Guidelines and Implementation

In individuals with dementia, as in any patient who presents with a psychiatric symptom, an initial assessment serves as a foundation for further evaluation and treatment planning (American Psychiatric Association 2015b). Assessing the type, frequency, severity, pattern, and timing of symptoms such as agitation and psychosis can help in identifying possible contributors and in targeting interventions to address symptoms and their causes. Pain is a common contributor to agitation or aggression and may signal other physical conditions, which may also need intervention. It is similarly important to determine the subtype(s) of dementia that is present, as this has implications for treating behavioral/psychological symptoms as well as providing information on likely disease course. The initial assessment also provides baseline information on symptoms, which is relevant to tracking of symptom progression or effects of intervention. Use of a quantitative measure to document information on symptoms in a systematic fashion can be helpful in monitoring the patient’s progress and assessing effects of treatment. A comprehensive treatment plan, as an outgrowth of the initial assessment, is beneficial in fostering a thorough review of the patient’s clinical presentation and in reviewing potential options for care that are person-centered and aimed at improving overall quality of life. Discussing the benefits and risks of possible treatments with the patient and surrogate decision makers is valuable in engaging them and helping them make informed decisions. Such discussions can also be beneficial by providing education on dementia and its symptoms and on available therapeutic options.

There are a number of potential advantages to including nonpharmacological interventions as a part of a comprehensive treatment plan. The most consistently effective interventions have focused on home-based caregivers and aim to develop their skills, improve their general well-being, and reduce their perceived burden (Adelman et al. 2014; Kales et al. 2015). These caregiver-related outcomes are predictive of whether a dementia patient is able to remain in the community or will be transitioned to institutional care (de Vugt et al. 2005; Miller et al. 2012). Other interventions can help in improving the culture and safety of the care environment and in conveying to patients and families that their needs and comfort are important. For most behavioral interventions there have not been a sufficient number of large-scale, well-controlled studies from which to draw conclusions about efficacy or safety in treating agitation or psychosis (Brasure et al. 2016). When studies with less rigorous designs and a broader range of target symptoms are also considered, modest benefits of behavioral interventions have been found (Brodaty and Arasaratnam 2012; Kales et al. 2015; Livingston et al. 2014). Among the specific benefits reported are reductions in agitation and aggression, alleviation of depression, improvement in sleep, and increased constructive activity. Studies of environmental modifications are even more limited than studies of behavioral interventions, and available data from clinical trials do not show significant effects (Kong et al. 2009). Nevertheless, anecdotal observations suggest that some individuals with dementia may benefit from reducing environmental clutter and ambient noise, optimizing lighting and walkways, providing cues to heighten orientation, and other environmental modifications.

Placebo-controlled trials of nonantipsychotic medications have not been reviewed in this practice guideline, and, thus, no recommendations are made about the appropriateness or sequence of their use based on their benefits and harms. In addition, no conclusions can be drawn from head-to-head comparisons between nonantipsychotic drugs (e.g., antidepressants, cholinesterase inhibitors, memantine) and antipsychotic drugs because of insufficient evidence (see “Review of Supporting Research Evidence” in Appendix A).

Expert consensus suggests that use of an antipsychotic medication in individuals with dementia can be appropriate, particularly in individuals with dangerous agitation or psychosis (see “Expert Opinion Survey Data: Results” in Appendix B), and can minimize the risk of violence, reduce patient distress, improve the patient’s quality of life, and reduce caregiver burden. However, in clinical trials, the benefits of antipsychotic medications are at best small (Corbett et al. 2014; Kales et al. 2015; see “Review of Supporting Research Evidence” in Appendix A) whether assessed through placebo-controlled trials, head-to-head comparison trials, or discontinuation trials. Effect sizes of second-generation antipsychotics (SGAs) range from nonsignificant to small depending on symptom domain (agitation, psychosis, and overall behavioral/psychological symptoms) and agent (see “Review of Supporting Research Evidence” in Appendix A). First-generation antipsychotics (FGAs) are deemed not different from SGAs in the management of agitation and overall behavioral/psychological symptoms, but the strength of the evidence for the comparisons is low, and haloperidol is the predominant agent that has been studied. There is not enough evidence to compare the effects of FGAs and SGAs on psychosis.

On the basis of both strength of the research evidence and effect size (moderate and small, respectively), the best evidence for SGA efficacy is in treatment of agitation, results that are driven by findings with risperidone treatment. Although evidence for the efficacy of SGAs suggests low utility (low strength of evidence for a very small effect) in the management of psychosis, the evidence for risperidone is substantially better than for the class (moderate strength of evidence for a small effect). Likewise, the efficacy evidence for SGAs in the management of overall behavioral/psychological symptoms also suggests low utility (high strength of evidence for a very small effect); the evidence for aripiprazole is substantially better than for the class (moderate strength of evidence for a small effect). For patients receiving treatment with an SGA as compared with placebo in the Clinical Antipsychotic Trials of Intervention Effectiveness for Alzheimer’s Disease (CATIE-AD), a modest reduction in caregiver burden was noted (Mohamed et al. 2012).

