Can health care claims data improve the estimation of the Medical CPI?

William D. Marder, PhDJune 28, 2004

Can health care claims data improve the estimation of the Medical CPI?

CRIW Conference on Price Index Concepts and Measurement

2 Copyright 2004 Thomson Medstat

Can health care claims improve the Medical CPI?

Acknowledgements

• Support for this project came from a contract with the Bureau of Labor Statistics.

• Coauthors:– Xue Song

– Onur Baser

– Bob Houchens

– Jon Conklin

– Ralph Bradley, BLS



Issues Specific to Medical Care

• Third party payment for healthcare creates challenges and opportunities for price measurement

– Consumer out-of-pocket spending is a fraction of the transaction price

– Medical claims capture information about transaction prices (for some transactions) with little or no respondent burden – can this improve the reliability of estimates

• Most of the money spent on healthcare is to treat illness. – Several researchers have investigated trends in the cost of

treatment episodes created from claims databases

– Committee on National Statistics (CNSTAT) recommendation

• We report results on each of these issues.



Data

• MarketScan data from Boston, New York, and Philadelphia 1998-2002.

• BLS data from Boston, New York, and Philadelphia PSUs for the same years.

• MarketScan is a convenience sample based on Medstat’s employer clients’ data

– Fully integrated medical and drug claims including outpatient Rx use data – Comprehensive patient level data from over 6 million covered lives

annually, tracking actual patients even as they change health plans– On-going data feeds from 50 major employers and over 100 health plans

covering all plan types and carve-outs– For some employers includes employer-sponsored Medicare supplemental

insurance with coordination of benefits– Fully HIPAA compliant



Characteristics of the claims database (1 of 2)

• Coverage does not match population size

• In 1998, MarketScan populations were:– Boston 146,000

– New York 43,520

– Philadelphia 104,901



Characteristics of the claims database (2 of 2)

Table 1. Number of Unique Pharmacies, Hospitals, and Physicians in Each PSU 1998 1999 2000 2001 2002 Boston Pharmacy 1,439 1,313 1,324 1,419 1,584 Hospital (Inpatient) 195 181 173 132 144 Hospital (Outpatient) 854 592 528 549 881 Physician 27,788 22,868 20,816 21,715 26,621 New York Pharmacy 1,406 1,219 1,279 1,329 1,460 Hospital (Inpatient) 69 58 47 53 66 Hospital (Outpatient) 321 117 101 113 250 Physician 11,905 6,677 6,048 5,636 8,536 Philadelphia Pharmacy 1,763 1,455 1,477 1,402 1,454 Hospital (Inpatient) 134 99 92 83 93 Hospital (Outpatient) 1,084 353 315 337 676 Physician 19,448 7,517 6,763 5,430 8,440 Source: Medstat MarketScan.



BLS Sample Sizes, 1998-2002

Pharmacy Physician Hospital

Boston 42 27 46

New York 41 35 59

Philadelphia 34 32 31



Comparability of claims and current BLS price levels

• Focus on outpatient prescription drug prices

• Prescription drugs identified by 11-digit NDC codes in both BLS procedures and claims data

• Compared price levels in the three cities for the specific NDCs in the current CPI data collection effort

– These particular drugs are carefully selected with a probability that depends on patient spending patterns



Drug prices lower in claims than BLS

• In each PSU separately, t-test statistics are -8.10 in Boston,

-17.76 in New York) and

-7.56 in Philadelphia,

All three PSU’s combined t = -16.78

• As expected, prices collected in the claims database are lower than the prices collected by the BLS because of discounts.



Comparability of Drug Price Trends - Philadelphia

80

90

100

110

120

130

BLS MarketScan

Disease-based price measures



Committee on National Statistics Recommendation 6.1

• BLS should select between 15-40 diagnoses from the ICD (International Classification of Diseases), chosen randomly in proportion to their direct medical treatment expenditures

• Use information from retrospective claims databases to identify and quantify the inputs used in their treatment and to estimate their cost.

• On a monthly basis, the BLS could re-price the current set of specific items (e.g., anesthesia, surgery, and medications), keeping quantity weights temporarily fixed.

• Then, at appropriate intervals, perhaps every year or two, the BLS should reconstruct the medical price index by pricing the treatment episodes of the 15 to 40 diagnoses—including the effects of changed inputs on the overall cost of those treatments.

• The frequency with which these diagnosis adjustments should be made will depend in part on the cost to BLS of doing so.

• The resulting MCPI price indexes should initially be published on an experimental basis. The panel also recommends that the BLS appoint a study group to consider, among other things, the possibility that the index will “jump” at the linkage points and whether a prospective smoothing technique should be used.



