Anda di halaman 1dari 23

Supplementary Appendix

This appendix has been provided by the authors to give readers additional information about their work. Supplement to: Mitchell JM. Urologists use of intensity-modulated radiation therapy for prostate cancer. N Engl J Med 2013;369:1629-37. DOI: 10.1056/NEJMsa1201141

Supplementary Materials for MS# 12-01141.R6 Urologists Use of Intensity Modulated Radiation Therapy for Prostate Cancer Author: Jean M. Mitchell, Ph.D., Professor of Public Policy, Georgetown University Table of Contents 1) 2) 3) 4) 5) Sample Selection Criteria Submitted to the Center for Medicare and Medicaid Services SAS Program Used to Extract Sample of Men with Newly Diagnosed Prostate Cancer SAS Program to Construct Variables Used in IMRT Analysis Specification of the Empirical Model Table S1: Linear Probability and Logistic Regression Difference-in-Differences Estimates Predicting Receipt of IMRT Treatments among Medicare Fee-for-Service Beneficiaries with Newly Diagnosed, Non-metastatic Prostate Cancer Analysis of Time-to-Treat by Self-referral Status Figure S1: Frequency of Use of IMRT among Men Age 65-79 with Newly Diagnosed Prostate Cancer-- Private Practice Self-referring Urologists versus Private Practice Non Self-referring Urologists Figure S2: Frequency of Use of IMRT among Men Age 80 Plus with Newly Diagnosed Prostate Cancer--Private Practice Self-referring Urologists versus Private Practice Non Self-referring Urologists Figure S3: Frequency of Use of IMRT among Men Age 65-79 with Newly Diagnosed Prostate Cancer--Private Practice Self-referring Urologists versus National Comprehensive Cancer Network Non Self-referring Urologists Figure S4: Frequency of Use of IMRT among Men Age 80 Plus with Newly Diagnosed Prostate Cancer--Private Practice Self-referring Urologists versus National Comprehensive Cancer Network Non Self-referring Urologists

6) 7)

8)

9)

10)

Sample Selection Criteria Submitted to the Center for Medicare and Medicaid Services The Center for Medicare and Medicaid Services algorithm to identify beneficiaries with prostate cancer is available on the Chronic Conditions Warehouse website (ww.ccwdata.org/web/guest/conditions-categories). Prostate cancer is identified by at least 1 inpatient or two carrier or hospital outpatient claims with ICD-9 diagnosis codes 185 or 233.4. My data request to CMS stipulated using the Chronic Conditions Warehouse algorithm to extract Medicare beneficiaries with prostate cancer who were continuously enrolled in fee-for-service during the time period 2005-2010. Only beneficiaries who resided in 26 designated states were selected. We attempted to identify newly diagnosed prostate cancer cases in each year by a date variable that identifies the first occurrence of prostate cancerthe variable CANCER_PROSTATE_EVER. The data came from 17 states where the 37 self-referring practices identified by the Wall Street Journal were located. We also requested data for some nearby states where there were no known selfreferring practices. These states include: Connecticut, Georgia, Massachusetts, Michigan, Minnesota, North Dakota, Tennessee, Utah and Washington. In the process of working with the CMS vendor to revise the parameters of the original data request to include additional states, Ohio was inadvertently excluded. Two of the self-referring practices identified by the Wall Street Journal were located in Ohio.

SAS Program to Extract Sample of Men with Newly Diagnosed Prostate Cancer We used the SAS program presented below to extract men with newly diagnosed prostate cancer from the initial data extract obtained. Men with metastatic prostate cancer were excluded. These cases were identified by ICD-9 codes reported as secondary diagnosis codes on the claims. These codes are: 198.5 (secondary neoplasm of bone and bone marrow) and 196 (secondary and unspecified malignant neoplasm of lymph nodes); ICD-9 code 196 has 8 sub codes which describe location. PROGRAM CODE %LET GROUP=; %LET PRED=; %LET POSTD=; %LET GRPTAX=; %LET GRPUPIN=; %LET GRPNPI=; %MACRO PYR(YR); *** check all claims for IDs from group of interest ***; DATA PRERADGFL_&YR (KEEP=BENE_ID GROUP GROUP_: PRE POST); SET INSASG.CMSRAD_&YR; BY BENE_ID; LENGTH GROUP $10.; IF LEFT(TRIM(TAX_NUM)) IN (&GRPTAX) OR LEFT(TRIM(PRF_PHYSN_UPIN)) IN (&GRPUPIN) OR LEFT(TRIM(PRF_PHYSN_NPI)) IN (&GRPNPI) THEN DO; GROUP_&GROUP=1; GROUP="&GROUP"; PRE=MDY(&PRED); POST=MDY(&POSTD); END; IF GROUP NE '' THEN OUTPUT;

