Arts & Architecture

Using administrative data to identify and stage breast cancer cases: implications for assessing quality of care.

Description
Using administrative data to identify and stage breast cancer cases: implications for assessing quality of care.
Published
of 9
All materials on our website are shared by users. If you have any questions about copyright issues, please report us to resolve them. We are always happy to assist you.
Related Documents
Share
Transcript
  See discussions, stats, and author profiles for this publication at: https://www.researchgate.net/publication/51709049 Using administrative data to identify and stagebreast cancer cases: Implications for assessingquality of care  Article  · July 2011 DOI: 10.1700/950.10393 · Source: PubMed CITATIONS 17 READS 578 8 authors , including:Daniel LouisThomas Jefferson University 81   PUBLICATIONS   1,439   CITATIONS   SEE PROFILE Vittorio MaioThomas Jefferson University 105   PUBLICATIONS   948   CITATIONS   SEE PROFILE All content following this page was uploaded by Daniel Louis on 28 December 2016. The user has requested enhancement of the downloaded file. All in-text references underlined in blue are added to the srcinal documentand are linked to publications on ResearchGate, letting you access and read them immediately.  Key words:  breast cancer, case find-ing, epidemiology, quality indicators.  Acknowledgments:  The authorswould like to thank the physicians andstaff of the cancer registries that pro-vided data for the study as well as thephysicians and staff of the hospitalswho created the  Scheda di dimissioneospedaliera  (SDO). Without their ef-fort and collaboration, the projectwould not have been possible. Wewould also like to thank the anony-mous reviewers from Tumori for theirhelpful comments which have im-proved the quality of the manuscript. Financial disclosure:  This work wassupported by funding from the  Agen- zia Sanitaria e Sociale Regionale , Re-gione Emilia-Romagna, Italy Correspondence to:  Elaine J Yuen,Suite 119, 1025 Walnut Street,Philadelphia, PA 19107, USAe-mail elaine.yuen@jefferson.eduReceived March 19, 2010;accepted April 4, 2011. Using administrative data to identify and stagebreast cancer cases: implications for assessingquality of care Elaine Yuen 1,2 , Daniel Louis 1 , Luca Cisbani 3 , Carol Rabinowitz 1 ,Rossana De Palma 3 , Vittorio Maio 1,2 , Maurizio Leoni 3,4 , and Roberto Grilli 3 1 Center for Research in Medical Education and Health Care,  2  Jefferson School of Population Health,Thomas Jefferson University, Philadelphia, Pennsylvania, USA;  3  Agenzia Sanitaria e SocialeRegionale, Regione Emilia-Romagna;  4 Ospedale Civile Ravenna, Regione Emilia-Romagna, Italy   ABSTRACT  Aims and background. The study evaluated the use of Italian hospital discharge data(SDO,  scheda di dimissione ospedaliera  ) for identifying women with incident breastcancer, determining stage at diagnosis and assessing quality of care. Study design.  Women aged 20+ years residing in the  Regione Emilia-Romagna  , Italy,between 2002 and 2005 were studied. Case identification using algorithms based onICD-9-CM codes on hospital discharge data were compared with AIRTUM-accredit-edcancerregistrydata.Sensitivity,specificityandpositivepredictivevaluewerecom-putedoverall,byageandcancerstage.Compliancewithguidelinesforradiationther-apy using registry and hospital data were compared. Results.  A total of 11,615 women was identified by AIRTUM-accredited cancer reg-istries as incident cases, whereas 10,876 women were identified by the SDO algo-rithm. Sensitivity was 84.8%, specificity was 99.9%, and the positive predictive value was 90.6%. Of the 1,022 who were false positives, 363 (35.5%) were women identifiedin registry data as having an incident case prior to 2002 and therefore were not in-cludedintheanalysis.Therewere1,761falsenegatives;nearly50%wereover70yearsofageordidnotundergoasurgicalprocedureandthereforewerenotincludedinourSDO-based case finding. Sensitivity declined as the patient population became older.However, we observed relatively good positive predictive value for all age groups. Al-gorithms using the SDO data did not clearly identify specific cancer stages. However,the algorithm may have utility where stages are grouped together for use in quality measures. Conclusions. Cases were identified with good sensitivity, specificity and positive pre-dictive