A number of studies have assessed the effects of discontinuing an antipsychotic medication in subjects with dementia, and the findings suggest a small effect of antipsychotic treatment. In individuals receiving placebo, there was a higher likelihood of symptom recurrence as compared with those continuing to receive an antipsychotic (moderate confidence), with some post hoc analyses showing that individuals who had higher baseline levels of symptoms or who were taking higher baseline doses of antipsychotic were more likely to have recurrent symptoms with discontinuation (Declercq et al. 2013; see “Review of Supporting Research Evidence” in Appendix A).

A dose-response effect, if present, can also provide suggestive evidence for a therapeutic benefit of a medication. The absence of a dose-response relationship is less informative; studies in which a dose-response relationship is absent are often underpowered, and a sufficiently wide range of doses is not always tested. Five published randomized controlled trials (RCTs) assessed differing doses of antipsychotic medications in managing behavioral and psychological symptoms of dementia, but these studies were of varying quality, had inconsistent findings, and often showed no therapeutic benefit at the highest dose (see “Review of Supporting Research Evidence” in Appendix A). There are no published studies on the optimal duration of antipsychotic treatment in individuals with dementia, and experts are divided in their opinion on optimal treatment duration (see “Expert Opinion Survey Data: Results” in Appendix B).

No studies have directly assessed harm from conducting an assessment or developing a comprehensive treatment plan. It is possible that questioning during an assessment may be upsetting to some patients and could increase rather than reduce agitation. Such worsening of symptoms is expected to be brief because the clinician will be able to curtail questioning or adjust the interview style and format to the patient’s responses. In an emergent situation, harm could result to the patient or others if interventions are delayed in order to complete assessment, treatment plan documentation, or discussions with the patient, family, or surrogate decision makers.

None of the available studies have reported direct harm to patients from behavioral interventions (Ayalon et al. 2006; Brasure et al. 2016). Reported risks associated with these interventions include falls and orthopedic injuries during physical activity, or worsening agitation and aggression with some approaches, particularly those involving physical contact between caregiver and patient (e.g., massage). Harm could also result to the patient or others if emergency interventions were to be delayed to complete trials of behavioral treatments. No direct comparisons of risk between behavioral and pharmacological therapies have been reported. No data are available on harms of environmental modifications or other nonpharmacological interventions, but again, the potential for harm is likely to be quite small.

With antipsychotic medications, the drugs’ potential for harms must be balanced against their modest evidence of benefit. As with any drug, this requires assessing the benefits and harms of prescribing the drug for an individual patient. No studies are available that assess the harms of withholding or delaying a trial of antipsychotic medication for individuals with agitation or psychosis in association with dementia. However, clinical observations suggest that such delays could lead to poorer outcomes for some individuals, such as physical injury to themselves or others, disruptions of relationships with family or other caregivers, or loss of housing due to unmanageable behavioral and psychological symptoms.

This estimation of benefits and risks should also consider clinical characteristics of the patient. For example, patients with Lewy body dementia or Parkinson’s disease dementia are at increased risk for adverse effects, which are typically more severe than in patients with other types of dementia and in some instances have been associated with irreversible cognitive decompensation or death. The risk of adverse effects may also be influenced by a history of falls or the presence of co-occurring medical conditions such as other neurological conditions, hypotension, diabetes, or cardiac or cerebrovascular disease.

The strength of evidence for harms of antipsychotic agents ranges from insufficient to high depending on the specific adverse effect; however, on the whole, there is consistent evidence that antipsychotics are associated with clinically significant adverse effects, including mortality (see section “Review of Supporting Research Evidence” in Appendix A). Harms data are rarely a primary outcome of randomized trials, and there is a paucity of randomized head-to-head comparisons of antipsychotic medications using equivalent doses of drug. In addition, the absolute number of serious adverse events in randomized trials is typically small, and this confounds statistical analysis. For example, pooled data from randomized placebo-controlled trials (Maglione et al. 2011) showed deaths in 8 of 340 (2.4%) individuals treated with aripiprazole as compared with 3 of 253 (1.2%) receiving placebo (pooled odds ratio [OR] = 2.37 from three studies; P = not significant [NS]), 2 of 278 (0.7%) treated with olanzapine as compared with 4 of 232 (1.7%) receiving placebo (pooled OR = 0.48 from two studies; P = NS), 5 of 185 (2.7%) treated with quetiapine as compared with 7 of 241 (2.9%) receiving placebo (pooled OR = 0.91 from two studies; P = NS), and 39 of 1,561 (2.5%) treated with risperidone as compared with 17 of 916 (1.9%) receiving placebo (pooled OR = 1.19; P = NS). For SGAs as a group, meta-analysis of the data from randomized placebo-controlled trials indicates that there is a statistically significant increase in mortality relative to placebo (Schneider et al. 2005).

