Clinical Epidemiology Dovepress open access to scientific and medical research Open Access Full Text Article ORIGINAL RESEARCH Validation Of Cancer Diagnoses In Electronic Health Records: Results From The Information System For Research In Primary Care (SIDIAP) In Clinical Epidemiology downloaded from https://www.dovepress.com/ by 77.231.195.187 on 08-Sep-2021 Northeast Spain This article was published in the following Dove Press journal: Clinical Epidemiology Martina Recalde, 1,2, * Cyntia B Background: Electronic health records are becoming an increasingly valuable resource for Manzano-Salgado, 1,2,* Yesika epidemiology but their data quality needs to be quantified. We aimed to validate twenty-five For personal use only. Díaz,1 Diana Puente, 1,2 Maria types of incident cancer cases in the Information System for Research in Primary Care Del Mar Garcia-Gil,1 Rafael (SIDIAP) in Catalonia with the population-based cancer registries of Girona and Tarragona Marcos-Gragera, 3,4 Josefa as the gold-standard. Ribes-Puig,5,6 Jaume Galceran, 7 Methods: We calculated the sensitivity, positive predictive values (PPV), and the time- Margarita Posso, 8 Francesc difference between the date of diagnosis entered into the SIDIAP and into the registries. We Macià, 8 Talita Duarte-Salles 1 added hospital discharge cancer diagnoses to the SIDIAP to assess sensitivity changes. 1 Fundació Institut Universitari per a la Results: We identified 27,046 incident cancer diagnoses in the SIDIAP from 2009–2015 Recerca a l’Atenció Primària de Salut Jordi among the 949,841 residents of Girona and Tarragona. The cancer types with the highest Gol i Gurina (IDIAPJGoL), Barcelona, Spain; sensitivity were breast (89%, 95% CI: 88–90%), colorectal (81%, 95% CI: 80–82%), and 2 Universitat Autònoma de Barcelona, Bellaterra (Cerdanyola del Vallès), prostate (81%, 95% CI: 80–83%). Trachea, bronchus and lung cancers had the highest PPV Cerdanyola del Vallès, Spain; 3Unitat (76%, 95% CI: 74%-78%) followed by stomach (72%, 95% CI: 68–75%) and pancreas d’Epidemiologia i Registre de Càncer de (71%, 95% CI: 67–75%). Most cancer diagnoses were reported with less than three months Girona (UERCG), Pla Director d’Oncologia, Institut Català d’Oncologia, of difference between the SIDIAP and the registries. More cases were registered first in the Institut d’Investigació Biomèdica de Girona registries than in the SIDIAP. By adding cancer diagnoses based on hospital discharge data, (IdIBGi), Universitat De Girona, Girona, sensitivity increased for all cancers, especially for gallbladder and biliary tract for which the Spain; 4CIBER de Epidemiología y Salud Pública (CIBERESP), Madrid, Spain; 5Catalan sensitivity increased by 21%. Cancer Plan, Department of Health of Conclusion: The SIDIAP includes 76% of the cancer diagnoses in the cancer registries but Catalonia, Barcelona, Spain; 6Department of Clinical Sciences, University of Barcelona includes a considerable number of cases that are not in the registries. The SIDIAP reports (UB), Barcelona, Spain; 7Registre de Càncer most of the cancer diagnoses within a three-month period difference from the date of de Tarragona, Fundació per a la Investigació diagnosis in the cancer registries. Our results support the use of the SIDIAP cancer diagnoses i Prevenció del Càncer (FUNCA), IISPV, Reus, Spain; 8Cancer Prevention Unit and for epidemiological research when cancer is the outcome of interest. We recommend adding Cancer Registry, Department of hospital discharge data to the SIDIAP to increase data quality, particularly for less frequent Epidemiology and Evaluation, Hospital del cancer types. Mar, Barcelona, Spain Keywords: validation studies, cancer, electronic health records, primary health care, *These authors contributed equally to this population-based cancer registries work Correspondence: Talita Duarte-Salles Introduction Fundació Institut Universitari per a la Cancer is one of the leading causes of morbidity and mortality worldwide.1 In 2018, Recerca a l’Atenció Primària de Salut Jordi Gol i Gurina (IDIAPJGoL) Gran Via Corts there were 18 million new cases and 9 million deaths.2 In Spain, cancer is a Catalanes, 587 Àtic, Barcelona 08007, Spain significant burden for the National Health System: cancer is the second most Tel +34 93 482 43 42 Email tduarte@idiapjgol.org frequent overall cause of death and results in more than 250,000 new invasive submit your manuscript | www.dovepress.com Clinical Epidemiology 2019:11 1015–1024 1015 DovePress © 2019 Recalde et al. This work is published and licensed by Dove Medical Press Limited. The full terms of this license are available at https://www.dovepress.com/terms. http://doi.org/10.2147/CLEP.S225568 php and incorporate the Creative Commons Attribution – Non Commercial (unported, v3.0) License (http://creativecommons.org/licenses/by-nc/3.0/). By accessing the work you hereby accept the Terms. Non-commercial uses of the work are permitted without any further permission from Dove Medical Press Limited, provided the work is properly attributed. For permission for commercial use of this work, please see paragraphs 4.2 and 5 of our Terms (https://www.dovepress.com/terms.php). Powered by TCPDF (www.tcpdf.org) Recalde et al Dovepress cancer cases every year.3 Therefore, conducting research population-based cancer registries that exist in Catalonia, focused on understanding cancer epidemiology is impor- the Girona and Tarragona cancer registries, as the gold- tant both at the national and international levels. standard. The SIDIAP includes information recorded in The use of databases of routinely