value with SDO data, with better rates than similar previously published algo-rithms based on Italian data.These hospital claims-based algorithms facilitate quali-ty of care analyses for large populations when registry data are not available by iden-tifying individual women and their subsequent use of health care services. Introduction Cancer registries are often used as a reliable source to identify populations withcancer. However, there has been increased interest in the utility of using administra-tive data to identify women with incident breast cancer 1-4 .Algorithms using hospitaldischargedata(s cheda di dimissione ospedaliera  ,SDO)toidentifywomenwithbreastcancer have been developed in an effort to examine national patterns of breast can-cer incidence and to assess the quality of care for these women. Whereas thesemethodologies have been developed and used in the United States, they are just be-ginning to be tested and applied in Italy  4 .This paper describes the use of SDO data toidentify women with incident breast cancer in the  Regione Emilia-Romagna   (RER),compares the rates of hospital discharge case findings with those of the regional can- Tumori,97:428-435,2011  cer registries, and explores the utility of these algo-rithms in assessing quality of care. Algorithms were first developed using US Medicareclaims data to identify incident breast cancer cases andtoassignacancerstagetothecases.Nattinger et al. 1 de-veloped algorithms using US Medicare claims and com-pared the methods to incidence from cancer registriesfor 7,700 US breast cancer registry subjects (based onUS Surveillance Epidemiology and End Results or SEERdata, with sensitivity ranging from 80.1-80.3 and speci-ficity over 99%) 1 . More recent reports have comparedsensitivity and specificity of different algorithms using data from Medicare, health maintenance organizationsand SEER, in some instances finding variations in ageand morbidity subgroups 2,3,5 .Italian researchers have used the same methodology  with Italian registries and hospital data from three Ital-ian regions, reporting similar results 4 . There has beenless success in using Medicare claims data to identify cancer stage 6 . In a comparison using Medicare claimsand SEER data to stage a variety of cancers, researchersfoundanoverestimationintheproportionofearlystagetumorsandunderestimationoflaterstagetumors,part-ly due to the lack of clinical diagnostic markers onclaims data, which are critical in assigning a cancerstage.In addition to the ability of administrative data toidentify populations with cancer, there is increasing in-terest in the use of these data in cancer care quality as-sessment. Recently released guidelines from the Ameri-canJointCommitteeonCancer(AJCC)forbreastcancercare specify how these quality indicators may be con-structed and measured 7-10 . Cancer care quality and out-comes are most appropriately studied with large, het-erogeneous populations, where differences are noted atthe population level, documenting variation by geogra-phy, race/ethnicity, and socio-economic indicators. Ad-ministrative data bases can help to explore these issues,as these data are routinely collected for large popula-tions,whereascancerregistrydatamaynotbeuniform-ly collected.Using administrative and cancer registry data fromItaly presents opportunities and challenges. As Italy does not have a national cancer registry, national esti-mates of cancer incidence have been calculated at apopulation level using a statistical model which com-bines mortality and survival data (MIAMOD, Mortality-Incidence Analysis MODel) 11 . Using Italian hospital da-ta and coding for breast surgeries, Piscitelli  et al. 12 re-ported cancer incidence in Italy to be higher than re-ported by the MIAMOD models. However, the modelsdo not identify singular women, so it is not possible tofollow their individual patterns of follow-up care or toassess cancer quality indicators. Although there have been initiatives that assess Ital-ian quality of cancer care measures for discrete popula-tions, there have been fewer population-based studies.  