Presentation on theme: "Understanding Patient Preferences in Individual Health Decisions: Conceptual and measurement considerations Sara J. Knight, PhD Interdisciplinary Program."— Presentation transcript:
Understanding Patient Preferences in Individual Health Decisions: Conceptual and measurement considerations Sara J. Knight, PhD Interdisciplinary Program to Improve Care for Veterans with Complex Comorbid Conditions
Goals Examine several conceptual models for understanding preferences Review approaches to the measurement of patient treatment preferences Discuss qualitative strategies for refining conceptual models of patient treatment preferences and health outcomes Identify types of studies used to provide evidence for the construct validity of measures of patient values
Characteristics of a Strong Measure of Patient Preferences Comprehensive in patient considerations Strong in psychometric properties Feasible in a busy clinic settings Appropriate for all educational backgrounds Related to health care recommendations and outcomes
What are possible contributions of improved patient preference assessment? Clarification of individual differences in decision making Understanding of how preferences may be constructed over time Improved patient-centered care Strengthened patient/physician relationship
What are preferences? Valuation of health care goods, services, and interventions Economic Theory Maximize subjective expected utility Utilities can be scaled in dollar-equivalence terms Based on assumptions of consistency and rationality with some limits
Preferences versus Values Preferences Economic theory Utilities Values Psychological theory Attitudes, preferences, interests, goals, needs
Preferences versus Health Status Health Status Patient experience, well-being, function Concurrent or retrospective Preferences Patient, consumer, or societal values and goals Prospective
Preferences Stated Choice Revealed Choice Demand Utilization Information Resources and Constraints U 1 = u(Attrib 11, Attrib 12, …, Attrib 1n ) Utility of Choice Alternatives U 2 = u(Attrib21, Attrib 22, …, Attrib 2n ) U 3 = u(Attrib 31, Attrib 32, …, Attrib 3n ) Economic Model
Another Economic Model: Adapted from McFadden Memory Motivation/Affect AttitudesPreferences Perceptions/Beliefs Process Experience Information Stated Perceptions Choice (Revealed Preferences ) Stated Preferences Attitude Scales Time and Dollar Constraints
Individual Decision Making International Patient Decision Aid Standards (IPDAS) Collaboration Recommendations on Values Measurement Describe procedures and outcomes to help patients imagine what it is like to experience their physical, emotional, and social effects Ask patients to consider which positive and negative features matter most Suggest ways for patients to share what matters most with others
IPDAS Effectiveness Recommendations: Does the patient decision aid ensure decision making is informed and values based? Recognize a decision needs to be made Know choice alternatives and their features Understand that values affect decisions Be clear about option features that matter most Discuss values with their practitioner Become involved in preferred ways
Individual Decision Making Model Information Attribute Value Attribute Value Attribute Value Attribute Value Attribute Value Physician Recommendation Treatment Choice Patient Preferences Decision Quality Knowledge, Satisfaction, Conflict, Regret Treatment Outcomes Functional Status, Quality of Life, Symptoms Preference/Choice Concordance Experience/Affect
Methods of Measurement Utilities Elicitation Standard Gamble Time Trade-Off Visual Analog Scale Rating Scales Attitude Questions Conjoint Analysis Combination Approaches
Utilities Value between 0 and 1, where 0 represents the value of being dead and 1 the value of living with perfect health Represents patient’s subjective value for choice attribute, such as a health state or a treatment characteristic
Example: Standard Gamble
Example: Time Trade-Off Imagine that you have two friends, Mr. Smith and Mr. Jones. Imagine that Mr. Smith’s health fits the following description; that it will stay the same for the rest of his life; and that he will live about 10 more years.
Time Trade-Off: The Health State Mr. Smith has mild difficulties or problems with urinating or bowel function. He is able to do most of his usual activities nearly all of the time. He is not overly tired and his energy level is pretty good. He usually has a good appetite. He has very little or no pain and it is easily controlled by medication. His ability to have sex and enjoy if has been mildly affected. He hardly ever feels tense, worried, irritable, sad, fearful or depressed.
