Experiences from building up the Danish National Clinical Cancer Database Anders Green, Professor and Consultant Epidemiologist, MD, DrMedSci Odense University Hospital and University of Southern Denmark Institute of Applied Economics and Health Research, Copenhagen KEY SYMPOSIUM ON QUALITY REGISTERS IN MEDICINE, December 8-9, 2014, STOCKHOLM
Conventional indicators of quality in cancer trajectories Survival status 1, 2, 5 years survival Start of treatment Direct waiting time for treatment Decision on treatment options Diagnostic conclusion of cancer Total waiting time for treatment Start of diagnostic evaluation Duration of diagnostic evaluation Referral for suspected malignancy Any cancer database must as a minimum be able to monitor these items!
The Danish National Clinical Cancer Database (DNCCD): Background 1. Centralized national health registers exist in Denmark, incl. The National Danish Patient Register (since 1977/1995) The Danish Pathology Register (computerized since 2004) The Danish National Civil Registration System (since 1968) The Danish Cancer Registry (since 1943) 2. Clinical databases exist in Denmark for most major cancer forms Deficiencies in reporting, particularly in oncology 3. Double/triple registration problem: Same clinical data reported several times in different systems
DNCCD: Underlying philosophies Conventional approaches in clinical cancer databases The individual patients are registered by the clinical units Data elements are entered by the individual clinical units independently of what is entered from other sites Cancer trajectories and milestones established post hoc The DNCCD approach Cancer trajectories and milestones established using a tailored DNCCD algorithm from data available from health registers Individual patient data presented up front for clinical units For approval or amendment For supplementation with data not available in central registers
The Danish National Clinical Cancer Database, DNCCD 1. Pilot project (lung cancer, colorectal cancer) performed 2011 2012, with positive evaluation: It is possible to use central health registers as the data catch for national clinical databases 2. Full project commenced autumn 2012 Steering committee (representation of main stakeholders) Danish National Board of Health Danish Ministry of Health The Association of Danish Multidisciplinary Cancer Groups Danish Regions Quality Development Programme Task force to interact with clinicians and competence centres Peter Gustav (Aldaco): Data management Erik Jakobsen (DLCR, DMCG): Clinical aspects and communications Anders Green (ApEHR): Chair, clinical epidemiology Budget provided from state finances
DNCCD: Sources used Danish National Patient Register Danish National Civil Registration System PIN/PIC Danish Pathology Register Danish Cancer Registry (for post validation ) PIN/PIC: Personal Identification Number / Personal Identification Code Source: Scandinavian Journal of Public Health 39 (suppl. 7), 2011
The Danish National Clinical Cancer Database: Principles Search in DNPR*: First activity qualified for diagnostic evaluation Search in DNPR*: First appearance of relevant ICD10-code, eg C34* Search in DNPR*: Date and type of first and subsequent treatment Time Start date ( Date of diagnosis ) of cancer trajectories Debute date of candidate patients Date of commencement of treatment Data enriched with pathology diagnoses (Danish Pathology Register) vital status, moves/migrations (Danish Civil Registration System) information on staging, treatment and diagnostic procedures (DNPR) Clinicians engaged to validate and supplement data elements * DNPR: Danish National Patient Register
DNCCD: Data flow for a given form of cancer USERS INPUT CENTRAL DATA MANAGEMENT and LINKAGE ANALYSIS Validation and suppl: Diagnostic activities Surgical activities Oncological activities Input/linkage from central registries: Danish Patient Register Danish Pathology Register Civil Registration System Research Users interface Central database Analysis (Stata; Excel;..) Closed network Open Data Base Connectivity Dynamic online reporting: Productivity statistics Error lists Dynamic online reporting: Management Information Systems Clinical units Fixed reporting: Annual full reports Ad hoc reports
DNCCD pilot phase, lung cancer: Patient identification, cases with diagnosis in 2010 DNCCD: Pathology conclusion Number DLCR and CAR with Dx in 2010 DLCR and/or CAR otherwise (c) Neither DLCR nor CAR Positive (a) 3621 (100.0%) 3162 (87.3%) 335 (9.3%) 124 (3.4%) Negative (b) 1138 (100.0%) 730 (64.1%) 258 (22.7%) 150 (13.2%) Total 4759 (100.0%) 3892 (81.8%) 593 (12.5%) 274 (5.8%) (a): Pathology conclusion accords with primary lung cancer (b): No pathology data or other specific tumor or unspecific malignancy (c): Identified in DLCR and/or CAR but not diagnosis in year 2010 for both these sources DNCC: Danish National Clinical Cancer database DLCR: Danish Lung Cancer Registry CAR: Danish Cancer Registry Domains with need of improved DNCCD algorithm
