Perceived Social Norms Guide Healthcare Decisions for Oneself and Others: A Cross-Sectional Experiment in a US Online Panel

Abstract

Background:

Global aging has increased reliance on surrogates to make healthcare decisions for others. We investigated differences between making healthcare decisions and predicting healthcare decisions, self-other differences for made and predicted healthcare decisions, and the roles of perceived social norms, emotional closeness, empathy, age, and gender.

Methods:

Participants (n=2,037) from a nationally-representative US panel were randomly assigned to make or to predict a healthcare decision. They were also randomly assigned to one of five recipients: themselves, a loved one 60 years or older, a loved one younger than 60, a distant acquaintance 60 or older, or a distant acquaintance younger than 60. Hypothetical healthcare scenarios depicted choices between relatively safe lower-risk treatments with a good chance of yielding mild health improvements versus higher-risk treatments that offered a moderate chance of substantial health improvements. Participants reported their likelihood of choosing lower vs. higher-risk treatments, their perceptions of family and friends’ approval of risky healthcare decisions, and their empathy.

Results:

We present three key findings. First, made decisions involved less risk taking than predicted decisions, especially for distant others. Second, predicted decisions were similar for others and oneself, but made decisions were less risk taking for others than oneself. People predicted that loved ones would be less risk taking than distant others. Third, perceived social norms were more strongly associated than empathy with made and predicted decisions.

Limitations:

Hypothetical scenarios may not adequately represent emotional processes in healthcare decision making.

Conclusions:

Perceived social norms may sway people to take less risk in healthcare decisions, especially when making decisions for others. These findings have implications for improving surrogate decision making.

Background

Population aging has increased the prevalence of surrogates who make medical decisions for adult patients.^1,2 Making healthcare decisions for incapacitated patients poses ethical challenges³ To address this, physicians call on surrogates to represent patients’ perspectives and may provide them with decision aids. Surrogates are asked to put aside their own preferences and apply the substituted judgment standard to make the healthcare decision that the patient would make for themselves, if they were able to do so.⁴ The substituted judgement standard assumes that surrogates accurately predict patients’ preferences, and make healthcare decisions that align with these predictions. However, surrogates’ predictions match patients’ actual preferences only 68% of the time. ³ Even when surrogates’ predictions match the preferences of the person they are deciding for (the recipient), they may not always decide in line with this knowledge.^2,5,6 For instance, hypothetical healthcare decisions for others reflected surrogates’ own preferences.^7,8 Here, we compared made and predicted healthcare decisions. We investigated whether there were differences in making healthcare decisions for oneself or another person, and in predicting healthcare decisions for oneself and for another person. We considered whether any differences depended on perceived social norms, empathy, participants’ age, gender, emotional closeness to recipients, and the recipient’s age. We used healthcare scenarios that depicted choices between relatively safe lower-risk treatments with a good chance of yielding mild health improvements versus higher-risk treatments that offered a moderate chance of substantial health improvements with a small chance of serious adverse physical side effects or death. Addressing these issues has implications for improving surrogates’ decision making.

Self-other differences in risky decisions

Much of the literature investigating how people make decisions for others examines self-other differences in risk taking.^9,10 Comparing the “other” condition to the “self” condition helps to identify factors associated with relatively more or less risk taking when making decisions for others vs. oneself. People decide differently when others, rather than they themselves, are at risk, but the direction of these differences depends on the domain.^9,10 For healthcare and similar decisions involving health and safety, people generally make less risky decisions for others than themselves.^6,11,12 For example, when doctors made hypothetical medical decisions for patients rather than for themselves, they were less likely to select the riskier of two treatments. ^5,13

Social values theory suggests that decisions for others are partly guided by perceptions of social expectations (injunctive norms), about how a person should decide.¹⁴ Increases or decreases in risk seeking for others relative to oneself are thought to reflect perceived social norms. For example, when social values prioritize caution, such as for medical treatments, hypothetical choices for others are less risky than choices for oneself.^5,6,13 Because perceived social norms are relatively more focal when making decisions for others than when making decisions for oneself, self-other differences in risky decisions emerge.¹⁵

Healthcare recipient characteristics

A recipient’s familiarity with the surrogate may influence the surrogate’s decision. Concerns about maintaining relationships and accountability for negative outcomes may foster greater risk aversion in decisions for close versus distant others.^2,16 Healthcare decisions made for a close relative were less risk seeking than those for a stranger, but the evidence was weak.⁹

Surrogates also consider patients’ ages when prioritizing healthcare.^17,18 Ageism can influence the treatments patients receive.¹⁹ Stereotypes of cautious older adults²⁰ could result in others selecting less risky treatments for older versus younger recipients.

