Presentation on theme: "General Introduction to health measurement (Note: I have added explanatory notes to many of the slides; to see these you will need to save the file and."— Presentation transcript:
General Introduction to health measurement (Note: I have added explanatory notes to many of the slides; to see these you will need to save the file and open in the normal view mode)
Defining Measurement Measurement consists of rules for assigning numbers to observable attributes so as to represent quantities of the attributes Oh gosh, lets try to sort that out… 1.The attributes can refer to dimensions, properties, characteristics, or behaviors (e.g., weight, density, cost, physical function) 2. If its not observable, its not measurable (but for health, lets define observable broadly: if they say they feel bad, you observed it). 3.Operational definitions indicate how to measure an attribute that is not directly observable (e.g., health, quality of care)
Classification of Health Measures 1.Measures can be classified by their purpose. For example: –Evaluative (e.g., outcome measures) –Diagnostic (e.g., BP, ESR) –Prognostic (e.g., Apgar; screening tests) –Discriminative (e.g., IQ tests) –Summary population measures (e.g., death rates)
Classification (2) 2. Or, measures may be classified descriptively: –Scope of the measure (e.g., specific or generic) –Qualitative vs. quantitative (some discussion needed here!) 3. Or methodologically: –Subjective vs. objective (how is the information collected?) –How administered (questionnaires; clinician ratings; laboratory tests) –Structured vs. semi-structured –How they are scored: indexes vs. profiles
Types of numerical scales Scale Type Example Uses NominalSex; blood type You can count these (how many females?) Ordinal House numbers; mild pain You can compare these ( apply) Interval°C; July 29, 2013 You can calculate absolute differences by adding & subtracting Ratio Weight; BP; # doctor visits You can calculate relative differences: X and ÷ apply Discrete vs. continuous variables: Mnemonic: NOIR
Ways to Present Scores The raw scores –Single index value or profile Norm-referenced: –Z-scores (or other types of standard scores : see next slide) –Percentiles Criterion-referenced: –Pass/fail –Clinical diagnosis
Raw Scores 1-100 scale: e.g., 3MS scores Population distribution (Note: we tend to assume, but dont really know, if the scale points are evenly spaced, as drawn) Maybe the scale spacing should be presented like this, to bell curve it?
Whats an Index? Standard, weighted, composite set of indicators –Weighted means that each element can receive a different salience in the overall score –Composite means theres some way to combine the elements Gives a broad-spectrum indication of overall level of a complex attribute Generally used for broad comparisons –Examples: consumer price index; hospital activity index; Health Utilities Index
Measuring vs. Classifying: some possible distinctions MeasuringClassifying QuantitativeQualitative studies Counting, analysingDescribing; diagnosing ScienceArt; policy Ordinal, interval, ratio scalesNominal, categorical StateTrait? TheoryPractice NomotheticIdiographic EvaluationDiagnosis DividingGrouping Being?Becoming?
Choosing & Applying Measurements Choosing Your purpose drives your choice. –What type: Specific or generic? –Objective or subjective? There are criteria for evaluating & comparing tests Off the peg, or design your own? Where do you get information on a scale? Applying Measures Practical issues: –Interview or self- administered? –Cost & difficulty –How to score it? –Analyzing scores –Interpreting scores
Old ways of administering health questionnaires may not be practical Face-to-face Interview$150 + Telephone Interview $ 50 + Self-administered postal$ 20 + Computerized via Web$ <5? (US $)
Cone of Measurement Demands: How much effort does it require of the respondent? IQ tests, etc. HRQOL. ADLs. EKGs Minimal effort Very demanding: lower response rate?
Practical issue: abbreviated versions Short Forms (same number of items) Measure too low (ceiling effect) but has good discrimination Broad spectrum, but coarse discrimination: may not show changes (broad range + fine discrimination: band width and fidelity) (Source: John Ware, October 2000) The ideal scale
Match the Instrument to the Application 1 2 3 4 1 2 3 4 1 2 3 4 Population Monitoring Outcomes Research Patient Management Source: John Ware, October 2000