The DICE approach — non-pharmacologic-first management of behavioral & psychological symptoms of dementia
Behavioral and psychological symptoms of dementia (BPSD) — agitation, wandering, aggression, psychosis, apathy, disinhibition — affect the great majority of people living with dementia at some point in their illness. The DICE approach (Describe, Investigate, Create, Evaluate), developed by Kales, Gitlin, and Lyketsos, begins not with a prescription pad but with careful, structured observation of the behavior itself.
BPSD is not one thing — it is an umbrella term covering agitation, aggression, wandering, apathy, depression, anxiety, disinhibition, irritability, sleep disturbance, and psychosis (hallucinations, delusions). Each has different likely causes and different management pathways, so the first DICE step is a structured description before any intervention is chosen.
A good description captures the ABC framework: Antecedent (what happened right before), Behavior (exactly what the patient did, in observable terms, not interpretation), and Consequence (what happened after, including how caregivers responded). "Patient was agitated" is not a description; "patient paced the hallway for 20 minutes after the TV was turned off, then pushed a chair when redirected" is.
Describing also means quantifying: frequency, duration, intensity, and the specific circumstances (time of day, who was present, environmental noise/light) that surround each episode. Tools such as the Neuropsychiatric Inventory (NPI) and the Cohen-Mansfield Agitation Inventory (CMAI) standardize this so severity can be tracked over time.
Roughly 90% of people with dementia will exhibit at least one behavioral or psychological symptom over the course of their illness (Lyketsos et al., Cache County Study), making BPSD the norm rather than the exception in dementia care.
DICE was designed as a collaborative, caregiver-inclusive decision-support framework, published by Kales, Gitlin & Lyketsos (BMJ, 2015) and endorsed by the American Geriatrics Society and Alzheimer's Association:
• Describe — caregiver and clinician jointly characterize the behavior, its context, and its impact using the ABC (antecedent-behavior-consequence) structure • Investigate — clinician reviews the patient for underlying medical, psychiatric, medication-related, and environmental contributors • Create — a treatment plan is co-created, prioritizing non-pharmacologic strategies matched to the identified cause(s) • Evaluate — the plan's effect is measured against baseline severity; if ineffective or if safety risk is too high, the cycle restarts
DICE reframes BPSD not as a symptom to suppress chemically by default, but as a communication — often the only remaining communication channel for a person with advanced cognitive impairment — that something is wrong and needs to be understood.
• Wandering/pacing — purposeful or purposeless ambulation, often peaking in late afternoon ("sundowning") • Verbal agitation — repetitive vocalizations, calling out, cursing • Physical agitation — restlessness, pacing, rummaging, disrobing • Aggression — physical (hitting, grabbing) or verbal (threats, shouting), usually reactive to a perceived threat during care tasks • Psychosis — visual/auditory hallucinations, paranoid delusions (commonly theft or infidelity themes)
Observation should also record what the caregiver already tried and how the patient responded — this becomes the raw material for the Investigate and Create steps that follow.
Once a behavior is well described, DICE moves to Investigate: a systematic search for the medical, situational, or environmental factors driving it. Behavior in dementia is rarely random — it is frequently the most effective way a person with impaired language and insight can express pain, fear, boredom, or an unmet need.
Investigation systematically screens five overlapping domains, since more than one factor is often present simultaneously:
• Medical — pain (often from arthritis, dental disease, or constipation and impossible for the patient to report verbally), urinary tract infection or other delirium-triggering illness, dehydration, hypoxia, sleep disorders • Pharmacologic — anticholinergics, benzodiazepines, corticosteroids, opioids, or new/discontinued medications; drug interactions; polypharmacy • Physical/basic needs — hunger, thirst, need to toilet, fatigue, temperature discomfort • Environmental — overstimulation (noise, crowding, glare), understimulation/boredom, unfamiliar surroundings, poor lighting contributing to misperception • Psychological/interpersonal — fear during care tasks perceived as threatening, communication mismatch, caregiver approach/tone, unresolved grief or loss of routine
A focused physical exam, review of vitals, urinalysis when infection is suspected, and full medication reconciliation are standard components of this step.
A new-onset behavioral change in dementia — especially if abrupt — should always trigger a delirium work-up. Delirium superimposed on dementia is common, frequently caused by infection (especially UTI), dehydration, or a new medication, and is reversible if the underlying cause is treated.
