11 Linda S. Gottfredson, Professor University of Delaware August 7, 2009 Presentation to accept 2008 George A. Miller Award for outstanding article across specialty areas, Division 1, APA Social Class Disparities in Health: A Vexing Puzzle with a Surprising Answer? American Psychological Association

2 2 1.What are “disparities”? 2.What’s the vexing puzzle? 3.Is human cognitive diversity key to solving it? 4.If yes, so what? Agenda Answers: All surprising

33 1.What are “disparities”? 2.Why such a vexing puzzle? 3.Is human cognitive diversity the key to solving it? 4.If yes, so what? Agenda Examples

44 “Disparity” = group differences on health outcome X “Explaining” between-group variation Means, rates, etc. 16 yrs 12 yrs 8 yrs Typical indicators of socioeconomic status (SES) Years education Occupational status Income But not clear what they really represent or have in common ?

5 Typical health disparities by education; in all races & sexes: % of non-ill 51-year-olds expected to have this chronic illness by age 63 (Hayward et al, 2000) Hypertension Diabetes white COPD black Cancer Men Women % Years

6 Typical health disparities by education; in all races & sexes: % of non-ill 51-year-olds expected to have this chronic illness by age 63 (Hayward et al, 2000) Hypertension Diabetes white COPD black Cancer Men Women % Fewer health problems in higher social classes (educ, occup, or $) True for all races, sexes Exceptions are rare (e.g., cancer morbidity) Years

7 Disparities in health behavior by education; all races & sexes: % who smoke, 2006 (age adjusted) (CDC, Health in the United States, 2008, Table 64) %

8 Typical course of behavior disparities over time, by education: % who smoke, , ages 25+ (age-adjusted) (CDC, Health in the United States, 2008, Table 64) % better, gap bigger %

99 Many families of health disparities HEALTH HABITS MORTALITY KNOWLEDGE CHRONIC ILLNESSES INJURIES INFECTIOUS DISEASES ADHERENCE

10 Many families of health disparities HEALTH HABITS MORTALITY KNOWLEDGE CHRONIC ILLNESSES INJURIES INFECTIOUS DISEASES ADHERENCE Outcomes for populations

11 a b c d d f g This is not about individual differences in health outcomes Not “explaining” within-group variation Within-group and between-group variance may arise from different mix of causes Often misunderstood!

12 Study of populations aided by epidemiological approach Outcomes oMeans, rates, relative risk, odds ratios for groups Predictors—classic trio oExposure to hazards, help (probability) oHost (susceptibility) oVector (virulence, burden)

13 Study of populations aided by epidemiological approach Outcomes oMeans, rates, relative risk, odds ratios for groups Predictors—classic trio oExposure (probability) oHost (susceptibility) oVector (virulence, burden) Missing 2/3 Current focus of SES disparities research

14 1.What are “disparities”? 2.Why such a vexing puzzle? But first, what exactly are we trying to explain?  Statistically Substantively 3.Is human cognitive diversity the key to solving it? 4.If yes, so what? Agenda Illustration

15 Illustration with 2 disparities # 1 # 2

Social class groupings Health (group mean or rate) Each disparity is a gradient, with a slope (ß) Statistically… # 1 ß1ß1

Social class groupings Health (group mean or rate) Each disparity is a gradient, with a slope Statistically… # 2 # 1 ß1ß1

Social class groupings Health (group mean or rate) Each disparity is a gradient, with a slope Statistically… # 2 # 1 ß1ß1 ß2ß2

Social class groupings Health (group mean or rate) Many families of health gradients (slopes): Morbidity, mortality, knowledge, prevention, adherence, etc. rare ß1ß1 ß8ß8 ß7ß7 -ß 6 ß5ß5 ß4ß4 ß3ß3 ß2ß2

20 So, to explain SES disparities: Explain the distribution of co-evolving gradients (ß, their standardized slopes) ß ß 3 ß 8 ßßßßß4ßßßßß4 ßßß2ßßß2 ß6ß6 ß ßßßß ßßßßß7ßßßßß7 ßßßßß2ßßßßß2 ßß1ßß1 ß 0 Slopes (steepness) of gradients negativepositive Common policy goal : All β = 0

21 1.What are “disparities”? 2.Why such a vexing puzzle? But first, what exactly are we trying to explain? Statistically  Substantively 3.Is human cognitive diversity the key to solving it? 4.If yes, so what? Agenda Examples

22 General puzzle: Health disparities are too general for SES mechanisms to explain They are pervasive, persistent and monotonic regardless of time, place, health system, disease, and behavior. Why??