From a methodological standpoint, data on harms generally come from studies that are less rigorous than randomized trials, such as observational or cohort studies. Administrative database studies are increasingly common and track associations between prescribed medications and diagnoses. This research cannot consider the effects of confounding variables such as dementia severity, co-occurring conditions, or the magnitude of agitation or psychosis. Nevertheless, administrative databases do permit study of large patient samples, which is important when looking at infrequent events. Some of these naturalistic studies have suggested a heightened risk of treatment with haloperidol and other FGAs and possible differences in risk among the other antipsychotic medications (see section “Review of Supporting Research Evidence” in Appendix A). However, as with studies of antipsychotic benefits, the limitations of existing research make it difficult to draw precise conclusions about the likely harms of treatment for an individual patient.

In addition to mortality, other serious adverse events of antipsychotic medications in individuals with dementia have been reported, including stroke, acute cardiovascular events, metabolic effects, and pulmonary effects (see section “Review of Supporting Research Evidence” in Appendix A). The strength of the evidence is low for stroke, but pooled analyses for risperidone and olanzapine suggest an increase in risk relative to placebo. The strength of the evidence on acute cardiovascular events is also low; however, there is some evidence of increased risk for all antipsychotics, with the risk being highest early in the treatment, and of a greater risk with risperidone and olanzapine than with other agents. Although the evidence on metabolic effects of antipsychotics (including weight gain, diabetes, dyslipidemia, and metabolic syndrome) is not as strong in individuals with dementia as it is in younger adults, the existing evidence is in keeping with what is largely known about this risk: highest for olanzapine and risperidone and lowest for aripiprazole and high-potency FGAs. Antipsychotic treatment in individuals with dementia also appears to carry an increased risk for pneumonia and for venous thromboembolism, but the strength of this evidence is low, with no apparent difference between FGAs and SGAs. Evidence is variable for other adverse effects, including cognitive worsening, sedation/fatigue, anticholinergic effects, postural hypotension, prolonged QTc intervals, sexual dysfunction, and extrapyramidal symptoms (e.g., parkinsonism, dystonia, tardive dyskinesia). However, case reports and observational data suggest a substantial increase in the likelihood of adverse effects when individuals with Lewy body dementia or Parkinson’s disease dementia receive antipsychotic treatment (Aarsland et al. 2005; Stinton et al. 2015). In some instances, these adverse effects have included irreversible cognitive decompensation or death. Less information is available for individuals with frontotemporal lobar degeneration, but a heightened sensitivity to antipsychotic medications has also been reported (Pijnenburg et al. 2003). No evidence is available that specifically addresses the possible harms of antipsychotic treatment in individuals being treated for chronic psychotic illness who subsequently develop dementia.

In terms of decisions about doses of antipsychotic medications, there is strong evidence that SGAs are associated with clinically significant dose-related adverse effects (Maust et al. 2015; see section “Review of Supporting Research Evidence” in Appendix A). Thus, if medications are begun at a low dose and increased gradually depending on clinical response, adverse effects may be minimized. On the other hand, it is possible that harms to the patient or others may occur if the response to treatment is delayed by underdosing of medication, particularly in emergency situations.

In terms of optimal treatment duration, the data suggest that the greatest risk of mortality occurs in the initial 120 days of antipsychotic use (Maust et al. 2015; see section “Review of Supporting Research Evidence” in Appendix A). The mechanisms by which heightened mortality could occur are unclear. In observational studies, unmeasured predisposing factors may lead both to a greater likelihood of antipsychotic treatment and to heightened mortality. However, although the greatest period of risk appears to occur with treatment initiation, the risk of adverse effects also persists with longer-term treatment. The cut-point of 120 days is, at least partially, an artifact of the designs of available research. Discontinuation studies suggest that antipsychotic medications can be tapered and stopped in many patients without return of symptoms (see section “Review of Supporting Research Evidence” in Appendix A). Expert consensus also suggests that an attempt at tapering an antipsychotic medication is indicated (see section “Expert Opinion Survey Data: Results” in Appendix B), with variation in the suggested timing of a taper attempt; however, only a small fraction of experts favored maintaining the dose of medication without a specific target date for a tapering attempt. Although some individuals will have recurrence of symptoms with antipsychotic discontinuation (moderate confidence), such risks can likely be mitigated by careful monitoring during treatment cessation with adjustments made in the medication tapering plan based on clinical response. However, there are no data on the most appropriate frequency for monitoring or the extent to which monitoring can reduce the severity or risk of symptom recurrence, which is unpredictable. There is insufficient evidence to determine whether individuals with more severe dementia, psychosis, or agitation will have a greater risk of symptom recurrence with discontinuation. There are also no data on whether symptom response is equivalent if antipsychotic medication is resumed after recurrence of symptoms.

No studies have examined the use of long-acting injectable antipsychotic medications in individuals with dementia. However, the longer duration of action of these medications suggests that they would be associated with an increased risk of harm relative to oral formulations or short-acting parenteral formulations of antipsychotic medications, particularly in frail elders.