collected electronic EHRs by health professionals during routine visits at 287 health records (EHRs) is becoming more common in epi- primary health care centers from the Institut Català de la demiology and clinical research. Due to their size, amount Salut (ICS, Catalan Health Institute).10,11 The SIDIAP has of data availability, representativeness, and long-term fol- anonymized records for more than seven million people low-up, EHR databases offer a great opportunity to con- and is representative of the Catalan population in terms of Clinical Epidemiology downloaded from https://www.dovepress.com/ by 77.231.195.187 on 08-Sep-2021 duct cancer research.4 Another advantage of large health age, sex, and geographic distribution.11 It includes infor- record databases is that they provide sufficient statistical mation on disease diagnoses (International Classification power to detect uncommon outcomes such as rare cancer for Diseases, 10th revision [ICD-10]), drug prescriptions types.5 However, validation processes are required to and dispensations in the primary care setting, and clini- quantify the correctness of the data and to increase the cally relevant parameters (eg, weight, blood pressure, reliability of large health record databases for use in sub- laboratory tests). It is also linked to a hospital discharge sequent observational studies.6 database for patients who attend ICS hospitals (30% of the The information recorded in EHRs by primary health SIDIAP population).12,13 The cancer registries of Girona care professionals in Catalonia – a region in Northeast (created in 1994) and Tarragona (in 1980) cover 20% of Spain with 7.5 million inhabitants (2017) – comprises the the Catalan population.14,15 They collect cancer diagnoses For personal use only. Information System for Research in Primary Care (SIDIAP) from public and private hospitals, anatomopathological platform.7 Since the SIDIAP aims to provide reliable infor- and hematological laboratories, mortality registries, and mation to support research in primary health care, validation other information sources.16–18 Both cancer registries com- studies are performed regularly.8 A previous study assessed ply with the International Agency for Research on Cancer the validity of lung, colon and rectum, prostate, breast, and quality requirements.19 cervix uteri cancers in the SIDIAP during the period 2009–2012 with sensitivities ranging from 64% (cervix Study Population And Cancer Case uteri) to 92% (breast).9 However, this study compared SIDIAP cancer cases with those from the registry of a Definition In the SIDIAP, incident cancer cases were identified as the single hospital in Barcelona. Although the data collection first cancer diagnosis from 2009 to 2015 among inhabi- for this hospital is rigorous for a specific area in Barcelona, tants of the provinces of Girona and Tarragona. We had the this area is not representative of the general population of number of incident cancer cases from the cancer registries Catalonia. Furthermore, the hospital does not have data during 2005–2015 for Girona and during 2005–2013 for available for research use on hematological cancers. A Tarragona available for reference. Cases registered during study validating more cancer types and using population- 2005–2008 were used to clean prevalent cases (Figure 1). based cancer registries as the gold-standard may increase The linkage between the SIDIAP and the cancer registries the scope of the validity of cancer diagnosis in the SIDIAP data was performed by a Trusted Third Party (the ICS in as well as its use in new areas of research. this study) using the unique personal identification number The aim of this study was to validate twenty-five types of patients. We obtained approval from the Clinical of incident cancer cases in the SIDIAP using the popula- Research Ethics Committee of the IDIAPJGol (project tion-based cancer registries of Girona and Tarragona as the code: P14/074) and the Research Ethics Committee of gold-standard and to assess the time-difference in the date the Hospital Doctor Josep Trueta (project code: 2017.024). of diagnosis between the SIDIAP and these cancer We used ICD-10 codes and date of diagnosis to iden- registries. tify the following 25 cancer types in adults (aged ≥18 years): head and neck (ICD-10 codes: C00-C14), esopha- Methods gus (C15), stomach (C16), colorectal (C18–21), liver Data Sources (C22), gallbladder and biliary tract (C23-24), pancreas We performed a cross-sectional study in the SIDIAP dur- (C25), larynx (C32), trachea, bronchus, and lung (C33- ing the years 2009–2015, using data from the two 34), bone and articular cartilage (C40-C41), malignant submit your manuscript | www.dovepress.com Clinical Epidemiology 2019:11 1016 DovePress Powered by TCPDF (www.tcpdf.org) Dovepress Recalde et al Cleaning Period Catchment of incident cancer cases Girona cancer registry Clinical Epidemiology downloaded from https://www.dovepress.com/ by 77.231.195.187 on 08-Sep-2021 Tarragona cancer registry SIDIAP data 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 For personal use only. Dec 31, 2013 Dec 31, 2015 Dec 31, 2008 Jan 1, 2009 Jan 1, 2005 Figure 1 Time period covered by each data source with respect to the duration of the study. Notes: Figure adapted from Margulis, A. et al. (2017). Validation of Cancer Cases Using Primary Care, Cancer Registry, and Hospitalization Data in the UK. Epidemiology, 29(2), 1. Abbreviation: SIDIAP, Information System for Research in Primary Care. melanoma of skin (C43), breast (C50), cervix uteri (C53), non-Spanish). Socioeconomic