Fewer Emergency Visits After Diabetes Education? Relative Risk, Odds Ratio and Number Needed to Treat From a Retrospective Cohort in a Rural Clinic, and What the Design Can and Cannot Claim
[Student Name]
University of Phoenix
DNP/701: Biostatistics and Epidemiology
Week 2 Assignment
[Instructor Name]
[Date]
The clinic and its figures are a composite written for a model paper.
Our rural clinic network offers a six-session diabetes self-management education program. Our director wants to know whether it is worth expanding. I used our health record to compare emergency department visits in the 12 months after enrollment between patients who attended at least four sessions and patients with diabetes who were referred but did not attend. This paper presents the design, the measures of association and their interpretation.
Study Design
Grimes and Schulz (2002a) classify clinical research as experimental, where investigators assign the exposure, or observational, where they do not. Observational studies with a comparison group are analytic and include cohort studies, which follow exposed and unexposed people forward to outcomes, and case-control studies, which start with outcomes and look back at exposures. My analysis is a retrospective cohort study: I identified exposed and unexposed groups from past records and compared later outcomes.
Why a Cohort Design Fits
Grimes and Schulz (2002b) note that cohort studies can establish that the exposure preceded the outcome, can calculate incidence and relative risk directly and can examine several outcomes, but are vulnerable to selection bias and loss to follow-up, and retrospective cohorts depend on the quality of existing records. For our question, the exposure, education, clearly came before the outcome, emergency visits, and our records capture both.
The Two-by-Two Table
Of 200 patients who attended education, 30 had at least one diabetes-related emergency visit in 12 months, or 15.0%. Of 300 who did not attend, 75 had a visit, or 25.0%. The table has four cells: attended with visit, 30; attended without visit, 170; did not attend with visit, 75; did not attend without visit, 225.
A 40% relative reduction and a number needed to treat of 10 describe the same data; only one of them tells the director how many patients must attend for one visit to be avoided.
Relative Risk
The relative risk is the risk in the exposed group divided by the risk in the unexposed group: 0.15 divided by 0.25, or 0.60. Patients who attended education had 60% of the risk of an emergency visit compared with those who did not. The relative risk reduction is 1 minus 0.60, or 40%.
Absolute Risk Reduction and Number Needed to Treat
Laupacis et al. (1988) argued that relative measures can make small benefits look large and proposed the number needed to treat as a more clinically useful measure, calculated as the reciprocal of the absolute risk reduction. Subtracting 15.0% from 25.0% gives an absolute difference of 10 percentage points, and its reciprocal, 1 over 0.10, gives a number needed to treat of 10. About 10 patients need to complete education for one fewer patient to have an emergency visit in a year.
Odds Ratio
The odds of a visit among attendees are 30 divided by 170, or 0.176. Among nonattendees, the odds are 75 divided by 225, or 0.333. The odds ratio is 0.176 divided by 0.333, or 0.53.
Why the Odds Ratio Differs
The odds ratio, 0.53, suggests a larger effect than the relative risk, 0.60. This happens because the outcome is common: 21% of the whole group had a visit. When outcomes are rare, odds and risks are similar and the two measures agree. In a cohort study where risks can be calculated directly, the relative risk is the more intuitive measure; the odds ratio is necessary in case-control studies, where risks cannot be calculated, and in logistic regression.
Relative and Absolute Measures Together
The same relative reduction can mean very different absolute benefits. If our baseline emergency visit rate were 2.5% instead of 25%, a relative risk of 0.60 would give an absolute reduction of only 1 percentage point and a number needed to treat of 100. Reporting only the 40% relative reduction would hide this difference. Laupacis et al. (1988) made this argument for treatment decisions, and it applies equally to program decisions.
If We Had Used a Case-Control Design
A case-control study would begin with patients who had an emergency visit, the cases, and compare their past education attendance with patients who did not, the controls. It would be efficient for rare outcomes but could only estimate the odds ratio, not the risk or number needed to treat, and it would be vulnerable to bias in how cases and controls are selected (Grimes & Schulz, 2002a).