RUN;

*** capture all claims for bene that had claim with group of interest - do per year ***; PROC SORT NODUPKEY DATA=PRERADGFL_&YR; BY BENE_ID GROUP; RUN; DATA PRERADGFL2_&YR; MERGE PRERADGFL_&YR (IN=A) INSASG.CMSRAD_&YR (IN=B); BY BENE_ID; IF A;

RUN; %MEND PYR; %PYR(05); %PYR(06); %PYR(07); %PYR(08); %PYR(09); %PYR(10);

%MACRO GRP(GROUP); DATA PRERAD&GROUP._ALLYR; SET PRERADGFL2_05

PRERADGFL2_06 PRERADGFL2_07 PRERADGFL2_08 PRERADGFL2_09 PRERADGFL2_10; IF GROUP="&GROUP"; RUN; PROC SORT DATA=PRERAD&GROUP._ALLYR OUT=INSASRAD.PRERAD&GROUP._ALLYR; BY BENE_ID CLM_THRU_DT; RUN;

%MEND GRP;

SAS Program to Construct Variables Used in IMRT Analysis DATA EPISODEBENEPRE_&GROUP (KEEP=BENE_ID BIOPWCANCDATE BIOPSYDATE RENAME=(BIOPWCANCDATE=NEWYEAR)); SET INSASRAD.PRERAD&GROUP._ALLYR; BY BENE_ID CLM_THRU_DT; RETAIN BIOPWCANCDATE BIOPSYDATE BIOPWCANCF; IF FIRST.BENE_ID THEN DO; BIOPWCANCF=0; BIOPWCANCDATE=.; BIOPSYDATE=.; END; *** flag claims with group ***; IF LEFT(TRIM(TAX_NUM)) IN (&GRPTAX) OR LEFT(TRIM(PRF_PHYSN_UPIN)) IN (&GRPUPIN) OR LEFT(TRIM(PRF_PHYSN_NPI)) IN (&GRPNPI) THEN GROUPCLAIM=1; *** flag claims for prostate cancer diag (including in situ) ***; ARRAY DIAG(*) $ ICD9_DGNS_CD1-ICD9_DGNS_CD8; DO I=1 TO DIM(DIAG); IF DIAG(I) IN ("185","2334") THEN DO; IF (CLM_THRU_DT-30) LE BIOPSYDATE THEN DO; BIOPWCANCF=1; IF BIOPWCANCDATE=. THEN BIOPWCANCDATE=CLM_THRU_DT; END; END; END; *** create flags for needle biopsy performed by group ***; IF HCPCS_CD="55700" AND GROUPCLAIM=1 THEN DO; IF (BIOPSYDATE=. OR BIOPWCANCF=0) THEN BIOPSYDATE=CLM_THRU_DT; END;