status was assessed using the corpus uteri (C54-C55), ovary (C56), prostate (C61), testis “Mortalidad en áreas pequeñas Españolas y Desigualdades (C62), kidney (C64), bladder (C67), brain, central nervous Socioeconómicas y Ambientales” (MEDEA) deprivation system, pituitary gland and pineal gland (C70-72, C75.1- index, which we categorized into quintiles for anonymization C75.3), thyroid (C73), Hodgkin lymphoma (C81), non- purposes. The 1st and the 5th quintiles represent the least and Hodgkin lymphoma (C82-C86, C96), multiple myeloma most deprived levels of the urban population in Catalonia, (C90), and leukemia (C91-95).20 We excluded other and respectively.22 We included a rural category since the unspecified malignant neoplasm of skin (C44). Other MEDEA index was not available for people living in these unspecified or very low-frequency cancers (n<100) were areas. excluded. Diagnosis in hospital discharge data was regis- tered using ICD-9 codes.21 We mapped diagnosis codes to Statistical Analysis ICD–10 using available conversion codes eCIEMaps We performed a descriptive analysis of the overall number v3.1.9, which we have provided in Supplementary of cancer cases in SIDIAP and of the confirmed (ie, Table S1. matched diagnoses between the SIDIAP and the cancer registries) vs non-confirmed cases (ie, in the SIDIAP but Other Variables not in the cancer registries) by sex, age, nationality, In the SIDIAP, we had information on the primary care center MEDEA deprivation index, and year of diagnosis, in to which individuals were assigned in 2016 (Girona, Girona and Tarragona, and we used a Chi-squared test to Tarragona), date of diagnosis, sex (women, men), age (18– assess for significant differences.23 We used the Catalonia 35, 36–50, 51–65, ≥66), and nationality (Spanish, Cancer Registries (CCRs, Girona and Tarragona submit your manuscript | www.dovepress.com Clinical Epidemiology 2019:11 1017 DovePress Powered by TCPDF (www.tcpdf.org) Recalde et al Dovepress combined) data as the gold-standard to calculate the sen- CI: 47–55%) and gallbladder and biliary tract (29%, 95% sitivity and the positive predictive values (PPVs) for each CI: 23–35%) (Table 2). cancer type (an illustration of our calculations is available Out of the 27,046 SIDIAP cancer cases present in in Figure S1. As secondary analyses, we stratified the Catalonia, 16,478 (61%) were also in the CCRs. The sensitivity and PPV analyses by province (Girona and trachea, bronchus and lung cancers had the highest PPV Tarragona) to assess if there were geographical differences (76%, 95% CI: 74–78%) followed by stomach (72%, 95% and by sex, nationality, age, and the MEDEA deprivation CI: 68–75%) and pancreas (71%, 95% CI: 67–75%) can- index to assess if there were differences for specific popu- cers (Table 2). On the other hand, bone and articular Clinical Epidemiology downloaded from https://www.dovepress.com/ by 77.231.195.187 on 08-Sep-2021 lation groups. We also checked if the sensitivities cartilage (23%, 95% CI: 15–31%) and cervix uteri (28%, improved after including cancer diagnoses from the hospi- 95% CI: 24–33%) cancers had the lowest PPVs (Table 2). tal discharge database. Most cancer diagnoses were reported within less than For the confirmed cases, we calculated the time differ- three months of difference between the SIDIAP and the ence (months) between the date of diagnosis registered in registries (Figure 2). More cases were reported first in the the SIDIAP and the date registered in the CCRs. cancer registries than in the SIDIAP. Only kidney cancer We used R version 3.5.0 for all the statistical analyses had more than twenty-five percent of cases reported first in and considered p-values <0.05 to be statistically the SIDIAP compared to the CCRs. significant. Secondary Analyses For personal use only. Results Overall, Girona had higher sensitivities than Tarragona, Sociodemographic Characteristics Of especially for cancers of the cervix uteri (68% vs 52%, for Girona and Tarragona, respectively), Hodgkin lym- SIDIAP And Confirmed Cases phoma (69% vs 56%) and head and neck (56% vs 45%) In the SIDIAP, we identified 496,356 inhabitants of Girona (Supplementary Table S2). The only cancer for which in 2016, of which 16,211 had a cancer diagnosis between Tarragona had a higher sensitivity than Girona was for 2009 and 2015, and 453,485 inhabitants of Tarragona, of bone and articular cartilage (56% vs 75%). Regarding which 10,835 had a cancer diagnosis between 2009 and PPVs, Tarragona had higher estimates than Girona, except 2013. There were more cancer cases registered in the for six cancer types. We observed the biggest differences SIDIAP among men (55%, 56% for Girona and for bladder (33% vs 69% %, for Girona and Tarragona, Tarragona, respectively), people aged 66 years or older respectively), colorectal (65% vs 77%) and larynx (52% vs (45%, 49%), Spanish citizens (94%, 95%), and people 63%) cancers. The cancer types for which Girona had the living in rural areas (32%, 37%) (Table 1). biggest differences in PPVs with Tarragona were gallblad- We confirmed 9,296 cancer cases in Girona and 7,182 der and biliary tract (56% vs 44%) and Hodgkin lym- in Tarragona. Compared to non-confirmed cases, con- phoma (71% vs 56%) (Supplementary Table S2). firmed cases had a higher proportion of men in Overall, sensitivity estimates differed by age groups, Tarragona (58% vs 52%) as well as people aged 51 to 65 and PPVs estimates differed by age, nationality and socio- in both provinces (35% vs 31% in Girona; 34% vs 27% in economic status. Those older than 66 years showed lower Tarragona) but