Mismatch between made and predicted decisions

To make a surrogate decision that matches a recipient’s preference, the surrogate has to accurately predict the choice a recipient would make for themselves, and then base their decision on this prediction.⁴ If recipients’ preferences are unknown, surrogates may choose what they think is best for the recipient, what is best for themselves, or use the “golden rule” to choose for the recipient as they would for themselves.^2,10 Surrogates have been found to project their own preferences onto recipients and also to make choices for others that are less risky than their predictions of the choices the recipient would make for themselves.^5,6,8,12 Potentially, the mismatch between made and predicted decisions may be smaller for close versus distant others.² Construal-level theory posits that decisions for close (versus distant) others reflect greater concrete (versus abstract) thinking.²¹ Concrete thinking may facilitate better prediction of decisions for close versus distant others.^2,22

Decision maker characteristics

Empathy and emotions may play a role in surrogate decision making. A mismatch between made and predicted decisions has been found to occur because people deliberate decisions, while relying on emotions for predictions.²³ This explanation derives from the “risk-as-feelings hypothesis.” ^24,25 Self-other differences may occur because of the inability to fully experience others’ emotions.²⁶ Smaller self-other differences when predicting decisions of familiar versus unfamiliar others may reflect greater empathy for familiar others.^{24, 27}

Age differences in risk taking vary by decision domain.^28,29 Age differences were not found across medical decisions involving surgery or medications.^30,31 Older age was, however, associated with reporting less risk taking for health and safety decisions, such as using sunscreen, and for social decisions, such as disagreeing with a friend.^{28,32,33,34,35}

Older adults’ lesser social risk taking has been suggested to reflect their motivation to prioritize emotionally meaningful relationships over peripheral ties as time left in life becomes more precious.^33,36,37,38 If willingness to take risks is reduced when making decisions for close versus more distant others,¹⁶ this may be stronger in older than in younger adults.

Men report taking more health and safety risks than women.^28,39,40 Gender differences were not found for social risk taking such as trusting others.^28,40 Gender differences in risky healthcare decisions for others have not been investigated.

The Current Study

Here, we presented participants with three healthcare decision scenarios that offered a choice between a higher-risk treatment and a lower-risk treatment. We randomly assigned participants to either make or predict hypothetical healthcare decisions, and then to one of five healthcare recipients: themselves, a loved one 60 years or older, a loved one younger than 60, a distant acquaintance 60 or older, or a distant acquaintance younger than 60. This design allowed comparisons between made and predicted decisions and investigation of self-other differences for made and predicted healthcare decisions. Building from previous work,^2,9 we further investigated whether self-other differences depended upon perceived social norms, the emotional closeness of the decision maker and the recipient, empathy, or the recipient’s age. We also considered the decision maker’s age and gender. Our research questions were:

1. Is there a mismatch between made and predicted healthcare decisions? If so, does the mismatch vary depending on the recipient’s emotional closeness and age?

2. Are there self-other differences in made and predicted healthcare decisions? If so, are such differences related to recipient (emotional closeness, age) and decision maker (age, gender) characteristics?

3. Are self-other differences in made and predicted healthcare decisions related to empathy or the perceived social value of healthcare risk taking?

Method

Participants

Participants were panel members of the University of Southern California’s internet-based Understanding America Study (UAS). The panel was recruited via mailed invitations to randomly-selected US addresses. Panellists were provided internet access and a computer or tablet, if needed. Panellists receive $20 for 30 minutes of survey time. Survey invitations were sent to 2,720 panellists; 2,037 (75%) completed the survey. All data and materials are available (UAS 151, https://uasdata.usc.edu/index.php). Participants were adults aged 18-100 (M=57.06, SD=15.02), 49.2% women, 85.2% Whites/Caucasians, and 94.8% Non-Hispanics/Latinos; 50% had an associate’s degree or higher, and 45.6% reported family income as $49,999 or less.

Procedure

We used a 2 (group) X 5 (healthcare recipient) between-subjects design. Participants were randomly assigned to either a “made decisions” or a “predicted decisions” group and to answer questions about one of five healthcare recipients: themselves, a loved one who was 60 years or older, a loved one who was younger than 60, a distant other—a person they knew, but with whom they were not emotionally close – who was 60 or older, or a distant other who was younger than 60 (Table 1). Age 60 was selected because it is marks of the end of midlife and beginning of later adulthood⁴¹ and because it increased the likelihood the respondent would be able to imagine a living loved one or distant other, compared to specifying an older age.⁴² Those assigned to answer questions about others indicated the person’s first name, gender, and their relationship (e.g., spouse, friend). As a manipulation check, participants reported the age of the person they had in mind and rated their emotional closeness to that person (1=not at all close to 6= very close). Both manipulations were successful (Supplementary Tables 1 and 2).

Table 1.