ADMINISTRATIVE DATA FOR BC IDENTIFICATION AND STAGING 429 Surveys of Italian radiation oncology centers studiedItalian radiotherapy practices and compared these tonational and international guidelines 13,14 . Other re-search on follow-up care has been based in selected on-cology units or hospitals, such as a recent single hospi-tal study of chemotherapy and radiation therapies forelderly women 15 . We evaluated the ability of Italian administrative datato identify women with incident breast cancer, as wellastheabilityofsuchdatatoidentifybreastcancerstage.The goals of the study were threefold: 1) to comparebreast cancer case-finding algorithms based on ICD-9-CM codes found in SDO with case findings based uponcancer registry data; 2) to compare the ability of theseICD-9-CM-based algorithms to stage breast cancer pa-tients; and 3) to demonstrate how such case-finding methodsmaybeusedtoconstructabreastcancerqual-ity measure – radiation therapy one year post diagnosis. Patients and methods  We used data from AIRTUM-accredited breast cancerregistries in RER as the gold standard for cancer identi-fication and staging. AIRTUM (  Associazione italiana dei registri tumori  ) is a national association of 28 Italian ac-credited cancer registries that cover approximately onequarter of the Italian population 16 . Data from RER hos-pital discharge files were used to develop alternative al-gorithmsthatidentifiedandstagedwomenwithcancer.The project was approved by the Institutional Review Board ofThomas Jefferson University. Case identificationusingRERcancerregistries   AIRTUM-accredited registries were used to identify  women with incident breast cancer between 2002 and2005 who were 20 years or older, as breast cancer is ex-tremely rare in women under 25 17 . AIRTUM-accreditedcancer registries cover the populations of a number of RER Local Health Units (LHU) (Parma, Modena, ReggioEmilia, Ferrara, Forlì, Cesena, Rimini, Ravenna andImola) and have been accredited by AIRTUM since1976. Data from the registries were used to determine agold standard of case identification. LHU are a majoradministrative unit within the Italian health care sys-tem. Each LHU is responsible for the organization andmanagement of health care of a geographically definedpopulation. Cancer registries in two other RER LHU were in the process of accreditation at the time of theseanalyses (Bologna and Piacenza).The data we received from the AIRTUM-accreditedregistries were anagraphic data, including a non-trace-able personal identification code, the community of residence at the time of diagnosis, and the progressivenumberofthetumorforeachindividual;characteristicsof the tumor such as topography, morphology; incident  430 EYUEN, D LOUIS, L CISBANI ET AL neoplasm of the breast (ICD-9-CM code 174.0–174.9).Records that met these criteria were sorted by the dis-charge date, and an incident record was identified foreach woman. To differentiate incident from prevalentcases, we evaluated SDO from the previous two years(2000-2001), and women with a principal or secondary diagnosis of breast cancer during this period were ex-cluded from our analysis. In addition, we excluded women if their incident records indicated a history of breast cancer, ICD-9-CM codeV10.3. Cancer staging:  ICD-9-CMdiagnosiscodesintheSDO were used to estimate AJCC cancer stages (Table1). ICD-9-CM codes identifying lymph node involve-mentandmetastaseswereincluded.AstheAJCCdefini-tions include tumor size, and the hospital administra-tive data does not have this information, some AJCCstages were not clearly identified. Cancer quality indicator:   Wetestedtheapplicabilityof our algorithm with a cancer quality indicator, radiationtherapy one year post diagnosis, which has been estab-lished as a quality measure by the U.S. Commission onCancer, American Society of Clinical Oncology and Na-tional Comprehensive Cancer Network  9 . Compliance with the radiation therapy guideline was assessed for women aged 20-69 at the time of incident cancer diag-nosis at AJCC stages I, II, or III, who did not undergo amastectomy, who were alive for 1 year (365 days) afterthe diagnosis, and who were residing in LHU with AIR-TUM-accredited registries. Statisticalanalysis  SAS ® version9.1 20  wasusedtocalculateratesofbreastcancer incidence by age group and cancer stage. Forcanceridentification,sensitivity,specificityandpositivepredictive value (PPV) were computed to compare theincidence estimated from AIRTUM-accredited reg-istries with incidence estimated from hospital dis-charge-abstract based algorithms 21 . We report case-finding comparisons using registry data from the AIRTUM-accredited registries as the goldstandard compared with algorithms