Roger Anderson, Ph.D. Professor Health Services Research Core Penn State Cancer Institute

Patterns of Care in Breast Cancer: On Care Patterns of Care in Breast Cancer: On Care Coordination in Underserved Populations and the Use Coordination in Underserved Populations and the Use

of Health Claims Data.of Health Claims Data.

Roger Anderson, Ph.D.Roger Anderson, Ph.D.ProfessorProfessor

Health Services Research Core Health Services Research Core Penn State Cancer InstitutePenn State Cancer Institute

Penn State Hershey Medical CollegePenn State Hershey Medical College

Research Roundtable Appalachia Community Cancer NetworkResearch Roundtable Appalachia Community Cancer NetworkSeptember 28, 2007 September 28, 2007 Hershey, PAHershey, PA

Research teamResearch team

PSU: Fabian Camacho, M.S.Gary Chase, Ph.D.

WFU: Wenke Hwang, Ph.D.Kristie Long-Foley, Ph.D.

Duke: Gretchen Kimmick, M.D.

CCR: Karen Knight

NCHS: Tim Whitmire, Ph.D.

North Eastern North CarolinaCounties

0

10

20

30

40

50

60

GatesHerfordPasquotankPerquimansCamdenNorth CarolinaUS

Source, North Carolina Cancer Registry, 2005,; SEER, National Cancer Institute, 2005

Age Age adjusted adjusted Breast Breast CancerCancerDeath Death

rate per rate per 100,000100,000

ObjectivesObjectives

Describe assembly of linked-Medicaid- North Describe assembly of linked-Medicaid- North Carolina cancer registry data. Carolina cancer registry data.

Describe the use of this dataset to identify unmet Describe the use of this dataset to identify unmet needs in cancer prevention and control needs in cancer prevention and control

Discuss applicability to ACCN Discuss applicability to ACCN

Discuss health services research implicationsDiscuss health services research implications

Medicaid and Rural HealthMedicaid and Rural Health

Medicaid is an important source of health insurance Medicaid is an important source of health insurance coverage for both rural residents and rural providers.coverage for both rural residents and rural providers.

Rural residents are more likely to live in poverty than Rural residents are more likely to live in poverty than urban residents, and are less likely to have employer urban residents, and are less likely to have employer sponsored health insurance coverage. sponsored health insurance coverage.

Rural residents are more likely to be covered by Rural residents are more likely to be covered by Medicaid than are urban residents. Medicaid than are urban residents.

Percent Residents with Medicaid Coverage During thePercent Residents with Medicaid Coverage During the Past Year in Urban and Rural Counties. Past Year in Urban and Rural Counties. CPS, 2004-2005 CPS, 2004-2005

StateState RuralRural UrbanUrban

KentuckyKentucky 17.417.4 12.112.1

MarylandMaryland 16.816.8 8.68.6

New YorkNew York 17.417.4 18.318.3

OhioOhio 11.211.2 12.312.3

Pennsylvania Pennsylvania 12.012.0 11.311.3

VirginiaVirginia 11.511.5 6.96.9

West VirginiaWest Virginia 18.218.2 11.311.3

North Carolina Project North Carolina Project

NC Tumor Registry NC Tumor Registry + + Incident casesIncident cases Clinical dataClinical data 11stst course of Tx course of Tx

(FORDS)(FORDS) Hospital RegistrarsHospital Registrars Other facilitiesOther facilities Merged dataMerged data CDC proficiency CDC proficiency

testingtesting

NC Medicaid ClaimsNC Medicaid Claims All medical services* All medical services*

(ICD-9/10, CPT)(ICD-9/10, CPT) Pharmacy (NDC)Pharmacy (NDC) [monthly eligibility][monthly eligibility] [no clinical data][no clinical data]

* Variable by state* Variable by state

DatabaseDatabase

1998-99 Medicaid Claims for North Carolina.1998-99 Medicaid Claims for North Carolina. N=1,401 female breast cancer (single primary, all N=1,401 female breast cancer (single primary, all

stages, 20% all cases) stages, 20% all cases)

Caveats of Medicaid claimsCaveats of Medicaid claims: : • Managed care organizations omitted Managed care organizations omitted • Dual eligibility - MedicareDual eligibility - Medicare• Continuous enrollment - pre-and-post diagnosis.Continuous enrollment - pre-and-post diagnosis.• Bundling of claims (date/services)Bundling of claims (date/services)• Completeness of claims (unbilled services?) Completeness of claims (unbilled services?)

Data SourcesData Sources

Caveats of Hospital Registry dataCaveats of Hospital Registry data: : • First course of treatment may be defined as 4-months First course of treatment may be defined as 4-months

post diagnosis.post diagnosis.• Treatment in physician offices may go unreported.Treatment in physician offices may go unreported.• Out-of state services often missingOut-of state services often missing• May exclude VHA casesMay exclude VHA cases• Comorbidity – added in 2003Comorbidity – added in 2003• Non-registry (mid-size) hospitals lower quality data Non-registry (mid-size) hospitals lower quality data

Data SourcesData Sources

Caveats of Medicaid dataCaveats of Medicaid data: :

• 9- 12 months of continuous eligibility is generally 9- 12 months of continuous eligibility is generally needed. needed.