a lower proportion of socioeconomically sensitivities than those aged between 36 and 65 years for deprived individuals in Girona (11% vs 14%) (Table 1). most cancer types (Supplementary Table S3). Overall, PPVs were lower in those aged between 18 and 35 years Overall Validation than in the rest of age groups, in non-Spanish than in the Out of the 21,559 cancer cases registered in the CCRs, Spanish population and in the most deprived compared to 16,478 (76%) were in the SIDIAP. The cancer types with the least deprived MEDEA quintiles (Supplementary the highest sensitivities in Catalonia were breast (89%, Table S4). Besides the abovementioned situations, we did 95% CI: 88–90%), colorectal (81%, 95% CI: 80–82%), not observe any other change in the sensitivity and PPVs and prostate (81%, 95% CI: 80–83%) (Table 2). Almost all according to sex, age, nationality, and socioeconomic sta- cancer types had sensitivities above 60% in both pro- tus, with exception of certain specific cancer types vinces. The exceptions were head and neck (51%, 95% (Supplementary Tables S3 and S4). submit your manuscript | www.dovepress.com Clinical Epidemiology 2019:11 1018 DovePress Powered by TCPDF (www.tcpdf.org) Clinical Epidemiology downloaded from https://www.dovepress.com/ by 77.231.195.187 on 08-Sep-2021 For personal use only. Powered by TCPDF (www.tcpdf.org) Table 1 Descriptive Characteristics Of The SIDIAP Population With A Cancer Diagnosis In Girona (2009–2015) And Tarragona (2009–2013) By Confirmation Status From The Cancer Registries Dovepress SIDIAP Cancer Cases, n (%) Girona Tarragona SIDIAP Cases Confirmed Cases Non-Confirmeda p-valueb SIDIAP Cases Confirmed Cases Non-Confirmeda p-valueb Clinical Epidemiology 2019:11 Characteristics N=16,211 N=9296 N=6915 N=10,835 N=7182 N=3653 Sex Women 7300 (45.0%) 4207 (45.3%) 3093 (44.7%) 0.515 4732 (43.7%) 2994 (41.7%) 1738 (47.6%) <0.001 Men 8911 (55.0%) 5089 (54.7%) 3822 (55.3%) 6103 (56.3%) 4188 (58.3%) 1915 (52.4%) Age (years)c 18–35 624 (3.8%) 298 (3.2%) 326 (4.7%) <0.001 375 (3.5%) 168 (2.3%) 207 (5.7%) <0.001 36–50 2830 (17.5%) 1634 (17.6%) 1196 (17.3%) 1636 (15.1%) 1058 (14.7%) 578 (15.8%) 51–65 5417 (33.4%) 3280 (35.3%) 2137 (30.9%) 3471 (32.0%) 2474 (34.5%) 997 (27.3%) ≥ 66 7340 (45.3%) 4084 (43.9%) 3256 (47.1%) 5353 (49.4%) 3482 (48.5%) 1871 (51.2%) Nationality Spanish 15,182 (93.7%) 8731 (93.9%) 6451 (93.3%) 0.110 10,328 (95.3%) 6930 (96.5%) 3398 (93.0%) <0.001 Non-Spanish 1029 (6.3%) 565 (6.1%) 464 (6.7%) 507 (4.7%) 252 (3.5%) 255 (7.0%) MEDEA deprivation indexd Quintile 1 1784 (11.0%) 1171 (12.6%) 613 (8.9%) <0.001 813 (7.5%) 528 (7.3%) 285 (7.8%) 0.022 Quintile 2 1461 (9.0%) 978 (10.5%) 483 (7.0%) 1062 (9.8%) 716 (10.0%) 346 (9.5%) Quintile 3 2180 (13.5%) 1152 (12.4%) 1028 (14.9%) 1201 (11.1%) 784 (10.9%) 417 (11.4%) Quintile 4 2540 (15.7%) 1209 (13.0%) 1331 (19.2%) 1664 (15.4%) 1097 (15.3%) 567 (15.5%) Quintile 5 2025 (12.5%) 1029 (11.1%) 996 (14.4%) 1366 (12.6%) 936 (13.0%) 430 (11.8%) Rural areas 5179 (31.9%) 3205 (34.5%) 1974 (28.5%) 4001 (36.9%) 2677 (37.3%) 1324 (36.2%) “Missing” 1042 (6.4%) 552 (5.9%) 490 (7.1%) 728 (6.7%) 444 (6.2%) 284 (7.8%) Year of diagnosis 2009 2354 (14.5%) 1151 (12.4%) 1203 (17.4%) <0.001 2143 (19.8%) 1150 (16.0%) 993 (27.2%) <0.001 2010 2329 (14.4%) 1342 (14.4%) 987 (14.3%) 2154 (19.9%) 1433 (20.0%) 721 (19.7%) 2011 2310 (14.3%) 1401 (15.1%) 909 (13.1%) 2046 (18.9%) 1443 (20.1%) 603 (16.5%) 2012 2374 (14.6%) 1438 (15.5%) 936 (13.5%) 2212 (20.4%) 1553 (21.6%) 659 (18.1%) 2013 2365 (14.6%) 1471 (15.8%) 894 (12.9%) 2280 (21.0%) 1603 (22.3%) 677 (18.5%) 2014 2259 (13.9%) 1383 (14.9%) 876 (12.7%) - - - DovePress submit your manuscript | www.dovepress.com 2015 2220 (13.7%) 1110 (11.9%) 1110 (16.1%) - - - Notes: aNon-confirmed cases either have different diagnoses in SIDIAP and the registry or were not available in the registry. bComparison of confirmed vs non-confirmed cases using Chi-squared of independence test. cAge in 2009. d Quintile 1 of the MEDEA Index represents the least deprived and quintile 5 represents the most deprived. Rural was included as a category since the index cannot be calculated for people living in rural areas. 1019 Recalde et al Abbreviations: SIDIAP, Information System for Research in Primary Care. Recalde et al Dovepress Table 2 Validity Of The ICD-10 Codes Used To Identify Incident Cancer Diagnoses Registered In The SIDIAP Database, Cataloniaa (2009–2015)b Cancer Type (ICD-10 CM) Cancer Cases, n Sensitivity, % (95% CI) PPV, % (95% CI) CCRs SIDIAP Confirmed Head and neck (C00-C14) 650 819 332 51.1 (47.2–54.9) 40.5 (37.2–43.9) Esophagus (C15) 211 255 157 74.4 (68.5–80.3) 61.6 (55.6–67.5) Stomach (C16) 673 633 455 67.6 (64.1–71.1) 71.9 (68.4–75.4) Colorectal (C18-C21) 3743 4329 3035 81.1 (79.8–82.3) 70.1 (68.7–71.5) Clinical Epidemiology downloaded from https://www.dovepress.com/ by 77.231.195.187 on 08-Sep-2021 Liver (C22) 561 625 364 64.9 (60.9–68.8) 58.2 (54.4–62.1) Gallbladder & biliary tract (C23-C24) 197 107 57 28.9 (22.6–35.3) 53.3 (43.8–62.7) Pancreas (C25) 578 590 419 72.5 (68.8–76.1) 71.0 (67.4–74.7) Larynx (C32) 337 403 226 67.1 (62.0–72.1) 56.1 (51.2–60.9) Trachea, bronchus & lung (C33-C34) 2152 2155 1631 75.8 (74.0–77.6) 75.7 (73.9–77.5) Bone and articular cartilage (C40-C41) 39 106 24 61.5 (46.3–76.8) 22.6 (14.7–30.6) Malignant melanoma of skin (C43) 550 962 417 75.8 (72.2–79.4) 43.3 (40.2–46.5) Breast (C50) 3325 4456 2958 89.0 (87.9–90.0) 66.4 (65.0–67.8) Cervix uteri (C53) 198 416 118 59.6 (52.8–66.4) 28.4 (24.0–32.7) Corpus uteri (C54-C55) 576 661 424 73.6 (70.0–77.2) 64.1 (60.5–67.8) Ovary (C56) 263 398 190 72.2 (66.8–77.7) 47.7 (42.8–52.6) For personal use only. Prostate (C61) 2820 3596 2286 81.1 (79.6–82.5) 63.6 (62.0–65.1) Testis (C62) 139 175 102 73.4 (66.0–80.7) 58.3 (51.0–65.6) Kidney (C64) 536 730 397 74.1 (70.4–77.8) 54.4 (50.8–58.0) Bladder (C67) 1456 2370 1108 76.1 (73.9–78.3) 46.8 (44.7–48.8) Brain and CNS (C70-C72, C75.1-C75.3)c 393 544 298 75.8 (71.6–80.1) 54.8 (50.6–59.0) Thyroid (C73) 395 432 264 66.8 (62.2–71.5) 61.1 (56.5–65.7) Hodgkin lymphoma (C81) 144 142 92 63.9 (56.0–71.7) 64.8 (56.9–72.6) Non-Hodgkin lymphoma (C82-C86, C96) 709 909 472 66.6 (63.1–70.0) 51.9 (48.7–55.2) Multiple myeloma (C90) 294 362 233 79.3 (74.6–83.9) 64.4 (59.4–69.3) Leukemia (C91-C95) 620 871 419 67.6 (63.9–71.3) 48.1 (44.8–51.4) Notes: aProvinces of Girona and Tarragona. bData from the Tarragona Cancer Registry was only available for 2009–2013. cInclude pituitary gland and pineal gland tumors. Abbreviations: CI, Confidence Interval; CNS, Central Nervous System; CCRs, Catalonia Cancer Registries; ICD-10, International Classification for Diseases, 10th revision; PPV, positive predictive values; SIDIAP, Information System for Research in Primary Care. When adding cancer diagnoses from hospital discharge specific cancer sites. More cases were first recorded in to primary care data, we observed an increase in sensitivity the cancer registries rather than in the SIDIAP, though for all cancer types. Gallbladder and biliary tract cancer for most cancer cases, the time difference between both had the most substantial change in sensitivity, changing data sources did not exceed three months. Including cancer from 29% to 50% (Supplementary Table S5). We also diagnoses from hospital discharge data considerably observed changes above 10% for larynx (67% to 83%), improved the reliability of the data for specific cancer head and neck (51% to 66%) and liver (65% to 78%) types. cancers (Supplementary Table S5). We observed a high sensitivity for the majority of cancer types. Breast, colorectal and prostate cancers had Discussion the highest sensitivities, which are some of the most inci- This study validated cancer diagnoses recorded in primary dent tumors and thoroughly screened cancers in systematic care using the data of the two provincial population-based programs (breast and colorectal) and strongly sought by cancer registries that exist in Catalonia as the gold-stan- opportunistic screening (prostate) in Catalonia.24,25 dard. We found that 23 out of 25 cancer types had sensi- Furthermore, these cancers take part in the rapid diagnos- tivities above 60%. PPV estimates were generally lower tic circuit program run in Catalonia, which could also than the sensitivities observed in most cancer types. The contribute to an increase in the accuracy of diagnosis in number of cancer cases in the SIDIAP that were not primary care.26 Previous studies conducted in the United confirmed by the cancer registries was high for some Kingdom (UK) that compared primary care data with submit your manuscript | www.dovepress.com Clinical Epidemiology 2019:11 1020 DovePress Powered by TCPDF (www.tcpdf.org) Dovepress Recalde et al Clinical Epidemiology downloaded from https://www.dovepress.com/ by 77.231.195.187 on 08-Sep-2021 For personal use only. Figure 2 Time-difference (months) in the date of cancer diagnosis recorded in the SIDIAP and the population-based Catalonia Cancer Registriesa (2009–2015)b. Notes: aPopulation-based cancer registries from the provinces of Girona and Tarragona. bData from the Tarragona Cancer Registry was only available for 2009–2013. Negative values indicate SIDIAP diagnosis before the registries’ diagnosis date. Brain and CNS include pituitary gland and pineal gland tumors. Abbreviations: CNS, Central Nervous System; m, months; SIDIAP, Information System for Research in Primary Care. hospital and cancer registry data also reported high sensi- are not aware of any previous national or international tivities for breast, prostate, and colorectal cancers, high- studies validating the primary care diagnosis of these lighting that these cancers are usually managed by general cancer types using external sources. Thus, our results practitioners.27,28 In Catalonia, a previous study compar- indicate that using SIDIAP cancer diagnoses for research ing SIDIAP cases with those registered in a hospital can- when cancer is the outcome of interest is reliable for most cer registry in Barcelona, also reported high sensitivities common cancer types in Catalonia but may be insufficient for breast, colorectal and prostate cancers.9 High sensitiv- for less frequent types. ities are important to enhance study inclusiveness and to PPV estimates were generally lower than the sensitiv- be able to ascertain common exposures.29 A high sensitiv- ities observed in most cancer types. The number of cancer ity paired with a high specificity (which is important for cases in the SIDIAP that were not confirmed by the cancer classifying outcomes) facilitates both the study of cancer registries was high for some specific cancer sites. A pre- as an outcome as well as the identification of the cases’ vious study validating only colorectal, lung, gastro-eso- common exposures. In our study, the lowest sensitivities phageal and urological cancer diagnosis in primary care were found for cancers that are less frequent and that are in the UK reported higher PVV estimates than in our study, more commonly managed in hospitals, such as gallbladder ranging from 92% to 98%.28 This study hypothesized that and biliary tract or bone and articular cartilage.9,24,30,31 We some of the reasons behind