Transcript 1999 2007 - Inter-American Development Bank

A Primary Care Approach to
Chronic Disease Control in Brazil
JAMES MACINKO, PHD
NEW YORK UNIVERSITY
Controlling chronic conditions requires a
broad-based approach
Copyright ©2004 BMJ Publishing Group Ltd.
Source: Epping-Jordan, J E et al. Qual Saf Health Care 2004;13:299-305
And health systems require re-orientation to
maximize their impact on chronic disease
Fonte: WHO, 2008
High quality primary health care is essential to
chronic disease prevention and control
 Primordial prevention


Mothers’ health and nutritional status during pregnancy may affect the chance that her
children will develop chronic disease (diabetes, hypertension, heart disease) later in life
Strengthen support for regulation of smoking and other environmental approaches to
reducing risk
 Primary prevention



Provide advice and support for physical activity, healthy diets, and non-smoking norms
Individual, school, and community-based health promotion
Regular screening for risk factors
 Secondary prevention




Monitoring of risk factors and complications
Individual and group support for increasing physical activity, changing dietary habits,
quitting smoking
Access to and strategies to increase adherence to medications (lipid-lowering, antihypertensive)
Coordination of specialist care
 Tertiary prevention

Rehabilitation and palliative care, coordination of specialist care, monitoring of adherence
to medications and complications
Brazil has a robust, universal, and comprehensive
approach to providing primary health care
Organization and Financing
• All citizens covered -access to care is a constitutional right (1998)
• No user fees for services
• Decentralization of management and provision to local levels
• Many medications provided free of charge
Primary care providers
• Traditional Public Facilities (health posts and centers)
• Private sector providers (minimal role in primary care ~<20%)
• Family Health Program (since 1994/6)
Note: in contrast to primary care, hospital and diagnostic services are
provided by a combination of public and contracted privatesector/nonprofit providers
The Family Health Strategy (FHS) is now the main
form of PHC in the country
• Multi-professional teams (at least 1 MD, 1 RN, 1 RN asst,
and 4-6 community health agents)
• Community orientation (organized by family and geographic
territory-3,540 people assigned to each team)
• Use local data to plan health activities and services
• Active promotion –home visits by community health
workers, conduct neighborhood health promotion activities
• Local health councils stimulate public participation,
accountability, and intersectoral actions
• Teams use evidence-based protocols (adapted from the
UK) for chronic disease diagnosis, prevention, and control
1. Note Brazil is NOT facing the same health professional crisis as many other developing countries
The FHS team has specific functions designed to
prevent, detect, and manage chronic diseases
Physicians
Nurses
Medical Assistants
Community Health
Agents
 Clinical consultations and
care;
 Confirm diagnosis,
evaluate risk factors,
identify comorbidities;
 Request diagnostics if
necessary;
 Team strategies for
patient education;
 Refer to secondary and
tertiary care
 Yearly referrals for all
diabetic patients to assess
complications;
 Work with patients on
treatment goals (e.g.
smoking, weight, blood
pressure)
 Assess risk factors,
adherence, complications;
referral to physician;
 Community health
promotion;
 Individual and group
activities for patients;
 Establish strategies to
promote adherence;
 Prescribe exams
established in clinical
protocols;
 Monitor medication of
clinically stable individuals;
 Manage MD referral
schedule based on risk
control, complications
 Check blood pressure,
weight, height and waist
circumference;
 Advise community on
lifestyle, diet, physical
activity;
 Community education on
CVD risk;
 Schedule appointments;
 Record information on
medical records.
 Manage referral requests
for complementary exams;
 Manage drug stock,
request replacements,
provide drugs in the
absence of pharmacist.
Home visits
Provide community
information on CVD risks
and their prevention;
 Identify and refer people
with risk factors;
 Monitor appointment
compliance and return of
test results;
 Ask if the individual is
following guidelines for diet,
physical activity, weight
control, smoking cessation,
alcohol, Rx adherence;
 Note in patient records:
CVD diagnosis, risk
factors.
Source: Brazilian Ministry of Health
Chronic disease mortality in Brazil, 1996-2005:
progress and challenges
Age-Standardized death rates (per 100,000)
65
60
60
56
55
55
53
50
50
49
48
49
48
35.2
34.9
35.1
35.4
35.1
35.3
20.7
20.3
21
21.1
22
22
2000
2001
2002
2003
2004
2005
50
45
40
35
35.5
30
35.3
35.8
35.4
25
20
16.7
17.2
17.5
1996
1997
1998
19.3
15
10
5
0
AMI
1999
Diabetes
Source: Saude Brasil 2008; Malta et al, 2006
Stroke
The effectiveness of PHC can be assessed using
Primary Health Care Sensitive Hospitalizations*
9
We are interested in PHCSCs because:
• Mortality data only provide a small proportion of the
total burden of disease in the population
• Analysis of hospital data can help to identify:
•
•
•
Access barriers to primary care (geographic, physical, financial,
cultural)
Variations in quality of care (poor quality acute or chronic care
management)
Overall performance of primary care services (e.g.
comprehensiveness of services provided, coordination of care
provided elsewhere)
 But, we need to distinguish between all hospitalizations
(trauma, etc.) and those sensitive to primary care

