Diabetic Data Analysis and Visualization: A Nurse Informaticist’s Guide (Case Study)
13 min read Case study: Diabetic data analysis and visualization
Scenario: You are a new nurse informaticist at a small-sized regional hospital that has been actively collecting diabetes treatment data over the past year. The hospital aims to evaluate the effectiveness of its diabetes management programs and identify areas for improvement, particularly in terms of patient compliance and overall treatment outcomes.
It is just your luck that the person who was to present the data to stakeholders had to have emergency surgery and will be out for at least a month. The much-anticipated stakeholder meeting is in 2 weeks. The task of presenting the data now falls on you. Luckily, the data collection is complete, but the analysis and presentation now fall on your shoulders.
Your task is to analyze the collected data, focusing on treatment compliance and its impact on patient outcomes such as HbA1c levels, complications, and hospital admissions related to diabetes. Your analysis will help the hospital understand which factors are contributing to successful treatment outcomes and where additional resources or interventions are needed.
Why diabetes data analysis matters
Diabetes is one of the most data-rich chronic conditions in healthcare. Every visit produces lab values, medication records, vital signs and education notes. Collecting data is only the first step, though. The value comes when a nurse informaticist turns that data into information stakeholders can understand and decisions they can act on. In this scenario, the questions are clear: Are patients taking their treatment as prescribed? Does adherence improve HbA1c, reduce complications and prevent admissions? Which patients need more support?
Evidence supports this focus. In a large cohort of patients with diabetes, medication non-adherence was associated with higher rates of all-cause hospitalization and mortality (Ho et al., 2006), and each 1% reduction in HbA1c was associated with fewer diabetes-related complications (Stratton et al., 2000). Showing these links in your own hospital’s data makes a compelling case for investment.
Questions, metrics, stakeholders
Clean, validate, de-identify
Who are the patients?
Adherent vs non-adherent
Which factors drive outcomes?
Tell the story, recommend action
Step 1: Define the questions, metrics and audience
Before opening the spreadsheet, write down what the stakeholders need to know. A two-week deadline leaves no time to analyze everything, so focus on the questions in the assignment.
| Question | Metric (operational definition) |
|---|---|
| How compliant are our patients? | Medication adherence as proportion of days covered (PDC): days with medication on hand ÷ days in the period. PDC ≥80% is the common threshold for “adherent.” Also: kept appointments, HbA1c tested at least twice a year, education sessions attended, self-monitoring logs. |
| Is glycemic control improving? | Mean and median HbA1c; % of patients with HbA1c <7% (a common goal for many non-pregnant adults); % with poor control, HbA1c >9% (American Diabetes Association [ADA], 2025) |
| Are complications being prevented? | % with documented retinopathy, nephropathy (eGFR, urine albumin-to-creatinine ratio), neuropathy, foot ulcers or amputation, cardiovascular events; also severe hypoglycemia |
| Are admissions decreasing? | Diabetes-related admissions per 100 patients (DKA, hyperglycemic hyperosmolar state, hypoglycemia, diabetic foot infection), 30-day readmissions, ED visits, length of stay |
| What drives success? | Factors such as age, diabetes type, insulin use, number of medications, insurance, language, health literacy, transportation, food insecurity, program participation |
Also identify your audience. Executives want cost, risk and return on investment. Physicians and nurse managers want clinical detail. Quality and finance teams want benchmarks. One clear message with supporting detail in an appendix serves everyone.
Step 2: Prepare the data (cleaning, validation and privacy)
Most analysis time is spent here, and skipping it leads to wrong conclusions in front of leadership.
- Understand the dataset: find the data dictionary. What does each column mean, which units are used, and what time period is covered? Ask the data owner or EHR analyst if anything is unclear.
- Check completeness: count missing values per variable. If 30% of patients have no follow-up HbA1c, say so. Missingness itself may be a finding, because patients without labs are often the least engaged.
- Check accuracy and outliers: an HbA1c of 71% is likely a typo for 7.1%; a birth date in 2030 is an error. Use plausible ranges, and document every correction.
- Remove duplicates and standardize formats (dates, units such as mg/dL vs mmol/mol, medication names).
- Create derived variables: adherence group (PDC ≥80% vs <80%), HbA1c change (last minus first), control category, age bands, any complication (yes/no), number of diabetes admissions.
- Protect privacy: follow HIPAA’s minimum necessary standard. Analyze a de-identified or limited dataset, store it on approved secure systems, and present only aggregated results. Suppress small cells (for example, groups under 10 patients) so individuals cannot be identified.
- Keep an audit trail: save the original file untouched and record every step, so another analyst can reproduce your work.
Step 3: Describe the population
Descriptive statistics answer “who are our patients?” and give stakeholders context before you show any comparisons.
- Categorical variables (sex, diabetes type, insulin use, insurance): counts and percentages.
- Continuous variables (age, HbA1c, BMI): mean and standard deviation if roughly normal, or median and interquartile range if skewed. HbA1c is often right-skewed, so the median is useful.