non-confirmed cases might be submit your manuscript | www.dovepress.com Clinical Epidemiology 2019:11 1021 DovePress Powered by TCPDF (www.tcpdf.org) Recalde et al Dovepress a disagreement in the type of cancer diagnosed in each for most cancer sites, with substantial improvements data source, or the possibility of suspicious symptoms observed particularly for less frequent cancer types. The being registered as cancer diagnoses in primary care.28 In use of multiple data sources is highly recommended when agreement with this hypothesis, we found that approxi- using EHRs for epidemiological research since the advan- mately 10% of the non-confirmed cases by the cancer tages of each database can overcome the limitations of the registries were due to disagreement in the type of cancer others.4,33 Specifically, the need to link primary care data- diagnosis between the data sources. The low PPV for bases to those from hospitals and cancer registries to cor- cervix uteri cancer (included in the rapid diagnostic circuit rectly identify certain cancer types has been proposed in the Clinical Epidemiology downloaded from https://www.dovepress.com/ by 77.231.195.187 on 08-Sep-2021 in Catalonia) could be due to detected suspicious symp- UK.27 Therefore, considering both SIDIAP and hospital toms recorded as cancers in SIDIAP; however, we did not discharge databases can improve the reliability in the results have the information needed to prove this hypothesis. of future research. This may be especially important for Another factor that can influence PPVs is the prevalence larynx, head and neck and liver cancers. For gallbladder of the cancer type which could partially explain the low and biliary tract cancer, despite the sizeable improvement in PPVs of bone and articular cartilage (106 cases registered sensitivity after adding hospital discharge to SIDIAP cancer in the SIDIAP) and gallbladder and biliary tract (107 cases diagnoses, the final sensitivity estimate (50%) seems insuf- registered). High PPVs are important when we want to ficient to perform future studies using this cancer type as an identify a cohort of people that only includes people with outcome. If data is available, future studies may consider the condition of interest but do not need to be representa- restricting their analyses to confirmed cases only to avoid For personal use only. tive of all cases.29 Therefore, the SIDIAP does not appear misclassifications and attain data robustness. to be an appropriate database to create a cohort of cancer The main strengths of this study are first, the use of the patients, except for certain cancer types (eg, trachea, SIDIAP database, which provides a large and representative bronchus and lung, stomach, pancreas or colorectal can- sample of the Catalonian population and increases external cers). More research needs to be conducted to understand validity.11 Second, the use of two population-based cancer the reasons behind non-confirmed cancer cases in SIDIAP. registries as the gold-standard allowed us to validate numer- Most cancer diagnoses were reported within less than ous cancer types. Third, we were able to calculate the sensi- three months of difference between the SIDIAP and the tivity of the SIDIAP cancer diagnoses, a type of measure that registries, and generally, the cancer registries reported the is often not reported in cancer validation studies. However, cases earlier than the SIDIAP. Our results are in line with our study has limitations. First, since the SIDIAP is a primary two previous studies in the UK which assessed the time care database, certain cancer types are harder to be detected difference between the date of cancer diagnoses registered at this level; nevertheless, we assessed the inclusion of hos- in the cancer registries and primary care databases. One pital discharge information to account for this limitation. study reported a median time difference in the date of Second, textual information in medical records could be of diagnosis of 11 days (range 6–30 days) between a UK value to distinguish cancer suspicions from actual diagnoses primary care database and the Cancer Registry in England in the SIDIAP, but this information was not available in this for colorectal, lung, gastro-esophageal and urological study. Third, for this study we were only able to add cancer cancers.28 The other study, also using information from diagnoses from hospital discharge from the ICS hospitals, the same UK primary care database and cancer registry but therefore we cannot confirm whether including information combining 11 cancer types, reported that 63% of cancer from all Catalan hospitals would permit better identification diagnoses were recorded with one month of difference of cases for the same cancer types we found. Finally, our between the data sources and 24% within one to three population of reference was the population of individuals months of difference. However, the authors did not specify assigned to a primary care center in Girona and Tarragona which source registered the diagnosis first.32 Although the provinces in 2016 and, thus, we could not account for time difference between the data sources was not substan- changes in patient address during the whole study period. tial in our study, investigators should be aware of it when addressing time-related research questions in the SIDIAP, Conclusion such as those in the cancer survival field. The SIDIAP includes 76% of the cancer diagnoses present In our study, the inclusion of hospital discharge data to in the cancer registries of Catalonia but also includes a SIDIAP