Scenarios by Healthcare Recipient Group

Healthcare Recipient
	Self	Loved One: ≥ 60yrs, < 60yrs	Distant Other: ≥ 60yrs, < 60yrs
Scenario 1	Imagine that you have suffered a moderately severe stroke. One arm and one leg are paralyzed. You have trouble speaking and trouble understanding when others speak. You rely on others for help with feeding, dressing, bathing, and toileting. Your doctor says that without treatment, you have a very slight chance of improvement. There are two treatments available: Treatment A (Therapy): Treatment A aims to reduce the severity of the symptoms through speech, physical, and occupational therapy. Your doctor says that there is a good chance (80%) that the therapy will improve quality of life because it will reduce the severity of your symptoms. There is a small chance (20%) there will be no change. Treatment B (Surgery): Treatment B aims to reduce the severity of the symptoms through surgery. Your doctor says that if successful, you will be completely cured after the surgery. However, the surgery is risky. There is a very small chance (5%) that it may be fatal. There is a 50-50 chance that after the surgery, you will not be cured and things may get worse	Imagine that your loved one has suffered a moderately severe stroke. One arm and one leg are paralyzed. Your loved one has trouble speaking and trouble understanding when others speak. Your loved one relies on others for help with feeding, dressing, bathing, and toileting. Your loved one’s doctor says that without treatment, your loved one has a very slight chance of improvement. There are two treatments available: Treatment A (Therapy): Treatment A aims to reduce the severity of the symptoms through speech, physical, and occupational therapy. Your loved one’s doctor says that there is a good chance (80%) that the therapy will improve quality of life because it will reduce the severity of your loved one’s symptoms. There is a small chance (20%) there will be no change. Treatment B (Surgery): Treatment B aims to reduce the severity of the symptoms through surgery. Your loved one’s doctor says that if successful, you will be completely cured after the surgery. However, the surgery is risky. There is a very small chance (5%) that it may be fatal. There is a 50-50 chance that after the surgery, your loved one will not be cured and things may get worse.	Imagine that the person has suffered a moderately severe stroke. One arm and one leg are paralyzed. The person has trouble speaking and trouble understanding when others speak. The person relies on others for help with feeding, dressing, bathing, and toileting. The person’s doctor says that without treatment, the person has a very slight chance of improvement. There are two treatments available: Treatment A (Therapy): Treatment A aims to reduce the severity of the symptoms through speech, physical, and occupational therapy. The person’s doctor says that there is a good chance (80%) that the therapy will improve quality of life because it will reduce the severity of the person’s symptoms. There is a small chance (20%) there will be no change. Treatment B (Surgery): Treatment B aims to reduce the severity of the symptoms through surgery. The person’s) doctor says that if successful, the person will be completely cured after the surgery. However, the surgery is risky. There is a very small chance (5%) that it may be fatal. There is a 50-50 chance that after the surgery, the person will not be cured and things may get worse.
Scenario 2	Now, instead, imagine a different situation in which you have colon cancer that has spread to the liver. You are tired and weak, needing some help with household chores. Your thinking and memory are not affected. You are not in pain. Your doctor says that without treatment, there is no chance of recovery and you would have about six months to live. There are two treatments available Treatment A (Radiation Therapy): Treatment A aims to reduce the size of tumors through radiation therapy. Your doctor says that there is a good chance (80%) that the therapy will improve your quality of life because it will reduce the severity of your symptoms and increase your chance of living longer than six months. There is a small chance (20%) there will be no change. Treatment B (Surgery): Treatment B aims to remove the tumors through surgery. Your doctor says that if successful, your quality of life will improve because the surgery will completely eliminate your symptoms and will increase your chance of living longer than one year. However, the surgery is risky. There is a very small (5%) chance that it may be fatal. There is a 50-50 chance that after the surgery you will not be cured and things may get worse.	Now, instead, imagine a different situation in which your loved one has colon cancer that has spread to the liver. Your loved one is tired and weak, needing some help with household chores. Your loved one’s thinking and memory are not affected. Your loved one is not in pain. Your loved one’s doctor says that without treatment, there is no chance of recovery and your loved one would have about six months to live. There are two treatments available Treatment A (Radiation Therapy): Treatment A aims to reduce the size of tumors through radiation therapy. Your loved one’s doctor says that there is a good chance (80%) that the therapy will improve your loved one’s quality of life because it will reduce the severity of your loved one’s symptoms and increase your loved one’s chance of living longer than six months. There is a small chance (20%) there will be no change. Treatment B (Surgery): Treatment B aims to remove the tumors through surgery. Your loved one’s doctor says that if successful, your loved one’s quality of life will improve because the surgery will completely eliminate your loved one’s symptoms and will increase your loved one’s chance of living longer than one year. However, the surgery is risky. There is a very small (5%) chance that it may be fatal. There is a 50-50 chance that after the surgery your loved one will not be cured and things may get worse.	