The Setup
Eight months ago, a regional hospital implemented a new discharge protocol. Every patient received a follow-up call within 48 hours of discharge — a nurse checking in, asking how they were feeling, whether they had questions.
The results looked encouraging:
- Patient satisfaction: up 15%
- Readmission rates: unchanged
The patients liked the call. But it wasn't preventing the readmissions it was designed to prevent. Something was missing. This is what the integrated analysis revealed.
The Data
What we had:
- 20 patient interview transcripts covering the discharge experience
patient_outcomes.csv: clinical metrics including 30-day readmission status, medication count, new device assigned, risk score, follow-up call completion, satisfaction scores
| Metric | Value |
|---|---|
| Follow-up call completion rate | 94% |
| Patient satisfaction (post-protocol) | 4.1/5 |
| Readmission rate — Medication Confusion theme | 42% |
| Readmission rate — Device Confusion theme | 38% |
| Readmission rate — Clear Instructions theme | 6% |
The Investigation
Qualitative Coding
Five themes emerged from the patient interviews:
- Medication Confusion — unclear instructions, unaware of drug substitutions, accidental doubling
- Device/Inhaler Confusion — new equipment with no demonstrated technique
- Clear Instructions — simple, explicit schedules that patients could follow
- Lives Alone — No Support — no one nearby to catch errors or notice decline
- Positive Discharge — patients who felt confident and prepared
Marcus's interview was the most instructive — and the most troubling.
"I was on four new medications. I didn't realise the new one replaced an old one."
"I ended up doubling up on something because I didn't realise the new one replaced an old one."
Marcus rated his discharge experience 4.2 out of 5. He thought everything was fine.
The 48-hour call asked how he was feeling. He said fine. He was readmitted 8 days later.
The call asked the wrong question.
David's interview revealed a different failure mode:
"The inhaler — I'd never used one before. No one showed me how."
"I was doing it wrong for three days."
David had used his inhaler incorrectly from the moment he left the hospital. The follow-up call asked how he was doing. He said he was managing. He was readmitted in 5 days.
No one asked him to demonstrate his technique.
Sandra's interview was the control:
"They gave me a clear schedule. One pill morning, one night."
One medication. An explicit schedule. She wasn't readmitted. Her satisfaction score: 4.5/5.
The difference wasn't the follow-up call. It was the complexity of the discharge instructions relative to the patient's support system.
The Joint Display: Where It Came Together
Individual Level: Three Patients, Three Outcomes
Using FableSense AI's Case-Level Analysis joint display, we examined each patient's qualitative profile alongside their clinical outcomes — simultaneously, in one view.
Marcus:
- Theme: Medication confusion
- Quote: "I ended up doubling up on something"
- New medications: 4 | Readmitted: Yes — day 8
- Follow-up call: Completed | Satisfaction: 4.2/5
David:
- Theme: Device/inhaler confusion
- Quote: "I was doing it wrong for three days"
- New device: Inhaler (first-time user) | Readmitted: Yes — day 5
- Follow-up call: Completed | Satisfaction: 3.8/5
Sandra:
- Theme: Clear instructions
- Quote: "They gave me a clear schedule. One pill morning, one night."
- New medications: 1 | Readmitted: No | Satisfaction: 4.5/5
Both Marcus and David had completed their follow-up calls. Both rated their experience positively. Both were readmitted within 10 days.
The call was measuring satisfaction. It wasn't measuring comprehension.
Population Level: The Integration Matrix
Across all 20 patients, the Integration Matrix confirmed the pattern was systematic:
| Theme | readmitted_30day | days_to_readmission | new_med_count |
|---|---|---|---|
| Medication Confusion | Strong positive | Strong negative | Strong positive |
| Device Confusion | Strong positive | Strong negative | Moderate positive |
| Clear Instructions | Strong negative | — | Strong negative |
| Lives Alone | Strong positive | Strong negative | — |
The aggregate satisfaction score (4.1/5) had hidden this entirely. Individual case-level data — qualitative profiles alongside clinical outcomes — made it visible.
The Recommendation
The generic discharge call stays — it demonstrably improves satisfaction, which matters for patient wellbeing and hospital reputation. But high-risk patients need a different kind of follow-up.
1. For patients with 3+ new medications: add pharmacist follow-up Not a nurse asking "how are you feeling?" — a pharmacist doing medication reconciliation. Confirming which old medications were replaced. Confirming dosing schedules. Checking for interactions.
Marcus's doubling-up incident was preventable with one structured conversation.
2. For patients with new devices: require teach-back before discharge Watch the patient use the inhaler (or injector, or monitor) before they leave the ward. Ask them to demonstrate — don't ask if they understand. Comprehension and competence are different things.
David would have been caught in 30 seconds of observation.
3. For patients who live alone with high risk scores: flag for day-3 follow-up The standard 48-hour call often comes before errors surface. For isolated high-risk patients, a day-3 or day-5 clinical check is more likely to intercept a problem while it's still reversible.
4. Redesign the call script for high-risk patients Replace "How are you feeling?" with "Walk me through your medications since you got home." One question change. Potentially significant impact on readmission rates. Costs nothing to test.
The Outcome
The discharge protocol had improved the right metric — satisfaction — but missed the outcome that mattered: readmissions. The integrated analysis revealed why: the call design optimised for how patients felt, not whether they understood what to do.
The qualitative showed the human cost of that gap. The quantitative proved how systematic it was.
Fixing the generic protocol wasn't the answer. Identifying which patients needed more than a call — and giving them that — was the answer. The data made that targeting possible.
The quantitative showed who was being readmitted. The qualitative showed why. The integration showed where to intervene.