risk factors, psychology, and communication · – fostering transparent risk communication •...
TRANSCRIPT
Hansjörg Neth Risk Management in Pharmacotherapy ISoP 2015, Prague | Oct. 30 2015
Risk factors, psychology, and communication
Abstract (ISoP 2015) ISOP-0055: AS01 - Risk Management in Pharmacotherapy
Risk factors, psychology and communication Hansjörg Neth Social Psychology and Decision Sciences Department of Psychology, University of Konstanz, Germany
• Statistical illiteracy in health––the inability to understand health statistics––is widespread among the general public and among medical experts. For many people, it is generally hard to accept uncertainty, and even if they do, to understand basic numerical information. The problem is aggravated when it comes to evaluating the benefits and harms of treatment options or to understanding test outcomes, which is a severe obstacle to an informed risk management strategy.
• Statistical illiteracy reflects not just a lack of education but often results from non-transparent framing of information that may be unintentional, but can also be a deliberate effort to manipulate people. Non-transparent framing of information seems to be the rule rather than the exception in health care: Patients have difficulties finding reliable and comprehensible information, be it online, in brochures on screening procedures, medical pamphlets, or media reports.
• Yet all these obstacles do not imply that nothing can be done. The most important mean of improvement consists in teaching the public statistical thinking, combined with training health care workers and journalists in transparent framing. Knowing what questions to ask, which information is missing, and how to translate non-transparent statistics into transparent ones can enable informed risk management strategies. More generally, a better understanding of risks will allow citizens to develop a more relaxed attitude towards health and render the hopes and anxieties of an informed society less manipulable.
Ideal vs. real health care
Desiderata: Status quo: evidence-based medicine eminence-based medicine
informed and shared decision-making
low numeracy, lack of statistical and graphical literacy
transparency in information and communication
biased reporting with persuasive intent
informed risk management: coping with risks
misunderstanding risks, aiming to avoid uncertainty
(Gigerenzer & Gray, 2011; Bodemer & Gaissmaier, 2012)
Goals of risk communication
Persuasion: Education/Information:
(Bodemer & Gaissmaier, 2012, Fig. 24.1, p. 626 ) . Fig. 24.1
Two different ways to inform women about mammography screening. The flyer on the left side by the American Cancer Society (Retrieved from www.
nysut.org/files/makingstrikes_070921_poster.pdf in April 2011) encourages women to participate in regular mammography screening without
providing information about benefits and harms of the screening program. It states that ‘‘mammograms save lives – there’s no doubt about it (. . .) Hope
for a cancer-free future starts with you.’’ The facts box on right side (Retrieved from www. http://www.harding-center.com/fact-boxes/
mammography-screening in April 2011) summarizes the most important results based on the current scientific evidence and informs rather than
persuades. It contrasts 2,000women aged 40 and olderwhoparticipate inmammography screening over 10 yearswith 2,000 of the same agewhodonot.
Besides the benefits of the screening program, the facts box also includes information about potential harms like overtreatment
626
4 sins of risk communication…
Relative risks Survival rates
Absolute risks Mortality rates
Single event probabilities Conditional probabilities
Reference class Natural frequencies
Official announcement: “Contraceptive pills double the risk of venous thromboembolism!”
(UK Committee on the Safety of Medicines, 1995)
“Contraceptive pills increase the risk of venous thromboembolism by 100%.”
Transparent format: “Contraceptive pills increase the risk of venous thromboembolism from 1 to 2 women out of every 7,000 women.”
160,000
170,000
180,000
190,000
1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000
Abt
reib
unge
n in
Eng
land
und
Wal
es
Jahr
"Antibaby-Pille erhöht Risiko einer Thromboembolie
um 100%"
Gigerenzer, Gaissmaier, Kurz-Milcke, Schwartz & Woloshin (2007). Psychological Science in the Public Interest
Num
ber o
f abo
rtio
ns in
Eng
land
and
Wal
es
“… double the risk”
Year
4 sins of risk communication
Relative risks Survival rates
Absolute risks Mortality rates
Single event probabilities Conditional probabilities
Reference classes Natural frequencies
Misleading statistics & anecdotes Diagnostic certainty
Fact boxes & icon arrays Managing uncertainty
“I had prostate cancer, five, six years ago. My chances of surviving prostate cancer—and thank God I was cured of it— in the United States, 82 percent. My chances of surviving prostate cancer in England, only 44 percent under socialized medicine.”