Now, instead, imagine a different situation in which the person has colon cancer that has spread to the liver. The person is tired and weak, needing some help with household chores. The person’s thinking and memory are not affected. The person is not in pain. The person’s doctor says that without treatment, there is no chance of recovery and the person would have about six months to live. There are two treatments available Treatment A (Radiation Therapy): Treatment A aims to reduce the size of tumors through radiation therapy. The person’s doctor says that there is a good chance (80%) that the therapy will improve the person’s quality of life because it will reduce the severity of the person’s symptoms and increase the person’s chance of living longer than six months. There is a small chance (20%) there will be no change. Treatment B (Surgery): Treatment B aims to remove the tumors through surgery. The person’s doctor says that if successful, the person’s quality of life will improve because the surgery will completely eliminate the person’s symptoms and will increase the person’s chance of living longer than one year. However, the surgery is risky. There is a very small (5%) chance that it may be fatal. There is a 50-50 chance that after the surgery the person will not be cured and things may get worse.
Scenario 3	Now, instead imagine a different situation in which you have Alzheimer’s disease. You have trouble remembering things and thinking clearly. You cannot always recognize people you know. Although you have no chance of getting better, it is not certain how fast things will get worse. Without treatment, your mental abilities may get worse quickly, or they may stay the way they are now for a long time. Your physical condition is not affected. There are two treatments available: Treatment A (Activities): Treatment A aims to reduce the severity of the symptoms through activities such as exercising, playing memory games, and solving puzzles. Your doctor says that there is a good chance (80%) that the activities will improve your quality of life because they will reduce the severity of your symptoms. There is a small chance (20%) there will be no change. Treatment B (Medication): Treatment B aims to reduce the severity of the symptoms and slow down the progression of the disease through medication. Your doctor says that if successful, the medication will improve your quality of life because it will reduce the severity of the symptoms and stop the disease from getting worse for at least six months. However, the medication is risky. There is a very small chance (5%) of serious, adverse physical side effects. There is a 50-50 chance that the medication will not work and things will get worse.	Now, instead imagine a different situation in which your loved one has Alzheimer’s disease. Your loved one has trouble remembering things and thinking clearly. Your loved one cannot always recognize people they know. Although your loved one has no chance of getting better, it is not certain how fast things will get worse. Without treatment, your loved one’s mental abilities may get worse quickly, or they may stay the way they are now for a long time. Your loved one’s physical condition is not affected. There are two treatments available: Treatment A (Activities): Treatment A aims to reduce the severity of the symptoms through activities such as exercising, playing memory games, and solving puzzles. Your loved one’s doctor says that there is a good chance (80%) that the activities will improve your loved one’s quality of life because they will reduce the severity of your loved one’s symptoms. There is a small chance (20%) there will be no change. Treatment B (Medication): Treatment B aims to reduce the severity of the symptoms and slow down the progression of the disease through medication. Your loved one’s doctor says that if successful, the medication will improve your loved one’s quality of life because it will reduce the severity of the symptoms and stop the disease from getting worse for at least six months. However, the medication is risky. There is a very small chance (5%) of serious, adverse physical side effects. There is a 50-50 chance that the medication will not work and things will get worse.	Now, instead imagine a different situation in which the person has Alzheimer’s disease. The person has trouble remembering things and thinking clearly. The person cannot always recognize people they know. Although the person has no chance of getting better, it is not certain how fast things will get worse. Without treatment, the person’s mental abilities may get worse quickly, or they may stay the way they are now for a long time. The person’s) physical condition is not affected. There are two treatments available: Treatment A (Activities): Treatment A aims to reduce the severity of the symptoms through activities such as exercising, playing memory games, and solving puzzles. The person’s doctor says that there is a good chance (80%) that the activities will improve the person’s quality of life because they will reduce the severity of the person’s symptoms. There is a small chance (20%) there will be no change. Treatment B (Medication): Treatment B aims to reduce the severity of the symptoms and slow down the progression of the disease through medication. The person’s) doctor says that if successful, the medication will improve the person’s quality of life because it will reduce the severity of the symptoms and stop the disease from getting worse for at least six months. However, the medication is risky. There is a very small chance (5%) of serious, adverse physical side effects. There is a 50-50 chance that the medication will not work and things will get worse.

