1 / 99

Measurements

Measurements. Measure of Variability, Scale Levels of Measurements, Descriptive Statistics, Measures of Central Tendency. Measurements need to:. Produce valid and reliable results be sensitive and specific be able to identify clinically important changes

astrid
Download Presentation

Measurements

An Image/Link below is provided (as is) to download presentation Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript


  1. Measurements Measure of Variability, Scale Levels of Measurements, Descriptive Statistics, Measures of Central Tendency

  2. Measurements need to: • Produce valid and reliable results • be sensitive and specific • be able to identify clinically important changes • have outcome measures and endpoints defined • be easy to interpret

  3. Reasons for errors in measurement: • Improper function or calibration of equipment • patients providing misleading or dishonest answers to verbal/written questions • Improper recording/transcribing of data • Investigators recording or making inaccurate measurements

  4. Types of Errors • Random error • Random in occurrence, often balancing out over course of study • mean or average of measurements still close to true value • Large patient size reduces random error

  5. Types of Errors • Systematic error • represents bias in measurements and does not tend to balance out over course of study. • Bias can be knowingly or unknowingly • Good study design minimizes systematic error.

  6. Measurement Terms • Validity- degree to which an instrument is measuring what it is intended to measure. • Predictive, Criterion, Face • Reliability- reproducibility of a test • Sensitivity- ability to measure a small treatment effect • Specificity- how well the test can differentiate between the effect resulting from treatment and random variation

  7. Validity terms used in association with measurements: • Predictive validity: • the extent to which a measurement or test actually reflects or predicts the true condition. • Criterion (construct) validity: • the degree to which a measurement or test agrees with or obtains the same results as other proven tests designed to measure the same. • Face validity • the extent to which a measure appears reasonable or sensible for measuring a desired outcome

  8. Reasons for False Positive Results • Patient related • patients weren’t as ill as originally believed, and drug was more effective in mildly ill pts. • Patients were much more ill than originally believed, and drug was more effective in severely ill patients. • A few patients had a very large response, which skewed the overall results. • Patients gradually improved independent of drug treatment.

  9. False Positive Results • Patient related • More medicine was absorbed than anticipated • Patients took excess medication. • Patients felt pressure to report a positive medicine effect • Concomitant non drug therapy or other drug therapy improved results

  10. False Positive Results • Study Design and Drug Related • Blinding was broken or ineffective • open label study can sometimes produce a larger positive response • no placebo control to help interpret • error occurred in dosing patients- gave more drug than intended • inadequate wash-out period, carry over effect • inappropriate clinical endpoints, tests or parameters were used

  11. False Positive Results • Investigator related • influenced response by great enthusiasm • chose inappropriate tests to measure • Results and Data Related • systematic error- reporting large drug effect • high percentage of non-responders dropped out • not all data was analyzed

  12. False Negative Results • Patient Related • were much more ill than realized • responded less to the drug than anticipated • study group had large number of non responders • non-compliance-- took fewer doses • concomitant medicines- interactions • exposed to conditions that interfered with study

  13. False Negative Results • Drug Related • not adequately absorbed • kinetics were different in study group than in other patient groups • Study Design Related • Too few of patients • inappropriate study design • insufficient drug dose was tested

  14. False Negative Results • Study Design Related (cont.) • Ineffective tests or parameters used • Inadequate wash-out period in previous treatment period • Concomitant non-drug therapy interfered • Investigator Related • influenced patients with skepticism displayed • chose inappropriate tests to measure effects

  15. False Negative Results • Results and Data Related • Patients who improved dropped out leaving higher number of non-responders • systematic error resulted in reporting of an inappropriately small drug effect.

  16. Outcome Measures • Example: A study is performed to compare the effects two antihypertensives, atenolol and propranolol in 2 groups of patients with mild high blood pressure. 2 types of outcomes measurements are selected for this study: measures of efficacy and measures of safety • Measures of efficacy: BP, HR, symptom relief • Measures of safety: adverse effects, blood glucose, electrolytes, serum lipids

  17. Criteria Used for Outcome Measures • Presence or Absence criteria: Is sign, symptom present or absent? • Graded or Scaled Criteria: the use of grading on a scale to measure clinical symptoms • Relative change criteria- measured changes • Global assessment criteria- Quality of Life • Relative effect criteria- change in time to effect.

