Work in Progress

Simulating the Local Dynamics of Cardiovascular Health and

Related Risk FactorsJack Homer

Homer Consultingjhomer@comcast.net

Bobby Milstein

Centers for Disease Control and Preventionbmilstein@cdc.gov

University of Michigan Tobacco Modeling MeetingMay 2008

This work was funded by the CDC’s Division for Heart Disease and Stroke Prevention and by the National Institutes of Health’s Office of Behavioral and Social Science Research. The

work was done in collaboration with the Health and Human Services Department of Austin/Travis County, Texas, and with Indigent Care Collaboration of Central Texas. The

external contractors are Sustainability Institute and RTI International.

ContributorsCore Design Team• CDC: Darwin Labarthe, Diane Orenstein, Bobby Milstein, Marilyn

Metzler, Rosanne Farris• Austin: Adolfo Valadez, Phil Huang, Karina Loyo, Rick

Schwertfeger, Cindy Batcher, Ella Pugo, Josh Vest • NIH: Patty Mabry• Consultants: Kristina Wile, Jack Homer, Justin Trogdon

Organizational Sponsors• Austin/Travis County Health and Human Services Department• CDC Division for Heart Disease and Stroke Prevention• CDC Division of Adult and Community Health• CDC Division of Nutrition, Physical Activity, and Obesity• CDC Division of Diabetes Translation • CDC Office on Smoking and Health• CDC NCCDPHP Office of the Director• Indigent Care Collaborative (Austin, TX)• NIH Office of Behavioral and Social Science Research• RTI International• Sustainability Institute• Texas Department of Health

Brief Background on System Dynamics Modeling

Compartmental models resting on a general theory of how systems change (or resist change) – often in ways we don’t expect

– Developed for corporate policies in the 1950s, and applied to health policies since the 1970s

– Concerned with understanding dynamic complexity• Accumulation (stocks and flows)

• Feedback (balancing and reinforcing loops)

– Used primarily to craft far-sighted, but empirically based, strategies• Anticipate real-world delays and resistance

• Identify “high leverage” interventions

– Modelers engage stakeholders through interactive workshopsForrester JW. Industrial Dynamics. Cambridge, MA: MIT Press; 1961.

Sterman JD. Business Dynamics: Systems Thinking and Modeling for a Complex World. Boston, MA: Irwin/McGraw-Hill; 2000.

Homer J, Hirsch G. System dynamics modeling for public health: Background and opportunities. American Journal of Public Health 2006;96(3):452-458.

Purpose of the Cardiovascular Risk Model

• How do local conditions affect multiple risk factors for CVD, and how do those risks, in turn, affect population health status and costs over time?

• How do different local interventions affect cardiovascular health and related expenditures in the short- and long-term?

• How might local health leaders better balance their policy efforts given limited resources?

Smoking

Obesity

Secondhandsmoke

Healthinessof diet

Extent ofphysical activity

Psychosocialstress

Diagnosisand control

First-time CVevents and

deaths

Access to and marketingof smoking quit products

and services

Access to andmarketing of mental

health services

Sources ofstress

Access to andmarketing of healthy

food options

Access to andmarketing of physical

activity options

Access to andmarketing of weight

loss services

Access to andmarketing ofprimary care

Particulate airpollution

Utilization ofquality primary

care

Tobacco taxes andsales/marketing

regulations

Smoking bans atwork and public

places

Junk food taxes andsales/marketing

regulations

Downwardtrend in CV

event fatality

Quality of primarycare provision

Chronic Disorders

Costs from CV and other riskfactor complications and

from utilization of services

Anti-smokingsocial marketing

High BP

Highcholesterol

Diabetes

The CDC is partnering on this project with the Austin (Travis County), Texas, Dept. of Health and Human Services. The model is calibrated to represent the

overall US, but is informed by the experience and data of the Austin team, which has been supported by the CDC’s “STEPS” program since 2004.