IF LAST.BENE_ID THEN OUTPUT; RUN; DATA EPISODEBENEPRE2_&GROUP (RENAME=(PRE=PREDATE POST=POSTDATE)); MERGE EPISODEBENEPRE_&GROUP (IN=A) INSASRAD.PRERAD&GROUP._ALLYR (IN=B); BY BENE_ID; *** 6 month episode ***; IF A AND B AND ((NEWYEAR-41) LE CLM_THRU_DT LE (NEWYEAR+183)); RUN; DATA EPISODEBENE_&GROUP (DROP=I J K) EPISODEBENE3_&GROUP (DROP=I J K) EPISODEALL_&GROUP (DROP=I J K); SET EPISODEBENEPRE2_&GROUP; BY BENE_ID CLM_THRU_DT; RETAIN &GROUP.COUNT FIRST&GROUP.DATE LAST&GROUP.DATE &GROUP.OFFVISF CDATE PROSTATEF FIRST&GROUP.CDATE BIOPWCANCF BIOPWCANCDATE &GROUP.CANCERF BIOPSYDATE CANCERONBIOPDATE CANCERONBIOPF &GROUP.BIOPSYDATE IMRTPF IMRTDF FIRSTIMRTPDATE FIRSTIMRTDDATE FIRST&GROUP.IMRTDATE &GROUP.IMRTPF &GROUP.IMRTDF NOT_&GROUP.IMRTDF NOT_&GROUP.IMRTDC PROSTATECTOMYF &GROUP.PROSTATECTOMYF BRACHYF &GROUP.BRACHYF HORMONEF &GROUP.HORMONEF CRYOF &GROUP.CRYOF RADIOF &GROUP.RADIOF TRANSF &GROUP.TRANSF TREATDATE TREATAGE IMRTPC IMRTDC &GROUP.IMRTPC &GROUP.IMRTDC &GROUP.DAYS &GROUP.DATE CANCERPROC;

IF FIRST.BENE_ID THEN DO; ARRAY INITZ(*) TREATAGE &GROUP.COUNT &GROUP.OFFVISF PROSTATEF BIOPWCANCF &GROUP.CANCERF CANCERONBIOPF IMRTPF IMRTDF &GROUP.IMRTPF &GROUP.IMRTDF PROSTATECTOMYF &GROUP.PROSTATECTOMYF BRACHYF &GROUP.BRACHYF HORMONEF &GROUP.HORMONEF CRYOF &GROUP.CRYOF RADIOF &GROUP.RADIOF TRANSF &GROUP.TRANSF IMRTPC IMRTDC &GROUP.IMRTPC &GROUP.IMRTDC &GROUP.DAYS CANCERPROC NOT_&GROUP.IMRTDF NOT_&GROUP.IMRTDC; DO J=1 TO DIM(INITZ); INITZ(J)=0; END; ARRAY INITM(*) FIRST&GROUP.DATE LAST&GROUP.DATE CDATE FIRST&GROUP.CDATE BIOPWCANCDATE BIOPSYDATE CANCERONBIOPDATE &GROUP.BIOPSYDATE FIRSTIMRTPDATE FIRSTIMRTDDATE FIRST&GROUP.IMRTDATE TREATDATE &GROUP.DATE; DO K=1 TO DIM(INITM); INITM(K)=.; END; END; *** flag claim if any id is with group of interest - also first date of claim with group, count of claims with group, and last claim date with group ***; IF TAX_NUM IN (&GRPTAX) OR PRF_PHYSN_UPIN IN (&GRPUPIN) OR PRF_PHYSN_NPI IN (&GRPNPI) THEN DO; &GROUP.CLAIM=1; IF &GROUP.COUNT=0 THEN DO; FIRST&GROUP.DATE=CLM_THRU_DT; &GROUP.DAYS+1; &GROUP.DATE=CLM_THRU_DT; END; &GROUP.COUNT+1; LAST&GROUP.DATE=CLM_THRU_DT; IF CLM_THRU_DT > &GROUP.DATE THEN DO; &GROUP.DAYS+1; &GROUP.DATE=CLM_THRU_DT; END; END; 8

*** flag claims for prostate cancer diag (including in situ) ***; ARRAY DIAG(*) $ ICD9_DGNS_CD1-ICD9_DGNS_CD8; DO I=1 TO DIM(DIAG); IF DIAG(I) IN ("185","2334") THEN DO; *** flag claim as prostate cancer, retain flag for bene with prostate cancer, note date of first diagnosis ***; PROSTATE=1; PROSTATEF=1; IF CDATE=. THEN DO; CDATE=CLM_THRU_DT; END; *** flag bene for having a cancer diagnosis on a claim by a group dr ***; IF &GROUP.CLAIM=1 THEN DO; &GROUP.CANCERF=1; IF FIRST&GROUP.CDATE=. THEN FIRST&GROUP.CDATE=CLM_THRU_DT; END; IF (CLM_THRU_DT-30) LE BIOPSYDATE THEN DO; BIOPWCANCF=1; IF BIOPWCANCDATE=. THEN BIOPWCANCDATE=CLM_THRU_DT; END; IF CLM_THRU_DT=BIOPWCANCDATE THEN CANCERPROC=1; END; END; *** flag claim and bene for having office visit with dr in group ***; IF SUBSTR(HCPCS_CD,1,3)='992' AND &GROUP.CLAIM=1 THEN DO; &GROUP.OFFVIS=1; &GROUP.OFFVISF=1; END; *** create flags for needle biopsy ***; IF HCPCS_CD="55700" THEN DO; IF (BIOPSYDATE=. OR BIOPWCANCF=0) THEN DO; BIOPSYDATE=CLM_THRU_DT; IF PROSTATE=1 THEN DO; CANCERONBIOPDATE=CLM_THRU_DT; CANCERONBIOPF=1; END; END; 9

