Background Cancer Screening Uptake Among Patients with Type 2 Diabetes Matthew Carr, Iain Buchan, Evan Kontopantelis, Tim Doran and Andrew Renehan (PI) National Awareness and Early Detection Initiative (NAEDI) Conference 2013 2.8 million patients with diabetes and 300,000 new cancer cases each year in the UK Diabetes is an established risk factor for incident cancer (independent of BMI): Renehan et al. (2010) Cancer screening uptake rates are lower in patients with type 2 diabetes compared with general population BUT these data are from North America no UK data 1
Financial Incentives Quality and outcomes framework (QOF) introduced in the UK in 2004 Designed to guide GP practice using financial incentives Objective of the scheme is to reduce inequalities in primary care services Diabetes care is incentivised but national cancer screening activities are not Project Aim Test the hypothesis that national programme cancer screening uptake is lower in patients with type 2 diabetes Objectives 1. Compare cancer screening utilisation (breast, cervical, colorectal) for individuals with & without type 2 diabetes 2. Assess whether QOF initiative shifted resources away from non-incentivised preventive services (i.e. national cancer screening) 2
Framework The Clinical Practice Research Datalink (CPRD) Data The Clinical Practice Research Datalink (CPRD) In the absence of a National Diabetes Registry, we elected to use the CPRD (formerly GPRD) CPRD not previously used to evaluate cancer screening As CPRD is not purpose-built to monitor cancer screening, we speculated that it would under-report (compared with data from National Cancer Screening Programme) However, we equally speculated that under-reporting would be systematic. Therefore, baseline analyses assessed fit for purpose Patient-level data Contains electronic medical records on patients attending primary care practices Data recorded by practitioners at the point of care By 2012 the CPRD contained: - 644 practices providing data of sufficient quality - data on more than 13 million patients Complete primary care coded medical records including: diagnoses, treatments, referrals, prescriptions, immunisation and tests Baseline patient data, external consultations and hospital admission information are also available 3
Methodology Identifying events in the database Baseline analysis Extract and clean the data Produce population-level summaries for verification of the data Identify patients with diabetes Classify the type of diabetes (type 1, 2 or other) Identify routine cancer screening events (breast, cervical and colorectal) Define the eligible population for cancer screening Calculate screening uptake rates for the general population and for patients with type 2 diabetes Future analyses Create matched patient cohorts and: 1) Apply case-control models with type 2 diabetes as the comparator 2) Apply time-series methods to assess the impact of the QOF initiative on clinical scores (pre and post-introduction) All events, (e.g. tests, referrals etc) are coded in the CPRD using Read/OXMIS codes (104,633 codes in current version of the database) Each code comes with a rubric (description) to assist the GP: Read/OXMIS code rubric C10..00 Diabetes mellitus C100011 Insulin dependent diabetes mellitus C100112 Non-insulin dependent diabetes mellitus C108.12 Type 1 diabetes mellitus Create list of text search terms (e.g. for diabetes: "diab", "mell", insulin", ) Search database to identify candidate Read/OXMIS codes Select relevant codes (clinical expertise) Search all CPRD event files (clinical, referral, test ) using selected codes 4
National Cancer Screening Programmes UK Breast Cancer Screening Testing the hypothesis that routine cancer screening uptake is lower among patients with type 2 diabetes Assess screening uptake rates for sub-section of population included in national screening guidelines: Type Screening Test Age Range (yrs) Frequency Breast Mammogram * 50 70 3 years Cervical Cervical smear 25 49 3 years 50 64 5 years Colorectal Faecal occult blood test (FOBT) 60 69 2 years NHS National Guidelines Summary stats 2010/11 Invited: 2.7 million women (50 to 70 years) Uptake rate: 73.6% Regional uptakes: similar with notable exception, London 59% 5-year age bands: 69% to 75% Key issue: using the coding to differentiate between routine and symptomatic screening * Extension of the mammogram programme age range to 47 74 years since 2009 5