23 Exposure hypothesis 1: “Wealth = health” (can afford good care) health wealth No leveling off when resources are more than sufficient REJECTED— Puzzle greater!

24 Experimental test of exposure hypothesis 1: Equalize access to care equalize health Time 1: Unequal access Time 2: After equal access (free care) Health disparities grow, not shrink FAILED—Puzzle greater! E.g., UK in 1950s

25 Experimental test of exposure hypothesis 2: Unequal education unequal health Time 1: Unequal knowledge of signs and symptoms Time 2: After public health campaign Knowledge disparities grow, not shrink FAILED—Puzzle greater!

26 Or disparities even reverse direction with new screening tests (e.g., death rates from breast cancer) Negative disparities for Outcome X at Time 1 Positive disparities for Outcome X at Time 2 ß -ß More educated women have higher death rates

27 Access matters, but so does utilization Even if equal access Unequal use & misuse Mammograms Adherence to treatment Seat belt use Etc. “Health literacy”

Social class groupings Health (group mean or rate) Summary of puzzle rare Exposure can’t explain why gradients: Virtually never = zero Virtually always positive All monotonic (~linear) For ~all health outcomes & behaviors Steepen when resources equalized What levers the gradients up or down? Can’t be material resources.

29 So, the field seeking more “fundamental cause” of SES disparities This cause must: obe pervasive & domain-general ohave linear (monotonic) effects onot be material Most popular suspect = inequality itself orelative deprivation chronic psychological stress damaging physiological process: “allostatic load” Stress important, but can’t explain: owhy adding resources increases disparities odisparities in non-biological outcomes

Biological mechanisms Involved here But not here First, physical illness is only one cause of injury & death: Causes of death, males by age (CDC, Health data interactive) Common theme—all are preventable

31 Example: Unintentional (“accidental”) death Odds ratios by neighborhood income ( ) 20 per 100, Reference group Odds = % affected Odds ratio = Odds for Group 1_______ % not Odds for reference group Just differential exposure??

32 Selected causes of “motor vehicle traffic” death, by neighborhood income/capita ( ) (Baker, O’Neill, Ginsburg, & Li, 1992) elderly adult men young men toddlers young men Primarily: Rate per 100,000

33 Selected “other” causes of unintentional death, by neighborhood income/capita ( ) (Baker, O’Neill, Ginsburg, & Li, 1992) Odds ratio Deaths per 100, infants, elderly rises with age young men toddlers, elderly young men infants, elderly Infants Primarily: Self-exposure Differential biological susceptivity

34 Prevention oIt’s our job oIt’s daily, unrelenting, life-long (hazards are everywhere) oIt’s complex It’s a highly cognitive, multi-step, active process oSpot & avoid hazards oRecognize signs of system veering out of control oTake action to regain control oLimit progression of illness/accident or damage it does oAdhere to treatment oLearn from experience to adjust future behavior The common mechanism for illness and injury? Passive-patient model is dead wrong

35 1.What are “disparities”? 2.Why such a vexing puzzle? 3.Is human cognitive diversity the key? 4.If yes, so what? Agenda IQ/g

36 Alternative hypothesis for disparities in health: “Intelligence (g) differences are the “fundamental cause” Two g–based levers ratchet up gradients * Bigger IQ differences (people) Heavier cognitive load (tasks) susceptibility burden * Based on extensive research in education & employment

37 Gaps in IQ/g (cognitive susceptibility-efficiency) Heavier cognitive load (g loading of tasks) Heavier cognitive load (g loading of tasks) ß ß ß ß ß ß ß ß ß ß ß ß ß Translated: A hypothesis about gradients

