In the first two articles, we highlighted why we care about comorbidity‐related fall risk and how we created a 24‐question, 1–5 survey to capture subtle yet critical mobility factors. Now, it’s time to step into the nuts and bolts: our data‐collection process, participant demographics, and the statistical approach we used.
Who Participated?
- Sample Size: We had 673 patients from a primary care setting.
- Timeframe: Data were collected in January 2025—a single point in time (cross‐sectional).
- Inclusion Criteria: Essentially any patient able to complete or have the survey completed by a provider. No strict age cutoff or comorbidity minimum.
- Diversity in Age & Conditions: Ages ranged widely, from younger adults with diabetes to older adults with multiple chronic conditions. This variety gave us a better sense of how different health issues might collectively influence fall risk.
Comorbidity Count: A Simple but Effective Measure
While our 24‐item survey examines functional fall‐risk factors, we also needed a quick, reliable way to quantify how many chronic conditions each participant had.
- Self‐Reported by Providers:
- We asked providers to list any diagnoses (like arthritis, heart disease, diabetes), which were recorded in a JSON‐like text field.
- Parsing:
- Using a short script in Python, we split each entry into a list.
- For example, [“Arthritis”,”High blood pressure”] → length = 2.
- Two Groups:
- Group 0: <2 comorbidities.
- Group 1: ≥2 comorbidities.
Why group them this way? Past research shows that having multiple comorbidities often leads to additive or synergistic effects on fall risk. We wanted to see if that applied in our data.
Data Collection & Management
- Survey Administration: Patients or providers completed the 24‐item questionnaire, rating each statement from 1 (Strongly Disagree) to 5 (Strongly Agree).
- De‐Identified Entries: We removed participant names and identifying info to safeguard privacy.
- Listwise Deletion: If a participant’s survey was incomplete (missing critical items), we omitted it from the analysis. In practice, that only applied to a small number of responses.
Analytical Toolkit: Python and a T‐Test
To compare Group 0 vs. Group 1 in terms of total fall‐risk score, we used a Welch’s t‐test. The workflow was:
- Summation:
- Each participant’s 24 answers summed into a total_score, ranging from 24 (all 1s) to 120 (all 5s).
- Group Splitting:
- One set of total scores for <2 comorbidities, another for ≥2.
- Welch’s t‐Test:
- Tested if the mean total_score differed significantly between groups.
- p‐value < 0.05 suggested the difference was unlikely by chance.
Why Welch’s? Unlike the traditional t‐test, Welch’s doesn’t require equal variances between groups, giving a more robust analysis when dealing with variable clinical data.
Ethical and Practical Considerations
- No IRB Review: Data were gathered for internal quality improvement and de‐identified, so it was exempt from formal IRB oversight.
- Cross‐Sectional Limits: Since this was a single time point, we can’t say a participant’s comorbidity count caused falls; we only know it’s correlated with higher fall‐risk scores.
- Deidentified: We used no personally identifying information to maintain strict patient confidentiality.
Looking Ahead
In Article 4, we’ll share the key findings from our 673 participants:
- Did the group with ≥2 comorbidities significantly score lower (higher fall risk)?
- How big was the difference, numerically and clinically?
- How does this shape our understanding of comorbidity and falls?
Stay tuned to see the data, interpret the results, and discover next steps for clinical practice.
Join the Conversation
- Done a Welch’s t‐Test or have a story about cross‐sectional data in healthcare? Drop a note below—let’s learn together!
- Want to replicate or adapt our methods? Reach out at researchinfo@ahcpllc.com.
Posted by:
Dr. Pariksith Singh, Dr. Manjusri Vennamaneni, and Dr. Carlos Arias (Authors)
with Ed Laughman and Nawtej Dosanjh (Editors)
and Lynda Benson (Research Associate)
in collaboration with Access Health Care Physicians & Vedere University