- Trends over time: monthly or quarterly averages show whether the program is moving the numbers.
Illustrative example (hypothetical data)
The numbers below are invented to demonstrate the method. Use your hospital’s real data for the assignment.
| Measure (12 months) | Adherent (PDC ≥80%) n = 248 | Non-adherent (PDC <80%) n = 152 | Total N = 400 |
|---|---|---|---|
| Mean age, years | 61 | 54 | 58 |
| Mean latest HbA1c | 7.2% | 9.0% | 7.9% |
| HbA1c <7% | 129 (52%) | 27 (18%) | 156 (39%) |
| HbA1c >9% (poor control) | 25 (10%) | 70 (46%) | 95 (24%) |
| At least one complication | 35 (14%) | 47 (31%) | 82 (21%) |
| At least one diabetes-related admission | 20 (8%) | 36 (24%) | 56 (14%) |
Step 4: Compare adherent and non-adherent patients
This is the heart of the scenario: does compliance affect outcomes? Match the statistical test to the type of data.
| Comparison | Data type | Test | Example |
|---|---|---|---|
| Mean HbA1c, two groups | Continuous vs 2 groups | Independent t-test (Mann-Whitney U if skewed) | Adherent 7.2% vs non-adherent 9.0% |
| HbA1c before vs after program, same patients | Paired continuous | Paired t-test (Wilcoxon signed-rank if skewed) | Baseline 8.4% vs 12-month 7.9% |
| % controlled, % admitted, % with complications | Categorical vs categorical | Chi-square test (Fisher’s exact for small cells) | 52% vs 18% with HbA1c <7% |
| HbA1c across 3+ groups | Continuous vs 3+ groups | ANOVA (Kruskal-Wallis if skewed) | Low, medium, high adherence |
| Relationship between two continuous measures | Continuous vs continuous | Pearson or Spearman correlation | Higher PDC, lower HbA1c |
Report effect sizes and confidence intervals, not just p-values. “Non-adherent patients were three times as likely to be admitted (24% vs 8%)” means more to a stakeholder than “p < .001.” Also remember that statistical significance is not the same as clinical significance: a 0.1% HbA1c difference may be significant in a large sample but meaningless clinically, whereas a difference of 0.5% or more is generally considered clinically meaningful.
Step 5: Explain which factors drive outcomes
Comparisons show that adherence matters. The next question is why some patients succeed and others do not, which tells leadership where to put resources.
- Stratify (break down) the results by age group, insurance, language, race and ethnicity, diabetes type, insulin use and clinic site. Disparities often appear only when results are stratified, and they are essential for health equity.
- Look at social determinants of health (SDOH): cost of medications, transportation, food insecurity and health literacy are common drivers of non-adherence.
- Use regression to control for confounders. Younger, sicker or uninsured patients may differ in many ways at once. Logistic regression estimates the odds of an outcome (such as admission) while adjusting for other factors; linear regression does the same for a continuous outcome such as HbA1c. For example (hypothetical): “After adjusting for age, insulin use and insurance, non-adherent patients had about three times the odds of a diabetes-related admission (adjusted OR 3.1, 95% CI 1.7–5.6).”
- Evaluate program components: compare outcomes for patients who attended diabetes self-management education and support (DSMES), had pharmacist medication reviews, or received follow-up calls against those who did not.
- Be honest about causation: this is observational data. Adherence is associated with better outcomes, but adherent patients may also differ in motivation or resources. Say “associated with,” not “caused.”
Need a statistics refresher? Excel’s Data Analysis ToolPak handles t-tests, chi-square and regression; SPSS, R or Python are better for larger datasets. Ask the hospital’s quality or analytics team to double-check your methods.
Step 6: Visualize the data
Good charts make complex results obvious in seconds. Choose the chart for the message, not for decoration (Knaflic, 2015).
| Message | Best chart | Diabetes example |
|---|---|---|
| Compare groups | Bar or clustered bar chart | % with HbA1c <7% by adherence group or clinic |
| Change over time | Line chart or run chart | Mean HbA1c by month or quarter; admissions per month |
| Distribution and spread | Box plot or histogram | HbA1c values in each group, showing outliers |
| Relationship between two measures | Scatter plot with trend line | PDC (x-axis) vs HbA1c (y-axis) |
| Process stability | Statistical process control (SPC) chart | Monthly admission rate with control limits (Provost & Murray, 2022) |
| Single key number | Big-number card with comparison | “39% of patients at goal, up from 31%” |
| Ongoing monitoring | Dashboard (Excel, Tableau, Power BI) | Adherence, HbA1c control, admissions, by unit and month |
Visualization best practices
- One message per chart, stated in the title: “Non-adherent patients were admitted three times as often,” not “Admissions by group.”
- Remove clutter: no 3D effects, heavy gridlines or unnecessary colors. Avoid pie charts with many slices.
- Use color with purpose: one highlight color for the group you want attention on, and gray or navy for the rest. Make sure colors are distinguishable for color-blind viewers.
- Add context: goal lines (HbA1c <7%), benchmarks, baseline values and sample sizes (n).