The CDC is partnering on this project with the Austin (Travis County), Texas, Dept. of Health and Human Services. The model is calibrated to represent the

overall US, but is informed by the experience and data of the Austin team, which has been supported by the CDC’s “STEPS” program since 2004.

Homer J, Milstein B, Wile K, Pratibhu P, Farris R, Orenstein D. Modeling the local dynamics of cardiovascular health: risk factors, context, and capacity. Preventing Chronic Disease 2008;5(2). Available at http://www.cdc.gov/pcd/issues/2008/apr/07_0230.htm

Direct Risk Factors

Smoking

Secondhandsmoke

First-time CVevents and

deaths

Particulate airpollution

Downwardtrend in CV

event fatalityChronic Disorders

High BP

Highcholesterol

Diabetes

Calculating First-Time CV Events & Deaths

Based on well-established Framingham approach for calculating probability of first-time events & deaths in individuals• CVD = CHD (MI, angina, cardiac arrest) + Stroke/TIA + CHF + PAD

Modifies individual-level risk calculator for use with populations• Uses prevalences of uncontrolled chronic disorders by sex/age group

• Introduces secondhand smoke and pollution as additional risk factors

• Combines risks multiplicatively to account for overlapping conditions

• Adjustment exponents reproduce synergies seen in individual-level calculator

• Adjustment multipliers reproduce AHA event and death frequencies for 2003

- Anderson et al, Am Heart J 1991 (based on Framingham MA population N=5573, 1968-1987)