using SDO fromthe same LHU. Comparisons were done for the overallcohort and by age group, and 95% confidence intervals were calculated. Age ranges were grouped by 10-yearcohorts (with the exception of women aged 20-49;breast cancer being less prevalent in this age group). Toexamine the ability of claims-based algorithms to iden-tifycancerstage,wereportonthesensitivityandPPVby integer stage, considering only those cases identified astruepositives(n=11,615).Wealsoincludeananalysisof sensitivity and PPV by LHU, including the two LHU whose registries are not currently accredited by AIR-TUM. Lastly, we use the quality measure of radiationtherapy one year post-diagnosis for women residing inLHU with AIRTUM-accredited registries to compare thetwo case-finding algorithms.date, and a field indicating history of breast cancer be-fore 1997; date and type of surgical intervention; TNMclassification of the tumor; breast cancer screening sta-tus; follow-up including deceased status and cause of death code.Breast cancer stage, calculated using the cancer reg-istry data, was based upon the American Joint Commit-tee on Cancer (AJCC) TNM system 18 . This methodology considers the size and spread of the tumor (T), spreadtothelymphnodes(N),andspreadtodistantorgans(Mfor metastasis). For our study, women whose tumors were not primary or single or those who had a history of breast cancer as documented by the registry data wereexcluded.Starting from a cohort of women with breast cancerresiding in Emilia-Romagna, a multistep procedure of deterministic record-linkage was performed.In order tolink the AIRTUM breast cancer registries to the SDO, ananonymous identifier code was attributed to the select-ed incident cases from the registries using an algorithmdeveloped by the RER. The algorithm, according to theprivacy protection rules, uses personal information,looks for each individual woman in a “master file” andalways assigns the same unique identifier. The same al-gorithm was applied to the SDO, and the linkage be-tween the two data bases was performed using thisunique identifier. For cases identified using cancer reg-istries, the patient’s LHU of residence at the incidencedate was mapped from the community of residence onthe registry files.  Algorithms usingSDO  The RER has developed a longitudinal populationhealthcare administrative data base which includes de-mographic information as well as hospital and specialty claims. The data base includes services received by resi-dents of RER in Italian facilities outside of the RER (ap-proximately5%ofsurgicalbreastcanceradmissionswereto hospitals outside of the RER region). The characteris-ticsofthisdatabasehavebeendescribedelsewhere 19 .Weidentifiedallwomenwhowereaged20+yearsasof31De-cember 2002. For cases identified in the SDO but not intheregistries,residencewasassumedtobetheLHUon31December of the incident year. No attempt was made tocheck residence on the incident day itself. Cancer incidence:   We used ICD-9-CM codes presentontheItalianSDOtoidentifywomenwithincidentcan-cer between 2002 and 2005, requiring that women havea diagnosis code for cancer as well as a principal or sec-ondary surgical code for lumpectomy or mastectomy:principal or secondary procedure indicating excision ordestruction of breast tissue (ICD-9-CM code 85.20-85.25) or mastectomy (ICD-9-CM code 85.41-85.48);and principal or secondary diagnosis of carcinoma  in situ   of the breast (ICD-9-CM code 233.0) or malignant   ADMINISTRATIVE DATA FOR BC IDENTIFICATION AND STAGING 431 Results Therewereatotalof1,379,459women20yearsorold-er identified using the RER population data base from2002-2005 residing in the areas covered by the 9 AIR-TUM-accredited registries (LHU Parma, Modena, Reg-gio Emilia, Ferrara, Forlì, Cesena, Rimini, Ravenna andImola). Of these, 11,615 women were identified by thesecancer registries as incident cases. For these LHU, the yearly breast cancer incidence rate was 2.10 per 1,000 women (Table 2).This incidence rate varied significant-lybyagegroup,cancerstageandLHU.Thehighestinci-dent rates of breast cancer were for stages I and IIA; thetwo stages represented over half of all breast cancer in-cident cases. Table 2 also includes cases and rates of breast cancer incidence for the LHU whose registries, atthe time of this study, were not accredited by AIRTRUM. When we compared case identification using the SDO with data from the