![](../Graphics/IntroTitle.gif) |
|
EPI9-6: THE QUALITY OF MEASUREMENTS: Precision Or Random Variation And Bias Or Systematic Variation |
OBJECTIVES |
At the end of this session you should be able to critically interpret findings in the literature. |
CHECKLIST FOR CRITICAL APPRAISAL OF EPIDEMIOLOGIC STUDIES:
Title - appropriate.
Abstract/summary - adequate and not misleading.
Literature review - adequacy, appropriate and relevant research problem.
Motivation
- Problem - what is it?
- Purpose - Why is it being done?
- Are there realistic implementation objectives?
Aim -Concisely stated.
Objectives - attainable and precisely defined, do they relate to aim?
Methods
- Definition of terms and variables adequate.
- Study Design clearly stated, appropriate and efficient.
- Study population adequately defined (tree diagram).
- Sampling of study group
- sample size adequate for power and significance;
- index subjects;
- referents;
- ratio of index/referents;
- good definition of index and reference subjects;
- sample representative (covers) the scope of study or target population;
- selection biases arising from source of index and referent subjects;
- diagnostic bias;
- self selection of the ill;
- healthy worker effect;
- sick community effect;
- confounding bias likely and measures taken for this;
- sample randomly selected or some other method;
- matching present and adequate;
- response rate acceptable and attempts to compare responders and
nonresponders;
- Measurements
- blindedness;
- quality control;
- precision/reliability/repeatability, any checks;
- sample size adequate;
- observers adequately trained;
- intraobserver difference;
- validity any checks made;
- misclassification considered;
- standardisation of procedures;
- recall bias considered/likely;
- calibration of equipment & laboratory QC;
- validation against other criteria or sources;
- sensitivity;
- specificity;
- predictive value - positive;
- negative;
- interobserver differences analysed, concordance, bias, kappa;
- data reading errors checked;
- Biological/subject variation controlled.
- Piloting
- pretesting of measurement instruments for appropriateness, precision and validity;
- implementation/data collection;
- ongoing quality control for precision and validity, missing or lost data, dropout and death.
- Logistics
- epidemiologic efficiency;
- cost efficiency.
- Statistical methods
- appropriate for data;
- appropriate for study design.
- Resources required
- Nature of the sponsors - interested parties or disinterested parties.
- Ethics
- informed consent for participation of subjects;
- feedback to individuals;
- referral for treatment or follow-up;
- feedback to interested parties;
- stakeholder preview for comment;
- release of scientific information;
- release of information to general public;
- other implementation objectives;
- agreements between researchers;
- stakeholder contracts.
Results:
- sufficient data description;
- population & study group description initially and after exclusions;
- univariate and bivariate statistics;
- computational mistakes;
- arithmetical mistakes;
- adequate treatment of variables;
- causes;
- confounders;
- effect modifiers;
- intermediate variables;
- effects;
- appropriate crude measures to estimate parameters;
- point and interval estimates;
- appropriate crude measures to estimate associations point and interval estimates;
- adjustments made for appropriate summary measures;
- confounding controlled for;
- interaction investigated;
- appropriate statistical testing and presentation of test results in terms of:
- point estimates of parameters;
- specified interval estimates of parameters;
- test statistic values;
- p-values for test statistics.
Discussion
- Awareness of limitations in methods and results especially biases:
- likely causes of negative studies;
- small sample size;
- nondifferential misclassification;
- inappropriate comparison groups;
- likely causes of false positive studies;
- differential misclassification;
- other biases above.
- Reasonableness of interpretation of own and other findings in literature.
- Application of guidelines for assessing causality:
- strength of association;
- consistency;
- specificity;
- time ordering;
- biologic gradient or dose-response;
- biologic plausibility;
- experimental evidence;
- analogy;
- coherence of above evidence.
- Generalizability of findings or External validity.
- Speculative comment identified as such.
- Implementation - appropriate or inappropriate recommendations
- clarity of implications;
- efforts made or suggested.>/li>
- Publication bias
- content area for negative or positive findings;
- nature of publication outlet - overseas, local.
CRITICAL APPRAISAL / CAUSAL ANALYSIS - SHORT CHECKLIST:
Application of rules of evidence to a study to assess the validity of the data, completeness of reporting, methods and procedures, conclusions, compliance with ethical standards, and so forth. The questions below are sequential.
- What were the findings?
- Are the findings internally valid?
- Are there non-causal explanations?
- Are there causal explanations?
- Are the findings externally valid?
- Was the study ethical?
1. What were the findings?
- What was the Population studied?
- What was the research Question?
- What were the Results?
- What was the Study type?
2. Are the findings internally valid?
- Are there non-causal explanations?
- Exposure/Effect: Was this determined equally for all groups?
- Disease/Determinants: Was this determined equally for all groups?
- Confounding.
- Bias Selection and Information bias.
- Analysis: Correct analyses? Type 1, Type 2 errors? Chance?
Are there causal explanations?
C.A.U.S.E.R.:
- Credible biologically; Coherence.
- Association strength.
- Ubiquitous/Universal.
- Sequence; Specificity.
- Exposure-effect gradient.
- Reversibility.
3. Are the findings externally valid?
Do they apply to:
- Eligible population.
- Extraction (source) population.
- External (target) populations.
4. Was the study ethical?
Internal and external ethics:
- Non-maleficence;
- Autonomy;
- Beneficence;
- Justice;
- Utility.
![](../Graphics/Line.GIF)
![](../Graphics/previous.gif)