Note. Scenarios were adapted from Bookwala, Coppola, Fagerlin, Ditto, Danks, and Smucker (2001). Participant were randomly assigned to think about one of five healthcare recipients. The Flesch-Kincaid Grade level for readability was 7.7 for Scenario 1, and 7.8 and 8.9 for Scenarios 2 and 3, respectively. For each scenario, Treatment A corresponds to the lower-risk treatment and Treatment B corresponds to the higher-risk treatment.

Risk taking in healthcare decisions.

Each participant saw three hypothetical scenarios used in prior research on surrogate decision making that depicted age-related diseases, stroke, cancer, and Alzheimer’s Disease (Table 1).^7,43,44 Following prior research,⁵ we created two treatment options for each scenario, a lower-risk treatment with a good chance of yielding mild health improvements and a higher-risk treatment that offered a moderate chance of substantial health improvements with a small chance of serious adverse physical side effects or death. The “made decisions” group was asked, “what decision would you make for yourself/your loved one/the other person?” The “predicted decisions” group was asked, “what do you think you/your loved one/the other person would decide to do in this situation?” Following previous research,⁶ we worded the “self” questions differently in the made decision and predicted decision groups to facilitate self-other comparisons within both made and predicted decisions.

For each scenario, participants indicated the decision they would make or predict by rating the likelihood of choosing the lower-risk (Treatment A) relative to the higher-risk (Treatment B) treatment on a scale ranging from 1= most likely to choose Treatment A to 6=most likely to choose Treatment B. The average score across the three scenarios was computed. Higher scores indicated greater risk taking in made decisions (Cronbach’s α=.56) or predicted decisions (Cronbach’s α= .59).

As a manipulation check, participants rated how risky, in general, they perceived the treatments to be on a scale ranging from 0= not at all risky to 4= very risky.⁴⁵ Participants perceived lower-risk treatments to be significantly less risky than their higher-risk counterparts (all ps<.001). Cohen’s D, an indicator of effect size, was 1.19, 1.09, and 1.07 for the difference between lower- and higher-risk treatments for Scenarios 1, 2, and 3, respectively.

Empathy.

We measured empathic concern using 7-items (e.g., I often have tender, concerned feelings toward people less fortunate than me) on a scale ranging from 0 = does not describe me well, to 4 = describes me very well (α=.73, M=3.04, SD=.64).⁴⁶

Perceived social norms.

We measured perceived social norms based on the idea that they constitute a shared belief system.⁴⁷ After indicating their made decision or predicted decision, all participants saw the three scenarios again, but scenarios depicted a gender-ambiguous surrogate and healthcare recipient, where “Charlie” was tasked with making a decision on behalf of themselves/loved one/distant other (matched to participants’ healthcare recipient group assignment), named “Pat.” After each scenario, two items assessed participants’ views of the social appropriateness of the lower- and higher-risk treatments. One item corresponded to the lower-risk treatment: “Charlie chose Treatment A. How do you think your friends and family would rate the appropriateness of Charlie’s decision?” The other, the higher-risk treatment: “Charlie chose Treatment B. How do you think your friends and family would rate the appropriateness of Charlie’s decision?”

Participants rated each item on a scale ranging from 0=not at all appropriate to 10 = completely appropriate. The difference between the appropriateness ratings of Treatment A and B indicated the direction and strength of the perceived social norm.^14,15,48 The average score of the three scenarios was computed. Higher scores indicated greater perceived appropriateness of the lower risk treatments relative to the higher risk ones (Cronbach’s α=.73, M=4.60, SD=8.39)

Demographic variables.

Gender, income, education, and race were assessed using previously-collected UAS data. Gender was coded 0=female, 1=male. Race was coded 0=non-white, 1=white.

Statistical analyses

Analyses were conducted with IBM SPSS version 27 (see Table 2 for descriptive statistics). As an initial analysis, we computed Pearson correlations to examine relations of study variables with demographic variables. Those demographic variables that showed significant correlations with study variables were then included as covariates in subsequent analyses.

Table 2.

Means (Standard Errors) of Risk Taking in Made Decisions and Predicted Decisions, and Perceived Social Norms for Risky Decisions by Healthcare Recipient, Gender, and Race

	Made Decisions		Predicted Decisions			Perceived Social Norms
	M	SE	M	SE	M	SE
Healthcare Recipient
Self	3.01	(.10)	2.96	(.09)	1.34	(.15)
Loved One 60 ≥	2.71	(.10)	2.84	(.10)	1.73	(.14)
Loved One 60 <	2.75	(.09)	3.00	(.09)	1.55	(.13)
Distant Other 60 ≥	2.79	(.09)	3.26	(.09)	1.55	(.13)
Distant Other 60 <	2.96	(.09)	3.39	(.09)	1.47	(.14)
Total	2.84	(.04)	3.09	(.04)	1.52	(.06)
Gender
Male	2.97	(.06)	3.18	(.06)	1.31	(.09)
Female	2.71	(.06)	3.00	(.06)	1.76	(.09)
Race
White	2.87	(.05)	3.09	(.05)	1.46	(.07)
Non-White	2.71	(.11)	3.08	(.11)	1.89	(.16)

Note. Risk taking in made decisions and predicted decisions are average scores across the three scenarios. Higher scores indicate more risk taking in made decisions and predicted decisions. Perceived social norms are average scores across the three scenarios. Higher scores indicate greater perceived appropriateness of lower-risk treatments relative to higher-risk ones.

1. Is there a mismatch between made and predicted healthcare decisions? If so, does the mismatch vary depending on the recipient’s emotional closeness and age?

We conducted a 2 (group: made vs. predicted decision) X 5 (healthcare recipient: self; loved one 60 or older; loved one younger than 60; distant other 60 or older; distant other younger than 60) analysis of covariance (ANCOVA) predicting risk taking. To follow up the significant interaction, we tested the simple effect of group (made vs. predicted decision) at each level of healthcare recipient. We used partial eta-squared (η_p²) as an indicator of effect size.

2. Are there self-other differences in made and predicted healthcare decisions? If so, are such differences related to recipient (emotional closeness, age) and decision maker (age, gender) characteristics?

We computed two linear regressions that respectively examined relations of risk taking in made decisions and predicted decisions with decision maker and healthcare recipient characteristics.