- Homer “Risk calculation in the CVD model” project document, June 19, 2007

- NHANES 1988-94 & 1999-04

- AHA Heart Disease and Stroke Statistics – 2006 Update

Indirect Risk Factors

Smoking

Obesity

Secondhandsmoke

Healthinessof diet

Extent ofphysical activity

Psychosocialstress

Diagnosisand control

First-time CVevents and

deaths

Particulate airpollution

Utilization ofquality primary

care

Downwardtrend in CV

event fatalityChronic Disorders

High BP

Highcholesterol

Diabetes

Tobacco and Air Quality Interventions

Smoking

Obesity

Secondhandsmoke

Healthinessof diet

Extent ofphysical activity

Psychosocialstress

Diagnosisand control

First-time CVevents and

deaths

Access to and marketingof smoking quit products

and services

Particulate airpollution

Utilization ofquality primary

care

Tobacco taxes andsales/marketing

regulations

Smoking bans atwork and public

places

Downwardtrend in CV

event fatalityChronic Disorders

Anti-smokingsocial marketing

High BP

Highcholesterol

Diabetes

Health Care Interventions

Smoking

Obesity

Secondhandsmoke

Healthinessof diet

Extent ofphysical activity

Psychosocialstress

Diagnosisand control

First-time CVevents and

deaths

Access to and marketingof smoking quit products

and services

Access to andmarketing ofprimary care

Particulate airpollution

Utilization ofquality primary

care

Tobacco taxes andsales/marketing

regulations

Smoking bans atwork and public

places

Downwardtrend in CV

event fatality

Quality of primarycare provision

Chronic Disorders

Anti-smokingsocial marketing

High BP

Highcholesterol

Diabetes

Interventions Affecting Stress

Smoking

Obesity

Secondhandsmoke

Healthinessof diet

Extent ofphysical activity

Psychosocialstress

Diagnosisand control

First-time CVevents and

deaths

Access to and marketingof smoking quit products

and services

Access to andmarketing of mental

health services

Sources ofstress

Access to andmarketing ofprimary care

Particulate airpollution

Utilization ofquality primary

care

Tobacco taxes andsales/marketing

regulations

Smoking bans atwork and public

places

Downwardtrend in CV

event fatality

Quality of primarycare provision

Chronic Disorders

Anti-smokingsocial marketing

High BP

Highcholesterol

Diabetes

Healthy Diet Interventions

Smoking

Obesity

Secondhandsmoke

Healthinessof diet

Extent ofphysical activity

Psychosocialstress

Diagnosisand control

First-time CVevents and

deaths

Access to and marketingof smoking quit products

and services

Access to andmarketing of mental

health services

Sources ofstress

Access to andmarketing of healthy

food options

Access to andmarketing ofprimary care

Particulate airpollution

Utilization ofquality primary

care

Tobacco taxes andsales/marketing

regulations

Smoking bans atwork and public

places

Junk food taxes andsales/marketing

regulations

Downwardtrend in CV

event fatality

Quality of primarycare provision

Chronic Disorders

Anti-smokingsocial marketing

High BP

Highcholesterol

Diabetes

Physical Activity & Weight Loss Interventions

Smoking

Obesity

Secondhandsmoke

Healthinessof diet

Extent ofphysical activity

Psychosocialstress

Diagnosisand control

First-time CVevents and

deaths

Access to and marketingof smoking quit products

and services

Access to andmarketing of mental

health services

Sources ofstress

Access to andmarketing of healthy

food options

Access to andmarketing of physical

activity options

Access to andmarketing of weight

loss services

Access to andmarketing ofprimary care

Particulate airpollution

Utilization ofquality primary

care

Tobacco taxes andsales/marketing

regulations

Smoking bans atwork and public

places

Junk food taxes andsales/marketing

regulations

Downwardtrend in CV

event fatality

Quality of primarycare provision

Chronic Disorders

Anti-smokingsocial marketing

High BP

Highcholesterol

Diabetes

Adding Up the Costs

Smoking

Obesity

Secondhandsmoke

Healthinessof diet

Extent ofphysical activity

Psychosocialstress

Diagnosisand control

First-time CVevents and

deaths

Access to and marketingof smoking quit products

and services

Access to andmarketing of mental

health services

Sources ofstress

Access to andmarketing of healthy

food options

Access to andmarketing of physical

activity options

Access to andmarketing of weight

loss services

Access to andmarketing ofprimary care

Particulate airpollution

Utilization ofquality primary

care

Tobacco taxes andsales/marketing

regulations

Smoking bans atwork and public

places

Junk food taxes andsales/marketing

regulations

Downwardtrend in CV

event fatality

Quality of primarycare provision

Chronic Disorders

Costs from CV and other riskfactor complications and

from utilization of services

Anti-smokingsocial marketing

High BP

Highcholesterol

Diabetes

Data Sources for Modeling CVD Risk

• Census– Population, deaths, births, net immigration, health coverage

• AHA & NIH statistical reports – Cardiovascular events, deaths, and prevalence (CHD, stroke, CHF, PAD)

• National Health and Nutrition Examination Survey (NHANES) – Risk factor prevalences by age (18-29, 30-64, 65+) and sex (M, F)– Chronic disorder diagnosis and control (hypertension, high cholesterol, diabetes)

• Behavioral Risk Factor Surveillance System (BRFSS)– Diet & physical activity– Primary care utilization– Lack of needed emotional/social support Psychosocial stress

• Medical Examination Panel (MEPS) / National Health Interview (NHIS) – Medical and productivity costs attributable to smoking, obesity, and chronic disorders

• Research literature– CVD risk calculator, and relative risks from SHS, air pollution, obesity, and inactivity– Medical and productivity costs of cardiovascular events

• Questionnaires for CDC and Austin teams (expert judgment)– Potential effects of social & services marketing on utilization behavior– Effects of behavioral services on smoking, weight loss, stress reduction– Relative risks of stress for high BP, high cholesterol, smoking, and obesity

A “status quo” baseline scenario

• Result: Past trends continue after 2000, but decelerate and level off

– Increasing obesity, high BP, and diabetes

– Decreasing smoking

– High cholesterol mixed bag by age and sex, flat overall

Obesity prevalence

0.4

0.3

0.2

0.1

0

1990 1995 2000 2005 2010 2015 2020 2025 2030 2035 2040Time (Year)

Uncontrolled high BP prevalence

0.3

0.2

0.1

0

1990 1995 2000 2005 2010 2015 2020 2025 2030 2035 2040Time (Year)

Smoking prevalence

0.3

0.2

0.1

0

1990 1995 2000 2005 2010 2015 2020 2025 2030 2035 2040Time (Year)

The model is calibrated to reproduce data from NHANES 1988-94 and 1999-2004 on risk factor

prevalences in the non-CVD population by age and sex.