The costs of assessment, treatment planning, and discussions with patients, family, or other surrogate decision makers relate to clinician time. Discussions with family or surrogate decision makers can also introduce direct or indirect costs to those individuals (e.g., lost work time, transportation). The feasibility of any treatment must also consider the unique situation of the patient and family, such as access to transportation, insurance status and coverage for specific services, and the effects of treatment requirements on the caregiver’s time or employment.

A small number of studies on the cost-effectiveness of behavioral treatments have consistently shown modest but favorable results for specific interventions (Gitlin et al. 2010). Prospective cost estimates for specific patients must take into account the need for individual therapists, the number and duration of required sessions, and the costs of home visits for community-based interventions (Brodaty and Arasaratnam 2012). Typically, such expenses have been assessed in terms of increased patient activities in the same setting and associated increases in personnel-related costs, but have not been weighed against the cost of pharmacological interventions, the cost of institutionalization for patients who cannot be managed at home or in less restrictive settings, or the cost of injuries to patients and caregivers during episodes of agitated or aggressive behavior.

The CATIE-AD trial (Rosenheck et al. 2007) examined the cost-effectiveness of antipsychotic treatment for outpatients with Alzheimer’s disease and psychosis, aggression, or agitation. Although individuals treated with an SGA showed no difference in quality adjusted life years or functional measures as compared with individuals receiving placebo, there were significantly lower costs in the placebo group. However, with the availability of generic SGAs, the costs of medication are likely to be less. We are not aware of studies on the cost-effectiveness of antipsychotic treatment for individuals with dementia in inpatient or nursing facilities or for severely agitated or aggressive individuals who require constant supervision.

Consensus on rating the strength of recommendations was high within the guideline writing group, and the statements were recommended unanimously. One group member (O.L.L.) chose not to vote on statements 7–15. The results of the expert opinion survey and input from the Alzheimer’s Association were incorporated in decisions about benefits and harms as noted below. Because costs of medications and other interventions vary widely, the guideline writing group did not consider cost-related considerations in weighing the benefits and harms of recommendations.

The strength of research evidence supporting these guideline statements is low to moderate. Statements 1, 2, 3, 4, 6, 7, 9, 11, and 13 are based on expert consensus that is derived from fundamental and generally accepted principles of medical ethics and medical practice, including elements of conducting an assessment, reviewing responses to prior treatments, and developing a plan of treatment. These statements also emphasize the importance of involving patients and surrogate decision makers, with input from family members and others. Perspectives of patients and their care partners highlight the need for such discussions and input at all steps of the decision-making and treatment-monitoring process to identify person-centered goals, values, and preferences that can shape care and enhance outcomes.

In statements 4 and 6, which address treatment planning and review of response to nonpharmacological interventions, the group chose not to comment on specific psychopharmacological medications other than antipsychotic medications. Although the guideline writing group only reviewed evidence on antipsychotic medications during the development process, available systematic reviews suggested that the harms of nonpharmacological interventions were minimal. In contrast, with other pharmacological treatments, more precise details on the balance of benefits and harms would have been needed before specific recommendations could be made.

In addition to the consensus-based recommendations described above, some specific recommendations are derived from more robust supporting evidence. For example, the recommendation for initiation of nonemergent antipsychotic treatment with a low dose of medication that is slowly titrated to the minimum effective dose (statement 8) is based on a substantial body of literature in geriatric pharmacology (Jacobson 2014; Lassiter et al. 2013; Mulsant and Pollock 2015; Wallace and Paauw 2015; Wooten 2012) as well as data suggesting that higher doses of antipsychotic medication are associated with a greater risk of harm in individuals with dementia (see section “Review of Supporting Research Evidence” in Appendix A). Statements 5, 8, 10, 14, and 15 are based on moderate-strength evidence in individuals with dementia that the benefits of antipsychotic medication are small. In addition, consistent evidence, predominantly from large observational studies, indicates that antipsychotic medications are associated with clinically significant adverse effects, including mortality, among individuals with dementia. The overall strength of evidence for these statements is graded as moderate on the basis of this balance of benefits and harms data and the fact that there were no studies that directly addressed all of the specific elements of each recommendation.

