week 12 health measurement Flashcards
WHO definition of health
a state of complete physical, mental, and social well-
being, and not merely the absence of disease or
infirmity
- exposure to something resulting in a deviation of health can then be considered a disease
how do we define health
commonly defined as a state free from illness or disease
what is health-related quality of life
- many use the WHO definition
- an agreement that premature mortality is undesirable, so an aspect of health should be avoidance of death
- encompasses physical, mental, and social well-being, not just the absence of illness or disease
- involved more than just avoiding diseases
- maximizing HRQL becomes critical for enjoyment of long life
psychometric approach
attempts to provide a separate measure for each dimension of HRQL
decision theory
weighs the different dimensions to provide a single expression of health status
- argues each dimension of health is not equally of importance
*combination of both decision and psychometric approaches are used
health measurement
All principles of health measurement apply the same, regardless of the topic!
health surveys
estimation of health (prevalence, incidence, risks) are routinely done using health surveys
- fatal diseases are collected on health via hospital records
- health surveys can be specific (ex. depression) or can be broad (community health )
what types of items are usually on health surveys
- Dichotomous (Yes/No)
- Likert-like (“In general, how do you rate your health?” –
“Excellent / Very Good / Good / Fair / Poor”
health assessment
dependent upon reliable and valid measures
in health, what is equally important, is the reliability and
validity in the classification of individuals
what validity is most important for health measures
criterion validity
medicine using strict categorization
- traditionally thought in terms of diagnoses and treatment
- if a patient has a disorder they are prescribed a treatment or not
example: depression
each symptom in the DSM-5 has its own criterion (threshold) for if they symptom is present
Under the categorical approach, “normal” people may have measurable levels of depression but it is ignored
- dimensional approach: a person can show severity in unique patterns over time
what are the two approaches
- categorical model
- dimensional model
measurement error
- mis measurement is present to some degree in every single health assessment
- no measure can be perfect
- diff tools to measure reliability and validity for classification
concordance
concordance measures the percentage of agreement between two sources
categorical variables (diseased/non diseased)
example: Use CES-D at a specific cut-off score to
classify as clinically depressed, compare findings to a diagnosis
- is the screener as reliable as the diagnosis
variance in the responses = reliability
- does not necessarily need to be against a “true” diagnosis
- seeing how often the pairs match values
Crude measure of reliability
- does not adjust for chance
Kappa
most widely used measure of reliability in classification
- designed to adjust for chance
- the amount by which the observed concordance exceeds what is expected by chance
- dependent on prevalence
Po is the observed concordance
Pe is the concordance by chance
- normally they fall between 0-1, become negative when observed is less than expected
difference bw concordance and Kappa
concordance measures percentage of agreement bw two sources - does not account for agreements that happen by chance
Kappa - more advanced that adjusts for chance measurements
pearson & spearman correlation coefficient
Correlation coefficients measure the degree pairs of observations fall on a straight line (ranging from -1 to 1)
Pearson correlation is for continuous variables while the
Spearman rank correlation is for ordinal variables
what is Intraclass Correlation Coefficient (ICC)
accesses the reliability of continuous measurements by analyzing the variance in scores
between person variance: how much people differ overall
within person variance: how much scores vary for the same person (ex. how diff they respond to the 2 surveys)
What does it mean if test has ICC=1?
test = PERFECTLY RELIABLE
- all variability in scores is due to differences bw individuals, no inconsistencies in how individuals are measured
What does it mean if a test has ICC=0?
NO RELIABILITY
- test scores are random and do not reflect consistent differences between individuals
validity in health
reliant on criterion validity
primary tools for validity: sensitivity and specificity
In validity, we break apart this agreement into two components: positive agreement and negative agreement
what is sensitivity
is the conditional probability that a true case is detected
The ability of a test to correctly identify true positives (e.g., detecting all people who have a disease).
High sensitivity = fewer false negatives.
what is specificity
the ability of a test to correctly identify true negatives (e.g. people who do not have the disease
High specificity = fewer false positives
should a test have both sensitivity and specificity?
High sensitivity and high specificity is desirable, but unrealistic
- The two cannot be considered in isolation when considering
validity - sensitivity and specificity of two measures can be compared
in order to help decisions on measure selection