Previous we saw how confidence interval is used. Lets look at examples of confidence intervals in public health and HEOR and learn how to interpret CIs critically to make better health and policy decisions.
Smoking Prevalence in a City
A public health survey in City A finds that 22 percent of adults are current smokers, based on a simple random sample of 2,000 people.
The 95% CI is:
This gives a range from 20.19% to 23.81%.The range is narrow, suggesting the estimate is precise. For planning smoking cessation programs, you can be confident the true prevalence is close to 22%. But remember, even with a narrow CI, you should consider whether the survey method reached all parts of the community equally, missing certain groups can bias the estimate.
Vaccination Coverage and Policy Targets
An immunization coverage survey of 1,500 children finds that 91 percent are fully vaccinated.
The 95% CI is:
This gives a range from 89.55% to 92.45%.
If your target is 90% coverage, the lower limit is below the target. That means you cannot claim with full confidence that the target was met, even though the point estimate is above 90%. This is where CI interpretation prevents overconfident conclusions in policy reports.
Average Hospital Stay for Stroke Patients
A hospital study finds that stroke patients have an average stay of 7.8 days, with a standard deviation of 3.2 days, in a sample of 400 patients.
The 95% CI is:
This gives a range from 7.49 days to 8.11 days.
The range is very narrow, meaning you can be confident about average stay length for this hospital. But if you are making regional policy, ask whether this hospital is representative, other hospitals might have longer or shorter stays depending on capacity and patient demographics.
Annual Treatment Cost in HEOR
A health economics study estimates the average annual cost of treating a patient with type 2 diabetes at $8,400, with a standard deviation of $1,200 in a sample of 500 patients.
The 95% CI is:
This gives a range from $8,294.79 to $8,505.21.
The range is extremely narrow, giving strong evidence about the average cost in the sample. For a payer deciding on reimbursement rates, this is valuable. But remember, costs can vary by region, insurance coverage, and care models, the CI reflects only the sampled population, not necessarily the whole country.
Hope you find this useful.