IF &GROUP.CLAIM=1 THEN &GROUP.BIOPSYDATE=CLM_THRU_DT; END; *** flag treatments - from group and from anyone ***; IF HCPCS_CD="77301" THEN DO; IMRTP=1; IMRTPF=1; IMRTPC+SUM(REV_CNTR_UNIT_CNT,LINE_SRVC_CNT); IF FIRSTIMRTPDATE=. THEN FIRSTIMRTPDATE=CLM_THRU_DT; IF &GROUP.CLAIM=1 THEN DO; IF FIRST&GROUP.IMRTDATE=. THEN FIRST&GROUP.IMRTDATE=CLM_THRU_DT; &GROUP.IMRTP=1; &GROUP.IMRTPF=1; &GROUP.IMRTPC+SUM(REV_CNTR_UNIT_CNT,LINE_SRVC_CNT); END; END; IF HCPCS_CD="77418" THEN DO; IMRTD=1; IMRTDF=1; IMRTDC+SUM(REV_CNTR_UNIT_CNT,LINE_SRVC_CNT); IF FIRSTIMRTDDATE=. THEN FIRSTIMRTDDATE=CLM_THRU_DT; IF &GROUP.CLAIM=1 THEN DO; IF FIRST&GROUP.IMRTDATE=. THEN FIRST&GROUP.IMRTDATE=CLM_THRU_DT; &GROUP.IMRTD=1; &GROUP.IMRTDF=1; &GROUP.IMRTDC+SUM(REV_CNTR_UNIT_CNT,LINE_SRVC_CNT); END; IF &GROUP.CLAIM NE 1 THEN DO; NOT_&GROUP.IMRTD=1; NOT_&GROUP.IMRTDF=1; NOT_&GROUP.IMRTDC+SUM(REV_CNTR_UNIT_CNT,LINE_SRVC_CNT); END; END; IF HCPCS_CD IN ("55801","55810","55812","55815","55821","55831","55840","55842","55845","55866","55899") THEN DO; PROSTATECTOMY=1; PROSTATECTOMYF=1; IF &GROUP.CLAIM=1 THEN &GROUP.PROSTATECTOMYF=1; END; IF HCPCS_CD IN ("55875","55860","55862","55865","77778","77784","77787") THEN DO; BRACHY=1; BRACHYF=1; if &GROUP.CLAIM=1 THEN &GROUP.BRACHYF=1; END; IF HCPCS_CD IN ("54520","J1950","J9217","J9218","J9219","J9202") THEN DO; HORMONE=1; HORMONEF=1; IF &GROUP.CLAIM=1 THEN &GROUP.HORMONEF=1; END; IF HCPCS_CD IN ("55873") THEN DO; CRYO=1; CRYOF=1; IF &GROUP.CLAIM=1 THEN &GROUP.CRYOF=1; END;