Hence, the need for a valid list of conditions sensitive to primary care
*Known elsewhere as Ambulatory Care Sensitive Hospitalizations (ACSHs)
Development and adaptation of the Brazilian
PHCSH list (2005-2010)
10
1.
2.
3.
4.
5.
6.
7.
8.
9.
Development of conceptual framework (2005)
Systematic literature review (2005-6)
Analysis of data from international studies (2006-7)
2 expert meetings (2005, 2006)
Consultation with the Brazilian Society of Family Physicians
(2007)
Public solicitation of comments (Ministry of Health, 2007)
Final revision of list (publications 2008/9)
Government adoption and publication of list (Portaria Nº 221, de
17 de Abril de 2008)*
Analysis of results and empirical validation (2008-2010)
*Government decree and list of conditions available:
http://www.saude.sp.gov.br/resources/profissional/acesso_rapido/gtae/saude_pessoa_idosa/u_pt_ms_sas_221_170408_condicoes_sensiveis.pdf
Source: Alfradaque, E. et al. (2009). Ambulatory Care Sensitive Conditions: elaboration of Brazilian list as tool for measuring health system performance. [In
Portuguese]. Cadernos de Saúde Pública. Jun;25(6):1337-49.
The final Brazilian list of Primary Health Care
Sensitive Hospitalizations (PHCSH)
1
Vaccine preventable conditions
11
Cardiac Arrest
2
Gastroenteritis
12
Stroke
3
Anemia
13
Diabetes mellitus
4
Nutritional deficiencies
14
Epilepsy
5
Ear, nose, and throat infections
15
Kidney and urinary tract infections
6
Bacterial Pneumonias
16
Skin and subcutaneous infections
7
Asthma
17
Female pelvic inflamatory disease
8
Lower respiratory infections
18
Gastrointestinal ulcer
9
Hypertension
19
Diseases related to pre-natal period
and birth
10 Angina pectoris
*Note: conditions marked in green are analyzed separately as “chronic disease PHCSH”
Source: Alfradaque, E. et al. (2009). Ambulatory Care Sensitive Conditions: elaboration of Brazilian list as tool for measuring health system performance. [In
Portuguese]. Cadernos de Saúde Pública. Jun;25(6):1337-49.
Data used to test whether FHS expansion was
related to changes in PHCSH rates over time
12
 Individual hospitalization files used for hospital reimbursement by the public
health system (known as the AIH).
 Specific condition (ICD-10 codes), patient information (age, sex,
municipality of residence), type of hospital, specific procedures, total costs
reimbursed, length of stay, use of intensive care, and in-hospital death.
 We link over 60 million registered hospitalizations since 1999 to information
on Brazil´s 27 states and 5,564 municipalities.
 Control variables (from IBGE, census, IPEA): supply of health services (public
and private hospital beds, percent of population covered by private health
plans, mean annual medical visits per capita), Family Health Program
availability, socio-economic conditions (median income, illiteracy rates,
availability of clean water) and population health needs (premature mortality).
 Population (denominator) data from the 2000 census and recent projections
made by IBGE/Datasus
FHS coverage increased each year to 30,000
teams in 2009, reaching 98 million people
>30,000 teams covering 98
million people (51% of
population)
Nearly 30,000 MDs and RNs,
about 225,000 CHWs
PHCSH and non-PHCSH *, 1999-2007,Brazil
Millions of hospitalizations (#)
6
5
4.6
4.76
4.96
5.02
5
4.95
4.98
5.03
4.72
9.5% increase
4
3
2
1.92
1.9
1.87
1.83
1.78
1.75
1.67
1.64
1.6
17% reduction
1
0
1999
2000
2001
2002
ACSH
*Excludes hospitalizations for births
2003
2004
2005
2006
Non-ACSH
Source: Dourado et al. 2010
2007
PHCSH rates* and trends over time vary by state
1999
RR
2007
RR
AP
AM
PA
MA
CE
PI
AC
RO
TO
MT
BA
RN
PB
PE
AL
SE
AP
AM
PA
MA
PI
AC
RO
TO
MT
BA
GO
GO
MG
MS
CE
MG
ES
SP
PR
SC
RS
RJ
ICSAP (taxas padronizadas por idade e sexo)
Por 10.000 habitantes
* Age standardized rates per 10,000.
Excludes hospitalizations for births.
Source: Dourado et al. 2010
> 217
189 - 217
159 - 188
130 - 158
< 130
MS
ES
SP
PR
SC
RS
RJ
RN
PB
PE
AL
SE
PHCSH rates, mean annual % change since 1999,
by state and sex
Men
Women
-1.00
-1.00
-3.92
-4.88
-3.92
-4.88
-2.96
-1.00
-4.88
-1.98
-3.92
-1.98
-4.88
0.00
-4.88
-3.92
-3.92
-2.96
-2.96
-1.98
0.00
-2.96
-1.98
2.02
6.18
2.02
-5.82
-10.00
-5.00
0.00
5.00
DF
GO
MT
MS
RS
SC
PR
SP
RJ
ES
MG
BA
SE
AL
PE
PB
RN
CE
PI
MA
TO
AP
PA
RR
AM
AC
RO
10.00
1.01
-1.00
-5.82
-4.88
-3.92
-5.82
-2.96
-1.98
-3.92
-2.96
-4.88
-2.96
-5.82
0.00
-4.88
-4.88
-4.88
-1.98
-3.92
-2.96
-1.98
-1.00
-1.98
4.08
6.18
3.05
-5.82
-10.00
-5.00
0.00
5.00
DF
GO
MT
MS
RS
SC
PR
SP
RJ
ES
MG
BA
SE
AL
PE
PB
RN
CE
PI
MA
TO
AP
PA
RR
AM
AC
RO
10.00
Results from Poisson regression of: log(rate)= β0+β1(year), rate1999 = 10,000 exp(β0). The annual % change from 1999 to 2007 = 100 [exp(β1)-1].
Source: AIH, population (states) from IBGE/DATASUS.
Inflation-adjusted $Reais (millions)