The model is calibrated to reproduce data from NHANES 1988-94 and 1999-2004 on risk factor

prevalences in the non-CVD population by age and sex.

Obese % of non-CVD popn

Uncontrolled hypertension %of non-CVD popn

Smoking % of non-CVD popn

• A straightforward starting point for “what if” analysis

– Assume no changes after 2000 in contextual factors or in risk factor inflow and outflow rates

– Any changes in risk prevalences after 2000 are due to “bathtub” adjustment and population aging

There are significant gains even at the least effective end

of the uncertainty range

There are significant gains even at the least effective end

of the uncertainty range

Testing All 19 Interventions Combined, with Uncertainty Ranges

Deaths 2015

Deaths 2040

All uncertain parameters at least impact

-15.0% -14.2%

All uncertain parameters at best-guess impact

-20.2% -19.3%

All uncertain parameters at most impact

-28.2% -26.3%

% Change from Base CaseDeaths from CVD per Capita*

4

2

01990 2000 2010 2020 2030 2040

All 19 Interventionswith uncertainty range

Base Case

Deaths from CVD if all risk factors = 0

20.2%

19.3%

Testing Selected Intervention Clusters

Primary Care interventions (N=3): • Quality of Primary care increased• Primary care services marketed• Access to primary care increased

Air Quality interventions (N=2)• Air pollution cut to half of recent value• Workplaces allowing smoking cut to zero

Tobacco interventions (N=4)• Tobacco tax and sales restrictions• Social marketing against smoking• Smoke quit services marketed• Access to smoking quit services increased

% Change vs. Base Run Intervention Cluster 2015 2040

Base case -- --

Primary care -11.5% -9.1%

Primary care + Air quality

-16.2% -12.3%

Primary care + Air quality +

Tobacco-18.6% -15.4%

All 19 interventions -20.2% -19.3%

CV Deaths

The 3 (or even just the first 2) clusters together provide a large fraction of the CV

deaths reduction of all 19 interventions, especially in the shorter term:

• 92% (80%) by 2015,• 80% (64%) by 2040.

The 3 (or even just the first 2) clusters together provide a large fraction of the CV

deaths reduction of all 19 interventions, especially in the shorter term:

• 92% (80%) by 2015,• 80% (64%) by 2040.

Intervention Effects on Smoking Inflows & Outflows

Age 18 smokers

Social marketing

Tax & sales restrict

0.5 [0.4-0.7]

0.5 [0.3-0.7]

Adult smoking initiation/relapse

Smoking quits

Adult Smokers

Use of quit services2.25 [1.5-3]

1.3 [1.2-1.5]1.85 [1.5-2.5]

0.7 [0.5-0.8]

0.6 [0.4-0.7]

Workplace ban(for those who work) 1.25 [1.2-1.4]

0.6 [0.5-0.7]

0.65 [0.55-0.75]

Sources: -Terry Pechacek CDC, personal correspondence, citing CPSTF (re: taxes and sales restrictions and re: social marketing).- Moskowitz et al, AJPH 2000 (re: workplace bans)- Glasgow et al, Tobacco Control 1997 (re: workplace bans)- Terry Pechacek, citing multiple studies and CPSG (re: quit services)- Abby Rosenthal CDC, personal correspondence (re: quit services)