Four orthogonal contrasts represented the five healthcare recipients to which participants were randomly assigned (Supplementary Table 3). The first, “self vs. other,” compared risk taking for oneself versus other people. The second, “loved one vs. distant other,” compared loved ones to distant others. The third, “older vs. younger,” compared those 60 years or older to those younger than 60. The fourth, “older vs. younger by loved one vs. distant other,” examined the interaction between healthcare recipient (60 years or older vs. younger than 60) and relationship (loved one vs. distant).

For the regression models, respondent characteristics, and the four contrasts were predictors. When building the models for made decisions and predicted decisions, we explored potential interactions between respondent age and each of the four contrasts, and respondent gender and each of the four contrasts. When significant, we included the interaction term in the model and used simple effects to interpret it.

3. Are self-other differences in made and predicted healthcare decisions related to empathy or the perceived social value of healthcare risk taking?

We computed Pearson correlations and then, two linear regressions that respectively examined correlates of risk taking in made decisions and predicted decisions. The regression models were the same as those used to address research question two except that perceived social norms and empathy were added to the equations.

Results

Initial analysis

Income, race, and education were correlated with key study variables of risk taking in predicted decisions, perceived social norms, empathy, and respondent age and were thus controlled in analyses that addressed our research questions (see Table 3, Supplementary Table 4). Controlling for these variables (vs. not) did not change the significance of the results.

Table 3.

Pearson Correlations of Risk Taking in Made Decisions and Predicted Decisions, Perceived Social Norms for Risky Decisions, Empathy, Respondent Age, Healthcare Recipient Age, and Respondent Gender, Race, Income and Education

	Made Decisions	Predicted Decisions	Perceived Social Norms	Empathy	Respondent Age	Healthcare Recipient Age	Gender	Race	Income	Education
1. Made Decisions			−.47**	−.12**	.00	−.01	.10**	.04	.04	−.02
2. Predicted Decisions			−.39**	−.06*	−.02	−.05	.07*	.00	.10**	−.01
3. Perceived Social Norms				.16**	.02	.03	−.08**	−.06*	−.04	−.02
4. Empathy					.07*	.04	−.25**	.05*	.01	−.01
5. Respondent Age						.15**	.04	.10**	.02	.05*
6. Healthcare Recipient Age							.01	.04	.04	.00
7. Gender								.02	.15**	.06**
8. Race									.18**	.04
9. Income										.44**
10. Education

Note.

^*p<.05

^**p< .001.

Risk taking in made decisions and predicted decisions are average scores across the three scenarios, collapsed across healthcare recipient groups (Supplementary Table 4 shows correlations of study variables by healthcare recipient group). Higher scores indicate more risk taking in made decisions and predicted decisions. Perceived social norms are average scores across the three scenarios, collapsed across healthcare recipient groups. Higher scores indicate greater perceived appropriateness of lower-risk treatments relative to higher-risk ones. Healthcare recipient age is a continuous variable in this analysis. For gender, 0=female, 1=male. For race, 0=non-white, 1=white. Correlations between binary and continuous variables correspond to a point-biserial correlation, correlations between two binary variables correspond to a phi coefficient; correlations for education correspond to Spearman’s rho.

1. Is there a mismatch between made and predicted healthcare decisions? If so, does the mismatch vary depending on the recipient’s emotional closeness and age?

Overall, made decisions (M=2.84, SE=.04) involved less risk taking than predicted decisions (M=3.09, SE=.04), F(1, 2036)=16.15, p<.0001, η_p²=.01 (Figure 1) seen in predicted decisions leaning more towards the higher-risk treatment than made decisions. Risk taking varied by healthcare recipient F(4, 2036)=5.51, p<.0001, η_p²=.01, which interacted with whether decisions were made vs. predicted F(4, 2036)=2.67, p=.03, η_p²=.01. For distant others, made decisions were less risk taking than predicted decisions, independent of whether distant others were 60 or older (p<.0001, η_p²=.03) or younger than 60 (p=.001, η_p²=.03). For loved ones, a similar trend was found, but only for recipients younger than 60 (p=.07 η_p²=.01). Risk taking in made and predicted decisions did not differ significantly for oneself (p =.60 η_p²=.00) and loved ones older than 60 (p=.31 η_p²=.00).

Figure 1.

Risk taking in made decisions and predicted decisions by healthcare recipient. Risk taking in made and predicted decisions are average scores across the three scenarios. Higher scores indicate more risk taking in made decisions and predicted decisions. Error bars indicate standard error of the mean. Asterisks indicate made decisions differed significantly from predicted decisions within healthcare recipients, p<.05

2. Are there self-other differences in made and predicted healthcare decisions? If so, are such differences related to recipient (emotional closeness, age) and decision maker (age, gender) characteristics?

Made decisions.

