Measurement Wu Gong, MS, MD Wu.gong@Vanderbilt.edu Department of Biostatistics Vanderbilt University Medical Center
Summary Considerations in selecting outcome measurements Type of outcome measurements
Why Measurement? Measurement supplies numbers used in analysis. Outcome variable, response variable, dependent variable Experimental variable, predictor, descriptor variable, independent variable Adjustment variable, confounder
Conceptual Models of Health Outcomes Health outcomes: medical conditions, impact on health-related or general quality of life, and resource utilization People of interest: patients, health care providers, and other decision- makers. Wilson and Cleary Conceptual Model: biomedical patient outcomes and measures of health-related quality of life. Economic, Clinical, Humanistic Outcomes Model: incorporating costs and economic outcomes.
Wilson and Cleary Conceptual Models Level Health Concepts Biological and physiological factors Genetic and molecular factors Symptoms Physical, psychosocial, emotional, and psychological symptoms Functional Status Physical, social, role, psychological, etc. General health perceptions Subjective rating of general health Overall quality of life Summary measure of quality of life
Economic, Clinical, Humanistic Outcomes Model
Properties of Outcome Measurement Reliability: the measure remains unchanged upon test and retest or across different interviewers or assessors Validity: the degree to which a measure assesses what it is intended to measure. Variability: the distribution of values associated with an outcome measure in the population of interest, with a broader distribution or range of values meaning more variability. Responsiveness: the ability of a measure to detect change in an individual over time.
Clinical Outcomes Prevent the occurrence of undesirable outcomes Delay disease progression Accelerate recovery or improve survival from disease Reduce the burden of chronic diseases
Temporal Aspects of Clinical Outcomes Incident: a first or new diagnosis of the condition of interest Prevalent: existing disease Recurrent: new occurrence of disease in a patient who has a previous diagnosis.
Subjective Versus Objective Assessment Laboratory tests: are not subject to individual interpretation, and are likely to be reliable measured across patients by different health care providers and over time. All-cause mortality: not subject to interpretation by individual health care providers. Depression: subject to interpretation. Type of Clinical Outcome Assessments: Patient-reported outcome assessment, observer-reported outcome assessment, and clinician-reported outcome assessment.
Humanistic Outcomes Health-related quality of life measures the impact of disease and treatment on the lives of patients and is defined as the capacity to perform the usual daily activities for a person’s age and major social life. It commonly includes physical functioning, psychological wellbeing, and social role functioning. Patient-reported outcomes are measurements based on reports that come directly from the patient about the status of the patient’s health condition without amendment or interpretation of the patient’s response by a clinician or anyone else.
Economic and Utilization Outcomes It measures health recourse utilization and represents the payer and societal perspective. Monetary Costs: direct cost of the medical treatment, and indirect cost caused by disability. Health resource utilization: number of inpatient or outpatient visits, total days of hospitalization, number of days treated with IV antibiotics, etc.
Measurement Scales How well the variable is designed to measure the phenomena of intereste? The goal is to minimize the measurement error Precise: free of random error Accurate: free of systematic error
Type of Measurement Numerical variables Categorical Vairables Continuous Discrete Categorical Vairables Dichotomous Nominal Ordinal
Continuous Variable It quantifies how much on an infinite scale. It has an unit and can take on any value on a range of same values. The precision in measuring the continuous variable is often limited by the instrument. Body weight in pound Height in inches Number of days to recover CD+ cell count Descriptive statistics: mean, standard deviation.
Discrete Variable It quantifies how many on a scale with fixed units, usually integers. It takes on a countable number of values. How many times a woman has been pregnant? Number of children Number of asthma attacks per week Number of antiretroviral therapy medicine pickups per year Number of outpatient visits per year
Dichotomous Variable It is also called binary variable and has only two possible values. It is often coded as zero and one. Dead or alive Improved or not improved Completed or failed Descriptive statistics: count, proportion.
Ordinal Variable The ordinal variable has an order such as severe, moderate, and mild pain. It does not specify a numerical difference between one category and the next, and has less information than that of discrete or continuous numeric variable. The quantitative differences between the categories are uneven. Likert Scale: Like, Like Somewhat, Neutral, Dislike Somewhat, Dislike Educational Level: None, Elementary, High School, College, etc.
Nominal Variable There is no natural ordering among the categories. There is no distance measure between two values. Gender: male and female Marital status: single, married, divorced, widowed Blood Type: A, B, O, AB
Repeated Measure Single time point Two time points for each patient before and after treatment Measured repeatedly at regular or varying time points
References Hulley, Stephen B., et al. Designing Clinical Research http://www.bjcancer.org/Sites/Uploaded/File/2016/11/30636160943115667 8937582961.pdf Velentgas, Priscilla, et al. Developing a Protocol for Observational Comparative Effectiveness Research: A User's Guide https://www.ncbi.nlm.nih.gov/books/NBK126190/?report=reader