Expenditures on hospitalizations rose in Brazil
from 1999-2007, but not as quickly for PHCSH
3500
Average expenditures per
hospitalization (in constant $Reais)
3000
Year
PHCSH
Non-PHCSH
1999
340.64
485.42
2000
320.84
437.23
2001
351.16
474.81
2002
337.29
458.94
2003
380.53
528.6
2004
402.65
543.94
2005
501.96
645.44
2006
478.82
624.61
2007
512.61
660.39
Non-PHCSH
2500
2000
1500
1000
PHCSH
500
Total expenditure on hospitalizations
in 2007: $4.14 Billion Reais
0
1999
2000
2001
*Excludes hospitalizations for births
2002
2003
2004
2005
2006
2007
Source: Macinko et al. AJPH (forthcoming)
Predicted impact of community health agents on
circulatory disease hospitalizations, Women, 1998-2002
18
115.0
Hospitalizations
/100,000
110.0
Without community health agents
105.0
100.0
95.0
With community health agents
90.0
1998
1999
2000
2001
2002
*Adjusted hospitalization rates from fixed effects regression models controlling for sociodemographics, health system variables, year and municipal fixed effects. R^2 = 0.84
2,448 municipalities included in sample.
Source: Guanais & Macinko, J Amb Care Mgmt 2009
Dynamic panel estimations of PHCSH rates,
Brazilian microregions, 1999-2007
*
70-79 years
*
*
60-69 years
*
*
*
FHP-high (75-100%)
FHP-mid (50-74%)
-60
-50
-40
Women only
*
*
FHP-low (25-49%)
-70
20-59 years
Men only
*
-30
-20
-10
All
0
Graphic presents beta coefficients. All models control for log income per capita, clean water, illiteracy, health insurance, medical
consultations per capita, premature mortality, and year effects. Dynamic models include 1- and 2-year lagged dependent
variables and treat all variables except for income as endogenous. M1 and M2 tests for the first-order and second-order serial
correlation in the first-differenced residuals and Sargan test of the over-identifying restrictions under the null of instruments’
validity (with two-step estimator) all confirm assumptions of the dynamic panel model.
*p<0.05
Source: Macinko et al. AJPH (forthcoming)
Predicted PHCSH rates, stratified by FHP
coverage and private/contracted hospital beds
*Excludes hospitalizations for births
Source: Macinko et al. AJPH (forthcoming)
Family Health Strategy coverage and chronic
disease PHCSH
Highest
FHP quartile 5 (vs quartile 1)
0.87
FHP quartile 4 (vs quartile 1)
0.89
FHP quartile 3 (vs quartile 1)
0.92
FHP quartile 2 (vs quartile 1)
0.96
0.85
0.9
0.95
Adjusted Prevalence Ratios
Lowest
1
Results from Negative binomial regression of the number of hospitalizations per municipality with log
population size as an offset. Regressions control for age, sex, hospital beds, water supply, literacy
rates, log municipal income, and municipal and year fixed effects. FHP coverage estimated using left
party mayor and percent municipal GDP spent on health and social services as instruments (R2=
0.15). Instrumented FHP variable transformed into quintiles.
*Excludes hospitalizations for births
Source: Macinko, Dourado, et al. 2010 Health Affairs (under review)
FHS coverage and specific chronic disease
PHCSH rates, 1999-2006
Fewer Hospitalizations
Greater Hospitalizations
Diabetes
1.02
1
COPD
0.99
0.96
0.96
Stroke
0.92
0.91
Other CVD
0.77
0.75
FHP quartile 5
0.83
0.8
0.87
0.85
FHP quartile 4
0.95
1.09
1.03
1.01
0.99
0.98
1.02
0.95
0.96
0.97
Hypertension
Asthma
1.06
1.05
0.98
0.95
0.9
0.95
FHP quartile 3
1
1.05
1.1
FHP quartile 2
Results from Negative binomial regression of the number of hospitalizations per municipality with log population size as an offset. Regressions control for
age, sex, hospital beds, water supply, literacy rates, log municipal income, and municipal and year fixed effects. FHP coverage estimated using left party
mayor and percent municipal GDP spent on health and social services as instruments (R2= 0.15). Instrumented FHP variable transformed into quintiles.
Source: Macinko, Dourado, et al, 2010. Health Affairs (under review)
Conclusions
A robust and universal approach to community-based primary health care can
lower hospitalizations related to a number of preventable and chronic diseases
 PHCSH rates declined by over 5 percent per year, independent of other
factors
 FHS expansion was associated with declines in PHCSH rates
 Hospitalizations for the main chronic diseases fell by 24% for men and 30%
for women.
 FHS expansion was associated with declines for cardiovascular diseases,
asthma, hypertension, and stroke.
However,
 FHS expansion was associated with greater hospitalizations for diabetes and
COPD (perhaps due to referrals?)
 Other health system factors were also related to PHCSH rates (such as
private/non-profit hospital beds which were associated with higher
hospitalization rates)
Study Limitations
 This approach requires good quality administrative data and
measures of PHC “exposure” that are absent in many countries

