Title A multilevel approach to health systems analysis using RISS - - PowerPoint PPT Presentation

A multilevel approach to health systems analysis using RISS (Reporting-by-Intranet Statistical System)

Carinci F, Corrado D, Dettorre A, Pellegrini F Monash Institute of Health Services Research, Australia Consorzio Mario Negri Sud, Italy

23 September 2001 4TH INTERNATIONAL CONFERENCE ON THE SCIENTIFIC BASIS OF HEALTH SERVICES Sydney, New South Wales, Australia

Title

Meta-learning Computer-Intensive Methods Neural Networks B a y e s i a n S t a t i s t i c s

Statistical Sciences

Data Mining

B i

• s

t a t i s t i c s M a c h i n e l e a r n i n g

Open Source Connectivity

J A V A C

• d

e r e u s a b i l i t y Object Oriented Programming

Internet

Distributed Computing CORBA

I n t r a n e t

G N U / L i n u x

M

• d

e r n E p i d e m i

• l
• g

y

Quality of Care Research O u t c

• m

e s R e s e a r c h Health Services Research Public Health Health Economics

Population-based Health Research

Evidence-Based Medicine

Clinical Epidemiology

Evidence-Based Medicine

Clinical Epidemiology

e.health services research

Information from multiple sources Multidisciplinary and Multidimensional Cross-sectional and Longitudinal Experimental and Observational Administrative and Ad Hoc Systematic and Non-Systematic

Population-based health research

“Data-Warehouse” Geographical Information Systems Disease Registries Research Evidence Hospital Episodes Drug utilization Individual/aggregated socio-economic indicators Primary care Individual/aggregated characteristics Health law/regulations Structures, Operators

Bias/Variance Sampling techniques Adjustment Correlation Subgroups analysis Causal pathways Sampling Techniques Markov Chains Monte Carlo Gibbs sampling Random Effects Models Generalized Mixed Models Bayesian Mixed Models Multilevel models Hierarchical Regression Models Multilevel linear models Marginal models Generalized Estimating Equations

Statistical Sciences

Public Access to Re-usable sources Software engineering Developers’ networks Standard browsing Remote access Free Software Foundation/ GNU’s not Unix (GNU) Language Compilers-C,C++ Linux Connectivity Web-server technology Java Browser technology Cross-platform applications

Open source connectivity

Ecological fallacy drawing inferences at the individual level based on group-level data Atomistic fallacy drawing inferences at the group level based on individual data Psychologistic fallacy assuming that individual-level outcomes can be explained only in terms of individual-level characteristics (opposite:sociologistic) “Confounding” Micro-macro level variables interaction !!

Population-based fallacies

Objective:

• Automating processes of data-warehousing and statistical analysis

for complex and fragmented databases

• Delivering results in the form of widely accessible and interpretable reports

Procedures:

• BROWSE
• REPORT/LIST
• REPORT/INDEX
• STANDARD
• MAPS
• GENMOD
• Supported by CMNS, Regione Puglia, SAS, SUN Microsystems
• Version 6.12, modules AF, BASE, STAT, GRAPH
• Standardized Browser Outputs (HTML+Javascript)
• Designed for remote analysis through SAS/CONNECT

Reporting-by-Intranet Statistical System

http://statbone.cmns.mnegri.it

Sites database: site code, description, IP+location address Specifications for linked operational database (.db file) Specifications for linked structure database (.st file) Network Script: .net file ASCII files: CSV record (pre-formatted, analytical form) Operational Database .db file Formatted document (input field) with structure data Structure Database .st file Environment data (linked to operational variables) Environment databases .env file Version specific SAS programs (overlays) SAS programs .mak file (main) .mak file (custom) Population and geographical data Population Datasets .sd2 file SAS datasets:

Open Source – Script “logic”

N2 fi

• m

( )

n

f f f ,..., ,

2 1

ϕ

Touch+Pile-Up+Meta-Analyze (TPA) Distributed/Fragmented, Multilevel Hierarchical Network

Operational Database

N1

Structure Data

fi

Cumulative meta-analysis

m

( )

n

f f f ,..., ,

2 1

ϕ

N2 fi N1 fi

• Parallel processing

An ‘admin’ RISS version: RISS-H

List Hospitals x Diabetes List Hospitals x Complications

Specialties (Nominal variables)

Complications Yes/No (Binary Variables) LOS classes (Categorical Variables) AGE (Continuous)

Regione Puglia – Southern Italy, 114 Hospitals

Diabetes

QUED Network RISS/MAPS

QuED

Qualità ed esito in Diabetologia

An ‘epi’ RISS version: RISS-Q

List Clinical Variables QuED centers Index SF-36 QuED centers

Public Health Data Burden of Disease Census data, AusDiab Operational Databases Hospital Databases Pharmaceutical prescriptions Regulations Case-Mix/DRG Geographical data Census, Boundaries

DHS SHCN Governance Governance

Health needs Vs Health services

LAYER 1 (MACRO LEVEL)

NHMRC SHCN

ln[π(x)/(1-π(x)]=β0+β1x

Research Research

Evidence-based Medicine Vs Clinical Practice

LAYER 2 (META-LEVEL)

Clinical guidelines Scientific Associations Cochrane Library Patient outcomes Medical records Clinical Variability Intranet Shared Information System Multilevel models

Diabetes Register

Service Mapping Provider resources Medical databases CARDIAB Ambulatory Data Diabetes Centres Educational programs Diabetes Australia Surveys PRE: AusDiab - POST: Ad-hoc Quality of Life in Diabetes (SF-36)

LAYER 3 (MICRO-LEVEL)

Implementation Implementation

Population-based Approach Vs Disease Management

DHS SHCN

e.health systems