AIRTUM-accredited registries, weobserved that overall, sensitivity was 84.8%, specificity  was 99.9%, and PPV was 90.6% (Table 3). We also ob-served variation in sensitivity and PPV by age group,and sensitivity declined as the patient population aged(Table 4). Although the sensitivity of our SDO-based al-gorithm was poor for those 80 years and older (63.1%), we observed relatively good PPV for all age groups.Case identification algorithms based on discharge ab-stract data were not as successful in correctly identify-ing breast cancer stage.Table 5 shows, by stage, the sen-sitivity and PPV for those cases identified as true posi-tives in the overall case-finding algorithm. For example,at AJCC stage I, the data from the AIRTUM-accreditedregistriesidentified3,899cases.Ofthese3,899cases,theSDO successfully identified 3,747, resulting in a sensi-tivity of 96.1%. However, the SDO identified a total of 7,376 stage I cases; of these, 3,747 were truly AJCC stageI based on the RER cancer registry, resulting in a PPV of 50.8%. Sensitivity ranged from a high of 96.1% for AJCCstage I cases to poor sensitivity and PPV for stage III(sensitivity = 0.6%; PPV = 46.2%) and stage IV (sensitivi-ty=22.5%;PPV=15.9%).Otherthanforstage0,PPVwaspoor due to high numbers of false positives.There was variation in sensitivity and PPV by LHU.Sensitivity ranged from a high of 94.6% in Bologna to alow of 79.2% in Piacenza; PPV ranged from a high of 94.0% in Modena to a low of 78.0% in Imola. AlthoughBologna and Piacenza did not have an AIRTUM-accred-ited registry at the time of this study and were not in-cluded in the gold standard, they are included inTable 6in order to show data from the entire region.Table 7 compares compliance with the radiation ther-apy quality indicator estimated from the SDO with com-pliance estimated from data from the AIRTUM-accredit-ed registries. For the radiation therapy quality indicator,therewere4,510womenidentifiedthroughtheAIRTUM-accredited registries who met criteria for the population Table 1 - Mapping hospital discharge ICD-9 codes to American Joint Committee on Cancer (AJCC) cancer stages AJCC stage & description ICD-9 CM codes & description0 Tis, N0, M0: Ductal carcinoma  in situ  233.0 Carcinoma  in situ  of breastI T1, N0, M0: Tumor 2 cm or less & has not spread to lymph nodes 174.0-174.9 Malignant neoplasm of femaleor distant sites breastII IIA – T0, N1, M0/T1, N1, M0/T2, N0, M0: Tumor is  ≤ 2 cm with minimal Stages 0-I PLUS 196.3 Secondary malignant neoplasm oflymph node involvement, or tumor is >2 cm & <5 cm but has not lymph nodes of axilla & upper limbspread to lymph nodesIIB – T2, N1, M0/T3, N0, M0: Tumor is >2 cm & <5 cm with minimallymph node involvement, or tumor is >5 cm but does not grow intothe chest wall or skinIII IIIA – T0-2, N2, M0/T3, N1-2, M0: Tumor is not >5 cm with moderate lymph Stages 0-II PLUS 196.0; Secondary malignant neoplasm ofnode involvement, or >5 cm but does not grow into chest wall or skin, 198.2 lymph nodes of head, face or neckwith moderate lymph node or mammary node involvement or skin of breastIIIB – T4, N0-2, M0: Tumor has grown into chest wall or skin & not spreadto lymph nodes, or has spread to 1-3 axillary lymph nodes and/or foundin internal mammary lymph nodes, or has enlarged the internalmammary lymph nodesIV Tumor has spread to distant organs, or to lymph nodes far from Stages 0-III PLUS 196.1-196.2; Secondary malignant neoplasm ofthe breast 196.5-196.6; 196.8-196.9; lymph nodes (intra-thoracic,197.0-197.8; 198.0-198.1; intra-abdominal, inguinal, lower198.3-198.7; 198.81-198.82 limb, intra-pelvic, multiple sites,unspecified site), respiratory ordigestive system, kidney, otherurinary organs, brain, spinal cord,other parts of the nervous system,bone, bone marrow, ovary,adrenal gland, breast, genitalorgans, other specified sites ordisseminated
Search
Similar documents
Related Search
We Need Your Support
Thank you for visiting our website and your interest in our free products and services. We are nonprofit website to share and download documents. To the running of this website, we need your help to support us.

Thanks to everyone for your continued support.

No, Thanks
SAVE OUR EARTH

We need your sign to support Project to invent "SMART AND CONTROLLABLE REFLECTIVE BALLOONS" to cover the Sun and Save Our Earth.

More details...

Sign Now!

We are very appreciated for your Prompt Action!

x