  18. Measurement Endpoints • Endpoints are measurable points used to statistically interpret the validity of a study. • Valid studies have appropriate endpoints. • Endpoints should be specified prior to start of study (should be included in study design) • Quality studies have simple, few and objective endpoints.

  19. Endpoints • Objective- based on actual or measurable findings or events (heart rate, BP, Temp.) • Subjective- based on thoughts, feelings, emotions (pain scale, mobility) • Morbidity- quality or condition at the present-- quality of life • Mortality- causing death or a death rate

  20. Endpoints Example • In a study determining the effects of clonidine on quality of life, the researchers determine the number of days a patient misses work. Each patient is also asked to complete a rating scale to describe the degree of fatigue they experience. • What type of endpoints are used? • What type of criteria are used?

  21. Surrogate Endpoints • These reduce the quality and validity of the study. • Surrogate or Substitute endpoint examples: • CD4/CD8 ratios instead of “survival” in studies for treatment of AID’s. • Measuring volume of acne instead of proportion of patient’s cleared of acne. • Determining cardiovascular disease or atherosclerotic disease instead of measuring blood pressure in a study of antihypertensive drug treatment

  22. Hawthorne Effect • Refers to the influence that a process of conducting a study may have on a subject’s behavior • Subject • Environment • Research design

  23. Reasons for Clinical Improvement in a Patient’s Condition • Natural regression to the mean (most acute and some chronic conditions resolve on their own • Specific effects of treatment (drug or intervention) • Non-specific effects- attributable to factors other than specific drug/intervention effect. • Called a Placebo Effect

  24. Placebo Effect • A placebo is an intervention designed to simulate medical therapy, but not believed to be a specific therapy for the target condition. • A placebo is used either for it’s psychological effect or to eliminate observe bias. • Placebo “response”= due to change in pt. Behavior following admin. of a placebo • Placebo “effect” = change in pt’s illness due to the symbolic importance of a treatment. • A placebo effect doesn’t require a placebo.

  25. Why do we see a Placebo Effect? • Three different theories: • 1. The effect is produced by a decrease in anxiety • 2. Expectations lead to a cognitive readjustment of appropriate behavior. • 3. The effect is a classical conditioned Pavlovian response.

  26. Placebo Effect • Expectations lead to behavior change • Patient’s and providers expectations • Patient’s positive attitude toward provider and treatment • Providers positive attitude toward therapy • Provider interest in patient (sympathy, time, positive attitude) • Compliant patients have better outcomes than noncompliant patients even with a placebo. • The placebo response is stronger when stronger drugs are used. • Crossover studies show a stronger placebo response when given in the 2nd period of study.

  27. Appropriate Statistical Tests • To determine whether appropriate statistical tests have been used, you must know 3 things: • 1. The specific research question or hypothesis being addressed. • The number of independent and dependent variables • The scales or levels of measurement used for the dependent variables

  28. Variables in a Study: • Dependent variables: • those variables whose value depends upon or is influenced by another variable. • It is the variable that is measured, and the one that changes as the result of a drug action. • Independent variables • Those variables which modify a dependent variable (drug treatment)

  29. Example • Patients given Lovastatin to lower cholesterol. • Dependent variable- lowering of cholesterol • Independent variable- Lovastatin • There can be more than one independent and dependent variable in a study.

  30. Dependent/Independent Variables • Example: A single blind study of 30 patients with poison ivy dermatitis were randomized to receive either topical hydrocortisone 1% or 2% and apply QID. Severity of the dermatitis was evaluated daily using a 5 point scale, where 5- severe and 0-none. • What is the independent variable? Dependent variable? • Example: A study was conducted to compare the efficacy of procainamide and quinidine for reducing ventricular arrhythmias. The number of ventricular ectopic depolarizations was determined in patients both before and during therapy with either drug. • What is the independent variable? Dependent variable?