Implementation of Recommendation 6-1

• Focused on three metropolitan areas

• Used COTS to construct disease-based episodes for all diseases

• Randomly selected 40 episodes for study

• Summarized all resources used in treatment episode (and the frequency of use)

• Repriced monthly within broad classes for hospital, physician and drug components of episodes:

– DRGs– BETOS– Drug therapeutic classes

• Respecified the bundle in the episode annually.



Medstat’s Episode Grouper (MEG)

• -Commercial Off-the-Shelf software tool that groups claims into episodes of care

• Comprehensive – covers all diseases in the ICD9

• Based on Disease Staging – originally developed for NCHSR (predecessor to AHRQ)



MEG Classification System

• Classify Diseases– Conditions defined by diagnoses

– Disease groupings based on conditions not treatments

– Stage of disease defined by severity of complications

• Build Episodes– Associate services and visits for same condition

– Associate non-specific diagnoses

– Require clean periods (look-back & episode-end)

• Associate Pharmacy Data– Array prescriptions temporally to ongoing episodes

– Assign drugs to appropriate episodes



From claims to episodes

Herniated Intervertabral

Disk

Acute Prostatitis

Office visitFeb 4

Office visitMay 18

PyelonephritisJun 1

Office VisitJan 10

X-rayFeb 28

MRIMar 15

HospitalApr 30

Office visitMay 8

EPISODE 1

EPISODE 2 EPISODE 3

Figure 1



Sampled Episodes

Conditions Sampled for Boston Episode Total Number Expected Group MarketScan of Times Number of Number Episode Label Payments Drawn Times Drawn 10 Angina Pectoris, Chronic Maintenance $27,424,386 4 2.690 374 Osteoarthritis $16,971,880 3 1.665 11 Acute Myocardial Infarction $16,192,922 1 1.588 13 Essential Hypertension, Chronic Maintenance $13,013,202 2 1.277 397 Cerebrovascular Dis with Stroke $11,187,732 2 1.097 187 Renal Failure $10,737,384 2 1.053 92 Cataract $8,905,881 1 0.874 500 Chronic Obstructive Pulmonary Disease $8,752,026 1 0.859 6 Arrhythmias $8,653,448 1 0.849 212 Neoplasm, Malignant: Breast, Female $8,542,333 1 0.838 348 Fracture: Femur, Head or Neck $7,267,359 2 0.713 426 Complications of Surgical and Medical Care $6,681,475 1 0.655 50 Diabetes Mellitus Type 2 and Hyperglycemic States Maintenance $5,312,060 1 0.521



Diabetes Trends

60

70

80

90

100

110

120

130

140

150

Philadelphia Boston New York



AMI Trends

60

70

80

90

100

110

120

130

140

150

Philadelphia Boston New York



Disease-based Index – Boston

0

20

40

60

80

100

120

140

160

Jan-99

Mar-99

May-99

Jul-99

Sep-99

Nov-99

Jan-00

Mar-00

May-00

Jul-00

Sep-00

Nov-00

Jan-01

Mar-01

May-01

Jul-01

Sep-01

Nov-01

Jan-02

Mar-02

May-02

Jul-02

Sep-02

Nov-02

Jan-03

MarketScan BLS



Disease-based Index – New York

0

20

40

60

80

100

120

140

160

Jan-99

Mar-99

May-99

Jul-99

Sep-99

Nov-99

Jan-00

Mar-00

May-00

Jul-00

Sep-00

Nov-00

Jan-01

Mar-01

May-01

Jul-01

Sep-01

Nov-01

Jan-02

Mar-02

May-02

Jul-02

Sep-02

Nov-02

Jan-03

MarketScan BLS



Disease-based Index – Philadelphia

0

20

40

60

80

100

120

140

160

Jan-99

Mar-99

May-99

Jul-99

Sep-99

Nov-99

Jan-00

Mar-00

May-00

Jul-00

Sep-00

Nov-00

Jan-01

Mar-01

May-01

Jul-01

Sep-01

Nov-01

Jan-02

Mar-02

May-02

Jul-02

Sep-02

Nov-02

Jan-03

MarketScan BLS



Decomposing the difference

• Claims prices differ even when methods and numbers of cases are the same

• Claims permit larger sample sizes at little or no incremental cost

• Disease-based views of claims represent a significant new analytic construct.

• How much of the difference is attributable to each of these steps?



Bootstrapping results are a work in progress

• Clear that the monthly changes are noisy enough that the differences are not statistically significant

• Cumulative effects are still being investigated.



Concluding remarks

• Previous research looked at disease-specific costs in a new way and found dramatically different trends

– Were the results specific to the diseases chosen?

– This initial investigation with a randomly selected group of diseases suggests that a more general effect could be at work.

• Limitations– Three cities with convenience samples that may not be

representative

– Significance testing not yet complete

• Next steps

Documents

Can health care claims data improve the estimation of the Medical CPI?