Time Trade-Off: The Choice Now imagine that your other friend, Mr. Jones, lives in perfect health, but will live somewhat less than ten years. If you had to be one of these 2 people, who would you rather be? Mark your answer below. 10 years as Mr. Smith in health state A or 5 years as Mr. Jones in perfect health?
Example: Visual Analog Scale
Advantages of Utilities Based on economic and psychological theory; meets assumptions of normative decision making Experimental task involving a choice Scores comparable across methods Useful in studies of health economics and policy
Limitations of Utilities Questionable reliability Logical inconsistencies within methods Limited evidence for validity Inconsistencies in scores across methods Not well accepted by patients Difficult for patients with low literacy and low numeracy Not related to treatment recommendations Souchek et al., 2000
Attitude Measures Attitudes are defined as a psychological tendency that is expressed by evaluating a particular entity with some degree of favor or disfavor Usual measurement is through ratings or rankings Eagly and Chaiken, 1996
Example: Attitude Measure If you were forced to choose between the term “faulty gene” and “altered gene”, which would you prefer? Please circle your response on the scale below: 1= I definitely prefer “altered gene” 2= I prefer “altered gene” 3= I am happy with either term 4= I prefer “faulty gene” 5= I definitely prefer “faulty gene” Wakefield et al., 2007
Evaluation of Attitude Measures Requires few assumptions about underlying choice processes Easy for participants Holistic approach Valuation of an attribute as a whole Does not allow evaluation of levels or different categories of an attribute Vulnerable to “halo” effects Phillips, Johnson, Maddala, 2002
Conjoint Analysis Allows attributes to be evaluated in relation to each other, or “conjointly” Yields utilities Assumptions Each good or service is a bundle of potential attributes Each individual has set of unique relative utilities weights for attribute levels Combining utilities for difference attributes provides an individuals overall relative utility Phillips, Johnson, Maddala, 2002
Conjoint Analysis ATTRIBUTESTEST ATEST B Accuracy The diagnosis is correct 80% of the time, 20% of the time it is incorrect The diagnosis is correct almost 100% of the time Privacy You, your doctor, and your insurance company will receive information about your genetics Only you will receive information about your genetics Cost $0$500 Would you choose Choice A or B?
Evaluation of Conjoint Analysis Allows the evaluation of each attribute separately Provides estimation of willingness to pay; standardized approach to quantifying preferences for economic evaluation Complex and difficult for respondents Vulnerable to simplification of judgments Possible inconsistent responses Phillips, Johnson, Maddala, 2002
Integrative Approaches Bockenholt has suggested that using a nested approach to understanding value judgments Comparative judgments are integrated with absolute judgments Scores contain information on both relative value and scale origin
Importance Ratings How important is… Extremely Important Very Important Moderately Important A Little Important Not Important Avoiding a treatment that causes problems for my relationship with my spouse/ partner
Best Worst Scaling Circle the concern that is most important to you in making your decision about your prostate cancer treatment: Urinary Function Responsibilities Survival
Developing Measures of Preferences Step 1 Identify a priori constructs Step 2 Establish comprehensive range of constructs and exemplars for each Step 3 Develop representative content for each construct Step 4 Generate items for each content area (items and subscales) Step 5 Select appropriate response choices Step 6 Refine measure based on pilot testing Step 7 Test measure properties
Construct Development Ideally should be based on the purpose of the effort to understand preferences Economic analysis? Individual decision making? Cost analyses may require a different conceptual framework for understanding preferences and values and a different method for assessment than needed for the measurement of patient preferences in shared decision making
Literature Review Descriptions of treatment or test alternatives Existing guidelines for care Patient education material Case studies Outcomes research Cost and utilization studies
Focus Groups Used to generate broad range of constructs Often 8 to10 participants; fewer can be justified Experienced moderator needed Stratify on key variables: treatment choice, ethnicity, gender Structured exercises Differences among constructs Rankings among constructs Characteristics that influence choice
Analytic Approach Content analysis using a priori constructs as a starting point Grounded theory approach may be used where an a priori model does not exist Software NVivo Ethnograph
Cognitive Interviews Think aloud Well, it depends on what you mean by family responsibilities. Do you mean money? Helping my wife with my grandkids? My wife says she’s OK with any treatment as long as I’m still alive.