With respect to statement 12, harms data suggest a continued risk with ongoing treatment, and discontinuation studies show that medications can be tapered in many patients without symptoms recurring (see section “Review of Supporting Research Evidence” in Appendix A). The guideline writing group members were unanimous in recommending that an attempt at tapering and withdrawing the antipsychotic medication should be done for individuals being treated for psychosis or agitation in the context of dementia. One guideline writing group member (M.H.-L.) felt that an attempt at tapering is indicated for all individuals, where the patient’s history of recurrence of symptoms during prior tapering attempts is an input to the tapering decision making along with other factors, as in statement 10. The strength of research evidence supporting statement 12 is rated as low because the precise timing of a tapering attempt was not studied in a randomized fashion and the recommendation to attempt a taper within 4 months was based on the timing of discontinuation in the available clinical trials and information from expert consensus (see sections “Review of Supporting Research Evidence” in Appendix A and “Expert Opinion Survey Data: Results” in Appendix B). Input from patients and their care partners, as well as comments from some geriatric psychiatrists, suggested that more flexible timing of a tapering attempt may be warranted. Some guideline writing group members also felt that a longer period of treatment may be justified in some patients before tapering is attempted because of the initial time needed to reach a clinically effective dose and the longer duration of psychosis in many patients as compared with the typical duration of agitated behaviors. It was also noted that for some patients, a medication taper could negatively affect quality of life or be dangerous for the patient or others. Some retrospective data also suggested that individuals with more severe symptoms may be at a greater risk of relapse with antipsychotic tapering, but the available research did not examine whether an a priori determination of such individuals would predict a high likelihood of symptom recurrence. Consequently, in the final guideline statement, the recommended attempt at tapering antipsychotics is accompanied by two additional recommendations. Statement 11 stresses the importance of patient, surrogate decision maker, and family input before a tapering attempt, as well as review of the clinical factors related to a tapering attempt, and statement 13 addresses the need for careful monitoring during tapering so that any recurrent symptoms can be addressed quickly.

For statement 14, the data on harms in observational and administrative database studies sometimes focused on specific medications and sometimes on the class of FGAs as compared with SGAs. Since haloperidol was the most commonly used agent among FGAs, it was difficult to determine whether other FGAs had a comparable risk of harms. For this reason, the group chose to recommend that haloperidol not be used as a first-line agent, rather than recommending against use of any FGA as a first-line agent.

For statement 15, there was an acknowledgement of potential benefits of a long-acting antipsychotic medication for adherence in some selected circumstances. Nevertheless, for the preponderance of patients, the potential harms of a long-acting formulation were viewed as greater than the potential benefits. However, there was recognition that under selected circumstances, this balance may shift. In particular, some individuals will have had a chronic psychotic disorder, such as schizophrenia, that preceded the onset of dementia, and clinical opinion suggests that these patients may have continuing benefits of long-acting antipsychotic medication.

In assessing the balance between the benefits and harms of these recommendations, there are a number of factors to note. As our knowledge of dementia and its treatment evolve, there may be shifts in the balance of benefits and harms for these recommendations. At present, however, studies are either not available or not designed to give precise guidance on many of the clinical questions. One example is the lack of studies that examine benefits of assessment or discussion with patients, surrogate decision makers, families, and others. Another example is the small number of head-to-head trials comparing different pharmacological and nonpharmacological treatments for agitation or psychosis in dementia and an even fewer number of trials with parallel placebo or sham treatment arms. With nonpharmacological interventions, there can be significant variations in methodology from study to study, and multiple interventions can be administered together, confounding the interpretation of findings. Trials often fail to examine quality of life or other outcomes that patients and families view as most important. Studies also have not assessed the optimal time at which an attempted tapering of antipsychotic medication is indicated. There is insufficient evidence to determine whether individuals with more severe dementia, psychosis, or agitation will have a greater risk of relapse with antipsychotic discontinuation. In terms of monitoring, studies have not examined optimal timing of assessment during antipsychotic treatment or after an attempt at tapering antipsychotic treatment. The optimal frequency of laboratory and physical assessments to detect metabolic or other side effects of treatment also requires study in patients with dementia. It is also not clear whether laboratory data or other findings could predict which patients are at the highest risk of stroke or mortality or whether other interventions could reduce such risks.

Other aspects of research design may introduce variability into the findings and affect the ability to compare studies. A key issue is the way in which behavioral and psychological symptoms are defined and measured, with the definition and measurement of agitation being particularly problematic (Geda et al. 2013). Rating scales for behavioral and psychological symptoms define and measure agitation and aggressive behaviors in different ways and often mix measures of symptom frequency with measures of severity. New, shorter scales are also needed for routine clinical use. When studies have examined adverse effects of antipsychotic treatment in patients with specific subtypes of dementia, these diagnoses are generally based on clinical grounds, and this can introduce substantial variability as compared with diagnoses established through structured criteria, biomarker confirmation, or neuropathology (Beach et al. 2012). Studies with heterogeneous samples may fail to find a benefit or harm of a specific treatment, even if one is present for a more homogeneous subset of the patients.

As another source of variability, patients with dementia who are enrolled in clinical trials are not likely to be representative of the full range of individuals for whom clinical use of an antipsychotic medication might be considered. Significant physical illness (e.g., cardiopulmonary or renal impairments, cancer), use of certain medications (e.g., anticoagulants), or severe aggression requiring emergent intervention will typically exclude a subject from such research. Other psychiatric disorders, including substance use disorders, are also common exclusion criteria. It is not clear whether these typical exclusion criteria or other factors contribute to the apparent mismatch between clinicians’ views of antipsychotic benefits and the limited benefits found in clinical trials. Nonetheless, these limitations of existing clinical trials make it hard to draw precise conclusions about the likely benefits of treatment for an individual patient.