cancer diagnoses improved the sensitivity estimates considerable number of cases that are not in the registries. submit your manuscript | www.dovepress.com Clinical Epidemiology 2019:11 1022 DovePress Powered by TCPDF (www.tcpdf.org) Dovepress Recalde et al Overall, the SIDIAP reports cancer cases later than the Discharges. We also thank Leonardo Mendez-Boo and registries but the time difference in the date of diagnosis Eduardo Hermosilla for their support on the data linkage between the databases is usually less than three months. processes and Matthew Bennett for revising the English Our results support the use of SIDIAP cancer diagnoses language composition. for national and international epidemiological research when cancer is used as an outcome, especially for the Funding most frequent cancer types. The inclusion of cancer diag- Funding [for grant number: 2017/1630] was obtained from noses from hospital discharge data is recommended to Wereld Kanker Onderzoek Fonds (WKOF), as part of the Clinical Epidemiology downloaded from https://www.dovepress.com/ by 77.231.195.187 on 08-Sep-2021 improve the reliability of certain cancer types such as World Cancer Research Fund International grant program. head and neck, liver, larynx, and leukemia. However, our TDS is funded by the Department of Health of the results do not support the use of SIDIAP data for all Generalitat de Catalunya, awarded on the 2016 call under cancer sites when the purpose of the study is to identify the Strategic Plan for Research and Innovation in Health a cohort of cancer patients. Further research is needed to (PERIS) 2016–2020, modality incorporation of scientists understand the cancer cases recorded in the SIDIAP that and technologists, with reference SLT002/16/00308. were not confirmed by the cancer registries. Abbreviations Disclosure CI, Confidence Interval; CCRs, Catalonia Cancer Registries; The authors report no conflicts of interest in this work. For personal use only. CNS, Central Nervous System; EHRs, Electronic Health Records; ICD-9,International Classification for Diseases, References 9th revision; ICD-10, International Classification for 1. International Agency for Research on Cancer. World Cancer Report Diseases, 10th revision; ICS, Institut Català de la Salut; 2014. Lyon; 2014. MEDEA, Mortalidad en áreas pequeñas Españolas y 2. Global Cancer Observatory. Cancer tomorrow. International Agency for Research on Cancer. https://gco.iarc.fr/tomorrow/home. Published Desigualdades Socioeconómicas y Ambientales; PPV, 2018. Accessed March 10, 2019. Positive Predictive Value; SIDIAP, Information System for 3. Galceran J, Ameijide A, Carulla M, et al. Cancer incidence in Spain, Research in Primary Care; UK, United Kingdom. 2015. Clin Transl Oncol. 2017;19(7):799–825. doi:10.1007/s12094- 016-1607-9 4. Cook JA, Collins GS. The rise of big clinical databases. Br J Surg. Ethics Approval 2015;102(2):93–101. doi:10.1002/bjs.9723 5. Haynes K, Forde KA, Schinnar R, Wong P, Strom BL, Lewis JD. Cancer The Clinical Research Ethics Committee of the incidence in the health improvement network. Pharmacoepidemiol IDIAPJGol (project code: P14/074) and the Research Drug Saf. 2009;18:730–736. doi:10.1002/pds.v18:8 Ethics Committee of the Hospital Doctor Josep Trueta 6. Nissen F, Quint JK, Morales DR, Douglas IJ. How to validate a diagnosis recorded in electronic health records. Breathe. 2019;15 (project code: 2017.024) approved this study. (1):64–68. doi:10.1183/20734735.0344-2018 7. Instituto Nacional de Estadística. Estadísticas Territoriales: cataluña. http://www.ine.es/FichasWeb/RegComunidades.do?fichas=49&busc_ Data Availability comu=&botonFichas=Ir+a+la+tabla+de+resultados. 2018. Accessed In accordance with current European and national law, the October 10, 2018. data used in this study is only available for the researchers 8. Ramos R, Balló E, Marrugat J, et al. Validity for use in research on vascular diseases of the SIDIAP (Information System for the participating in this project. Thus, we are not allowed to Development of Research in Primary Care): the EMMA study. Rev distribute or make publicly available the data to other Española Cardiol (English Ed. 2012;65(1):29–37. doi:10.1016/j. recesp.2011.07.017 parties. However, researchers from public institutions can 9. Garcia-Gil M, Elorza J-M, Banque M, et al. Linking of primary care request data from the SIDIAP and other sources (eg, Cancer records to census data to study the association between socioeco- Registries) if they comply with certain requirements. nomic status and cancer incidence in Southern Europe: a nation-wide ecological study. PLOS ONE. 2014;9(10):e109706. doi:10.1371/jour- Further information is available online (https://www.sidiap. nal.pone.0109706 org/index.php/menu-solicitudes-en/application-proccedure) 10. Bolíbar B, Fina Avilés F, Morros R, et al. Base de datos SIDIAP: la historia clínica informatizada de Atención Primaria como fuente de or by contacting Anna Moleras amoleras@idiapjgol.org). información para la investigación epidemiológica. Med Clin (Barc). 2012;138(14):617–621. doi:10.1016/j.medcli.2012.01.020 Acknowledgements 11. García-Gil MDM, Hermosilla E, Prieto-Alhambra D, et al. Construction and validation of a scoring system for selection of We thank the Registre del conjunt mínim de bases de high quality data in a Spanish population primary care database dades (CMBD), for providing data on Hospital (SIDIAP). Inf Prim Care. 2012;20(2):1. submit your manuscript | www.dovepress.com Clinical Epidemiology 2019:11 1023 DovePress Powered by TCPDF (www.tcpdf.org) Recalde et al Dovepress 12. Generalitat de Catalunya. Conjunt mínim bàsic de dades (CMBD). ciudades españolas (Proyecto MEDEA). Gac Sanit. 