Rudy Giuliani New Hampshire Radio advertisement, October 29, 2007
Survival vs. mortality rates
Gigerenzer et al. (2007); Bodemer & Gaissmaier (2012)
Five-year survival rate ‘‘The 5-year survival rate for people diagnosed with prostate cancer is 98% in the USA vs. 71% in Britain.’’ Annual mortality rate ‘‘There are 26 prostate cancer deaths per 100,000 American men vs. 27 per 100,000 men in Britain.’’
N diagnosed with X & alive 5 years later 5-year survival rate =
N diagnosed with X in study population
N who die from X over 1 year annual mortality rate =
N in study population
Misleading survival rates
Gigerenzer et al. (2007); Bodemer & Gaissmaier (2012)
WITHOUT SCREENING
WITH SCREENING
Cancer starts
Cancer starts
Cancer diagnosed because ofsymptoms at age 67
Dead at age 70
Dead at age 70
Cancer diagnosed becauseof screening at age 60
5-year survial = 0 %
5-year survial = 100 %
1,000 patientswith
progressivetumors
1,000 patientswith non-
progressivetumors
500 dead
500 dead
1,500 alive
500 alive
1,000 patientswith
progressivetumors
5 years later
5 years later
5 - years survival = 500 = 50 %1000
5 - years survival = 1500 = 75 %2000
a
b
. Fig. 24.2
Shortcomings of 5-year survival rates: The figure illustrates the two potential biases of 5-year
survival rates (modified from Gigerenzer et al. 2007). (a) Lead-time bias: The arrows illustrate the
course from the beginning of a disease to death. In the group without screening, cancer is
diagnosed at age 67, in the screening group at age 60. However, in both groups, patients die at
the same age (age 70). Whereas in the non-screening group the 5-year survival rate is 0%, it is
100% in the screening group. (b) Overdiagnosis bias: (1) A group of 1,000 patients with
progressive tumors is monitored over 5 years. After 5 years, 500 are still alive; the survival rate is
50%. (2) The same group of 1,000 patients with progressive tumors is monitored over 5 years.
Additionally, the screening detects patients with nonprogressive, indolent tumors. Again, after
5 years 500 patients died (500 out of 1,000 with progressive cancer). However, in the calculation
of the 5-year survival rate, those 1,000 with nonprogressive tumors are also included hence the
5-year survival rate is 75%
643 1) lead-time bias
Gigerenzer et al. (2007); Bodemer & Gaissmaier (2012)
WITHOUT SCREENING
WITH SCREENING
Cancer starts
Cancer starts
Cancer diagnosed because ofsymptoms at age 67
Dead at age 70
Dead at age 70
Cancer diagnosed becauseof screening at age 60
5-year survial = 0 %
5-year survial = 100 %
1,000 patientswith
progressivetumors
1,000 patientswith non-
progressivetumors
500 dead
500 dead
1,500 alive
500 alive
1,000 patientswith
progressivetumors
5 years later
5 years later
5 - years survival = 500 = 50 %1000
5 - years survival = 1500 = 75 %2000
a
b
. Fig. 24.2
Shortcomings of 5-year survival rates: The figure illustrates the two potential biases of 5-year
survival rates (modified from Gigerenzer et al. 2007). (a) Lead-time bias: The arrows illustrate the
course from the beginning of a disease to death. In the group without screening, cancer is
diagnosed at age 67, in the screening group at age 60. However, in both groups, patients die at
the same age (age 70). Whereas in the non-screening group the 5-year survival rate is 0%, it is
100% in the screening group. (b) Overdiagnosis bias: (1) A group of 1,000 patients with
progressive tumors is monitored over 5 years. After 5 years, 500 are still alive; the survival rate is
50%. (2) The same group of 1,000 patients with progressive tumors is monitored over 5 years.
Additionally, the screening detects patients with nonprogressive, indolent tumors. Again, after
5 years 500 patients died (500 out of 1,000 with progressive cancer). However, in the calculation
of the 5-year survival rate, those 1,000 with nonprogressive tumors are also included hence the
5-year survival rate is 75%
643
2) overdiagnosis bias
Misleading survival rates
4 sins of risk communication
Relative risks Survival rates
Absolute risks Mortality rates
Single event probabilities Conditional probabilities
Reference classes Natural frequencies
Misleading statistics & anecdotes Diagnostic certainty
Fact boxes & icon arrays Managing uncertainty
Single-event probabilities require reference classes
Gigerenzer et al. (2007); Bodemer & Gaissmaier (2012)
Example: Side effects of taking Prozac Single-event probability: ‘‘The probability that you will experience sexual problems is 30–50% (or: 3 to 5 chances out of 10).’’ Frequency statement with reference class: ‘‘Out of every 10 of my patients, 3–5 experience a sexual problem.’’
