What If a Single-Case Baseline Never Stabilizes?

When a single-case baseline never stabilizes, first check whether measurement, opportunity, context, health, access, or implementation changes explain the pattern. Genuine variability may be clinically important rather than an error to erase. The team can collect better-matched observations, revise the measure, choose a design or analysis suited to the pattern, proceed for care with stated inference limits, or stop the study question.

Audit the measurement before the person

Review definitions, observers, timing, devices, opportunity counts, scale limits, missingness, and data entry before attributing fluctuation to the client.

Map context to each point

Track task, partner, setting, schedule, health, sleep, medication when relevant, access supports, and competing events.

Decide whether variability is the finding

Some targets vary meaningfully across conditions. Preserve that pattern and consider a question that models those conditions.

Set an endpoint for waiting

Name a maximum review point or date and the decisions available then. Open-ended collection invites burden and outcome-guided timing.

Choose a defensible next route

Options include a repaired measure, stratified display, additional design replication, a different design, descriptive monitoring, or clinically necessary care with limits.

Build Priya's unstable-baseline investigation log

For the what to do when single-case baseline never stabilizes question, create a versioned unstable-baseline investigation log. Record the target, measurement unit, eligible opportunities or observation time, scale, desired direction, raw series, missing and invalid states, context, graph version, timing mechanism, decision rule, reviewers, clinical gates, participation response, decision time, actual change, deviations, and follow-up. The record should let another qualified reviewer reconstruct why Priya's phase decision occurred without relying on memory.

Work through Priya's baseline example

Priya's independent help requests are 2 of 10, 7 of 12, 3 of 8, 8 of 14, and 4 of 9 opportunities. Counts and denominators both change, so the percentages range from 20% to 58.3%. Review finds that platform access and task difficulty also varied. Collecting more of the same sessions would not repair those design differences. Show the raw series and all denominators before adding summaries. This fictional telepractice help-request study example illustrates one decision pattern and does not create a universal stability percentage, point count, phase duration, treatment rule, or causal conclusion.

Audit the evidence available for Priya

Priya's log retains each numerator, denominator, task level, connection quality, AAC availability, partner, health note, and exclusion. It separates true response variation from shifting opportunity and system conditions and records which fixes are feasible without manufacturing uniformity. The audit also checks data-entry history, definition version, observer training, agreement when needed, graph axes, session spacing, exclusions, phase labels, source timestamps, and access to original records. Unresolved discrepancies remain visible and pause any claim that depends on them.

Address the main interpretation risk for Priya

Waiting indefinitely can delay useful care and selectively stop the baseline at a convenient moment. Immediate phase change can also leave a weak comparison. Priya's team sets a review date and chooses among measurement repair, design revision, clinical action, or closure rather than promising eventual stability. The review describes the observed pattern in original units, names plausible alternatives, and separates a methods judgment from a clinical recommendation. Software may calculate, graph, and surface missing evidence; qualified people interpret the series and make decisions within their authority.

Use the later phase responsibly for Priya

After access and task definitions are repaired, Priya's data may become interpretable or remain genuinely variable. Either result informs the next decision. The report explains what changed and avoids combining the original and revised measures as one uninterrupted series. Visual analysis examines level, trend, variability, immediacy, overlap, and consistency across comparable phases or tiers. A supplemental statistic can summarize a defined feature. It cannot establish clinical importance, functional relation, consent, authorization, or treatment fit on its own.

Protect Priya's access and participation

Keep Priya's augmentative and alternative communication, interpreters, mobility, food, water, bathroom use, prescribed care, health support, rest, relationships, and emergency help available throughout the telepractice help-request study. Use accessible consent and assent processes when applicable and respond to withdrawal, dissent, or distress. Necessary safety or clinical action proceeds through qualified authority even when it changes the planned phase timing.

Apply current sources to Priya's decision

Priya's page uses the waiting-for-stability literature as a caution about response-guided timing rather than a rule to ignore unstable data. The BACB ethics hub and CASP public summary provide professional context, while the BCBA Test Content Outline identifies examination content on measurement and single-case design. The WWC Version 5.0 handbook is a research-review standard. A single-case design review describes within-phase and between-phase visual analysis. Research on waiting for baseline stability examines response-guided timing, and work on systematic visual protocols, current analytic reflections, and visual-analysis software clarifies methods and limits. ASHA supports continuous AAC access.

Rehearse Priya's review before live use

Run the unstable-baseline investigation log with a fictional series before it governs Priya's data. Test minimum and maximum observations, missing values, equal timestamps, trend in both directions, extreme points, floor and ceiling values, a failed access gate, assent withdrawal, delayed implementation, and an amended rule. Store expected decisions, reviewer rationale, screenshots or graph versions, software version, and correction history. The rehearsal tests mechanics while leaving the live judgment to qualified reviewers.

Close Priya's phase-decision review

Review the unstable-baseline investigation log with Priya, the responsible clinician, and a methodologist familiar with the design. Preserve raw data, graph, rule, decision snapshot, reviewers, client input, access and safety evidence, phase implementation, deviations, sensitivity checks, later outcomes, and inference limits. Keep the page draft and noindex until the manifest-named reviews are complete.

What to Do When a Single-Case Baseline Never Stabilizes