The significant “self vs. others” contrast indicated that people were less risk taking, as seen in being less likely to choose a higher-risk treatment, when making a decision for another person rather than for themselves (Table 4, Model 1—Made Decisions). Men were significantly more risk taking than women. The gender by “loved one vs. distant others” interaction was significant. Tests of simple effects indicated that men were more risk taking when making a decision for a distant other compared to a loved one. Women’s decisions for a loved one and distant other did not differ significantly. Neither respondent age, nor interactions between respondent age and any of the four contrasts were significant.

Table 4.

Beta coefficients (Standard Errors) for Linear Regressions Predicting Risk Taking in Made Decisions and Predicted Decisions

Predictor Variable	Made Decisions				Predicted Decisions
	Model 11		Model 22		Model 13		Model 24
	Beta	SE	Beta	SE	Beta	SE	Beta	SE
Self vs. Others	−.07**	(.02)	−.03	(.02)	.05	(.02)	.05	(.02)
Loved One vs. Distant Other	−.05	(.05)	−.04	(.04)	−.13**	(.05)	−.13**	(.04)
Older vs. Younger	−.04	(.05)	−.02	(.04)	−.05	(.05)	−.05	(.04)
Older vs. Younger X Loved One vs. Distant Other	−.03	(.05)	−.02	(.04)	.00	(.05)	−.01	(.04)
Respondent age	−.00	(.00)	.02	(.00)	−.02	(.00)	−.01	(.00)
Gender	.10*	(.08)	.07*	(.07)	.06*	(.08)	.02	(.08)
Gender X Loved One vs. Distant Other	−.07*	(.05)	−.06*	(.04)
Men: Loved One v. Distant Other	−.08**	(.06)	−.0.7*	(.06)
Women: Loved One v. Distant Other	.01	(.07)	.01	(.06)
Empathy			−.02	(.06)			.00	(.06)
Perceived social norms			−.46**	(.01)			−.39**	(.01)
Control Variables
Household income	.04	(.01)	.04	(.01)	.12**	(.01)	.11**	(.01)
Education	−.04	(.02)	−.05	(.02)	−.07*	(.02)	−.07*	(.02)
Race	.04	(.12)	−.02	(.11)	−.02	(.12)	−.02	(.11)

Note.

^*p <.05

^**p < .001.

For risk taking in made decisions and predicted decisions, Model 1 corresponds Research Question 2. For risk taking in made decisions and predicted decisions, Model 2 corresponds to Research Question 3. Risk taking in made and predicted decisions are average scores across the three scenarios. Higher scores indicate more risk taking in made decisions and predicted decisions. Perceived social norms are average scores across the three scenarios. Higher scores indicate greater perceived appropriateness of lower-risk treatments relative to higher-risk ones. For gender, 0=female, 1=male. For race, 0=non-white, 1=white. Excluding demographic control variables did not change the significance of any of the coefficients.

¹F(10,1002) =2.76, p =.002, R²=.02

²F(9,1031)=4.60, p <.001, R²=.03

³F(12,1002)=25.79, p <.0001, R²=.23

⁴F(11,1031)=21.35, p <.001, R²=.18

Predicted decisions.

The significant “loved one vs. distant others” contrast indicated that individuals predicted that loved ones would be less risk taking than distant others (Table 4, Model 1—Predicted Decisions). Compared to women, men’s predicted decisions were significantly more risk taking. There were no significant interactions between the four contrasts and respondent age or gender.

3. Are self-other differences in made and predicted healthcare decisions related to empathy or the perceived social value of healthcare risk taking?

For each of the five healthcare recipient groups, perceiving that one’s social group valued avoiding healthcare risks was significantly correlated with less risk taking in made decisions and predicted decisions (Supplementary Table 4). Greater empathy was significantly correlated with less risk taking in made decisions for distant others and with less risk taking in predicted decisions for loved ones older than 60 yrs.

Made decisions.

Model 2 was identical to Model 1, except that perceived social norms and empathy were included in the regression equation. Perceived social norms were significantly associated with risk taking in made decisions—those who perceived that avoiding healthcare risks was valued by their social groups were less risk taking (Table 4, Model 2—Made Decisions). Gender, and the significant gender by “loved one vs. distant other” interaction described above, were significant. Empathy was not significantly associated with risk taking. The “self vs. others” contrast described above was no longer significant when perceived social norms was added to the model.

Predicted decisions.

Perceived social norms were significantly associated with risk taking in predicted decisions—those who perceived that avoiding healthcare risks was valued by their social groups predicted that others would be less risk taking (Table 4, Model 2—Predicted Decisions). Empathy was not significant. The “loved one vs. distant other” contrast described above remained significant.

Discussion

Population aging and the associated increase in surrogate decision making increase the need to understand factors that influence healthcare decisions for others.² In a national US sample, we investigated differences in participants’ responses when they were asked to make or predict hypothetical healthcare decisions, for oneself and for various others. We report on three key findings.

First, decisions made for others involved less risk taking than decisions predicted for others, as seen in avoiding higher-risk treatments. The mismatch between made decisions and predicted decisions was especially apparent for distant others, suggesting that made and predicted decisions may diverge as psychological distance increases.²¹ Greater psychological closeness to loved ones (vs. distant others) may increase familiarity and confidence that predictions match patients’ preferences.² Thus, surrogate decision makers who are less close to the patient may struggle to align treatment decisions to the patients’ preferences.

