Project Homepage |Â Synthetic Dataset Creation | Spreadsheet Cleaning and Analysis | SQL Summary and Analysis | Tableau Summary and Analysis
Below is the documentation for the Tableau Dashboard creation. Here is a link to the visualization for increased viewing real estate.
This project uses fully synthetic patient-level fall data. Hospital names are used only as fictional labels for the exercise; the results do not describe actual patients, hospitals, or health-system performance.
Hospital Falls Analysis and Recommendations — Interactive Dashboard (Synthetic Dataset)
I organized the Tableau deliverable as a three-dashboard story so users can move from a broad overview, to risk-pattern exploration, to recommendations based on the synthetic findings.
The Overview dashboard summarizes overall fall volume, provider-presence status, hospital and unit incident counts, and the relationship between Morse Fall Scale score, post-fall mobility change, and injury severity.
The Risk Identification dashboard explores:
Hour of fall
Fall reason
Cognitive status
Synthetic preventability tier
Repeat fallers
The final dashboard translates the strongest patterns identified in the synthetic sample into practical fall-prevention themes while acknowledging where additional operational data would be required before making a real clinical or staffing decision.
SUM(
  IF [Cleaned Injury Severity] IN ('Major')
  THEN 1
  ELSE 0
  END
) / COUNT([Incident ID])
Overall before filters:
20 of 85 incidents
23.5%
SUM(
  IF [Cleaned Injury Severity] IN ('Sentinel Event')
  THEN 1
  ELSE 0
  END
) / COUNT([Incident ID])
Overall before filters:
3 of 85 incidents
3.5%
Severe Falls %
SUM(
  IF [Cleaned Injury Severity] IN ('Major', 'Sentinel Event')
  THEN 1
  ELSE 0
  END
) / COUNT([Incident ID])
Overall before filters:
23 of 85 incidents
27.1%
SUM(
  IF [Cleaned Provider Present] = 'No'
  THEN 1
  ELSE 0
  END
) / COUNT([Incident ID])
Overall:
64 of 85 incidents
75.3%
CASE MAX([Injury Severity Score])
WHEN 3 THEN 'Sentinel Event'
WHEN 2 THEN 'Major Injury'
WHEN 1 THEN 'Minor Injury'
ELSE 'No Injury/Undocumented'
END
This displays the highest ordinal injury-severity value within the current filtered view.
The following filters are applied across the Tableau story:
Unit Type
Hospital Name
Fall DateTime
Cleaned Injury Severity
Cleaned Cognitive Status
These allow users to narrow the same synthetic dataset by hospital, unit, date, injury level, and cognitive status.
The Overview dashboard displays:
Total Falls
Provider Not Present %
TUG Loss After Fall
Severe Injury % — Major + Sentinel Event
A heatmap compares incident volume across the four hospital labels and four unit types.
Because the dataset does not include patient-days, admissions, or another exposure denominator, the visualization represents incident counts, not true hospital or unit fall rates.
A horizontal bar chart displays fall counts by reason and is colored by:
Provider present
Provider not present
Provider presence not documented
The visualization shows that toileting and balance-related falls were overwhelmingly in the provider-not-present category.
Transfer and orthostatic events included a larger number of incidents where a provider was documented as present.
A scatterplot compares:
X-axis: Morse Fall Scale score
Y-axis: Post-fall TUG percent change
Color: Injury severity
The visualization shows a strong relationship within the synthetic dataset between higher Morse scores, larger increases in post-fall TUG time, and more severe outcomes.
Because the data are synthetic, the strength of this relationship should not be interpreted as a real clinical effect size.
A bar chart displays incident counts by hour of day and is colored by fall reason.
Falls occurred throughout the 24-hour period.
The largest individual hourly count was:
03:00 — 6 incidents
Most other hours contained between 2 and 4 incidents.
The distribution does not provide enough evidence to conclude that a particular shift requires additional staffing.
A horizontal bar chart displays incident counts by cognitive status, with color representing the synthetic preventability tier.
This provides a way to compare fall characteristics across:
Alert & Oriented
Mild Confusion
Severe Dementia
Sun-downing
Not Documented
The preventability tiers are part of the synthetic source lookup and are not a validated clinical taxonomy.
