, and how can it
  be corrected?
• http//www.herc.research.va.gov/resources/
  faq_e02.asp

29
Limitations

• Relies on similar cost structures of external and
  study (internal) data.
• Reduces the number of outliers.
• Can create statistical anomalies.

30
Microcost method 5

• Estimating Medicare reimbursements

31
Medicare Reimbursements

• Part A -- Prospective Payment for Inpatient Stays
• Part B -- Payment for Physician Services to
  Inpatients

32
Medicare Inpatient Facility Payment

• DRG-based payments adjusted by
• Disproportionate share payments
• Indirect medical education
• Geographic adjustments
• Outlier payments for unusual cases
• Direct medical education

33
Medicare Payments

• Medicare pays flat rate per DRG, regardless of
  length of stay (except for outliers)
• Cost analysis may wish to capture effect of
  length of stay on cost

34
Medicare Pricer Software

• Computer application for calculating facility
  payment
• Requires
• 6-digit hospital PPS (identifier)
• DRG
• Admission and discharge dates (LOS)

35
Medicare Outpatient Payment

• Payment based on CPT procedure codes
• Provider payment and facility payment (if done in
  hospital)
• See documentation for HERC Outpatient Average
  Cost data www.herc.research.med.va.gov/
  methods_data/va_cost_methods_ac.asp

36
Outpatient Medicare Payments

• Some CPTs have no APC
• Paid on cost pass-through basis
• Paid through another APC (e.g., anesthesia)
• Paid through a separate cost list
• Multiple CPTs assigned to a single group-APC
• Some surgery procedures are discounted

37
Selecting a Method

• Data available?
• Method feasible?
• Assumptions appropriate?
• Method accurate Will it capture the effect of
  the intervention on resource use?

38
Direct Measurement

• Assumptions
• Activity survey and payroll data are
  representative
• May assume all utilization uses the same amount
  of resources
• Advantages
• Useful to determine cost of a program that is
  unique to VA
• Disadvantages
• Limited to small number of programs
• Cant find indirect costs
• Cant find total health care cost

39
Pseudo-bill

• Assumptions
• Schedule of charges reflects relative resource
  use
• Cost-adjusted charges reflect VA costs
• Advantages
• Captures effect of intervention on pattern of
  care within an encounter
• Disadvantages
• Expense of obtaining detailed utilization data

40
Reduced List Costing

• Assumptions
• Items on reduced list are sufficient to capture
  variation in resource use
• Cost of items on reduced list is accurate
• Advantages
• Requires less data than pseudo-bill
• Disadvantages
• Needs to find data on cost associated with items
  on reduced list

41
Cost Regression

• Assumptions
• Cost-adjusted charges accurately reflect resource
  use
• The relation between cost and utilization is the
  same in the current study as in the previous
  study
• Advantages
• Less effort to obtain reduced list of utilization
  measures than to prepare pseudo-bill
• Disadvantages
• Must have detailed data
• Data from prior study may have error or bias

42
Estimate Medicare payments

• Assumptions
• Medicare payments reflect average cost for a
  population your sample is generalizable
• RVU captures effect of intervention on resources
  used
• Advantage easy to understand
• Disadvantages
• Accuracy limited VA may have different cost
  structures from average non-VA facilities
• Inpatient doesnt reflect variation in resources
  beyond DRG (or LOS)

43
Combining Methods

• No single method may fill all needs, even within
  a single study
• Hybrid method may be best
• Direct method or pseudo-bill on utilization most
  affected by intervention
• Cost regression or Medicare payment for other
  utilization

44
Discussion

• CSP 519 compares
• Separate PTSD and smoking cessation visits
• Combined PTSD and smoking cessation visits
• What are some costs that you could estimate by
  an average-costing approach?
• Is there anything that might need to be
  measured directly?

45
Reference

• Barnett PG. Determination of VA health care
  costs. Medical Care Research and Review 200360(3
  Suppl.)124S-141S.
• www.herc.research.med.va.gov/
• publications/supplement_mcrr_2003.asp

46
Other Resources

• HERC web site FAQ responses, technical reports
  (click on Publications tab)
• HERC Help Desk (herc_at_med.va.gov)

47
HERC email list

• To join the HERC email list, send a request to
  herc_at_va.gov.

48
Next session

• Wednesday, 5/16/2006, 2 p.m. ET
• Estimating the Cost of Health Care VA Costs
• Paul Barnett, PhD
• Reading for next session
• M Gold et al. Cost-Effectiveness in Health and
  Medicine
• pp. 199-210. Available for purchase at
  http//www.oup.com/us/ or http//www.amazon.com
• PG Barnett. Medical Care Research and Review
  60(3), pp. 124S-141S. Download from
  http//www.herc.research.med.va.gov/
  publications/supplement_mcrr_2003.asp