Radulescu et al listed criteria for evaluating a prevalence survey. This can help determine how much reliance can be placed on the findings of the study.
Key criteria:
(1) target population specified
(2) sampling method adequate (random sample vs whole population)
(3) sample size adequate
(4) response rate adequate
(5) information provided on non-responders
(6) disease definition valid and repeatable
(7) way to minimize observer bias
Additional criteria:
(1) inclusion criteria specified
(2) information provided on individuals in the study
(3) valid instruments used
(4) epidemiologic terms used correctly
(5) confidence intervals or standard errors calculated
Khumalo et al supplied an additional measure:
(1) percent of participants lost to followup acceptable
I would add an additional measure:
(1) exclusion criteria specified (in specifying the target population the exclusion of an ethnic group is mentioned, but not other exclusions)
Criteria |
Adequate Response |
target population |
population defined, with rationale for selection and explanation for why groups were excluded |
sampling method |
random |
sample size |
whole population ideal; else as large a sample as possible after subtraction of nonresponders |
response rate |
>= 70% |
non-responders |
followup to characterize and determine reason for non-participation |
disease definitions |
valid and repeatable, either for several observers or a single observer |
minimize observer bias |
attempt to minimize bias (training, standardized methods, blinding, comparison of results if multiple observers, no evidence of an interest in a particular outcome) |
inclusion criteria |
adequately specified |
information on participants |
age range, gender distribution, ethnic group, etc. |
instruments employed |
valid |
epidemiologic terms |
used correctly |
confidence intervals or standard error |
narrow 95% confidence interval and/or small sample error to indicate precision |
percent lost to followup |
< 20% |
exclusion criteria |
adequately specified |
Scoring can be done as 0 if absent or 1 if present. For the implementation I used:
Criteria Met |
Points |
No |
0 |
partial |
0.5 |
Yes |
1 |
total score =
= SUM(points for all 14 items)
Interpretation:
• minimum score: 0
• maximum score: 14
• The ideal survey should have a score of 14.
ICD-10: ,