In terms of harms data, typical administrative database studies are unable to show the temporal sequence between treatment and a specific outcome. Thus, an individual with dementia may fracture a hip, become delirious, and receive antipsychotic medication. An administrative database study would associate the hip fracture or a subsequent pulmonary embolus with antipsychotic medication even without a causal relationship. Alternatively, the presence of psychiatric symptoms such as agitation may result in both a greater risk of falls and an increased likelihood of receiving an antipsychotic medication (Lopez et al. 2013). In the future, prospective collection of harms data using registry reporting or electronic health record data analytics may help delineate the temporal sequence of antipsychotic use and adverse outcomes.

In individuals with dementia who exhibit psychosis or agitation, initial assessment includes determining the type, frequency, severity, pattern, and timing of symptoms. Gathering this information typically requires multiple approaches, including interview and observation of the patient and review of relevant medical records. Flexibility is needed in adapting questions to the level of the patient’s understanding and being sensitive to signs of frustration or cognitive overload (e.g., with formal cognitive testing) during the interview. The ability to answer questions can also be affected by language skills, educational achievement, or unrecognized impairments in hearing. Given that memory and other cognitive functions are impaired in individuals with dementia, it will probably not be feasible to obtain information on recent symptoms from direct questioning. On the other hand, a patient may minimize his or her difficulties or give a seemingly coherent response to a question about recent events despite having no actual recall. Thus, it is also important to obtain information from family members and other caregivers, including other treating clinicians and nursing facility or hospital staff.

Quantitative measures provide a structured replicable way to document the patient’s baseline symptoms and determine which symptoms (if any) should be the target of intervention based on factors such as frequency of occurrence, magnitude, potential for associated harm to the patient or others, and associated distress to the patient. The exact frequency at which measures are warranted will depend on clinical circumstances. However, use of quantitative measures as treatment proceeds allows more precise tracking of whether nonpharmacological and pharmacological treatments are having their intended effect or whether a shift in the treatment plan is needed. Examples of available quantitative measures include the Neuropsychiatric Inventory Questionnaire (NPI-Q), which is Form B5 of the National Alzheimer’s Coordinating Center Uniform Data Set (https://www.alz.washington.edu/NONMEMBER/UDS/DOCS/VER2/IVPforms/B5.pdf) (Kaufer et al. 2000) and Section E (Behavior) of the Minimum Data Set (MDS)—Version 3.0 of the Centers for Medicare & Medicaid Services Resident Assessment and Care Screening instrument (http://www.cms.gov/Medicare/Quality-Initiatives-Patient-Assessment-Instruments/NursingHomeQualityInits/index.html?redirect=/NursingHomeQualityInits/25_NHQIMDS30.asp), and the Brief Psychiatric Rating Scale (BPRS; Overall and Gorham 1988; Ventura et al. 1993), each of which incorporates measurement of agitation and psychosis. Alternatively, for individuals who are agitated but do not show evidence of psychosis, measures include the Cohen-Mansfield Agitation Inventory (CMAI; Cohen-Mansfield et al. 1989) or the four-item Modified Overt Aggression Scale (Kay et al. 1988). Although these measures and others have been used for reporting purposes as well as research (Gitlin et al. 2014), it remains unclear whether routine use of these scales in clinical practice improves overall outcomes. However, it is clear that each rating scale defines and measures psychosis, agitation, aggression, and other symptoms differently (Geda et al. 2013), making it preferable to use a consistent approach to quantitive measurement for a given patient. The extent of the assessment, including the use of quantitative measures, will be mediated by the urgency of the situation and by the time that is available for evaluation. Depending on the clinical circumstances, printed or electronic versions of quantitative scales may not be readily available or information may not be available to complete all scale items. If time constraints are present, the clinician may wish to focus on rating of relevant target symptoms (e.g., on a Likert scale). Another approach is for family or nursing facility staff to keep a log of target behaviors such as aggression and track the number of episodes that occur. In emergent circumstances, safety of the patient and others must take precedence; the initial assessment may need to be brief, with a more detailed assessment obtained once the acute clinical situation has been stabilized. If collateral sources of information are not immediately available, treatment may also need to proceed, with adjustments in the plan, if indicated, as additional knowledge is gained.