The repeat-faller table contains:
Hospital name
Patient MRN
Cognitive status
Unit type
Fall count
Patient age
Morse Fall Scale score
Pre-fall TUG
TUG change in seconds
TUG percent change
Four fictional MRNs appear more than once, accounting for 9 total incidents.
The original version of this dashboard made several recommendations more definitively than the dataset supports.
I would revise the dashboard so recommendations are tied directly to the observed synthetic patterns while avoiding causal, staffing, or treatment claims that cannot be validated from 85 incidents alone.
Why this is included
Unassisted toileting / urgency was the most common documented fall reason:
26 incidents total
25 occurred when a provider was documented as not present
Recommendation
Prioritize purposeful rounding, individualized toileting schedules, call-light response, and assisted toileting for patients identified as high risk.
Evaluate whether bed/chair alarms or other supervision tools are appropriate according to local policy and the patient's individual needs.
Why this is included
Balance-related falls were common in:
Geriatric Psych: 9 incidents
Med-Surg: 10 incidents
The Severe Dementia and Sun-downing groups also had higher average Morse scores and more severe outcomes within the synthetic sample.
Recommendation
Use individualized mobility assistance, clear cueing, environmental setup, and supervision plans for patients with cognitive impairment.
Involve PT or OT when formal mobility assessment, transfer training, or mobility recommendations are needed.
Why this is included
Bed/chair transfer slippage was the leading Acute Rehab fall reason:
7 of 14 incidents
Recommendation
Reinforce documented transfer-assistance levels, call-light use, equipment setup, and patient/staff transfer education.
Review whether transfer plans are clearly visible and consistently followed.
Why this is included
Medication-induced orthostatic hypotension was the leading ICU fall reason:
7 of 14 incidents
Recommendation
For patients experiencing symptoms or undergoing medication changes, review medication effects and assess orthostatic tolerance according to clinical policy before unassisted mobility.
Why this is included
Falls occurred throughout the day.
The largest hourly count was 6 incidents at 03:00, but the differences between hours and days were modest.
Recommendation
Do not make a staffing change based on this dataset alone.
A larger dataset containing staffing levels, patient census, patient-days, and shift-level exposure would be needed to determine whether specific shifts carry greater fall risk.
Why this is included
The dataset contains only fall incidents and does not include a post-intervention period.
Recommendation
Document implemented interventions, establish clear outcome measures and exposure denominators, continue collecting data, and compare pre- and post-intervention performance before concluding that a change improved fall outcomes.
75.3% of incidents, or 64 of 85, were documented with no provider present.
All 23 major or sentinel outcomes occurred in the provider-not-present category, but the synthetic sample cannot establish causation.
Med-Surg had the largest incident count with 39. Without a patient census or patient-day denominator, this should not be described as the highest fall rate.
Geriatric Psych had the largest average relative increase in TUG completion time after a fall at 56.10%.
The validated Morse distribution was 60 high-risk scores, 24 medium-risk scores, and 1 missing score.
Falls were not concentrated strongly enough by hour or day to support a confident overnight-staffing recommendation.
Four repeat fallers accounted for 9 incidents, showing the value of examining patient-level recurrence in addition to aggregate counts.
For the final portfolio version, I would make the following wording changes:
Rename Unwitnessed Fall Rate to Provider Not Present % or Falls With No Provider Documented Present.
Replace statements that provider presence "prevented" severe injury with: Severe outcomes were observed only in the provider-not-present category in this synthetic sample.
Remove any statement that all patients were high risk. The validated distribution is 60 high risk, 24 medium risk, and 1 missing Morse score.
Display the SQL None injury category as No Injury / Undocumented unless the underlying SQL field is corrected to Not Documented.
Do not compare hospital incident counts as though they were fall rates.
Keep a visible synthetic-data statement on the dashboard or story header so the results cannot be mistaken for actual hospital performance.
This Tableau story explores 85 synthetic inpatient fall incidents across four fictionalized Chicago hospital labels. Use the filters to compare unit, cognitive-status, injury, timing, and provider-presence patterns. The data are synthetic and the views show incident counts rather than hospital fall rates. Findings are intended to demonstrate analytical workflow and should not be interpreted as real clinical performance or causal evidence.