38 Background fact #1 Great cognitive diversity is a biological fact about all populations IQ

39 Background fact #2 IQ ≈ g (general mental ability factor) g is no longer a black box g is a domain-general facility for learning, reasoning, spotting & solving novel problems oHigher g reduces susceptibility to error oGives bigger edge as task complexity (cognitive load) increases oAllows one to exploit resources more fully & effectively (e.g., classroom instruction, medical treatments)

40 Background fact #2 IQ ≈ g (general mental ability factor) g is no longer a black box g is a domain-general facility for learning, reasoning, spotting & solving novel problems oHigher g reduces susceptibility to error oGives bigger edge as task complexity (cognitive load) increases oAllows one to exploit resources more fully & effectively (e.g., classroom instruction, medical treatments)

41 Gives an edge in planning; anticipating problems

42 Background fact #3 Mean IQs differ by occupation level and years education Unskilled Semiskilled Skilled Manager, Cler, Sales Professional & Tech Occupation: Years education: WAIS-R IQ (mean + 1 SD), US adults ages IQ

43 Background fact #4: Some SES indicators correlate more with IQ.8 Standardized academic achievement.6 Years education.5 Occupation level.3-.4 Income (prior) IQ All moderately heritable, & overlap genetically with IQ

44.8 Literacy.8 Standardized academic achievement.6 Years education.5 Occupation level.3-.4 Income (prior) IQ Excellent Good Weak Background fact #4: Conversely, some are better surrogates for IQ Better surrogates for g show larger health disparities

Better surrogates for g show larger health disparities (steeper gradients) 45 income occupation education “literacy”

46.8 Literacy.8 Standardized academic achievement.6 Years education.5 Occupation level.3-.4 Income Excellent Good Weak Background fact #4: Conversely, some are better surrogates for IQ (prior) IQ Cannot “control” for SES without controlling away much of (genetic) g itself

47 Gaps small when learning & reasoning demands are light Gaps large when learning & reasoning demands are heavy Common in schools & jobs Background fact #5: Task complexity increases gaps in performance

48 Gaps small when learning & reasoning demands are light Gaps large when learning & reasoning demands are heavy Common in schools & jobs Background fact #5: Task complexity increases gaps in performance Cognitive load brings out differences in cognitive susceptibility

49 New technologies make life increasingly complex, which puts yet higher premium on g Preventive & curative care becoming increasing complex

50 Background fact #6: People differ more than often assumed U.S. Dept of Education 1993 survey of adult functional literacy (nationally representative sample, ages 16+, N=26,091) NALS Level % pop. Simulated Everyday Tasks 5 3% Use calculator to determine cost of carpet for a room Use table of information to compare 2 credit cards 4 17% Use eligibility pamphlet to calculate SSI benefits Explain difference between 2 types of employee benefits 3 31% Calculate miles per gallon from mileage record chart Write brief letter explaining error on credit card bill 2 27% Determine difference in price between 2 show tickets Locate intersection on street map 1 22% Total bank deposit entry Locate expiration date on driver’s license Routinely able to perform tasks only up to this level of difficulty

51 NALS Level % pop. Simulated Everyday Tasks 5 3% Use calculator to determine cost of carpet for a room Use table of information to compare 2 credit cards 4 17% Use eligibility pamphlet to calculate SSI benefits Explain difference between 2 types of employee benefits 3 31% Calculate miles per gallon from mileage record chart Write brief letter explaining error on credit card bill 2 27% Determine difference in price between 2 show tickets Locate intersection on street map 1 22% Total bank deposit entry Locate expiration date on driver’s license Difficulty based on “process complexity” Difficulty based on “process complexity”  level of inference  abstractness of info  distracting information Not reading per se, but “problem solving” Background fact #6: People differ more than often assumed U.S. Dept of Education 1993 survey of adult functional literacy (nationally representative sample, ages 16+, N=26,091) Cognitive load brings out cognitive susceptibilities

52 Item at NALS Level 1 Literal match One item Little distracting info 22% of US adults 78% of adults do better 80% probability of correctly answering items of this difficulty level * *