A careful assessment of the type, frequency, severity, pattern, and timing of symptoms will also serve as the foundation for determining potentially modifiable contributors to the patient’s symptoms and identifying factors, such as the subtype of dementia (American Psychiatric Association 2013), that may influence choice of treatment. For example, pain is a common contributor to agitation (Bradford et al. 2012; Husebo et al. 2011; Kunik et al. 2010) but is not easily recognized because of sensory confusion and communication deficits (Pieper et al. 2013). Thus, priority should be given to identifying any source of pain and alleviating it through nonpharmacological and pharmacological approaches, as clinically indicated (American Geriatrics Society 2009). The pattern and timing of agitation may also suggest that the individual is becoming upset when he or she is hungry, fatigued, or cold or when there is a high amount of noise, clutter, or overstimulation in the environment. Vision or hearing deficits in combination with environmental factors can yield additive difficulties in an individual’s ability to understand and cope with a situation. Interactions with caregivers may also have a temporal association with behavioral dyscontrol if the caregiver asks cognitively challenging questions, rushes the patient in carrying out tasks, or communicates his or her sense of anxiety or frustration, directly or indirectly. If the patient is being assisted with bathing, dressing, or other activities of daily living, rejection of care and agitation may be an outgrowth of many factors, including overstimulation, pain with particular movements, or the patient’s sense of loss of control (Volicer et al. 2007). Attention to patient privacy needs is particularly important in assisting with activities of daily living. Constipation, incontinence, and other bowel or bladder issues can also prompt discomfort and distress. Other unmet needs may include, but are not limited to, relief from sensory deprivation, boredom, and loneliness (Cohen-Mansfield 2001).

Both precipitants and mitigating factors for agitation should be considered in the context of the patient’s unique facets. These include the patient’s likes and dislikes, lifestyle, hobbies, personality traits, intimacy and relationship patterns, spiritual and cultural beliefs, and past and current life circumstances. It can also be helpful to elicit information on prior aggressive behaviors (including associated legal problems), impulsivity, gambling, and problems with use of alcohol or other substances use. Using a person-centered approach calls for clinical staff to develop an understanding of the unique illness experience of the person and his or her care partners. This entails recognizing how individuals interpret the meaning of and navigate the difficult terrain associated with dementia and its symptoms. Input from the patient, his or her family members, and others (e.g., nursing facility or senior program staff) can give insights into patient preferences and the meaning of the behavior for the individual. It can also help in identifying approaches that have been helpful in managing agitation in the past and are therefore likely to be calming (Cohen-Mansfield et al. 2007). A person-centered approach also includes collecting information about previous traumatic experiences (e.g., childhood abuse, jail or prison experiences, domestic violence, combat experience, surviving the Holocaust, elder abuse) and possible triggers that may provoke inappropriate behaviors. Past life events, including traumas, are also relevant in terms of the resilience of the patient and his or her family as well as their previous approaches to coping with stress, loss, and decision making.

When interpreting the timing of symptom onset or worsening, clinicians should also consider changes in the patient’s physical status such as a recent fall (e.g., associated with head injury or pain), onset of a medical condition (e.g., urinary tract infection, pneumonia), evidence of other psychiatric symptoms (e.g., depression, anxiety), or recent change in medications. Individuals may not take medications in prescribed doses at home; changes in adherence (e.g., due to forgetfulness, admission to or discharge from a hospital) may be associated with altered clinical response or toxicity. The Beers criteria provide a useful checklist of medications, such as benzodiazepines or anticholinergic agents, that may be particularly likely to cause side effects or toxicity in older individuals (American Geriatrics Society 2015 Beers Criteria Update Expert Panel 2015). In inpatient or nursing home facilities where medications have to be re-ordered at designated intervals, it is not uncommon for a medication to be inadvertently stopped. Given the sizable numbers of medications that many older adults are prescribed, it is also important to be mindful of the potential for drug-drug interactions or prescribing of multiple similar drugs. Toxicity with associated psychosis or agitation can develop with seemingly minor dose changes or medication additions. Furthermore, the reduced metabolism, altered distribution, and diminished clearance of medications in older individuals means that the time to achieve steady-state levels will be longer than in younger patients. With drugs that have a long half-life or a long-half-life active metabolite (e.g., aripiprazole, fluoxetine, clonazepam, diazepam), the full effects of a dose change may not be apparent for several weeks, and this fact should be considered before titration or tapering of such medications. The use of long-acting intramuscular depot formulations of medications can be particularly problematic in frail, older individuals because of the longer duration of effect and the inability to stop the medication if an adverse effect occurs.

Another important step is determining the exact nature of the symptom. For example, in an individual with visual or hearing impairments, sensory illusions and other perceptual distortions may occur; these must be distinguished from true hallucinations and delusions before decisions about interventions are made. Also, benzodiazepine use can be associated with disinhibition; restlessness or pacing may reflect medication-related akathisia. Whether a symptom such as psychosis or agitation will require intervention is dependent on how frequently the symptom occurs and whether it is associated with significant distress to the patient or potential harm to the patient or others. To determine the degree of distress and the severity of symptoms, the treating clinician will synthesize information from multiple sources, such as direct observations of behavior, verbalizations by the patient, and input from family members, others involved with the patient, and nursing facility staff (if relevant), to arrive at a clinical judgment.