It could be adapted to smaller areas or by using linked survey data
 Cause of hospitalizations cannot be independently verified
 Existing studies suggest the quality of coding is adequate and improving, but
more work needed to validate hospital claims
 Ecological measures do not permit assessment of causal
relationship at the individual level

However, we use rigorous methods to assess the relationship at the municipal
and micro-regional levels
 We did not assess cancer or mental health--two additional
important chronic conditions

These two conditions are important but are not currently managed in primary
care in Brazil (besides some forms of screening and primary prevention)
 Further work needs to be done to establish the overall costs and
cost-effectiveness of the Brazilian PHC approach
Applicability to other countries
 Countries might want to think about PHCSH when considering





investments to be made in data infrastructure and PHC evaluation
strategies
Work with individual-level risk adjustment, admissions versus
readmissions (via data linkage), changes in length of stay, and variation
by type and size of hospital
City-wide household surveys could link households to their primary
care providers and correlate this with levels of PHCSC at the census
tract (Belo Horizonte)
Demonstration projects to reduce PHCSHs in municipalities with high
rates
Further analysis of expenditures and cost-effectiveness
Training municipal health authorities to use and interpret the PHCSH
measures
Acknowledgments
 Many thanks to the IDB Division of Health and Social Protection
 This study was financed by the Inter-American Development Bank and
the Ministry of Health of Brazil
 Co-authors include the “Projeto ICSAP” team: Ines Dourado, Veneza
Berenice de Oliveira, Maria Fernanda Lima-Costa, Maria Turci,
Guadalupe Medino, Eduardo Mota, Rosana Aquino, and Palmira
Bonolo, and Frederico C. Guanais.
Thank you!
QUESTIONS?
[email protected]