4 sins of risk communication
Relative risks Survival rates
Absolute risks Mortality rates
Single event probabilities Conditional probabilities
Reference classes Natural frequencies
Misleading statistics & anecdotes Diagnostic certainty
Fact boxes & icon arrays Managing uncertainty
~10%
10 cancer
990 no cancer
9 positive
89 positive
1 negative
901 negative
1000 women
Natural frequencies:
p( cancer | positive )
= 9 9 + 89
Conditional probabilities:
.01 x .90 .01 x .90 + .99 x .09
p( cancer | positive )
=
Prevalence: p( breast cancer ) = 1%
Sensitivity:
p( positive | cancer ) = 90% False alarm rate:
p( positive | no cancer ) = 9%
Probability ( breast cancer | positive mammogram )?
4 sins of risk communication
Relative risks Survival rates
Absolute risks Mortality rates
Single event probabilities Conditional probabilities
Reference classes Natural frequencies
Misleading statistics & anecdotes Diagnostic certainty
Fact boxes & icon arrays Managing uncertainty
… vs. transparent representations
Relative risks Survival rates
Absolute risks Mortality rates
Single event probabilities Conditional probabilities
Reference classes Natural frequencies
Misleading statistics & anecdotes Diagnostic certainty
Fact boxes & icon arrays Managing uncertainty
General remedies?
mind environment
(Gigerenzer et al., 1999; Gigerenzer & Gaissmaier, 2011; Todd et al., 2014; Neth & Gigerenzer, 2015)
match?
Designing for ecological rationality:
!"#$%&'($)*#"'+$",-'.#&#*/0)'!"#$%$$&'(%)*"#+,(--./.'##01$!-(+#2&(#3&$-.#%'-4#56#"-%(+#&(#&74-(#3*&#)%(8,/)%9-4#/.#+,(--./.'#2&(#:6#"-%(+#&(#$&(-#
##!#)#1&%'
23444'506#)'57&809&'%*"##)7):'
23444'506#)'57&8'
%*"##)7):';&3#$%."#3&$-.#4/-4#2(&$#!(-%+9#,%.,-(<# 5# =#;&3#$%."#3&$-.#4/-4#2(&$#%77#9")-+#&2#,%.,-(<## >:# >:#
;$"6%';&3#$%."#3&$-.#3/9*&19#,%.,-(#-?)-(/-.,-4#2%7+-#%7%($+#&(#!/&)+/-+<##
@# :66#
;&3#$%."#*-%79*"#3&$-.#3-(-#4/%'.&+-4#%.4#9(-%9-4#2&(#!(-%+9#,%.,-(#1..-,-++%(/7"<##
@# 5#
A&1(,-B#CD9E+,*-F#GHF#ID('-.+-.F#JI#K>6:LMN#!"#$%&'()*&+&,&-()".)/0-+(1&2#)3(45(6-#KOMB#HP66:QRRN##01$!-(+#/.#9*-#2%,9+#!&?#%(-#(&1.4-4N#S*-(-#.%9%#2&(#3&$-.#%!&T-#56#"-%(+#&2#%'-#%(-#%T%/7%!7-F#.1$!-(+#(-2-(#9&$-.#%!&T-#=6#"-%(+#&2#%'-N#333N*%(4/.'U,-.9-(N$)'N4-####
!"#$%& ($)*#"'+$",-'.#&#*/0)!" $%$$&'(%)*" +,(--./.'+,(--./.'##01$!-(+#2&( 3&$-.3&$-. %'-4#56#"-%(+ &( &74-( 3*& )%(8,/)%9-4#/.#+,(--./.'+,(--./.' 2&(#:6#"-%(+&( $&(-
!#)#1&%
23444'506#)506#)57&809&%*"##)7):%*"##)7):
23444'506#)57&8
%*"##)7):;&3 $%." 3&$-.3&$-. 4/-4 2(&$ !(-%+9 ,%.,-(< 5# =#;&3 $%." 3&$-.3&$-. 4/-4 2(&$#%77#9")-+ &2 ,%.,-(<# >:# >:#
;$"6%';&3 $%." 3&$-.3&$-. 3/9*&19 ,%.,-( -?)-(/-.,-4 2%7+-%7%($+ &( !/&)+/-+<#
@# :66#