10

IF HCPCS_CD IN ("77435","G0339","G0340") THEN DO; RADIO=1; RADIOF=1; IF &GROUP.CLAIM=1 THEN &GROUP.RADIOF=1; END; IF HCPCS_CD IN ("53850","53852","53853", "52601") THEN DO; TRANS=1; TRANSF=1; IF &GROUP.CLAIM=1 THEN &GROUP.TRANSF=1; END; *** note first date of treatment ***; IF TREATDATE=. AND (PROSTATECTOMY=1 OR BRACHY=1 OR IMRTD=1) THEN DO; TREATDATE=CLM_THRU_DT; TREATAGE=BENE_AGE_AT_END_REF_YR; END; LABEL PROSTATE="claim has malignant neoplasm/in situ of prostate" PROSTATEF="bene has malignant neoplasm/in situ of prostate" SECONDF="bene has a secondary treatment - prostatectomy, brachy" TERTIARYF="bene has a tertiary treatment - cryo, radio, trans" IMRTDF="bene had IMRT delivery (77418)" IMRTPF="bene had IMRT planning (77301)" BIOPWCANCF="bene has a biopsy and a cancer diagnosis w/in 30 days" CANCERONBIOPF="bene has a cancer diagnosis on a biopsy claim" EPLENGTH="length of time from first date with group to last date" CANCTOTREAT="length of time from cancer diagnosis to treatment" AFTERTREATMENT="bene had first claim with group after their first treatment date" BEFORETREATMENT="bene had last claim with group before their first treatment date"; IF LAST.BENE_ID THEN DO; LASTDATE=CLM_THRU_DT; EPLENGTH=LAST&GROUP.DATE-FIRST&GROUP.DATE; CANCTOTREAT=TREATDATE-(MAX(BIOPWCANCDATE,CANCERONBIOPDATE));

IF FIRST&GROUP.DATE>TREATDATE AND TREATDATE NE . THEN AFTERTREATMENT=1;

11

IF LAST&GROUP.DATE<TREATDATE AND TREATDATE NE . THEN BEFORETREATMENT=1; IF (&GROUP.PROSTATECTOMYF=1 OR &GROUP.BRACHYF=1 OR &GROUP.IMRTDF=1) THEN &GROUP.TREATMENT=1; ELSE &GROUP.TREATMENT=0; IF TREATDATE NE . THEN TREATMENT=1; ELSE TREATMENT=0; *** newly diagnosed if bene has biopsy with cancer diag w/in 30 days, but not first diag on biopsy ***; IF BIOPWCANCF=1 AND CANCERONBIOPF=0 THEN NEWDIAG=1; ELSE NEWDIAG=0; IF PROSTATECTOMYF=1 OR BRACHYF=1 THEN SECONDF=1; ELSE SECONDF=0; IF CRYOF=1 OR RADIOF=1 OR TRANSF=1 THEN TERTIARYF=1; IF IMRTDF=1 AND SECONDF NE 1 THEN IMRTDONLY=1; ELSE IMRTDONLY=0; ***** do not include benes that start in last 6mo *****; IF NEWDIAG=1 AND BIOPWCANCDATE>MDY(6,30,2010) THEN NEWDIAG=0; IF &GROUP.PROSTATECTOMYF=1 OR &GROUP.BRACHYF=1 THEN &GROUP.SECONDF=1; IF &GROUP.CRYOF=1 OR &GROUP.RADIOF=1 OR &GROUP.TRANSF=1 THEN &GROUP.TERTIARYF=1; IF &GROUP.IMRTDF=1 AND SECONDF NE 1 THEN &GROUP.IMRTDONLY=1; ELSE &GROUP.IMRTDONLY=0; IF NOT_&GROUP.IMRTDF=1 AND SECONDF NE 1 THEN NOT_&GROUP.IMRTDONLY=1; ELSE NOT_&GROUP.IMRTDONLY=0;

IF TREATDATE NE . AND TREATDATE LE PREDATE THEN PRE=1; ELSE IF TREATDATE NE . AND TREATDATE GE POSTDATE THEN POST=1; ELSE IF BIOPWCANCDATE LE PREDATE THEN PRE=1; ELSE IF BIOPWCANCDATE GE POSTDATE THEN POST=1;

12

*** 6 month requirement for pre period ***; IF PRE=1 AND NEWDIAG=1 AND BIOPWCANCDATE>(PREDATE-183) THEN NEWDIAG=0; IF FIRSTIMRTPDATE NE . AND FIRSTIMRTDDATE NE . THEN IMRTP2D=FIRSTIMRTDDATEFIRSTIMRTPDATE; *** do not include benes with <30 imrtd ***; IF IMRTDONLY=1 AND IMRTDC<30 THEN NEWDIAG=0; *** output dataset for all benes, plus for benes with 3 or more days of claims with group ***; IF &GROUP.DAYS > 2 THEN OUTPUT EPISODEBENE3_&GROUP; OUTPUT EPISODEBENE_&GROUP; END; OUTPUT EPISODEALL_&GROUP; RUN;