- Start bar charts at zero and keep scales consistent across charts so comparisons are fair.
- Label directly on the chart and note the data source, time period and any limitations.
Step 7: Tell the story to stakeholders
A stakeholder presentation is not a data dump. Use a simple structure that moves from the problem to the action you want leaders to take.
- Purpose (1 slide): why we collected the data and what decisions it supports.
- Bottom line up front (1 slide): three key findings in plain language.
- Who we serve (1 slide): the population at a glance.
- Adherence and outcomes (2–3 slides): HbA1c, complications, admissions by adherence group, with one chart each.
- What drives success (1–2 slides): key factors and disparities.
- Recommendations (1 slide): specific, resourced actions with expected impact.
- Limitations and next steps (1 slide): data gaps, plan for monitoring with a dashboard and PDSA cycles.
Sample findings and recommendations (based on the hypothetical data)
Key findings
- 62% of patients were adherent to their diabetes medications (PDC ≥80%).
- Adherent patients were nearly three times as likely to reach HbA1c <7% (52% vs 18%).
- Non-adherent patients had twice the complication rate (31% vs 14%) and three times the rate of diabetes-related admissions (24% vs 8%).
- Non-adherence was concentrated among younger, uninsured patients and those reporting medication cost or transportation barriers.
- Overall mean HbA1c improved from 8.4% to 7.9% over the year, but the non-adherent group did not improve.
Recommendations
- Screen all patients with diabetes for medication cost and SDOH barriers, and link them to financial counseling, patient assistance programs and lower-cost formulary options.
- Expand nurse- or pharmacist-led follow-up calls and medication reviews for patients with PDC <80% or HbA1c >9%.
- Increase referrals to DSMES, with evening, telehealth and language-concordant options.
- Build a live diabetes dashboard in the EHR to track adherence, HbA1c control and admissions monthly.
- Test changes with PDSA cycles and report back to stakeholders in 6 months.
Address limitations openly
Stakeholders trust analysts who are transparent. Mention missing data, that pharmacy refills show medications were picked up but not necessarily taken, that one year of data from one small hospital may not generalize, that admissions to other hospitals may be missed, and that associations do not prove causation.
A 2-week plan for the new nurse informaticist
| Days | Tasks |
|---|---|
| 1–2 | Meet the project sponsor; confirm the questions and audience; get the data dictionary and secure access |
| 3–5 | Clean and validate the data; create derived variables; document every step |
| 6–8 | Run descriptive statistics, group comparisons and stratified analyses; have a colleague check the methods |
| 9–10 | Build charts and draft the slides around three key messages |
| 11–12 | Review with clinical and quality leaders; refine the recommendations |
| 13–14 | Rehearse, prepare a one-page summary handout and anticipate questions |
The nurse informaticist’s role
Nurse informaticists bridge clinical practice and data science. Their clinical knowledge helps them notice implausible values, understand why a patient might miss refills, and translate numbers into realistic interventions. Their informatics knowledge helps them manage data securely, choose sound methods and design clear visualizations. This combination turns a year of data collection into better diabetes care.
Related reading: Type 2 diabetes case study (C.B.), PDSA practice improvement plan, Reflection on informatics and knowledge work, Triple and Quadruple Aim and Agencies that collect healthcare data.
Conclusion
An effective diabetes data analysis starts with clear questions and well-defined metrics, depends on careful data cleaning and privacy protection, and moves from description to comparison to explanation. Matching statistical tests to the data, reporting effect sizes and acknowledging limitations keeps the analysis credible. Clear, purposeful visualizations and a story that ends in specific recommendations help stakeholders see how treatment adherence affects HbA1c, complications and admissions, and where to invest so more patients reach their goals.
References
American Diabetes Association Professional Practice Committee. (2025). 6. Glycemic goals and hypoglycemia: Standards of Care in Diabetes—2025. Diabetes Care, 48(Suppl. 1), S128–S145. https://doi.org/10.2337/dc25-S006
Ho, P. M., Rumsfeld, J. S., Masoudi, F. A., McClure, D. L., Plomondon, M. E., Steiner, J. F., & Magid, D. J. (2006). Effect of medication nonadherence on hospitalization and mortality among patients with diabetes mellitus. Archives of Internal Medicine, 166(17), 1836–1841. https://doi.org/10.1001/archinte.166.17.1836
Knaflic, C. N. (2015). Storytelling with data: A data visualization guide for business professionals. Wiley.
Provost, L. P., & Murray, S. K. (2022). The health care data guide: Learning from data for improvement (2nd ed.). Jossey-Bass.
Stratton, I. M., Adler, A. I., Neil, H. A. W., Matthews, D. R., Manley, S. E., Cull, C. A., Hadden, D., Turner, R. C., & Holman, R. R. (2000). Association of glycaemia with macrovascular and microvascular complications of type 2 diabetes (UKPDS 35): Prospective observational study. BMJ, 321(7258), 405–412. https://doi.org/10.1136/bmj.321.7258.405