A widely used conceptual model (Cohen-Mansfield's "unmet needs" theory) frames agitation as the behavioral expression of needs the person can no longer articulate verbally: physical discomfort, social contact, meaningful activity, or environmental fit. Under this model, agitation is a symptom to be decoded, not simply suppressed.
Caregiver-related contributors are also investigated: task-focused rather than person-focused care approaches, rushing, multiple unfamiliar caregivers, and mismatched communication style (speaking too fast, too much information at once, or failing to make eye contact) are well documented triggers for reactive aggression during activities of daily living such as bathing and dressing.
Investigation is the hinge of the entire DICE cycle: the Create step that follows is only as good as the causes identified here. Treating agitation from untreated pain with a sedating antipsychotic, for example, masks the symptom while leaving the patient in pain and exposed to medication risk with no benefit to the underlying problem.
When no single modifiable cause is found, this is itself useful information — it shifts the Create step toward broad environmental and person-centered strategies (routine, structured engagement, sensory calm) rather than a single targeted fix.
Every major dementia care guideline — the American Geriatrics Society, the Alzheimer's Association, and the American Psychiatric Association — designates non-pharmacologic strategies as first-line treatment for BPSD, reserving medication for situations where these approaches are insufficient and safety is at risk. The Create step of DICE turns the causes found in Investigate into a concrete, individualized plan.
Interventions are matched to the causes and preferences uncovered during Investigate:
• Person-centered care — care approaches tailored to the individual's history, preferences, and remaining strengths rather than institutional routines; slowing down, explaining each step, and offering choices during care tasks • Environmental modification — reducing noise and clutter, improving lighting to reduce misperception/shadows, creating safe wandering paths, using contrasting colors to aid orientation • Structured routine — consistent daily schedule for meals, activity, and rest to reduce unpredictability-driven anxiety • Meaningful engagement/redirection — music therapy (often highly effective, including for patients with advanced language loss), reminiscence, simple structured activities, redirecting attention away from a distressing stimulus rather than confronting it • Light therapy — bright light exposure, particularly for sundowning and disrupted circadian rhythm • Sensory and comfort strategies — addressing pain proactively, ensuring hunger/thirst/toileting needs are anticipated on a schedule • Caregiver training/education programs — e.g., REACH II, STAR-C, WeCareAdvisor — coaching caregivers in communication techniques and behavioral strategies, consistently shown to reduce both patient agitation and caregiver burden
Music-based interventions and structured person-centered care approaches have some of the strongest evidence bases among non-pharmacologic strategies for agitation, with effect sizes in several trials comparable to or exceeding those reported for antipsychotics — without the medication's mortality risk.
Non-pharmacologic treatment is not passive — like a medication, it has a "dose": frequency, duration, consistency, and fit to the individual all determine effect. A single one-off attempt at redirection is unlikely to work as well as a sustained, consistently-applied plan that the whole care team follows.
Intensity should be titrated: mild agitation may respond to simple environmental fixes (turning down the TV, adjusting lighting), while more entrenched or frequent behaviors typically require a combination approach — routine plus engagement plus caregiver-approach coaching — applied consistently over 2–4 weeks before being judged as effective or insufficient.
Non-pharmacologic strategies fail to fully control behavior in a meaningful minority of cases, particularly when safety risk to the patient or others is high (physical aggression causing injury, dangerous wandering, severe psychosis with resulting danger). In these situations, DICE does not abandon non-pharmacologic care — it continues alongside a cautious, time-limited pharmacologic trial, which is addressed at the Evaluate/escalation step.
When non-pharmacologic strategies have been adequately trialed and behavioral risk to the patient or others remains high, a cautious, time-limited pharmacologic trial may be justified. Antipsychotics are the most studied pharmacologic option for severe BPSD, but they carry an FDA black-box warning for increased mortality in patients with dementia-related psychosis — making this decision gate the most consequential step in the DICE cycle.
In 2005, the FDA required a black-box warning on atypical (second-generation) antipsychotics after a meta-analysis of placebo-controlled trials in elderly patients with dementia found a roughly 1.6–1.7-fold increase in mortality among those randomized to antipsychotics, mostly from cardiovascular events (heart failure, sudden death) and infections (notably pneumonia, possibly related to sedation-associated aspiration). In 2008, the warning was extended to conventional (first-generation) antipsychotics as well, based on similar or greater risk signals.
This is not a theoretical risk: it is drawn directly from randomized trial data in patients with dementia-related psychosis and agitation — the exact population these drugs are used to treat — which is why the decision to prescribe is treated as a genuine risk/benefit trade-off rather than a routine choice.