13

Specification of the Empirical Model

The dependent variable receipt of IMRT treatments equals 1 if: 1) the beneficiary was seen by a non self-referring urologist in either the preownership or ownership periods and he received IMRT treatments; or 2) the beneficiary was seen by a self-referring urologist in the preownership period and he received IMRT treatments; or 3) the beneficiary was seen by a self-referring urologist in the ownership period and he received IMRT treatments performed and billed by the self-referring urology group. For all other observations the dependent variable equals zero. This includes beneficiaries seen by a self-referring urologist during the ownership period that underwent IMRT treatments performed and billed by a non self-referring provider (about 6% of the primary sample). Although these beneficiaries received IMRT treatments, assigning a value of 1 to these cases would bias upward the coefficient on the self-referral variable. The unadjusted analyses only control for: 1) whether the case was treated in the preownership or ownership period, 2) self-referral status and 3) whether the patient was treated by a self-referring urologist after acquiring ownership. The unadjusted analyses do not control for any other confounding factors. Each matched pair has the same preownership and ownership period. The regression adjusted models included controls for patient age, comorbid conditions, year of diagnosis and indicator variables identifying the urology group that treated each beneficiary. Patient age and coexisting conditions are included because these factors are likely to be taken into account in selecting the most appropriate treatment. For example, monitoring approaches are probably more appropriate for older men as the risks of undergoing IMRT are likely to outweigh the benefits. A similar argument can be made regarding coexisting conditions. We used a modified version of the methodology developed by Anne Elixhauser and co-authors.1 They developed coexisting condition indicators using only inpatient claims whereas our approach relies on inpatient, hospital outpatient and carrier files claims. Year of cancer diagnosis was included to recognize that IMRT, a relatively new technology, was 14

rapidly adopted during the time period of study, despite uncertainty regarding its relative effectiveness.2 Indicator variables identifying the 70 urology groups that comprise the primary sample were included to control for variations in physician practices patterns that are likely to exist but are unrelated to selfreferral status. References 1. Elixhauser, A, Steiner, C, Harris, DR and Coffey RM. Comorbidity Measures for Use with Administrative Data. Medical Care. 1998; 36(1): 8-27. 2 . Jacobs BL, Zhong Y, Skolarus, TA, Hollenbeck, BK. Growth of High-Cost Intensity Modulated Radiation Therapy for Prostate Cancer Raises Concerns about Overuse. Health Affairs, 2012; 31(4): 750-759.

15

Table S1. Linear Probability and Logistic Regression Difference-in-Differences Estimates Predicting Receipt of IMRT Treatments among Fee-for-Service Medicare Beneficiaries with Newly Diagnosed Non-metastatic Prostate Cancer a Private Practice Self-referring Urologists vs. Private Practice Non Self-referring Urologists c Beneficiary Treated by Selfreferring Urologist during Ownership Period (n= 38,765)
Marginal EffectLinear Probability Model Marginal EffectLogistic Regression Model Odds RatioLogistic Regression Modele Adjusted R2Linear Probability Model Pseudo R2Logistic Regression Model 16.4% points (p<.001) 16.9% points (p<.001) 2.79 (2.53, 3.08) (p<.001) 0.10 0.12

Private Practice Self-referring Urologists vs. NCCN Non Self-referring Urologists d Beneficiary Treated by Selfreferring Urologist during Ownership Period (n=6,773)
29.3% points (p<.001) 30.0% points (p<.001) 6.18 (4.77, 8.01) (p<.001) 0.19 0.19

Notes: a The dependent variable Receipt of IMRT Treatments equals 1 if: 1) the beneficiary was seen by a non
self-referring urologist in either the preownership or ownership periods and he received IMRT treatments; or 2) the beneficiary was seen by a self-referring urologist in the preownership period and he received IMRT treatments; or 3) the beneficiary was seen by a self-referring urologist in the ownership period and he received IMRT treatments performed and billed by the self-referring urology group. For all other observations the dependent variable equals zero. This includes beneficiaries seen by a self-referring urologist during the ownership period that underwent IMRT treatments performed and billed by a non self-referring provider. Although these beneficiaries received IMRT treatments, assigning a value of 1 to these cases would bias upward the coefficient on the selfreferral variable.
b