;&3 $%." *-%79*"*-%79*" 3&$-. 3-(- 4/%'.&+-4 %.49(-%9-4 2&( !(-%+9 ,%.,-( 1..-,-++%(/7"<#
@# 5#
A&1(,-B#CD9E+,*-F#GHF#ID('-.+-.ID('-.+-.F#JI#K>6:LMN#F#JI#K>6:LMN#!"#$%&'()*&+&,&-()".)/0-+(1&2#)3(45(6-!"#$%&'()*&+&,&-()".)/0-+(1&2#)3(45(6-#KOMB#HP66:QRRN###KOMB#HP66:QRRN##01$!-(+#/.#9*-#2%,9+#!&?#%(-#(&1.4-4N#S*-(-#.%9%#2&(#3&$-.#%!&T-#56#"-%(+#&2#%'-#%(-#%T%/7%!7-F#.1$!-(+#(-2-(#9&$-.#%!&T-#=6#"-%(+#&2#%'-N#333N*%(4/.'U,-.9-(N$)'N4-
Remedies: Fact boxes
https://www.harding-center.mpg.de/en/health-information/fact-boxes/mammography
Remedies: Icon arrays !"#$%&'($)*#"'+$",-'.#&#*/0)'!"#$%$$&'(%)*"#+,(--./.'##01$!-(+#2&(#3&$-.#%'-4#56#"-%(+#&(#&74-(#3*&#)%(8,/)%9-4#/.#+,(--./.'#2&(#:6#"-%(+#&(#$&(-#
1222'304#)'35&607&'%*"##)5)89'' 1222'304#)'35&6'%*"##)5)89''
B
;&$-.#3*/-4#2(&$#!(-%+9#,%.,-(<##
;&$-.#3*/-4#2(&$#%77#9")-+#&2#,%.,-(<##
;&$-.#3*-%(.-4#%=-(#%#!/&)+"#9*%9#9*-/(#4/%'.&+/+#3%+#%#2%7+->)&+/8?-<##
;&$-.#3*-(-#4/%'.&+-4#%.4#9(-%9-4#2&(#!(-%+9#,%.,-(#1..-,-++%(/7"<##
@-$%/./.'#3&$-.<###
5##
A:##
#B##
#B##
CDC#
E##
A:##
#:66#
#
#5##
FDE#
:07"*#9''GH9I+,*-J#KLJ#MH('-.+-.J#NM#OA6:PQR#!"#$%&'()*&+&,&-()".)/0-+(1&2#)3(45(6-#OSQ<#LT66:FDD#01$!-(+#/.#9*-#2%,9+#!&U#%(-#(&1.4-4R#;*-(-#.%9%#2&(#3&$-.#%!&?-#56#"-%(+#&2#%'-#%(-#%?%/7%!7-J#.1$!-(+#(-2-(#9&$-.#%!&?-#E6#"-%(+#&2#%'-R##333R*%(4/.'>,-.9-(R$)'R4-#
B B B B
!"#$%& ($)*#"'+$",-'.#&#*/0)!" $%$$&'(%)*" +,(--./.'##01$!-(+#2&( 3&$-. %'-4#56#"-%(+ &( &74-( 3*& )%(8,/)%9-4#/.#+,(--./.' 2&(#:6#"-%(+ &( $&(-
1222'304#)'35&607&'%*"##)5)89'' 1222'304#)'35&6'%*"##)5)89''
B B
;&$-.#3*/-4#2(&$#!(-%+9#,%.,-(<#;&$-.#3*/-4#2(&$#%77#9")-+#&2#,%.,-(<#
;&$-.#3*-%(.-4#%=-(#%#!/&)+"#9*%9#9*-/(#4/%'.&+/+#3%+#%#2%7+->)&+/8?-<#
;&$-.#3*-(-#4/%'.&+-4#%.4#9(-%9-4#2&(#!(-%+9#,%.,-(#1..-,-++%(/7"<#@-$%/./.'#3&$-.<###
5#A:#
B#
B#CDC#
E#A:#
:66#
5#FDE#
:07"*#9''GH9I+,*-J#KLJ#MH('-.+-.J#NM#OA6:PQR#!"#$%&'()*&+&,&-()".)/0-+(1&2#)3(45(6-#OSQ<#LT66:FDD#01$!-(+#/.#9*-#2%,9+#!&U#%(-#(&1.4-4R#;*-(-#.%9%#2&(#3&$-.#%!&?-#56#"-%(+#&2#%'-#%(-#%?%/7%!7-J#.1$!-(+#(-2-(#9&$-.#%!&?-#E6#"-%(+#&2#%'-R##333R*%(4/.'>,-.9-(R$)'R4-#
B B B B B B B B B B B B
https://www.harding-center.mpg.de/en/health-information/fact-boxes/mammography
Conclusion & challenges
• Risk can be managed:
– teaching statistical literacy – fostering transparent risk communication
• Uncertainty is inevitable: – communicating uncertainty (e.g., of diagnostic tests) – avoiding defensive decision making
You step outside, you risk your life. You take a drink of water, you risk your life. And nowadays you breathe, and you risk your life. Every moment you don‘t have a choice. The only thing you can choose is what you‘re risking it for.