  31. Scales of Levels of Measurement • Nominal Level • variables are grouped into mutually exclusive categories. • Gender as female or male • cured and not cured • response and no response • include histograms (bar graphs) • weakest level of measurement • referred to as dichotomous data

  32. Scales of Levels of Measurement • Ordinal level • ranked or ordered categories • 1-2-3-4 • severe, moderate, mild, none • always, sometimes, never • stronger level than nominal • not measured quantitatively, but qualitatively • distance between groups need not be equal

  33. Scale Levels of Measurement • Continuous Measurement • Interval level: exact difference between two measurements is known and constant • has arbitrary zero point • highest level of measurement • quantitative data • Examples: BP (mm Hg) serum Theo levels (ug/ml), WBC (cells/cu mm)

  34. Continuous Level of Measurement • Ratio level: • exact differences between measurements is known and constant • true zero point (Centigrade temp scale) • can make ratio statements (2:1) that denote relative size • Can be converted to an ordinal scale (but ordinal scale can’t convert to interval)

  35. Scale levels of Measurements Baseline Pain Assessment 0 1 2 3 (absent) (mild) (mod) (severe) Placebo 0 2 18 14 PainawayR 0 4 12 16 (number of subjects in each group with varying degrees of baseline pain intensity. What scale level of measurement?

  36. Scale level of Measurements Infectious Outcome Among 46 Patients Infection No infection Total Oxacillin 2 20 22 Placebo 0 24 24 Column total 2 44 46 What scale level of measurement?

  37. Types of Interval/Ratio Data • Discrete scale of data (non-continuous): when a measurement has the interval characteristics but can only be assigned integer values. (HR, number of patients admitted to hospital/day) • Non discrete (continuous) scale of data: each data point falls on a continuum with an infinite number of possible subdivisions (temp, BP, BG, weight)

  38. Data Distributions • Once data is collected, it can be organized into a distribution, or graph of frequency of occurrence, or chart of the number of times that each measurement value occurs. • Bar Graphs • Bar Chart (Histogram) • Line Graphs

  39. Data Distributions • Nominal and Ordinal level data use histograms (Bar charts) because data classified into distinct categories • Continuous level data are distributed in the form of curves and line graphs (normal distributions and non-symmetrical distributions)

  40. Bar Chart (Histogram)

  41. Continuous Level DataNormal DistributionGaussian Curve

  42. Non-Normal DistributionsBi-Modal Curve Weights of American Adults (women and men)

  43. Non-symmetrical distributionsNon-normal distributions

  44. Continuous Distribution Examples: The distribution of GPA’s of college students: 1.0 2.5 4.0

  45. Continuous Distribution Example Distribution of the ages of patients taking Digoxin 20 40 60 80

  46. Descriptive Statistics Measures of Central Tendency

  47. Measures of Central Tendency • Mean- • mathematical average of a set of numbers. • Affected by extreme data points (outliers) • Useful for continuous level data (interval/ratio). • Ex: uric acid concentrations: 8,6,5,4,3,2,2,2. Total number of samples =8. Sum of measurements = 32. 32/8= 4 (mean).

  48. Measures of Central Tendency • Median • “Middle” number of a group of numbers in which an equal number of responses above and below that point exist. (called 50th percentile) • Not affected by outliers. Useful for ordinal, interval and ratio data and non-symmetrical. • Ex: Uric acid concentrations: 8,6,5,4,3,2,2,2. Since even number, median lies between 4 and 3 or median= 3.5.

  49. How to Recognize “skewed” data • If the magnitude of the difference between the mean and median is none or small, the data is approaching normal (symmetrical) distribution. • If the difference between the mean and median is large, the data usually prove to be skewed.

  50. Measures of Central Tendency • Mode • The most commonly or frequently occurring value(s) in a data distribution. • Useful for nominal, ordinal, interval/ratio data. • Only meaningful measure for nominal data. • Can have more than one mode in set of data • Ex: Uric acid concentration: 8,6,5,4,3,2,2,2. The mode = 2.

More Related