Psychometric measures versus utilities? Psychometric measures of attitudes yield scores interpreted from a population reference point and utility elicitation yields judgments scaled in terms of an absolute reference point (death, perfect health) Psychometric concepts of content validity and predictive validity may not be applicable to utilities where construct-irrelevant variance and construct under representation may be more relevant concepts Lenert and Kaplan, 2000
Evaluating Measurement Properties: Accumulating Evidence Reliability Internal Consistency Test-retest Validity Face Content Criterion Construct Construct irrelevant variance Construct under- representation Consistency within Consistency over time Correct concept Adequate coverage of concept External criterion Construct being measured Minimal spurious influence Representative constructs
Reliability: Internal Consistency InterpretationThe consistency of responses within a scale or within a subscale Statistical MethodCronbach’s alpha (α) is the most common method. It is based on the average correlation over items weighted by variances InterpretationIt is the percent of variance that the current scale explains of the unidimensional underlying construct being measured Acceptable valuesValues range from with higher scores reflecting more accurate measure of the underlying trait. Values below.70 suggest a need for evaluation of each item within the scale, values above.90 for use of a measure with individuals.
Reliability: Test-Retest InterpretationThe stability in responses across assessment occasions Experimental MethodThe same instrument is administered twice to the same sample on two occasions usually 1-2 weeks or up to one month apart Statistical MethodIntraclass reliability coefficient for continuous valued measures, kappa for discrete valued measures Acceptable valuesValues range from with higher scores reflecting more accurate measure of the underlying trait. Values below.60 suggest low consistency and potential future difficulty in detecting differences due to real effects
Validity: Face InterpretationDetermination that the instrument is measuring the construct of interest Assessment MethodPrior to development, focus groups can inform the content. After the development of the instrument, cognitive interviews can be used to assess understanding of the content of the questionnaire Statistical MethodNo specific statistical method is associated with documenting face validity Population considerations The face validity of an instrument may vary across patient populations
Validity: Content InterpretationDetermining the adequacy of coverage of the instrument Experimental MethodExperts in the area judge whether the instrument captures all domains of the construct of interest Statistical MethodThough rarely reported, the Content Validity Index or the Content Validity Ratio indicates the extent of expert agreement. Determination of Experts Choose a panel of experts that informs rather than limits your interpretation of the breadth of your construct.
Validity: Criterion InterpretationThe association between the construct and some external criterion Experimental Methods Concurrent validity: both validation and construct tools are administered on one occasion. Predictive validity: construct is measured first; criterion is measured later Statistical Method Tests of association are used to assess construct validity such as a Pearson r for continuous valued scale scores, or Chi-square for categorical scale scores. Acceptable valuescoefficients of.70 or greater
Validity: Construct InterpretationEstablishes the ability of the instrument to measure the construct and to distinguish varying levels of the presence of that construct Experimental Methods 1) Collection of data from many subjects from the patient population of interest to determine latent trait(s) underlying the assessment instrument. 2) Collection of data from two groups known to vary on the construct (known-groups technique). 3) Multi-Trait Multi-Method—measure should be more strongly related to other measures of similar constructs and to measures of dissimilar constructs Statistical Methods 1) A statistical method to explain variability in responses (e.g. factor analysis, component analysis). 2) A statistical method to detect differences between groups (e.g. t-test)