Comparing preferences for return of genome sequencing results assessed with rating and ranking items

1 |. INTRODUCTION

Genome sequencing is being used in research and clinical settings for identification of genetic variations (Biesecker & Green, 2014). Because sequencing technologies may yield secondary findings, which are genetic variations unrelated to the indication for sequencing, the American College of Medical Genetics and Genomics recommended the evaluation and reporting of secondary findings based on its minimum gene list (Blackburn et al., 2015; Kalia et al., 2017). This list currently contains 59 medically actionable genes and will be updated based on new research findings (Kalia et al., 2017). To inform policy regarding return of secondary findings further, physicians and researchers have explored the preferences of patients for the return of different types of sequencing results. Existing evidence suggests that patients most often prefer to receive results for medically actionable genes, although they may also want to receive other types of results (Kaphingst et al., 2016).

Among surveys designed to assess preference, rating is usually preferred over ranking due to its convenient implementation (Harzing et al., 2009). Rating scale questions use Likert response options to assess each individual item, this is an absolute assessment; while a ranking scale asks participants to rate their preference for each item compared to the others, this is a relative measure (Hein, Jaeger, Tom Carr, & Delahunty, 2008). Absolute scales provide information about preferences for each item, however, when several of the ratings are similar, relative scales provide information about the preferred importance of each item (Mccarty & Shrum, 2000). Here, the two methods are compared with respect to determining preferences for return of results from genome sequencing.

In a previous study using a dataset generated by a survey of 1,080 women diagnosed with breast cancer at a young age, researchers explored the association between psychological and clinical predictors and preference for return of seven different types of genome sequencing results as assessed with Likert Scales as a rating method for level of interest in return (Kaphingst et al., 2018). The results suggested that psychological factors were related to the preferences; participants with more knowledge about sequencing benefits, higher worry about genetic risks, and stronger health information orientation were more likely to be very interested in learning each type of result while clinical factors largely were not related to these preferences (Kaphingst et al., 2018). Here, we examine the association between psychological and clinical predictors and preferences for return of results using the ranking of the seven types of sequencing results and compare similarities and differences to the previous analysis using the rating scale (Hong, Biesecker, Ivanovich, Goodman, & Kaphingst, 2019; Kaphingst et al., 2018).

The study population was a nationwide cohort (Young Women’s Breast Cancer Program [YWBCP]) of women diagnosed with breast cancer at age 40 or younger. YWBCP participants were surveyed between June and December 2014. Among contacted individuals (n = 1,778), 1,080 (61%) women completed the survey. Only participants with non-missing responses for the ranking question were included (n = 1,045; 97%). All participants reviewed a consent information sheet and gave consent to participate. Primary data collection was approved by the Human Research Protection Office at Washington University in St. Louis. This secondary analysis study was approved by the New York University Institutional Review Board, University Committee on Activities Involving Human Subjects.

Ranked Preference was the outcome variable, indicating participants’ interest in learning different types of results that might be generated using genome sequencing, which was assessed by asking them to rank their interest in order from the most (marked as 1) to the least interested (marked as 7). There were seven types of possible results: 1) increase individual’s risk for preventable or treatable diseases, 2) increase individual’s risk for unpreventable or untreatable diseases, 3) affect individual’s response to medication, 4) affect children’s health, 5) affect health of other relatives, 6) have uncertain meaning (VUS), and 7) are unrelated to health (e.g., ancestry, physical traits). We examined the top choice (rank 1) for each participant and classified it into three groups: Actionable (types 1 or 3), Carrier (types 4 or 5), Non-actionable (types 2, 6, or 7).