All regression models include beneficiary age at cancer diagnosis, indicator variables to distinguish year of diagnosis, indicator variables to identify the presence of absence of specific coexisting conditions, and indicator variables to control for the urology group that treated each beneficiary.
c

Sample includes 38,765 beneficiaries treated by 1) private practice physicians from 35 self-referring urology groups that began billing Medicare for IMRT at some point during 2005-2010, and 2) private practice physicians from 35 matched non self-referring urology groups that did not bill Medicare for IMRT.
d

Sample includes 6,773 beneficiaries treated by 1) private practice physicians from 11 self-referring urology groups that began billing Medicare for IMRT at some point during 2005-2010, and 2) salaried physicians from 11 matched non self-referring urology groups employed by a National Comprehensive Cancer Network center.
e

For the logistic regression model, 95% confidence interval is reported in parentheses below the odds ratio.

16

Analysis of Time-to-Treat by Self-referral Status Self-referring urologists contend that the vertically integrated urology-radiation oncology practice reduces time-to-treat, that is, time from date of cancer diagnosis to initiation of definitive treatment. If so then one would expect to find that men treated by self-referring urologists will experience a reduction in time-to-treat compared to men who obtained care from non self-referring urologists. Results reported below show this is not the case. After controlling for type of definitive treatment received (brachytherapy, IMRT only or prostatectomy) and other confounding factors, timeto-treat does not vary by self-referral status. Table S2. Length of Time from Date of Prostate Cancer Diagnosis to Date of Receipt of Definitive Treatment (Brachytherapy, Prostatectomy, IMRT) Sample Difference-in-Differences (Self-referral) Effect (Unadjusted) -3.0 (p<.001) -6.4 (p<.001) Difference-in-Differences (Self-referral) Effect (Adjusted)b 1.29 (p = 0.12) 1.85 (p = 0.39)

Private Practice Self-referring Urologists Versus Private Practice Non Self-referring Urologists Private Practice Self-referring Urologists versus National Comprehensive Cancer Network Non Self-referring Urologists

Notes: a Time-to-treat was defined as length of time from date of prostate cancer diagnosis to date of receipt of definitive treatment (brachytherapy, IMRT only or prostatectomy,). b Other control variables included in the regression were indicator variables to identify which group treated each beneficiary, type of definitive treatment (brachytherapy, IMRT only or prostatectomy,), age, year of cancer diagnosis, and indicator variables to identify the presence or absence of specific coexisting conditions.

Private Practice Self-referring Urologists versus Private Practice Non Self-referring Urologists Table S2 reports the unadjusted and regression-adjusted difference-in-differences results on time-to-treat. Row 1 shows the results based on private practice self-referring and private practice non self-referring urologists. The unadjusted difference-in-differences estimator implies self-referral was associated with a three day reduction in time-to-treat (p<.001).

17

In addition, we estimated a regression where the dependent variable was length of time in days from date of diagnosis until receipt of definitive treatment controlling for treatment by a self-referring urologist in the ownership period, urology group fixed effects, type of definitive treatment (brachytherapy, IMRT only or prostatectomy), age, year of cancer diagnosis and comorbid illnesses. The coefficient on the self-referral variable was 1.29 (p = 0.12) and was not statistically significant. Thus, the regression-adjusted results show there was no difference in length of time from date of cancer diagnosis to receipt of definitive treatment associated with self-referral. Private Practice Self-referring Urologists versus National Comprehensive Cancer Network Non Self-referring Urologists Difference-in-difference results from time-to-treat comparisons of men treated by selfreferring urologists and their non self-referring counterparts employed by cancer centers are reported in row 2 of Table S2. The unadjusted difference-in-differences estimator or self-referral effect was a decline in time-to-treat of 6.4 days (p<.001). Again, we employed regression analysis to evaluate whether self-referral is linked to reduction in time-to-treat. The regression adjusted difference-indifferences analysis shows that after controlling for other confounding factors, self-referral had no impact on time from date of cancer diagnosis to receipt of definitive treatment. The coefficient on the self-referral indicator was 1.85 (p = 0.39) and was statistically insignificant.

18

19

20

21

22

Anda mungkin juga menyukai