2008;22 https://catsalut.gencat.cat/ca/proveidors-professionals/registres-cata (3):179–187. doi:10.1157/13123961 legs/registres/cmbd/index.html#googtrans(ca%7Ces. 2017. Accessed 23. Subirana I, Vila J, Sanz H, Lucas G, Penafiel J, Gimenez D. Building March 5, 2019. bivariate tables: the comparegroups package for R. J Stat Softw. 13. Instituto Nacional de Estadística. Hospital morbidity Survey: general 2014;57(12):1–16. doi:10.18637/jss.v057.i12 methodology. http://www.ine.es/en/daco/daco42/sanitarias/notaemh_ 24. Pla director d’Oncologia de Catalunya. El Càncer a Catalunya. en.htm. 2019. Accessed March 6, 2019. Barcelona; 2016. 14. Josep MB, Izquierdo A, Vilardell L, et al. Cancer Incidence in Girona 25. Instituto Catalán de Oncología. Prevención del cáncer. http://ico.gencat. (2008-2012). In: Bray F, Colombet M, Mery L, et al., editors. Cancer cat/es/el-cancer/programes_de_deteccio_precoc/. Accessed April 17, Incidence in Five Continents. Vol. XI (electronic version). Lyon: 2019. International Agency for Research on Cancer; 2017. Available 26. Generalitat de Catalunya. Circuit de diagnòstic ràpid. http://canalsa Clinical Epidemiology downloaded from https://www.dovepress.com/ by 77.231.195.187 on 08-Sep-2021 from: http://ci5.iarc.fr. Accessed March 10, 2018. lut.gencat.cat/ca/salut-a-z/c/cancer/recursos-per-a-professionals/diag 15. Galceran J, Carulla M, Ameijide A, et al. Cancer incidence in nostic/circuit-de-diagnostic-rapid/index.html. 2018. Accessed April Tarragona (2008–2012). In: Bray F, Colombet M, Mery L, et al., 16, 2019. editors. Cancer Incidence in Five Continents. Vol. XI (electronic 27. Margulis AV, Fortuny J, Kaye JA, et al. Validation of cancer cases version). Lyon: International Agency for Research on Cancer; 2017. using primary care, cancer registry, and hospitalization data in the United Kingdom. Epidemiology. 2018;29(2):308–313. doi:10.1097/ Available from: http://ci5.iarc.fr. Accessed March 10, 2018. EDE.0000000000000786 16. Unitat d’Epidemiologia i Registre de Càncer de Girona. El Càncer a 28. Dregan A, Moller H, Murray-Thomas T, Gulliford MC. Validity of Girona 2010–12: Projeccions De La Incidència 2017. Girona: cancer diagnosis in a primary care database compared with linked cancer Institut Català d’Oncologia; 2016. registrations in England. Population-based cohort study. Cancer 17. Navarro C, Martos C, Ardanaz E, et al. Population-based cancer Epidemiol. 2012;36(5):425–429. doi:10.1016/j.canep.2012.05.013 registries in Spain and their role in cancer control. Ann Oncol. 29. Chubak J, Pocobelli G, Weiss NS. Tradeoffs between accuracy mea- 2010;21(Supplement3):iii3–iii13. doi:10.1093/annonc/mdq094 sures for electronic health care data algorithms. J Clin Epidemiol. 18. International Agency for Research on Cancer. Indices of data Quality: 2012;65(3):343–349.e2. doi:10.1016/j.jclinepi.2011.09.002 For personal use only. All sites except non-melanoma skin (C00-96 exc. C44). Cancer 30. American Cancer Society. Can gallbladder cancer be found early? Incidence in Five Continents Volume XI. http://ci5.iarc.fr/CI5-XI/ https://www.cancer.org/cancer/gallbladder-cancer/detection-diagno PDF/INDICES/21.pdf. 2017. Accessed March 10, 2018. sis-staging/detection.html. 2018. Accessed March 5, 2019. 19. International Agency for Research on Cancer. Chapter 5: Data 31. American Cancer Society. Can bone cancer be found early? https:// Comparability and Quality. Cancer Incidence in Five Continents www.cancer.org/cancer/bone-cancer/detection-diagnosis-staging/ Volume XI. http://ci5.iarc.fr/CI5-XI/Pages/Chapter5.aspx. 2017. detection.html. 2018. Accessed March 5, 2019. Accessed March 10, 2018. 32. Boggon R, Van Staa T, Chapman M, Gallagher AM, Hammad TA, 20. World Health Organization. ICD-10 : International Statistical Richards MA. Cancer recording and mortality in the general practice Classification of Diseases and Related Health Problems: Tenth research database and linked cancer registries. Pharmacoepidemiol Revision. 2nd ed. Geneva: World Health Organization; 2004. Drug Saf. 2013;22:168–175. doi:10.1002/pds.v22.2 21. World Health Organization. Ninth revision of the International 33. Herrett E, Thomas SL, Schoonen WM, Smeeth L, Hall AJ. Validation Classification of Diseases. 1976. and validity of diagnoses in the general practice research database: a 22. Domínguez-Berjón MF, Borrell C, Cano-Serral G, et al. Construcción systematic review. Br J Clin Pharmacol. 2010;69(1):4–14. doi:10.11 de un índice de privación a partir de datos censales en grandes 11/bcp.2010.69.issue-1 Clinical Epidemiology Dovepress Publish your work in this journal Clinical Epidemiology is an international, peer-reviewed, open access, systematic reviews, risk & safety of medical interventions, epidemiol- online journal focusing on disease and drug epidemiology, identifica- ogy & biostatistical methods, and evaluation of guidelines, translational tion of risk factors and screening procedures to develop optimal pre- medicine, health policies & economic evaluations. The manuscript ventative initiatives and programs. Specific topics include: diagnosis, management system is completely online and includes a very quick prognosis, treatment, screening, prevention, risk factor modification, and fair peer-review system, which is all easy to use. Submit your manuscript here: https://www.dovepress.com/clinical-epidemiology-journal submit your manuscript | www.dovepress.com Clinical Epidemiology 2019:11 1024 DovePress Powered by TCPDF (www.tcpdf.org)