Second, people predicted that others would make decisions similar to their own, but were less risk taking when making decisions for others than decisions for themselves. Although this pattern of results has been found in other situations,^5,6,12 our study suggests that this finding generalizes to a surrogate decision-making context. Social values theory proposes that self-other differences in decision making occur because people rely more on perceived social norms when making decisions for others than for themselves.¹⁴

Third, in line with social values theory,¹⁴ we found that perceived social norms explained self-other differences in made decisions. These findings align with an emerging literature that shows perceived social norms are central to people’s decisions in a variety of contexts.^51,52,53 Surrogate decision makers may rely on social norms because it dampens their concerns about being held accountable for bad outcomes.² This suggests that if surrogates’ perceptions of social norms do not match patients’ preferences, they may not make decisions patients want.

Social norms played a greater role in made and predicted decisions than did empathy. Theorists have suggested that empathy and the decision maker’s relationship to the treatment recipient are central to surrogate decision making.² In our study, greater empathy was correlated with less risk taking in made decisions for distant others, and with less risk taking in predicted decisions for loved ones older than 60. However, empathy was not significantly related to made or predicted decisions after controlling for the other variables in regression models.

Our study also uncovered demographic differences. Men made riskier treatment decisions than women did. Men report worrying less about risks when making healthcare decisions⁴⁹, and risk taking is consistent with stereotypical masculine roles.⁵⁰ Yet, in our study, men’s decisions were not uniformly risky, but depended on the healthcare recipient. Men were less risk taking in made decisions for loved ones compared to distant others. Thus, surrogates’ gender may need to be taken into account.

Risk taking in healthcare decisions did not vary based on the decision maker’s age, or on the healthcare recipient’s age. Prior research found no age differences in healthcare decisions about surgery and medication either, though older adults did make less risky decisions about preventative health behaviors.^31,35 Prior research did suggest that decision makers consider patients’ ages when making healthcare decisions.^17,18 However, in our study, risk taking in healthcare decisions did not differ for healthcare recipients who were 60 or older versus younger than 60. We instructed participants to imagine a specific person, whereas prior research asked about an anonymous “patient.” Age may be less relevant than social norms^54,55 when making decisions about a known person versus an abstract other.²¹

All studies have limitations. First, our findings do not address the accuracy of decision makers’ predictions of healthcare recipient’s preferences. Yet, even when surrogates’ predictions are accurate, their decisions may not match their predictions,^5,6 making it important to identify factors related to this mismatch, such as those we investigated. Second, hypothetical scenarios are unlikely to elicit intense emotions that may affect real-world decisions for others.⁵⁶ Even so, perceptions of one’s friends’ and family members’ approval of decisions are likely to be important when decisions have real consequences. Third, we examined the role of perceived social norms but did not assess actual social norms. Yet, research suggests that perceived social norms predict people’s decisions over and above actual social norms.^57,58 Finally, the effect sizes corresponding to significant differences were often small. Effect sizes are best interpreted within the context of the experimental design and variables included in each analysis.

Our findings advance understanding of self-other differences in healthcare decisions by demonstrating the importance of perceived social norms. Perceived social norms were consistently correlated with decisions, irrespective of the healthcare recipient. Understanding the power of social norms in shaping surrogates’ decisions may help physicians to engage surrogates in shared decision making. Surrogate’s beliefs about social norms and risky healthcare choices for others can be applied to improve decision aids.^59,60

Ultimately, surrogates are in charge of making decisions that patients would have made if they had had the decision-making capacity to do so. Our findings suggest that surrogates may need help aligning healthcare decisions with the preferences they believe recipients to have. Encouraging individuals to document their healthcare preferences in advance and communicate them to potential surrogates may enable surrogates to make decisions that align with recipients’ preferences.^61,62 Such discussions may also prevent common problems due to surrogates being unaware of patients’ preferences.³ Additionally, acknowledging that perceived social norms may favor a decision that conflicts with a patient’s preferences may help to reduce surrogates’ stress and guilt, especially when there are family conflicts over healthcare decisions.⁶³ Nurse-led interventions that acknowledge surrogates’ emotions and remind them to consider patients’ preferences may further improve surrogates’ decision making.⁶⁴

Article highlights.

• People made less risky healthcare decisions for others than themselves, even though they predicted others would make decisions similar to their own. This has implications for understanding how surrogates apply the substituted judgement standard when making decisions for patients.

• Perceived social norms were more strongly related to decisions than treatment-recipient (relationship closeness, age) and decision-maker (age, gender, empathy) characteristics. Those who perceived that avoiding healthcare risks was valued by their social group were less likely to choose risky medical treatments.

• Understanding the power of perceived social norms in shaping surrogates’ decisions may help physicians to engage surrogates in shared decision making.

• Knowledge of perceived social norms may facilitate the design of decision aids for surrogates.