We included psychological, clinical and demographic predictors in our model. For psychological predictors, we had Genetic Causal Beliefs (The degree to which respondents believed that breast cancer was caused by genes), assessed using one item answered on a five-point scale (Weinman, Petrie, Moss-Morris, & Horne, 1996). Answers were dichotomized into ‘Low causal belief’ (‘1-Not at all’; ‘2-A little’; ‘3-Somewhat’) and ‘High causal belief’ (‘4-Mostly’; ‘5-Completely’) for analysis. We assessed two sub-scales: Health Consciousness (five items) and Health Information Orientation (eight items) (Dutta-Bergman, 2003). The average scores for each sub-scale were modeled as continuous variables. Worry about genetic risk, we assessed genetic worry with the average score from three items (Biesecker et al., 2009). Worry about cancer, we assessed worry related to cancer with three items (Gotay & Pagano, 2007). The average score was included as a continuous variable. Genome Sequencing Knowledge, we used two sub-scales: knowledge about sequencing benefits and knowledge about sequencing limitations (Kaphingst et al., 2012). In the analysis, average sub-scale scores were included as two continuous variables. Decision-making preference, we utilized two items adapted from the Control Preferences Scale (Lillie et al., 2007). We categorized the responses into passive decision-making (‘1-I prefer to leave the decision to my doctor’; ‘2-I prefer my doctor makes the decision, but seriously considers my opinion’); shared decision-making (‘3-I prefer my doctor and I share responsibility for deciding’); and active decision-making (‘4-I prefer to make my own decision after seriously considering my doctor’s opinion’; ‘5-I prefer to make my own decision regardless of my doctor’s opinion’). Family communication responses were assessed using two items inquiring having shared genetic test results with family members, which were categorized as ‘0-did not share/not sure’; ‘1-shared with relatives’; ‘2- spouse/partner/other/not specified’ (Elrick et al., 2017).

For clinical predictors, we had Prior Genetic testing. We combined the responses to ‘Have you ever received genetic testing as part of your clinical care?’ and ‘Which of these types of clinical tests have you received? Check all that apply.’ into ‘Did not have testing/not sure/other genetic test’ and ‘Had BRCA1/2 genetic testing.’

Finally, for demographic characteristics, we included Having biological children (yes/no), and education (Graduate degree or more/College or less), which were included as dichotomous variables; Employment status (Full-time/Part-time/Other) was a categorical variable.

Multinomial logistic regression analysis was used to examine relationships between predictors and the top choice category (actionable/carrier/non-actionable). Psychological variables were tested for entry into the multivariable model if they were statistically significant in bivariate analyses. The final model contained psychological variables that remained significant in the multivariable model, controlling for clinical (prior genetic testing) and sociodemographic variables (age, biological children, education and employment). We conducted the analyses using StataSE version 15 (StataCorp.). Statistical significance was assessed as p < .05.

Participant characteristics are shown in Table 1. In Table 2, we examined associations between predictors and top choice in the ranking. Controlling for education attainment, employment status, and age, compared to women without biological children, women with biological children were significantly more likely to rank learning about carrier status results as their top choice (RRR = 9.084 [3.56–23.18]).

TABLE 1.

Characteristics of participants in analytic sample (n = 1,025)

Continuous variables	Mean(SD)	Range
Age	45.8(9.0)	26–78
Categorical variables	N	Percent(%)
aTop choice (Dependent variable)
Actionable	822	80.2
Carrier	104	10.1
Non-actionable	39	3.8
Prior genetic testing
Did not have testing/not sure/had other genetic testing	149	14.5
Had BRCA1/BRCA2 testing	876	85.5
Educational attainment
College degree or less	559	54.5
Graduate level	466	45.5
Employment status
Full-time	601	58.6
Part-time	199	19.4
Other	225	22.0
Having biological children
No biological children	323	31.5
Have biological children	702	68.5
Race/ethnicity
Black/multiracial	13	1.2
White, non-Hispanic	945	92.2
Hispanic	31	3.1
Other	36	3.5

Abbreviation: SD, standard deviation.

^aTypes of genetic variants: associated with (1) preventable/treatable diseases; (2) non-preventable/non-treatable diseases; (3) response to medication; (4) affect children; (5) affect relatives; (6) had uncertain meaning; (7) not related to health.

TABLE 2.

