Behaviour Incident Pattern Analysis: Turning Data Into a Better BSP

Most disability and NDIS providers are good at recording behaviour incidents. The forms get filled in, the antecedents get noted, the duration gets logged. What happens far less often is the step after that: someone sitting down with a month of records and asking what they mean together. That step is where a behaviour support plan stops being a compliance document and starts changing outcomes for a person.

This guide is about pattern analysis, the discipline of reading across many incidents to find the recurring triggers, likely functions, and strategy failures that no single record reveals. It assumes you are already collecting good data. If you are not there yet, start with our companion guide on NDIS behaviour data collection, then come back here to make that data earn its keep.

Why does one incident at a time never tell the full story?

Reading incidents individually is like reading one frame of a film. Each record is accurate, but the meaning lives in the sequence. A single episode of a participant refusing to leave the house looks like a one-off. Fifteen episodes, all clustered on the mornings a particular staff member is rostered, is a pattern that points somewhere specific.

The problem is that human memory is a poor aggregator. Support workers remember the dramatic incidents and forget the quiet ones. Managers remember last week and not last quarter. Without a deliberate process, the loudest incidents drive the plan, not the most frequent or most preventable ones. Pattern analysis replaces recall with evidence.

The Positive Behaviour Support model that underpins the NDIS approach is built on this idea. As the NDIS Commission's Positive Behaviour Support Capability Framework sets out, quality practice means understanding the function of behaviour and changing the environment around the person, not just responding to episodes as they happen. You cannot understand function from a single data point.

What patterns actually matter in behaviour data?

Not every trend is meaningful, and chasing noise wastes clinical time. There are five families of pattern that consistently inform a stronger plan. Work through them in order.

1. Antecedent clusters and co-occurring triggers

An antecedent is what happened immediately before the behaviour. On its own, one antecedent is weak evidence. The signal appears when you group antecedents across many incidents and weight them by frequency. If loud environments precede 60 per cent of episodes and transitions between activities precede another 25 per cent, you have two candidate triggers ranked by strength.

The more advanced move is spotting co-occurrence. Loud environments alone might be tolerable, and transitions alone might be tolerable, but the two together might account for nearly every serious episode. That interaction is invisible unless you analyse antecedents in combination, not in isolation.

2. The likely function of the behaviour

Function is the purpose the behaviour serves for the person. In behaviour support the common functions are to gain something (attention, a tangible item, sensory input) or to escape or avoid something (a demand, a person, an environment). A plan that misreads function will fail, because it addresses the wrong need.

Pattern analysis informs function by looking at what reliably happens after the behaviour. If a behaviour is consistently followed by a demand being withdrawn, escape is the likely function. If it is consistently followed by one-to-one staff attention, that points elsewhere. The word "likely" matters here. Function should always carry a confidence level, and low-confidence hypotheses should be flagged for a practitioner to test, not treated as fact.

3. Time and duration patterns

When behaviour happens is often as informative as what triggers it. Concentration by time of day can reveal fatigue, hunger, medication timing, or a specific recurring activity. Concentration by day of week can reveal staffing patterns, community access schedules, or the absence of a preferred routine on weekends.

Duration is a separate and underused dimension. Episodes that are getting longer over time suggest an escalating situation or a strategy that is losing its grip. Episodes that are getting shorter can be quiet evidence that an intervention is working, which is exactly the kind of positive signal that rarely gets celebrated because nobody is measuring it.

4. Strategy effectiveness and habituation

A behaviour support plan lists strategies. The obvious question, rarely answered with data, is whether those strategies still work. Pattern analysis lets you tag each incident with the strategy used and the outcome, then track effectiveness over time.

The subtle failure mode here is habituation, where a strategy that worked at first gradually loses effect as the person adapts to it. A distraction technique that de-escalated 80 per cent of episodes in January and 30 per cent by June is not broken, it has simply run its course. Without trend analysis, teams keep using a strategy long after it has stopped helping, because it "used to work".

5. Data quality gaps

The final pattern to look for is the absence of data. If half your incidents have no recorded antecedent, or the duration field is routinely blank, any conclusion you draw is built on sand. Honest analysis reports its own gaps. A finding based on twelve well-documented incidents is worth more than one based on fifty half-completed forms, and the analysis should say so.

