Projecting Future Mortality Using Information on Health Behaviors David M. Cutler, Edward L. Glaeser, and Allison B. Rosen Questions Is the US population healthier than in the past? – Yes: Smoking has declined greatly – No: We are more obese – Yes: We treat disease better Will these trends continue? Restrictions We focus on health behaviors. – Smoking – Drinking – Weight – Taking medications Actual Causes of Death in the United States, 2004 Tobacco Obesity Alcohol Microbial agents Toxic agents MVAs Guns Sexual behaviors Illicit use of drugs Source: Mokdad et al., 2004. 18% 15% 4% 3% 2% 2% 1% 1% 1% Methodology 1. Relate risk factors to subsequent mortality – NHANES I data (1971-75) linked to subsequent mortality 2. Evaluate change in risk factors, 1971-75 vs. 1999-2002 3. Consider forecasts about risk factors in the future. Data NHANES I (1971-75) and NHANES 1999-2002 Population 25-74 and 55-74 Demographics Self reported smoking, alcohol Physical measures of BP, cholesterol, BMI Rough relationships Demographics weight Behaviors Risk: BP, Cholesterol smoking Mortality Mortality Equation: 10 year mortality as a function of… Age/sex (10 year age x sex) Race (white/black/other) Education (<HS, HS, Some College, College+) Smoker (current/former/never) BMI (low, normal, overweight, obese) Alcohol (heavy, light, never) Blood pressure (normal, pre-HTN, Stage 1, Stage 2) Cholesterol (low, borderline, high) Summary Statistics: Education NHANES I NHANES 1999-2002 <HS 20% College 28% <HS 34% College 55% HS 38% HS 25% Summary Statistics: Smoking NHANES I NHANES 1999-2002 Current 25% Current Never 39% 40% Former 21% Never 49% Former 26% Summary Statistics: Drinking NHANES I Non 21% NHANES 1999-2002 Heavy 7% Heavy 4% Non 30% Light 72% Light 66% Summary Statistics: BMI NHANES I Obese 15% NHANES 1999-2002 Under weight 3% Under weight 2% Obese 33% Optimal 30% Optimal Over weight 35% 47% Over weight 35% Summary Statistics: Hypertension NHANES I Stage 2 16% Normal 22% NHANES 1999-2002 Stage 2 5% Stage 1 13% Stage 1 24% Pre-HTN 38% Pre-HTN 39% Normal 43% Summary Statistics: High Cholesterol NHANES I High 30% NHANES 1999-2002 Normal 35% High 18% Normal 48% Border line 35% Border line 34% Effect of Risk Factors on 10 Year Mortality Variable Odds Ratio Variable Odds Ratio Race: Black 1.40* BMI: Underweight 2.41* Other .25 Educ: <HS 1.27* College 1.06 BP: Pre-HTN 0.90 2.13* Stage 1 1.13 1.23 Stage 2 1.54* Smoke: Current Former Drink: Heavy Light 1.02 0.77* Overweight 0.76* Obese 1.28 Chol: Borderline High 1.03 1.15 Predicted 10 Year Mortality Risk Risk in: Ages 25-74 Ages 55-74 1971-75 9.8% 25.7% 1999-02 8.4 21.7 Change -1.4 [-14%] -3.9 [-15%] Percentage Point Change in Risk Due to… Population 25-74 1.2% 0.8% 0.4% 0.0% I M B in g rin k D on tr ol C hl C at io n Ed uc Co n P B Sm ok in g tro l -0.4% Percentage Point Change in Risk Due to… Population 55-74 2.4% 1.6% 0.8% 0.0% I M B in g rin k D on tr ol C hl C at io n ok in g Sm Ed uc B P Co n tro l -0.8% Forecasts Simulate 20 years from now Not totally clear what explains these behaviors. – Smoking: Taxes (a bit); Beliefs; Peer effects – Obesity: Lower (time) price of food – Assume these are still playing out Forecasts Smoking – Know ever smoking for many cohorts (guess for others) – Assume trend reduction in current smoking continues – Smoking rate falls from 25% to 15% Per Capita Cigarette Consumption, US 1900-2000 5,000 4,000 3,000 2,000 1,000 0 1900 1910 1920 1930 1940 1950 1960 1970 1980 1990 2000 Forecast Assumptions Drinking – Continued reduction in heavy and light drinking BMI – Same change in BMI over past 20 years as previous 20 years – Increase of about 10 lbs. Implications of higher BMI: BP and Cholesterol Use 1959-62 NHANES to relate BP and cholesterol to obesity – Essentially no treatment Predict BP and cholesterol 20 years hence – Includes random error term Assume same share of people treated and same impact of treatment – Draw BP and cholesterol from distribution among treated. Control of Adverse Risk Factors, 1999-2000 100% 80% 60% 40% 20% 0% Hypertension Aware Treated High Cholesterol Controlled Impact of Changes on Population Mortality Rate Smoking forecast %age points -0.7% Percent -8% Drinking forecast 0.1% 1% BMI forecast 1.4% 17% - If everyone takes meds 1.0% 12% - More effective meds / greater adherence 0.2% 2% Impact of 10% reduction in mortality on expected age at death for people alive at age 95 0.5 years 0.9 years 90 1.2 years 1.1 years 85 80 75 25 30 35 40 45 50 55 Baseline 60 65 Alternate 70 75 80 85 Results Impact of continued increase in obesity could be enormous. Key is risk factor control – Get more people treated – Improve effectiveness of therapy – better drugs, and taken more regularly Qualifications Impact on mortality is not necessarily the same as impact on disability – Important for DI and Medicare/Medicaid forecasts