DNCCD, pilot phase, lung cancer, diagnosis in 2010: Identification of first treatment, cases in DNCCD and DLCR DNCCD: First treatment modality DLCR: First treatment modality Resection Oncology No treatment Total Resection 698 11 20 729 Oncology 11 2117 164 2292 No treatment 0 319 853 1172 Total 709 2447 1037 4193 DNCCD: Danish National Clinical Cancer database DLCR: Danish Lung Cancer Registry Domains with need of improved DNCCD algorithm
DNCCD: Challenges and lessons learned Concerning data availability and coding in the central source registers Limited data validation where data is born The source registers cover activities rather than clinical trajectories Need for optimizing coding culture and enforcing coding discipline New working routines must be adapted for the clinicians Help desk and other central support functions are needed DNCCD is an ever ongoing process requiring Continuous commitment from clinical environment Continuous commitment from support structures Preparedness to modify/adjust system as needed in the future
And the winners are: Clinicians: Feed-back increases ownership to otherwise boring patient data Reduction of work load (double registration etc.) Owners, regulators and managers of health care systems: One data source one authorized version of the state of art The customers and other stakeholders: One data source one authorized version of the state of art Epidemiologists, data managers and statisticians: Satisfaction! New research opportunities... Linkage studies... Bridging the gap between registers and clinical databases...
DNCCD: Status ultimo 2014 Fully implemented: Lung cancer In process of transformation: Cancer of the brain Cancer of the prostate Cancer of the bladder Pipeline for 2015 Renal cancer Cancer of the pancreas Cancer of the stomach
Lung cancer in Denmark: Epidemiological forecasting of prevalence Based on automated data available in central health registers Source: A. Green, Unpublished, preliminary results
Monitorering quality of lung cancer care in Denmark: Trends in 30-days post-surgical survival Threshold value for acceptance Clinical audit is performed for every post-surgical death registered Source: Danish Lung Cancer Registry/National Indicator Project
Danish Lung Cancer Register: Indicator IIA: Post-operative survival, 28 days after resection DLCR Annual Report 2011 Unadjusted analysis Pattern: Improvement and convergence, ie reduction in variation across clinical units
Lung cancer in Denmark: Relative distribution of duration (in days) from start of diagnostic evaluation until first oncological intervention 100% 90% 80% 70% 60% 50% 40% 30% 20% Massive political intervention! 64+ 56-63 49-55 42-48 <42 10% 0% 2010, Qtr3 2010, Qtr1 2009, Qtr3 2009, Qtr1 2008, Qtr3 2008, Qtr1 2007, Qtr3 2007, Qtr1 2006, Qtr3 2006, Qtr1 2005, Qtr3 2005, Qtr1 2004, Qtr3 2004, Qtr1 2003, Qtr3 2003, Qtr1 Source: Danish Lung Cancer Registry/National Indicator Project
Lung cancer survival in Denmark after resection 2007-2009, analysis updated April 2011
Health registers, clinical databases, Big data... Dad says you can access all our data and spy on us He s not your dad Thanks for your attention!
Health registers and clinical databases in Denmark Regional and national biobanks Statistics Denmark Data Bank Employment Income and social data And many other data Danish National Civil Registration System Births Deaths Migrations and moves Danish National Patient Register Danish Pathology Register PIN/PIC Danish Cancer Registry Danish National Health Service Register Danish National Prescription Registry Danish Register of Causes of Death National Diabetes Register 50+ national clinical databases PIN/PIC: Personal Identification Number / Personal Identification Code Source: Scandinavian Journal of Public Health 39 (suppl. 7), 2011
Cancer registry domain versus clinical database domain Survival status Information on treatment Cancer registry domain Diagnostic conclusion of cancer Clinical database domain Start of diagnostic evaluation Referral for suspected malignancy
The Danish National Clinical Cancer Database: The stakeholders Danish Quality Model : Overall framework and structure Danish Regions: Owners Danish National Board of Health: Authorities; central health registers Scientific societies: Medical profession Improving quality of care Patients lay associations: Representation of the patients and the public Nurses association etc: Health care providers Clinical environment: Data providers Danish Regions Quality Development Program: Competence centre network to support databases