This document discusses a study that used multiple correspondence analysis to help identify health inequities across regions in Portugal. The study analyzed linked electronic health records of adults admitted to hospitals for acute myocardial infarction between 2012-2013. It found disparities between regions in recording health indicators like blood pressure and cholesterol. Regions with higher rates of no recording for indicators also had higher mortality rates. The multivariate analysis generated visual outputs that clearly showed regional inequities in mortality and quality of electronic health records, helping pinpoint areas that need policy action.
Poster presented at ISPOR Vienna 2016 LBacelar-Nicolau
1. Multiple
Correspondence
Analysis Factorial
Plan 1-2
Registration of each health
indicator by region Association between Mortality
and No registration of health
indicators by Region (1)(2)
HELPING DECISION-MAKERS VISUALIZE
INEQUITIES IN HEALTH IMPACT ASSESSMENT:
LINKED ELECTRONIC RECORDS, MORTALITY
AND REGIONAL DISPARITIES IN PORTUGAL
Bacelar-Nicolau L Rodrigues T Fernandes E Lobo MF Nisa C
Azzone V Teixeira-Pinto A Rocha-Gonçalves F Azevedo LF
Freitas A Normand SL Costa-Pereira A Pereira Miguel J
Methods
″ Show how multivariate methods
outputs visualization may help
decision-makers identify inequities
through health impact assessment
(HIA)
″ Show how a policy promoting quality
of linked Electronic Health Records
(EHR) regarding Acute Myocardial
Infarction (AMI) patients may be
associated to mortality and regional
inequities.
″ Data
Adults (>=20 years) admitted in Portuguese
NHS Hospitals with AMI diagnosis during the
2nd semester 2012, followed regularly in NHS
Primary Care (PC) during 2013
Source: SPMS/ACSS
″ Variables
• Mortality outcome at December 31 2013
• 4 PC registered indicators - Blood Pressure,
Cholesterol, Triglycerides, Waist
Circumference, Body Mass Index (BMI)
• 5 regions - Norte, Centro, LVT, Alentejo,
Algarve
• Adjustment variables - sex, age, AMI
hospital readmissions during the 2nd
semester 2012, surgery during hospital
admission, mean admission duration>6
days, hospital diagnosis>=6
″ Statistical Procedures
• Qui-Square and Fisher Exact tests
• Logistic Regressions simple and stratified by
region, estimating crude and adjusted ORs
(and 95%CI)
• Multivariate Correspondence Analysis
(MCA)
Aims Findings
″ Algarve and LVT have the highest no-registration percentages for every
indicator. Norte, usually followed by Algarve and Centro, have the lowest
percentages.
″ Globally, patients followed regularly in PC, but without registered
information for all health indicators, are more than twice as likely to be
deceased, than patients with information registered (OR>2.0).
″ Mortality is more associated with no records for: Cholesterol, Triglycerides
(OR>13.0) and Blood pressure (OR>3) in Norte; BMI in Alentejo (OR
near7); Cholesterol and Triglycerides (3.17 and 4.41) in Centro.
″ MCA summarizes previous findings with a clear visual graphical aid.
Main Conclusions
″ Disparities between regions are found regarding Electronic Health Records of AMI patients, but also concerning
association patterns of mortality and registration of these health indicators.
″ Complex multivariate methodologies generate visual outputs easily interpreted by decision-makers, exposing
regional inequities regarding mortality and PC recording policies and pinpointing where action priorities may lay.
Disclosures: Nothing to disclose. Funded by FCT, QREN, COMPETE (HMSP-ICT/0013/2011).
Acknowledgments: SPMS/ACSS for kindly making available the data that made this research possible
CUTEheart
Comparative Use of Technologies
for Coronary Heart Disease
Harvard Medical School – Portugal Program
http://www.cuteheart.med.up.pt
LNICOLAU@MEDICINA.ULISBOA.PT
ISPOR 19th Annual European
Congress 2016 Vienna, Austria