Criteria to assess quality of observational studies evaluating the inc

On September 15, 2009, Tatyana Shamliyan made this presentation at the 2009 Annual Conference. Select to access the PowerPoint® presentation (265 KB) (Plugin Software Help).

Slide 1

Criteria to assess quality of observational studies evaluating the incidence, prevalence, and risk factors of chronic diseases

Minnesota EPC
Clinical Epidemiology Methods Centre, Ottawa Health Research Institute, Ottawa
Tufts University Medical Center, Boston
RTI International - University of North Carolina, Chapel Hill
Blue Cross and Blue Shield Association, Chicago
Centers for Disease Control and Prevention, Atlanta
Southern California EPC; RAND Corporation, Santa Monica
German Institute for Quality and Efficiency in Health Care, K�ln, Germany
Johns Hopkins University, Baltimore

Slide 2

Analytical tasks

Analytical tasks included 4 steps:

• Step I- Systematic literature review:
• Step II- Development of two checklists based on the results from literature review
• Step III- Collaboration with the participating experts:
• Step IV- Finalizing the checklists

Slide 3

Therapeutic vs. non therapeutic studies

Slide 4

Systematic literature review

• Systematic reviews and meta-analyses of observational nontherapeutic studies published in English in core clinical journals (145)
• Published tools (scales or checklists) for quality assessment of observational studies (96)
• Annotated bibliography of the methodological literature to identify biases and strategies to reduce bias in observational studies

Slide 5

Development of Checklists

• Separate evaluation of external and internal validity
• Exhaustive range of criteria
• Discriminating reporting and methodological quality
• Predefined major and minor flaws
• Study and hypotheses level analyses
• Automated reports
• Templates for the protocols of quality evaluations with predefined research specific quality standards
• Manual, help files, instructions

Slide 6

Collaboration with participating experts

• Criteria evaluation: face and content validity
• Pilot reliability testing (10 randomly selected articles): overall, by topic, by article, by domains of external and internal validity, and by quality component
• Discriminant validity: testing hypotheses of detected differences in quality and reporting vs. methodological quality

Slide 7

Finalizing checklists

• Checklist for studies of incidence or prevalence of chronic diseases (MORE)
• Checklist for studies of risk factors of chronic diseases (MEVORECH)
• Synthesis of evidence from the studies with major flaws:
  • Level A exclusion from synthesis
  • Level B full review and quality abstraction
  • Level C separate limited synthesis if major flaws detected

Slide 8

Quality criteria

Incidence or prevalence studies

Aim of study
Study design

External Validity

1. Sampling of the subjects by the investigators
2. Assessment of sampling bias
3. Estimation of sampling bias
4. Exclusion rate from the analysis
5. Sampling bias is addressed in the analysis
6. Subject flow

1. Source to measure outcomes
2. Definition of outcomes
3. Measurements of outcomes
4. Outcomes in race, ethnic, age, or gender subpopulations
5. Reporting of outcomes

Risk factors studies

Aim of study
Objectives
Study design

External Validity

1. Sampling of the subjects by the investigators
2. Assessment of sampling bias
3. Estimation of sampling bias
4. Exclusion rate from the analysis
5. Sampling bias is addressed in the analysis
6. Subject flow

1. Source to measure outcomes
2. Definition of outcomes
3. Measurements of outcomes
4. Definition of the exposure
5. Measurements of the exposure
6. Confounding factors
7. Loss of followup
8. Masking of exposure status
9. Statistical analysis
10. Appropriateness of statistical models
11. Reporting of tested hypothesis
12. Precision of the estimates
13. Sample size justification

Slide 9

Major flaws

External validity

• Sampling in clinics for incidence in the general population;
• Sampling of controls from different population as cases;
• Response rate <40%;
• Exclusion of >10% of eligible subjects from the analyses
• Reasons to exclude from the analyses differ for exposed and not exposed

• Severity of the outcome can be relevant but not assessed in the study;
• Non valid methods were obtained to measure the outcomes, exposure, or confounding factors;
• Different methods to measure exposure (risk factors, independent variable) in cases and controls;
• No strategies to reduce bias

Slide 10

Minor Flaws

External validity

• Random sampling restricted to geographic area
• Convenient sampling
• The authors did not assess sampling bias
• Sampling bias was not addressed in the analysis
• Sampling of controls from health care related sources

• Outcomes proxy reported
• Obtained from medical records (data mining) or administrative database
• Inter-methods validation
• Subjective judgment of reliability
• Intensity/dose of exposure can be relevant but not assessed in the study

Slide 11

Example of the report

Slide 12

Conclusions

• We propose two checklists for transparent and standardized quality assessment in systematic reviews and evidence-based reports
• Reasonable validity
• Poor random overall and good component reliability in a pilot testing of the articles from different topics

Slide 13

Future Research with Quality Assessment Using Proposed Checklists

• Protocols of systematic reviews of nontherapeutic observational studies with justified definitions of research specific quality standards and preplanned reliability testing
• Synthesis of evidence incorporating quality assessment (sensitivity of the results to quality)
• The evaluation of the level of evidence from several observational nontherapeutic studies
• Quality assessment of primary studies should improve quality of systematic reviews and evidence-based reports