• Medicare files m,ay be needed for dually insured.Medicare files m,ay be needed for dually insured.• Policies on covered services may vary by state. Policies on covered services may vary by state.

MethodsMethods

Test Population (1998-99 cases)Test Population (1998-99 cases): 1,401 cases : 1,401 cases single primary breast cancer in NC registry years single primary breast cancer in NC registry years 1998 - 1999.1998 - 1999.

Test sampleTest sample: 845 (60%) cases enrolled in : 845 (60%) cases enrolled in Medicaid 1 month prior and 12 months post Medicaid 1 month prior and 12 months post registry date of diagnosisregistry date of diagnosis

ApproachApproach: : • 1) Assume registry data is accurate if treatment is listed 1) Assume registry data is accurate if treatment is listed

as provided (not missing or indeterminate). as provided (not missing or indeterminate). • 2) Else, replace data with Medicaid claims (if 2) Else, replace data with Medicaid claims (if

discordant).discordant).• 3) Validate by performing record review on sample of 3) Validate by performing record review on sample of

cases. cases.

Accuracy of Radiation data in BCS sampleAccuracy of Radiation data in BCS sample

BCS sample BCS sample Sensitivity Specificity Sensitivity Specificity Registry 84% 100%Registry 84% 100% Claims 95% 93%Claims 95% 93%

CombinedCombined 97% 98%97% 98%

Table 3. Adjusted Odds of Registry Codes for Radiation and Chemotherapy when Medicaid Claims are Present

Sample NSample N

Caucasian vs OtherCaucasian vs Other

# with radiation claims: 279# with radiation claims: 279

0.89 (0.48, 1.66)0.89 (0.48, 1.66)

# with chemotherapy claims: 236# with chemotherapy claims: 236

0.57 (0.31, 1.05)0.57 (0.31, 1.05)

Dually Eligible No vs YesDually Eligible No vs Yes 0.72 (0.37,1.40)0.72 (0.37,1.40) 2.33 (1.24, 4.41)2.33 (1.24, 4.41)

N/A vs 5 + cmN/A vs 5 + cm 1.10 (0.36,3.31)1.10 (0.36,3.31) 0.62 (0.17,2.35)0.62 (0.17,2.35)

0-1 cm vs 5+ cm0-1 cm vs 5+ cm 2.39 (0.67, 8.54)2.39 (0.67, 8.54) 0.27 (0.05,1.65)0.27 (0.05,1.65)

1-2 cm vs 5+ cm1-2 cm vs 5+ cm 1.71 (0.60, 4.84)1.71 (0.60, 4.84) 0.65 (0.18,2.27)0.65 (0.18,2.27)

2-5 cm vs 5+ cm2-5 cm vs 5+ cm 0.74 (0.29,1.94)0.74 (0.29,1.94) 0.35 (0.12,1.09)0.35 (0.12,1.09)

Lymph Nodes RemovedLymph Nodes Removed 1.15 (0.58,2.25)1.15 (0.58,2.25) 0.94 (0.45,1.95)0.94 (0.45,1.95)

Class of Case 1 or 2 present 4Class of Case 1 or 2 present 4 5.27 (0.54,51.14)5.27 (0.54,51.14) 7.74 (1.22,49.09)7.74 (1.22,49.09)

0.11 (0.037, 0.35)0.11 (0.037, 0.35) 0.43 (0.18,1.05)0.43 (0.18,1.05)Registry FacilityRegistry Facility7.77 (2.80,21.58)7.77 (2.80,21.58) 4.31 (2.08,8.94)4.31 (2.08,8.94)

Days to first claims since dx Days to first claims since dx (Rad column, Chem column)(Rad column, Chem column) 0.989 (0.985,0.993)0.989 (0.985,0.993) 0.994 (0.988, 0.998)0.994 (0.988, 0.998)

Tum

or S

ize

Tum

or S

ize

CCR Agreement on RadiationCCR Agreement on

Chemotherapy

Examples of Application to Answer Patterns of CareExamples of Application to Answer Patterns of Care And Outcomes Research Questions And Outcomes Research Questions

Total Radiation No Radiation p

N= 344 N=242(70.4%)

N=102(30.0%)

Age group at time of diagnosis

65+ years 159 (46.2%) 58.49 41.51 <.0001 <65 years 185 (53.8%) 80.54 19.46

Race/ethnicity White 175 (50.9%) 69.71 30.29 0.7931 Other 169 (49.1%) 71.01 28.99

Charlson Comorbidity Score (Excluding Cancer dx)0 166 (48.26%) 75.90 24.10 0.09311 52 (15.12%) 65.38 34.62

