Presentation on theme: "Martin KenneallyBrenda Lynch Integrating Administrative Data in Health Studies: A Case Study."— Presentation transcript:
Martin KenneallyBrenda Lynch email@example.com@ucc.ie Integrating Administrative Data in Health Studies: A Case Study
Objectives Profile the Health Status of Irish Regions (2010) Link Regional Health Profiles to Regional Prescribing. Incorporate Regional Demographics, Community Drug Scheme Coverage rates and Prescribing Norms Simulate effects of Health Status, GMS coverage & Demographics on Prescribing Rates & Cost outcomes. Identify correlates of health status
Profile Regional Health Status Construct a Composite Health Index to; (a) Calibrate the health status of 8 Irish regions below (b) Use Ireland as Standard with base value of 100. RegionsCounties 1. EastDublinKildareWicklow 2. MidlandsLaoisOffalyLongfordW. Meath 3. Mid WestClareLimerickN Tipperary 4. North EastCavanLouthMeathMonaghan 5. North WestDonegalLeitrimSligo 6. South EastCarlowKilkennyS TipperaryWaterford & Wexford 7. South WestCorkKerry 8. WestGalwayMayoRoscommon
Profiling Steps 1. Select Health Indicators: (mainly morbidity rates) for each type of major health condition 2. Standardized Rates = national/regional morbidity rate *100 3. Aggregate Rates: using Prescription Weights. 3(a) Aggregate separately for GMS, DP and LTI drug schemes in each region. 3(b) Then Aggregate across schemes in a region (using coverage weights) to obtain that region’s Regional Health Index (i) Use Index to Benchmark Regional Health Status (ii) Use Index to Benchmark Regional Health Gaps (iii) Simulate selected policy outcomes (iv) Identify Correlates of Health Status & Health Gaps
Official/Administrative Datasets Available Official : CSO e.g. QNHS – Health Module (Used) Admin: non-CSO, respondent centred, dedicated focus e.g. IPH: General. Morbidity Prevalence Rates by County (Used) PCRS: Drug Scheme Coverage & Prescribing data (Used) Health Atlas: Focus on Public Patients Tilda: Focus on Over 50s (Ageing) SLAN: Focus on Lifestyle, Attitudes & Nutrition. Periodic. CME: General, 2008/9 only [BNF not ICD codes used]. Technical Challenges Missing Definitions, sources and methods Disjoint Definitions/Concepts e.g. CSO vs PCRS ‘regions’
Health Data Gaps QNHS Health Module covers 19 Health Conditions: (i) ‘Adults only’ (excludes under 18s) (ii) Rates refer to “at any time in a respondent’s past” (iii) Excludes GERD & pregnancy/immunization services and (iv) Does not cross-tabulate conditions by region/medical cover. PCRS: Publishes prescribing by scheme, region, age & gender but does not publish allied morbidity rates on the same basis. Upshot, we don’t know; (i)How much morbidity rates of ‘public’ & ‘private’ patients differ (ii)How much GP visit rates reflect ill-health v’s type of health cover (iii)How much prescribing rates reflect ill-health v’s health cover
Community Drug Schemes Incorporated Incorporated 1. GMS (General Medical Services) Means tested. Income adjusted for mortgage/housing, childcare, travel costs, savings etc. Discretionary Medical Card also granted to avoid “undue hardship”. 2. DP (Drug Payment Scheme) – Not means tested. Person/family pays first €144/month; HSE pays any excess 3. LTI (Long Term Illness) Not means tested. Schedule includes - Cerebral Palsy, Spina Bifida, Epilepsy Acute Leukaemia, Multiple Sclerosis, Diabetes & *************************************************************************************** Not incorporated HTD (High Tech Drug Scheme) – mainly hospital originated anti-rejection drugs for transplants and chemotherapy
Methodology 1) Select 28 health indicators (18 prevalence rates/10 others) 2) Assign to 6 ATC Health Categories/Dimensions: Alimentary, Cardio, CNS, Respiratory, Various & Other 3) Break down 6 ATC categories into 24 Therapeutic Drug Groups 4) Construct (prescription weighted) Composite Health Indices for the 6 ATC categories under each scheme in each region. 5) Aggregate scheme-specific Indices into Regional Indices (using scheme coverage weights). Base Value is Ireland = 100.
Health Gaps & Weights in East & Midlands Regions
Simulated Prescribing & Cost Outcomes We constructed & validated a simulation model. Simulation Model incorporates regional health status, scheme coverage & prescribing norms; Simulates number & type of drug prescribed in each region in 2010 with high (97%) accuracy Prescribing semi-elasticity w.r.t. GMS coverage is twice semi- elasticity w.r.t. health-status Pattern and causes of regional unit drug cost variations still under investigation.
Income, Demographics, Coverage & Health Status Region CSO Disposable Income 2010 Percentage Aged over 65 % Covered by GMS Composite Health Index East 20,30010.00%28%106.14 Republic of Ireland 19,30011.10%35%100.00 South 19,20012.00%36%100.48 Mid West 19,10011.80%38%103.92 West 18,50012.30%41%96.52 South East 18,10012.00%41%97.05 North East 17,30010.00%38%102.31 North West 17,30013.00%49%94.04 Midlands 17,10011.00%38%91.89
Unanswered Questions & Policy Issues 1) Macro-causality pattern of regional health status remains “Smudged” 2) “Ground-clearing”: lacking the with precision of, say, Kabir et al. 2013 on CHD 3) North-West v’s North East (for example): Why is GMS cover in NW so much higher? (Equity) Is poorer NW health status due to poor demographics? How much do other factors contribute to health status?
Recommendations to Increase Usability Working Party of Official & Admin Groups to Agree; Common and individual domains Common base observation unit (e.g. DEDs for SAPS & Census) Common publication unit (NUTS3 or NUTS4) Harmonised methodologies. ‘Definitions, Sources and Methods’ manual (IPH/PCRS) Linked prescribing and morbidity data (for policy analysis) Published accessible anonymised (Statbank style) Archive Tables Increased professional statistical input The End