Presenting to the Director
I will present the result this way: "Among patients who attended education, 15 of every 100 had a diabetes emergency visit in the next year, compared with 25 of every 100 who did not attend. For every 10 patients who complete education, about one emergency visit is avoided." Natural frequencies and the number needed to treat are easier for decision makers to interpret than ratios.
Cost Implications
If an emergency visit costs our patients and the system about $1,500 on average and the program costs $300 per patient, 10 patients cost $3,000 to prevent one visit costing $1,500. This rough calculation suggests the program would not pay for itself on emergency visits alone, though it may improve A1c and other outcomes not counted here.
What the Design Cannot Claim
Patients who choose to attend education may differ from those who do not: they may be more motivated, have better transportation or have fewer competing demands. These differences, not the education, may explain part or all of the difference in emergency visits. A cohort design shows an association, not proof of cause. I will examine this problem in detail next week.
A Stronger Design
A randomized trial, in which patients are assigned to education or usual care, would balance motivation and other factors between groups (Grimes & Schulz, 2002a). A practical alternative for our clinic might be a stepped-wedge rollout, in which sites begin education at randomly assigned times.
Loss to Follow-Up
Twenty-two patients left the clinic during the follow-up year, 8 attendees and 14 nonattenders. If those who left had different emergency visit rates, the comparison would be biased. I repeated the analysis assuming all who left had a visit, then assuming none did; the relative risk ranged from 0.57 to 0.66, so the conclusion held.
Confidence Intervals
Point estimates without confidence intervals overstate certainty. With our sample, the 95% confidence interval for the relative risk is approximately 0.41 to 0.88, which excludes 1 but is wide. I will report intervals with every measure.
Conclusion
In our retrospective cohort, attending diabetes education was associated with a relative risk of 0.60 for emergency visits, a 10-point absolute reduction and a number needed to treat of 10. The odds ratio of 0.53 overstates the effect because the outcome is common. The cohort design supports an association but not a causal claim, because patients who attend may differ from those who do not.
References
Grimes, D. A., & Schulz, K. F. (2002a). An overview of clinical research: The lay of the land. The Lancet, 359(9300), 57-61. https://doi.org/10.1016/S0140-6736(02)07283-5
Grimes, D. A., & Schulz, K. F. (2002b). Cohort studies: Marching towards outcomes. The Lancet, 359(9303), 341-345. https://doi.org/10.1016/S0140-6736(02)07500-1
Laupacis, A., Sackett, D. L., & Roberts, R. S. (1988). An assessment of clinically useful measures of the consequences of treatment. New England Journal of Medicine, 318(26), 1728-1733. https://doi.org/10.1056/NEJM198806303182605
How this DNP 701 Week 2 example is structured
The DNP/701 Week 2 work usually covers measures of association and study designs. This paper builds a two-by-two table from practice data, calculates each measure with the working shown, interprets them in plain language and states what a retrospective cohort design can support. Students search this week as DNP 701 Week 2, DNP701 Wk 2 or DNP/701 Wk 2; all three are the same assignment.
DNP/701 Week 2 questions, answered
What does DNP/701 Week 2 usually ask for?
Many sections ask students to calculate and interpret measures of association, such as relative risk and odds ratio, and to identify study designs and their strengths and limitations.
When do the odds ratio and relative risk differ?
When the outcome is common, the odds ratio moves further from 1 than the relative risk; they are similar only when the outcome is rare, usually below about 10%.
What is the number needed to treat?
The reciprocal of the absolute risk reduction; it states how many patients must receive an intervention for one additional patient to benefit, which clinicians find easier to interpret than relative measures.
Write yours, or have the desk draft it
This paper is an original model document written by our desk, not a submitted student paper and not an official University of Phoenix document. Read it for the moves, then write your own to the instructions in your classroom. If you want one built to your exact prompt and rubric, the first custom sample is free and arrives in 24 to 48 hours.