The FDA black-box warning specifically covers antipsychotic use for dementia-related psychosis: elderly patients with dementia treated with antipsychotics are at increased risk of death compared with placebo, primarily from cardiovascular and infectious causes — not from acute toxicity.
Guidelines (AGS, APA, Alzheimer's Association) converge on a narrow set of circumstances where a cautious antipsychotic trial is appropriate:
• The behavior poses a genuine safety risk to the patient or to others (severe aggression causing or threatening injury, dangerous agitation) • Non-pharmacologic strategies have been adequately trialed and are insufficient, or the situation is too urgent to allow time for them to take effect alone • Psychotic symptoms (hallucinations, delusions) are themselves causing significant distress or driving dangerous behavior • The lowest effective dose is used, for the shortest necessary duration, with a specific target symptom and explicit informed consent discussion covering the mortality risk
Antipsychotics are not indicated for wandering alone, mild agitation, or simply to make caregiving more convenient — appropriate use is reserved for more severe, safety-relevant presentations after the earlier DICE steps.
Because of the mortality signal, guidelines including the American Geriatrics Society Beers Criteria and CMS nursing-home regulations require that antipsychotic use in dementia be periodically reassessed for continued need — typically with a gradual dose-reduction (deprescribing) attempt every 3–6 months once behavior has stabilized, unless a documented clinical reason exists to continue.
This closes the loop with the Evaluate step: even when pharmacologic treatment is started, non-pharmacologic strategies continue in parallel, and the plan is to taper and discontinue medication as soon as it can safely be done — antipsychotics are conceived as a bridge, not a destination.
| Product | Indication | Trial Design | Key Result |
|---|---|---|---|
| Continue non-pharm only | Low-moderate severity, no safety risk | Sustain/intensify person-centered care, environment, routine, engagement | No medication risk; addresses root cause directly |
| Intensify + reassess | Moderate severity, cause partly identified | Combine interventions, revisit Investigate step, short reassessment interval | Avoids premature escalation |
| Cautious antipsychotic trial | High safety risk despite adequate non-pharm trial | Lowest effective dose, explicit target symptom, informed consent | Reduces acute danger when other options exhausted |
| Deprescribing attempt | Behavior stabilized on medication | Gradual taper every 3–6 months per Beers Criteria/CMS guidance | Minimizes cumulative mortality/adverse-event exposure |
The final Evaluate step measures whether the chosen plan — non-pharmacologic, pharmacologic, or both — actually reduced behavior severity relative to the documented baseline. DICE is explicitly iterative: an insufficient response does not mean escalating medication further by default, it means returning to Describe and Investigate with fresh eyes.
Evaluation compares current behavior severity, frequency, and caregiver-reported distress against the baseline captured in the initial Describe step, using the same tools (e.g., NPI, CMAI) so the comparison is apples-to-apples rather than subjective impression alone.
A successful outcome shows a downward trend in severity and frequency, reduced caregiver burden, and no new safety concerns. Improvement should be attributed carefully: if multiple interventions were layered simultaneously, it can be hard to know which one worked — a reason to introduce and evaluate interventions in a stepwise, trackable fashion where feasible.
If behavior severity is not improving, the DICE model calls for looping back to Describe/Investigate rather than reflexively increasing medication dose or adding a second psychotropic. Common reasons for an apparent non-response include:
• An underlying medical cause (pain, infection, constipation) that was missed or has recurred • Non-pharmacologic strategies that were not actually delivered consistently by all caregivers • A mismatch between the intervention and the true unmet need • Disease progression introducing a new symptom domain not previously present
Revisiting these questions before escalating pharmacologically prevents the common failure mode of stacking sedating medications onto a problem that was never actually a medication-responsive one.
DICE is a loop, not a line: even after a successful pharmacologic trial, the cycle continues — non-pharmacologic strategies persist, response is tracked, and a deprescribing attempt is planned once behavior has been stable, closing the cycle back toward the least-medicated sustainable state.
Sustained non-pharmacologic, person-centered management is associated with reduced caregiver burden and depression, delayed nursing-home placement, and — because it avoids unnecessary psychotropic exposure — a lower cumulative risk of falls, sedation, stroke, and mortality associated with antipsychotic use in this population.
Successful long-term BPSD management is rarely a single decisive fix; it is the accumulated result of repeated, disciplined cycling through Describe, Investigate, Create, and Evaluate as the disease — and the person's needs — continue to change over time.