53 Item at NALS Level 2 X Simple inference Simple inference Little distracting information Little distracting information 27% of US adults 51%22%

54 Another item at NALS Level 2 27% of US adults Match two pieces of info 51%22%

55 Item at NALS Level 3 31% of US adults Cycle through complex table Irrelevant info 20%49%

56 Item at NALS Level 4 More elements to match More elements to match More inferences More inferences More distracting information More distracting information 3%80% 17% of US adults Solved Or,

57 Item at NALS Level 5 97% Search through complex displays Multiple distractors Make high-level text-based inferences Use specialized knowledge 3% of US adults

58 NALS Level % pop. Simulated Everyday Tasks 5 3% Use calculator to determine cost of carpet for a room Use table of information to compare 2 credit cards 4 17% Use eligibility pamphlet to calculate SSI benefits Explain difference between 2 types of employee benefits 3 31% Calculate miles per gallon from mileage record chart Write brief letter explaining error on credit card bill 2 27% Determine difference in price between 2 show tickets Locate intersection on street map 1 22% Total bank deposit entry Locate expiration date on driver’s license US Dept of Education: People at levels 1-2 are below literacy level required to enjoy rights & fulfill responsibilities of citizenship Could teach these individual items, but not all such tasks in daily life Background fact #6: People differ more than often assumed U.S. Dept of Education 1993 survey of adult functional literacy (nationally representative sample, ages 16+, N=26,091)

59 1.What are “disparities”? 2.Why such a vexing puzzle? 3.Is human cognitive diversity the key to solving it? 4.If yes, so what?  Mine the other 2/3 (cognitive susceptibility & cognitive load) Agenda

Passive exposure matters SES differences predicted Current SES stress model Alternative g stress model PredictorsTime 1Time 2Time 1Time 2 ExposurePassiveEp++ ActiveEa SusceptibilityBiologicalSb0+ CognitiveSc BurdenBiologicalBb CognitiveBc Health outcomesPhysiologicalYp0+ BehavioralYb mechanismY = ∑Ep 60

But so does g-based self-exposure, susceptibility, & cognitive load SES differences predicted Current SES stress model Alternative g stress model PredictorsTime 1Time 2Time 1Time 2 ExposurePassiveEp++++ ActiveEa++ SusceptibilityBiologicalSb0+?+ CognitiveSc++ BurdenBiologicalBb?? CognitiveBc?+ Health outcomesPhysiologicalYp0+?++ BehavioralYb+ + mechanismY = ∑Ep Y = ∑E(S)(B) 61

SES differences predicted Current SES stress model Alternative g stress model PredictorsTime 1Time 2Time 1Time 2 ExposurePassiveEp ActiveEa SusceptibilityBiologicalSb CognitiveSc BurdenBiologicalBb CognitiveBc Health outcomesPhysiologicalYp BehavioralYb mechanismY = ∑Ep Y = ∑E(S)(B) 62 Internal External Some are multiplicative 6 (not 1) generators of health disparities, and multiplicative besides

2 new points of leverage SES differences predicted Current SES stress model Alternative g stress model PredictorsTime 1Time 2Time 1Time 2 ExposurePassiveEp ActiveEa SusceptibilityBiologicalSb CognitiveSc BurdenBiologicalBb CognitiveBc Health outcomesPhysiologicalYp BehavioralYb mechanismY = ∑Ep Y = ∑E(S)(B) 63 Internal External #1 #2 Respect diversity of needs Lighten the load

64 Need appreciate differential cognitive needs #1

65 Need appreciate size of cognitive burdens Example: Do job analysis of chronic diseases Diabetes? #2

66 Guidance for providers? E.g., Matrices of cognitive risk IQ Lo Hi Lo Hi LoHi Some errors more dangerous But all cumulate Triage Task complexity Error rates to expect by patient susceptibility task cognitive load #1 #2 #1 #2

67 Conclusions Key mechanisms unrecognized Mechanisms highly exploitable Huge opportunity costs o For national policy o For clinic practice o For vulnerable populations American Psychological Association

68 Thank You Linda S. Gottfredson, Professor University of Delaware (302) American Psychological Association