10.0%

20.0%

34.3%

0%

10%

20%

30%

40%

Baseline If Full Access If Full Access,Max Mktg & Max

Quality Care

% o

f s

mo

ke

rs u

sin

g q

uit

sv

cs

Use of Quit Services by Smokers

Sources: - MEPS spending analysis, re: baseline use of quit services and products- Terry Pechacek CDC, personal correspondence, citing Group Health Cooperative study, re: effects of marketing and quality primary care

Effects of Interventions on Smoking Prevalence

In the base run, smoking prevalence among the non-CVD population declines from 17.7% in 2010 to 13.5% in 2040. The AirQ2 intervention

cluster reduces the 2040 value to 12.9% (due to the effect of indoor smoking laws), and then adding the Tob4 cluster reduces it to 4.5%.

Smoking Prevalence (non-CVD population)

0.3

0.2

0.1

01990 2000 2010 2020 2030 2040

All19

PC 3

PC 3 + AirQ 2 + Tob 4

PC 3 + AirQ 2

Base

Effects of Interventions on Secondhand Smoke Exposure

Fraction Nonsmokers SHS Exposure (non-CVD population)

0.6

0.4

0.2

01990 2000 2010 2020 2030 2040

All19

PC 3

PC 3 + AirQ 2 PC 3 + AirQ 2 + Tob 4

Base

In the base run, the fraction of non-smokers with significant secondhand smoke exposure declines from 19.1% in 2010 to 15.4% in 2040, tracking the decline in smoking. The AirQ2 intervention cluster reduces the 2040 value to 4.2% (due to the effect of indoor smoking

laws), and then adding the Tob4 cluster reduces it to 1.5%.

Deaths from CVD per 1000 (non-CVD population)4

3

2

1

01990 2000 2010 2020 2030 2040

All19

Deaths from CVD if all risk factors = 0

PC 3

PC 3 + AirQ 2 + Tob 4

PC 3 + AirQ 2

Base

Effects of Interventions on CVD deaths

Note the increasing impacts over time for Tobacco4 as well as the other “gradual impact” interventions included in All19:

Physical activity, Nutrition, Weight loss, Stress

Effects of Interventions on Smoking-related Cancer & COPD Deaths

Includes smoking-related deaths from cancers and respiratory diseases, based on 2001 data from SAMMEC (http://apps.nccd.cdc.gov/sammec/).

SAMMEC = Smoking Attributable Mortality, Morbidity and Economic Costs. Male = 273,665 cancer and respir deaths due to smoking; Female = 135,296.

NonCVD deaths from smoking complications

400,000

300,000

200,000

100,000

0

1990 2000 2010 2020 2030 2040

All19

PC 3

PC 3 + AirQ 2

PC 3 + AirQ 2 + Tob 4

Base

Effects of Interventions on Preventable Deaths (2010-2040 cumulative)

Cumulative deaths 2010-2040 (in non-CVD population) from CV and other risk factor complications, in millions

From From Other CV Complications CombinedBase 19.6 10.8 30.4PC3 17.7 10.4 28.2PC3AirQ2 17.1 10.3 27.4PC3AirQ2Tob4 16.6 6.8 23.5All19 16.1 6.5 22.6

Over 30 years, the “Tob4” intervention cluster reduces CV deaths by 0.5m, and reduces other deaths (cancers & respiratory) by 3.4m, for a total reduction of 3.9m. Note that the CV deaths are based on the Framingham methodology, whereas the smoking-related deaths from other complications are based on

the SAMMEC methodology.