73 74 75 76 22/04/2013 Trends in National Breast Cancer Screening Methods: breast cancer screening from CPRD Calculating screening uptake rates Uptake Rate (%) 2000 2005 2010 Years 2.8 2.4 2.0 1.6 Women invited (millions) Uptake rate defined as the proportion of women attending routine screening that satisfy the following eligibility criteria: registered with GP practice for the whole of the year issued a call/recall for screening in the current or previous year did not attend for screening in the previous 2 years (3 year interval) in the relevant age band (50-70 years) no history of breast cancer no prior mastectomy no presentation of relevant symptoms in the previous year (suggesting nonroutine screening) not defined as otherwise high-risk 6
CPRD Breast Cancer Screening Summary stats 2010/11 Uptake rate: 43% (50 to 70 years) Regional uptakes: similar with notable exception, London: 31% 5-year age bands: 38% 0.1 to 45% 0 0.6 0.5 0.4 0.3 0.2 Uptake rates by age band 50-54 y 55-59 y 60-64 y 65-70 y 2005 2007 2009 2011 General under-reporting compared with National Data but patterns are consistent Identifying patients with diabetes High-sensitivity approach used to identify patients with prevalent diabetes Utilising diagnostic codes and supporting evidence of diabetes: - clinical session records, - referrals, - test records, - diabetes treatments (product records), - home glucose monitoring, - elevated glycated haemoglobin levels. Dictionaries moderated and validated internally Events record search of clinical, referral, test and therapy data 7
Classifying the type of diabetes Differentiating between the different types of diabetes Initial classification criteria Patients with type 2 diabetes do not meet the definitions for type 1 or other forms of diabetes The prevalence of type 2 diabetes Prevalence: proportion of the population with type 2 diabetes Eligibility: at least 2 years since initial diabetes diagnosis Age distributions in 2000 and 2010 Algorithm for differentiation based on: age at diagnosis, treatment (e.g. diet only or oral hypoglycaemic), and insulin use: Step 1: Define patients with type 1 (or other forms of) diabetes Step 2: Re-classify using contradictory drug data (suggestive of type 2 diabetes): Non-metformin orals for more than 1 year (at any time in history) Concurrent prescription of nonmetformin oral and insulin (at any time) Non-metformin orals prescribed more than 18 months after diagnosis No insulin or insulin syringe mention yet at least 18 months worth of drug data available 0.07 0.06 0.05 0.04 0.03 0.02 0.01 male female 0 1999 2001 2003 2005 2007 2009 2011 Year The prevalence of diabetes in adults aged 35+ Women diagnosed at an earlier age than men (5 years on average) 8
Breast Cancer Screening Uptake Uptake rate in patients with and without diabetes (age range: 50-70 yrs): Uptake rate (%) 0.60 0.55 0.50 0.45 0.40 General under-reporting 2000 2005 2010 Year Diabetes population Non-diabetes population Preliminary findings are not consistent with our hypothesis that patients with type 2 diabetes underutilise screening services Using CPRD Summary We have successfully identified and quantified patients with diabetes We have successfully quantified the proportions of those with type 2 diabetes, consistent with the literature For cancer screening CPRD is associated with an under-reporting of utilisation of screening for breast cancer (and cervical and colorectal, not shown) However, this misclassification appears to be systematic Against this background, our analyses to-date suggest that breast cancer screening is elevated in patients with type 2 diabetes compared with the general population The separation from the general population occurred around the time of the introduction of QoF incentivization 9
Research Tool Implications CPRD is a fit for purpose research tool to explore hypotheses that require linkage of a chronic disease (here, diabetes) and cancer screening Future approaches will require (yet untested) linkage between the National Diabetes Registry(under development) and National Cancer Screening data Acknowledgements Investigators Dr Andrew Renehan (University of Manchester) Professor Iain Buchan (University of Manchester) Dr Evangelos Kontopantelis (University of Manchester) Dr Tim Doran (University of York) Collaborators Dr Christine Campbell (University of Edinburgh) Professor Tony Howell (University of Manchester) Dr Lorraine Lipscombe (University of Toronto) Professor Martin Gibson (University of Manchester) GP incentivization QoF incentivization may be associated with positive side-effects, including enhanced cancer screening uptake These approaches warrant further research Algorithms Dr Nada Khan, University of Oxford Dr Daniel Levin, University of Dundee 10