How do you run the analysis without a spreadsheet nightmare?

The traditional method is a support coordinator or team leader exporting incident records into a spreadsheet, tagging antecedents by hand, building pivot tables, and eyeballing the result. It works, but it is slow, it is inconsistent between people, and it quietly discourages anyone from doing it more than once a quarter. The analysis that would help most is the one nobody has time for.

CareIQ was built to close that gap. When incidents are recorded through the platform, the structured fields (including behaviour start and end times, antecedent quality, and the incident narrative) become the input for an automated behaviour analysis that a practitioner then reviews. The same passive monitoring approach that flags risk in day-to-day progress notes feeds into this deeper, plan-level view.

The CareIQ Assist analysis produces a structured report covering the patterns above:

• Antecedent analysis that groups triggers, weights them by frequency, and highlights co-occurrence between them
• Functional behaviour analysis that proposes likely functions with an explicit confidence level for each
• Duration analysis derived from recorded start and end times, showing whether episodes are lengthening or shortening
• Strategy effectiveness tracking, including habituation detection where a strategy appears to be losing effect over time
• Data quality assessment that names the gaps so nobody over-reads a thin dataset
• Clinical considerations that are prioritised and framed as prompts, never as prescriptions

The point is not to remove the clinician. It is to hand the clinician a pre-sorted, honestly-caveated view of the data so their time goes into judgement rather than into pivot tables.

How is participant privacy protected during analysis?

Behaviour data is among the most sensitive information a provider holds. Any system that processes it needs to treat privacy as a design requirement, not an afterthought. In CareIQ, personal identifiers are removed before any incident text is sent for external processing. The participant's name is replaced with a neutral reference such as "the individual", and staff names are pseudonymised to labels like Staff A and Staff B.

This matters for two reasons. Practically, it reduces the exposure of identifiable information. Clinically, it keeps the analysis focused on behaviour and environment rather than on personalities, which is exactly where a positive behaviour support lens should sit. If the automated service is unavailable, CareIQ falls back to a rule-based analysis rather than failing silently, so the report is always produced from the same de-identified inputs.

How does pattern analysis feed a compliant behaviour support plan?

The output of good analysis maps directly onto the sections of a behaviour support plan. The antecedent clusters inform the setting events and triggers section. The function hypothesis drives the entire strategy design, because proactive strategies exist to meet the underlying need in an appropriate way. The duration and strategy-effectiveness trends feed the review section, giving the practitioner evidence of what to keep, change, or retire.

This alignment also supports your obligations. Under the NDIS framework, implementing providers must work from a behaviour support plan and, where regulated restrictive practices are involved, meet strict authorisation and reporting duties. The NDIS Commission's guidance for implementing providers sets out these responsibilities, and its behaviour support resources reinforce that plans should be reviewed as new evidence emerges. A plan that has never been reviewed against fresh incident data is hard to defend. One that shows a clear analysis-to-review loop is exactly what an auditor wants to see.

If your plan involves restrictive practices, the analysis becomes even more important. The goal of positive behaviour support is to reduce and where possible eliminate restrictive practices over time, and you can only demonstrate that reduction with longitudinal data. Our guide to restrictive practices in the NDIS covers the documentation and authorisation side in detail.

What does a good analysis cadence look like?

Pattern analysis is not a one-off event before a plan review. The teams that get the most value build it into a rhythm.

1. Weekly: a quick scan for any sharp change in frequency or duration, so an escalating situation gets attention before the next scheduled review
2. Monthly: a full pattern read across antecedents, function, timing, and strategy effectiveness, shared with the practitioner
3. Before every plan review: a consolidated analysis over the full plan period, with data quality gaps named explicitly
4. After any significant change: a fresh read when a strategy, medication, environment, or staffing pattern changes, to catch both improvement and regression early

The habit is what compounds. A provider that reviews behaviour patterns monthly will spot habituation, celebrate genuine improvement, and adjust plans while the evidence is fresh. A provider that only looks at the data when a plan is due will keep discovering last quarter's problems a quarter too late.