Hershel, The Walking Dead
Thanks for your attention! Comments, panel discussion, feedback... Dr. Hansjörg Neth Social Psychology and Decision Sciences Tel.: +49 (0) 75 31/88 - 2972 Fax: +49 (0) 75 31/88 - 2899 [email protected] http://www.spds.uni-konstanz.de
Key references Risk perception and risk communication
• Bodemer, N., & Gaissmaier, W. (2015). Risk perception. In H. Cho, T. Reimer, & K. A. McComas (Eds.). The Sage Handbook of Risk Communication (pp. 10–23). Thousand Oaks, CA: Sage.
• Gigerenzer, G., Gaissmaier, W., Kurz-Milcke, E., Schwartz, L. M., & Woloshin, S. (2007). Helping doctors and patients make sense of health statistics. Psychological Science in the Public Interest, 8, 53–96.
Heuristic decision making
• Gigerenzer, G., & Gaissmaier, W. (2011). Heuristic decision making. Annual Review of Psychology, 62, 451–482.
• Neth, H., & Gigerenzer, G. (2015). Heuristics: Tools for an uncertain world. In R. Scott & S. Kosslyn (Eds.), Emerging trends in the social and behavioral sciences: An interdisciplinary, searchable, and linkable resource (pp. 1–18). New York, NY: Wiley Online Library.
Available at http://www.spds.uni-konstanz.de/publications/
Examples
Bodemer & Gaissmaier (2012), p. 655
lessons learned in health risk communication can be adapted to other domains as well.
Transparency and statistical literacy help people evaluate financial, environmental, and tech-
nological risks, and enable society to competently meet future challenges.
References
Allen M, Preiss R (1997) Comparing the persuasiveness
of narrative and statistical evidence using meta-anal-
ysis. Commun Res Rep 14:125–131
Ancker JS, Kaufman D (2007) Rethinking health numer-
acy: a multidisciplinary literature review. J Am Med
Inform Assoc 14:713–721
Ancker JS, Senathirajah Y, Kukafka R, Starren JB
(2006) Design features of graphs in health risk com-
munication: a systematic review. J Am Med Inform
Assoc 3:608–618
Baesler JE (1997) Persuasive effects of story and statistical
evidence. Argument Advocacy 33:170–175
. Table 24.3
Nontransparent versus transparent communication of risks: Four examples of how risks can be
communicated to mislead and misinform the public and their transparent counterparts
How to communicate risks nontransparently How to communicate risks transparently
Relative risks‘‘The new generation of the contraceptive pillincreases the risk of thrombosis by 100%.’’
Absolute risks‘‘The new generation of the contraceptive pillincreases the risk of thrombosis from 1 in 7,000to 2 in 7,000.’’
Conditional probabilities– The probability of breast cancer is 1% fora woman at age 40 who participates in routinescreening (this is the prevalence or base rate)– If a woman has breast cancer, the probability is90% that she will get a positive mammography(this is the sensitivity or hit rate)– If a woman does not have breast cancer, theprobability is 9% that she will also get a positivemammography (this is the false-positive rate)
What is the probability that a woman at age 40who had a positive mammogram actually hasbreast cancer?
P H Djð Þ ¼ 0:9$0:010:9$0:01þ0:09$0:99 ¼ 0:092
Natural frequencies– Ten out of 1,000 women at age 40 whoparticipate in mammography screening havebreast cancer (prevalence or base rate)– Of these 10 women, 9 have a positivemammogram (sensitivity or hit rate)– Out of the 990 women who do not have breastcancer, about 89 will have a positivemammogram nonetheless (false-positive rate)
Now imagine a representative sample of 1,000women age 40 who participate in breast cancerscreening. How many of these women witha positive test result actually have breast cancer?
P H Djð Þ ¼ 99þ89 ¼ 9:2
Five-year survival rate‘‘The 5-year survival rate for people diagnosedwith prostate cancer is 98% in the USA vs. 71% inBritain.’’
Annual mortality rate‘‘There are 26 prostate cancer deaths per100,000 American men versus 27 per 100,000men in Britain.’’
Single-event probability‘‘If you take Prozac, the probability that you willexperience sexual problems is 30–50% (or: 30 to50 chances out of 100).’’
Frequency statement‘‘Out of every 10 of my patients who take Prozac,3–5 experience a sexual problem.’’
655