4.7 million lives saved

due to air quality & tobacco

interventions

Effects of Interventions on Costs of CV Events and Related Risk Factor Complications

An average of $321 per capita could be saved—and justified for intervention spending—due

to air quality and tobacco interventions

Complication and Risk Factor Management Costs per Capita

3,000

2,000

1,000

0

1990 2000 2010 2020 2030 2040

PC 3

All risk factors = 0

All19

PC 3 + AirQ 2 PC 3 + AirQ2 + Tob 4

Base

EXTRAS

Comparing Air Pollution vs. Tobacco-5 Interventions on Deaths and Costs

CVD and NonCVD deaths from RF complications per 10008

6

4

2

01990 2000 2010 2020 2030 2040

Base

Air Pollution

Tobacco-5

Complication and Risk Factor Management Costs per Capita3,000

2,000

1,000

01990 2000 2010 2020 2030 2040

Base

Air Pollution

Tobacco-5

All Risk Factors = 0

Smoking

Obesity

Secondhandsmoke

Healthinessof diet

Extent ofphysical activity

Psychosocialstress

Diagnosisand control

First-time CVevents and

deaths

Access to and marketingof smoking quit products

and services

Access to andmarketing of mental

health services

Sources ofstress

Access to andmarketing of healthy

food options

Access to andmarketing of physical

activity options

Access to andmarketing of weight

loss services

Access to andmarketing ofprimary care

Particulate airpollution

Utilization ofquality primary

care

Tobacco taxes andsales/marketing

regulations

Smoking bans atwork and public

places

Junk food taxes andsales/marketing

regulations

Downwardtrend in CV

event fatality

Quality of primarycare provision

Chronic Disorders

Anti-smokingsocial marketing

High BP

Highcholesterol

Diabetes

How Smoking is Modeled

SmokingAdults

Newly smokingadults

Quitting or dying

2040

0

0.3% Smokers

1990

• Historical estimates of current smoking prevalence among non-CVD popn from NHANES 1988-94 and 1999-2004 by sex and age group.

• Smoking prevalence in adults is modeled as a stock affected by flows of initiation and quitting, by the inflow of teen smokers turning age 18, and by deaths (related to CVD and otherwise).

• Historical estimates of Age 18 smoking fraction by sex from YRBSS.

• Baseline rates of adults quitting smoking based on Mendez & Warner AJPH 2007 and Sloan et al MIT Press 2004 (Fig. 2.1)

• Baseline rates of adult initiation/relapse adjusted to reproduce NHANES adult smoking trends by sex and age.

• Historical estimates of current smoking prevalence among non-CVD popn from NHANES 1988-94 and 1999-2004 by sex and age group.

• Smoking prevalence in adults is modeled as a stock affected by flows of initiation and quitting, by the inflow of teen smokers turning age 18, and by deaths (related to CVD and otherwise).

• Historical estimates of Age 18 smoking fraction by sex from YRBSS.

• Baseline rates of adults quitting smoking based on Mendez & Warner AJPH 2007 and Sloan et al MIT Press 2004 (Fig. 2.1)

• Baseline rates of adult initiation/relapse adjusted to reproduce NHANES adult smoking trends by sex and age.

Effects of Interventions on Youth Smoking

Smoking fraction of age 18

0.4

0.3

0.2

0.1

01990 2000 2010 2020 2030 2040

All19

PC 3

PC 3 + AirQ 2

PC 3 + Air` 2 + Tob 4

Base

Effects of Interventions on Smoking Quits

Smoking quit rate (combining all sex/age groups)

0.08

0.06

0.04

0.02

01990 2000 2010 2020 2030 2040

All19

PC 3

PC 3 + AirQ 2

PC 3 + Air` 2 + Tob 4

Base

Effects of Interventions on Use of Quit Services & Products by Smokers

Use of Quit Services & Products by Smokers

0.4

0.3

0.2

0.1

01990 2000 2010 2020 2030 2040

All19

PC 3

PC 3 + AirQ 2

PC 3 + Air` 2 + Tob 4

Base

Simulating the Local Dynamics of Cardiovascular Health and Related Risk Factors

Homer J, Milstein B, Wile K, Pratibhu P, Farris R, Orenstein D. Modeling the local dynamics of cardiovascular health: risk factors, context, and capacity. Preventing Chronic Disease 2008;5(2). Available at http://www.cdc.gov/pcd/issues/2008/apr/07_0230.htm