2 126 (36.63%) 65.08 34.92

Patient County of Residence

Non-metropolitan county 116 (33.7%) 64.66 35.34 0.0991 Metropolitan county 228 (66.3%) 73.25 26.75

County Medicaid Density

High (> 50 percentile) 167 (48.55) 68.26 31.74 0.4107 Low 177 (51.45) 72.32 27.68

Assisted Living1

Yes 80 (23.3%) 52.50 47.50 0.0001 No 264 (76.7%) 75.76 24.24

Hospital Size2

Large / Medium 302 (72.52%) 72.52 27.48 0.0182 Small 42 (12.21%) 54.76 45.24

Surgery Hospital Volume3

High/Medium (> 5000 patients discharges) 301 (87.50%) 72.43 27.57 0.0257 Low 43 (12.50%) 55.81 44.19

Breast cancer patient volume (Medicaid)4

Higher >50% 312 (90.70%) 72.44 27.56 0.0081 Lower ≤ 50% 32 (9.30%) 50.00 50.00

1. Based on presence of any paid claims from nursing home (location of service=T), home healthcare services (Q) or skilled nursing facility (cos = 35,36). 2. Tertile distribution of number of beds reported by American Hospital Directory. 3. Based on patient discharges reported by American Hospital Directory 4. Median split of all breast cancer cases with Medicaid enrollment.

Correlates of Under Use of Radiation Treatment with BCS in North Carolina Medicaid

Table 3. Interaction Graph showing unadjusted proportions of Radiation Treatment in Table 3. Interaction Graph showing unadjusted proportions of Radiation Treatment in BCS patients by Metropolitan Status and Hospital SizeBCS patients by Metropolitan Status and Hospital Size

0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1

L/M SHospital Size

% BCS who received radiation

Metro NonMetro

Kaplan-Meier Survival Curves of All-Cause Mortality by Radiotherapy Treatment

0.00

0.25

0.50

0.75

1.00

0 500 1000 1500 2000 2500

Days since diagnosis

No Radiation Radiation

Cormorbidity among Women with Breast Cancer in NC MedicaidCormorbidity among Women with Breast Cancer in NC Medicaid

N= 1,401N= 1,401

● ● 55 % had at least one other comorbid condition defined in Charlson 55 % had at least one other comorbid condition defined in Charlson comorbidity index comorbidity index

● ● Among those with comorbidity, > 50% had multiple conditions.Among those with comorbidity, > 50% had multiple conditions. ● ● The top three comorbid conditions were: The top three comorbid conditions were: Diabetes (26%) Diabetes (26%) Congestive heart failure (18%), Congestive heart failure (18%), Chronic pulmonary disease (11%). Chronic pulmonary disease (11%).

North Carolina Medicaid enrollees with a diagnosis of diabetes with and without breast cancer:

Medication Possession Ratio: diabetes medicines/ insulin products 12 month days supply

mean days (SD) ____________________________________________________

Women without cancer 300 (19) With breast cancer 205 (95)

Hormone Medication Persistence. NC Medicaid Breast CancerHormone Medication Persistence. NC Medicaid Breast Cancer

Monali Bhosle OSUMonali Bhosle OSU

Cumulative nonpersistence rate based on number with + ER status who started therapy. No significant association between race and medication persistence adjusting for a type of index therapy and other confounders (hazard ratio (SE) [95%CI]: 1.13 (0.30) [0.68-1.89].

0.00

0.25

0.50

0.75

1.00

Cum

ulat

ive

pers

iste

nce

rate

to in

dex

med

icat

ion

0 50 100 150 200Days on Therapy

White Black

Kaplan-Meier Estimates for Matched Cohort

Current Multi-Site StudyCurrent Multi-Site Study

● ● CDC Patterns of Care Study (Breast and CDC Patterns of Care Study (Breast and Prostate CA)Prostate CA)

7 States + PSU7 States + PSURegistry data + CMS Registry data + CMS Centralized data processing (5 sites).Centralized data processing (5 sites).Pooled data analysisPooled data analysis

ModelModelNCCN guideline concordant careNCCN guideline concordant careHealth system and patient characteristicsHealth system and patient characteristicsCare coordinationCare coordination

ProposalProposal

● ● Develop set of cancer prevention and control outcomes consistent Develop set of cancer prevention and control outcomes consistent with CDC pattern of care studies.with CDC pattern of care studies. ● ● Seek funding to link 7 ACCN States Registry data to Seek funding to link 7 ACCN States Registry data to Medicaid/Medicare.Medicaid/Medicare. ● ● PSCI Health Services Research Core serve as Data and Support PSCI Health Services Research Core serve as Data and Support Center to provide:Center to provide:

IRB templatesIRB templatesData acquisition and linkage Data acquisition and linkage ArchiveArchiveAnalysis supportAnalysis support