Objective: Train the clinician to read a patient’s trajectory, not just their current snapshot — by applying time-series thinking and rate-of-change analysis to clinical data. Most bedside teaching focuses on the value at a single time-point: a creatinine of 3.2, an SpO₂ of 88%, a GCS of 14. This module trains the discipline of asking how fast did we get here, is this stabilising or accelerating, and what does the velocity of this parameter tell me that the value alone does not. The AI acts as a clinical trend analyst, prompting the learner to derive trajectory-based clinical reasoning from longitudinal data.
Indication: Any case with serial investigations, observations, or scoring data available over time — deteriorating inpatients, step-down monitoring, chronic disease progression reviews, post-operative surveillance, ICU trend analysis, or any point in Module 5 or Module 7 where the learner senses something is changing but cannot articulate the pattern precisely.
[!IMPORTANT] Data Scope. This module is designed for anonymised or de-identified time-stamped data only. Before pasting any table, chart, or longitudinal entry into a prompt, confirm all patient identifiers have been removed per your institution’s data governance policy. The AI’s role is to help the learner reason about a pattern — not to store, transfer, or act on real patient data.
Phase 1 · Initiation → Phase 2 · Execution → Phase 3 · Closure / Review
Prompt:
#VibeRounds You are a clinical trend analyst for this session. Your job
is not to interpret a single data point — it is to help me reason about
how values are changing over time: direction, rate, acceleration, and
pattern. Ask me to paste the longitudinal data I want to analyse (a table
of serial observations, investigation results, or monitoring parameters
with timestamps). Do not offer any interpretation until you have seen the
data and I have confirmed it is de-identified. Confirm you understand
this role before we begin.
[!NOTE] Application Note: The explicit instruction to withhold interpretation until data is seen prevents the common failure mode of generic commentary based on a hypothetical — the AI must be anchored to the actual series before reasoning begins.
Prompt:
#VibeRounds Here is the de-identified longitudinal data for analysis:
[paste de-identified time-stamped table here]
Before interpreting anything, do the following:
1. Confirm the number of data points and the time-span covered.
2. Flag any gaps in the series where data is missing — these absences
are themselves clinically meaningful.
3. Flag any values that appear to be artefacts or transcription outliers
rather than true readings.
4. Ask me to confirm the units and whether the time axis is regular
(fixed intervals) or irregular (as-needed sampling).
Do not interpret trend yet. This step is cleaning only.
[!NOTE] Application Note: Data-cleaning before trend interpretation is a discipline borrowed from quantitative research — missing data points in clinical series often correspond to clinical events (discharge, deterioration, patient refusal), and treating them as random noise rather than signal is a common analytical error.
Prompt:
#VibeRounds Now analyse the direction and slope of each parameter in
the series:
- Is each parameter trending up, down, or oscillating?
- Is the trend linear (constant rate of change) or non-linear
(accelerating or decelerating)?
- For each parameter, give me one sentence describing what this
direction and slope tells me clinically — and one sentence describing
what I would need to see in the next 24–48 hours for this trend to be
considered reassuring versus concerning.
[!NOTE] Application Note: The 24–48-hour forward projection is deliberately included as a clinical reasoning anchor — it converts a retrospective description of trend into a prospective clinical decision criterion, which is the operationally useful output.
Prompt:
#VibeRounds For the parameter I am most concerned about, calculate the
average rate of change between each measurement pair (delta per unit
time). Then ask me: is this rate of change faster or slower than what
is typical for this condition or pathophysiology? If I do not know the
typical expected velocity, help me reason about what biological process
could produce this rate — acute versus subacute versus chronic, and
what that implies for the urgency of intervention.
[!NOTE] Application Note: Rate-of-change reasoning (e.g. a creatinine rising 0.5 mg/dL per 8 hours versus 0.5 mg/dL per 3 days) is taught in nephrology and cardiology subspecialty training but rarely formalised in general medical education — this step imports that discipline into the general ward reasoning toolkit.
Prompt:
#VibeRounds Look across all parameters for patterns beyond simple
trend:
- Is there a periodicity (values worsening at a regular interval,
e.g. overnight, post-dose, post-procedure)?
- Is there clustering of abnormal values around a specific event or
time-window?
- Is there a lag effect visible — one parameter changing before
another responds, suggesting a causal or physiological sequence?
For any pattern identified, ask me what clinical intervention or
event in the patient's timeline could explain it, before offering
your own hypothesis.
[!NOTE] Application Note: The Socratic inversion — asking the learner for an explanation before offering one — is a consistent architectural feature of the VibeRounds stack. In trend analysis, this matters especially because pattern-recognition in time-series data is prone to apophenia (finding patterns in noise). Requiring the learner to propose a causal mechanism forces a plausibility check before the pattern is accepted as real.
Prompt:
#VibeRounds Based on the full series, identify any parameters that
show early-warning signal characteristics — values that crossed a
threshold before a clinical deterioration event (if visible in the
data), or that are currently trending toward a threshold without
having reached it yet. For each, tell me:
- What threshold (validated or clinical rule-of-thumb) is relevant?
- How many time-steps away from that threshold is the current
trajectory, at the present rate of change?
- What is the single earliest actionable intervention that should be
considered now versus at threshold breach?
[!NOTE] Application Note: This step operationalises the concept of lead time in clinical monitoring — the window between detectable signal and outcome event — which is the core premise of early-warning scoring systems such as NEWS2. Applying it to a real patient’s own data, rather than a generic score, develops a more sophisticated and transferable version of the same skill.
Prompt:
#VibeRounds Now look across parameters rather than within each one:
Which pairs of parameters appear to be moving in synchrony or in
opposition? For any apparent correlation, help me reason about whether
this is mechanistically plausible (a true pathophysiological coupling)
or coincidental (parallel but unrelated trends). Name one clinical
decision this cross-parameter relationship should influence.
[!NOTE] Application Note: Cross-parameter correlation is the analytical step most commonly skipped at the bedside — clinicians typically review each investigation column separately rather than reading across the row. This step trains the more computationally demanding but more informative cross-parameter pattern.
Prompt:
#VibeRounds Synthesise the full time-series analysis into a three-part
trajectory summary suitable for a clinical handover or ward round
presentation:
1. Trend headline: one sentence on overall trajectory direction.
2. Critical velocity flag: the single parameter whose rate of change
is most clinically urgent, and why.
3. Decision trigger: the specific observation, in the next monitoring
window, that should prompt escalation versus reassurance.
Write this in the terse, information-dense style of a good handover
note — not a paragraph of hedged prose.
[!NOTE] Application Note: The handover output format connects directly to Module 4, Step 4.9 — the goal is that the output of this module is immediately deployable in the clinical context where trend reasoning actually matters.
Prompt:
#VibeRounds Apply a critical awareness check to this session:
1. Is there any risk that I imposed a trend on data that is actually
within normal biological variability — i.e. read a pattern into
noise?
2. Was any rate-of-change conclusion driven by two outlier data points
rather than a true sustained trend?
3. What additional data (a missing time-point, a concurrent clinical
event log) would most change the interpretation produced in this
session?
4. If a colleague looked at this series and disagreed with my trend
reading, what is the most defensible alternative interpretation?
Be constructive but honest about where this analysis could be wrong.
[!NOTE] Application Note: Extends Framework D’s critical awareness framework to the specific risks of time-series analysis — namely, the tendency toward over-interpretation of noisy biological data and the vulnerability of rate-of-change calculations to outlier sensitivity.
Previous: ← Module 21 — Evidence Frontier Search