Multinomial logistic regression model to predict choosing actionable, carrier, non-actionable genetic results as top choice (Reference: Non-Actionable; n = 1,025)

Reference: non-actionable	Actionable		Carrier
	RRR	95% CI	RRR	95% CI
aGenetic causal belief	0.415	0.21–0.81	0.388	0.18–0.85
aPrior genetic testing
Had BRCA1/BRCA2 testing	1.012	0.36–2.86	0.586	0.17–2.09
aHaving biological children
Have biological children	0.816	0.26–2.52	9.084	3.56–23.18
aEducation attainment
Graduate degree or higher	0.689	0.90–3.63	0.696	0.32–1.49
aEmployment status
Part-time	0.816	0.32–1.95	0.757	0.27–2.09
Other	0.640	0.23–1.45	0.453	0.17–1.19
Age	1.026	0.98–1.07	0.987	0.94–1.04

Abbreviations: CI, confidence interval; RRR, relative risk ratio; Bold coefficients indicate significance (p < .05).

Reference: Actionable: ‘preventable/treatable diseases’ or ‘response to medication’ as top choice; Carrier: ‘affect children’ or ‘affect relatives’ as top choice; Non-actionable: ‘non-preventable/non-treatable diseases’ or ‘had uncertain meaning’ or ‘not related to health’ as top choice.

^aCausal belief: Low genetic causal belief (REF); Prior genetic testing: Did not have testing/not sure/had other genetic testing (REF); Having biological children: No biological children (REF); Education attainment: College or less (REF); Employment status: Full-time (REF).

Compared to people with lower genetic causal beliefs, people with higher genetic causal beliefs were more likely to rank learning non-actionable results as their top choice over actionable or carrier status results (RRR = 0.415 [0.21–0.81]; 0.388 [0.18–0.85]).

We found differences in the results based on rating scale (absolute) with those of the ranking scale (relative). Using the Likert scale rating system, participants were able to select all types of genetic results as ‘very interested’, whereas the ranking system forced participants to state a preference order for genetic results. In the rating system analysis, knowledge about sequencing benefits, worry about genetic risks, and health information orientation were found to be significant predictors of interest in learning all types of genome sequencing results. In this analysis, these variables were not significantly associated with the ranked outcome, suggesting that variables affecting being very interested in return of a particular result are different from those affecting prioritization of results.

The results presented here suggest that, among these women diagnosed with breast cancer at a young age, patients with biological children most valued carrier status results as secondary findings, even more than secondary findings of actionable results related to their own health. This is somewhat consistent with previous findings using the interest rating scale, which showed that having biological children was significantly associated with strong interest in learning at least one type of sequencing result (Kaphingst et al., 2018).

In another analysis using rating scale data from this survey, high genetic causal belief was found to be related to strong interest in learning genetic risk information for results associated with preventable and treatable diseases (Hong et al., 2019). In this analysis with the ranking scale, we found that participants with higher genetic causal beliefs were more likely to rank non-actionable results as their top choice for return than either actionable or carrier status results. Taken together, these results indicate that patients with strong genetic causal beliefs may be interested in return of a broad range of possible secondary findings.

These findings should be considered in light of several study limitations. The study population included breast cancer patients who were diagnosed at age 40 or younger, and thus, findings are not generalizable to other patient populations. Moreover, the majority of respondents were Caucasian (92%) with high educational attainment (46% graduate degree), which resulted in limited variability of some covariates (Yu, Crouch, Jamal, Bamshad, & Tabor, 2014). Lastly, though the mandatory order of ranking provided us information on preference for options other than the most preferred, we assessed the preference using the top choice.

This study showed that ranking and rating scales give different and complementary information about patients’ preferences for return of secondary findings generated using genome sequencing. In designing methods of determining patient preferences, genetic counselors may want to consider using both types of items to give a complete picture of preferences. We also found that having biological children was related to having a preference for actionable information, and having higher genetic causal beliefs for breast cancer was related to prioritizing information other than actionable results. Future studies should use both ranking and rating measurements to examine factors associated with preferences for genetic information in other contexts and populations.