With agitation and with psychotic symptoms, there can be considerable variability in manifestations and potential for risk. For example, a patient may respond very differently to a delusion that belongings have been stolen as compared with a delusion that his or her loved one has been kidnapped and replaced by an imposter. Irritability may presage verbal threats, pacing, or emotional outbursts, whereas in other individuals episodes of rage and severe physical aggression may develop without apparent warning. The potential risk to the patient or others of a particular set of symptoms may vary with the circumstances. Thus, the same behavior may be riskier in a patient residing at home with a frail spouse than in a well-staffed nursing facility.

Given the complexities of addressing agitation and psychosis in individuals with dementia, it is important to develop and document a comprehensive plan of treatment that is an outgrowth of the assessment described above. Such a plan does not need to adhere to a defined development process (e.g., face-to-face multidisciplinary team meeting) or format (e.g., time-specified goals and objectives), but it should give an overview of the identified clinical and psychosocial issues along with a specific plan for further evaluation, ongoing monitoring, and nonpharmacological and pharmacological interventions, as indicated. Depending on the urgency of the initial clinical presentation, the availability of caregivers, and the time for assessment, the initial plan may need to be augmented over several visits and as more details of the history and treatment response are obtained.

If a symptom is rare, reassurance and redirection, with education of family and other caregivers, are likely to be sufficient, with other time-limited interventions used if needed. In some instances family members or other caregivers may find a symptom upsetting even when the patient is not distressed by it. For example, some patients experience visual or auditory hallucinations that are pleasant to them and not associated with anxiety or agitation. Other patients become verbally aggressive at times without physical aggression. Providing education and support to caregivers may aid them in coping with these symptoms (Brodaty and Arasaratnam 2012; Livingston et al. 2014).

If symptoms are more frequent and specific contributors to symptoms have been identified, these factors can be targeted for direct intervention. Common steps include treating underlying physical causes of psychosis or agitation and providing treatment for pain (Husebo et al. 2014). Mobility support, hearing aids, or eyeglasses should be used, when indicated. Some patients may respond positively to particular interventions (e.g., hand massage, pet therapy, music listening), whereas other patients may find the same nonpharmacological interventions upsetting or overwhelming, depending on their personal preferences and domains of cognitive impairment. Modifications to the environment can also be helpful such as optimizing lighting, reducing clutter, and removing items that the patient finds upsetting or that could be thrown or used as a weapon while agitated.

When individuals with dementia are residing in the community, behavioral symptoms such as agitation and psychosis can be extremely challenging for family and other caregivers to address (van der Lee et al. 2014). The associated impact on interpersonal relationships and increased caregiver burden can increase agitation and aggressive behaviors even further (Kunik et al. 2010). Psychosocial interventions that include individualized interpersonally based education and support for caregivers also appear to reduce the use of antipsychotic therapies in persons with dementia-related agitation (Richter et al. 2012). Education should increase knowledge, skills, and attitudes related to unmet needs, environmental regulation, and respect for individual preferences. One example of such an educational approach is the “Bathing without a Battle” training program (Gozalo et al. 2014). Clear communication of intended tasks, modification of caregiving strategies (e.g., bed baths vs. tub baths), or use of distraction to minimize the focus on caregiving can reduce combativeness and rejection of care (Galindo-Garre et al. 2015; Ishii et al. 2010). Additional strategies include use of therapeutic communication techniques (Cohen-Mansfield et al. 2007) and other approaches to challenging behaviors (Alzheimer's Association 2015; Glenner et al. 2005; Mace and Rabins 2011) that are appropriate for the person’s level of impairment. Additional supports can be facilitated by treating clinicians and can be invaluable (Jensen et al. 2015; Tam-Tham et al. 2013), although their availability may depend on factors such as geographical accessibility of resources, financial constraints, insurance limitations, or other obligations of the caregiver (e.g., to work, to young children in the home).

Training in reflective practice can increase self-awareness and improve care by having staff or caregivers reflect on behavioral incidents in terms of what occurred, their own thoughts and feelings, their assessment of positives and negatives of the experience, their interpretations of possible contributors to the incident, and their conclusions about adaptations to make in the future. Frameworks for understanding agitated behavior (Cohen-Mansfield 2001) may suggest a focus on other factors such as unmet needs, positive rather than negative behaviors, reduced stimulation, or promotion of relaxation. In inpatient settings and nursing home facilities, attention to the culture of the treatment setting and availability of a sufficient number of staff will also be important if staff is to participate in education, develop new skills, and be able to apply them. When staff and caregivers learn to view and respond to agitation and aggression in a way that is less emotionally charged, it may also help offset compassion fatigue and burnout, which are often consequences of working with individuals with dementia.

In addition to nonpharmacological interventions, the treatment plan may include pharmacological interventions to address physical conditions or symptoms such as pain or constipation. Although outside the scope of this practice guideline, cholinesterase inhibitors or memantine for dementia, and medications for other psychiatric disorders such as depression or anxiety disorders, may also be part of the treatment plan. Monitoring of physiological parameters (e.g., weight, blood pressure), point-of-care testing (e.g., glucose fingersticks), or laboratory testing may be included when indicated. Other elements of the treatment plan will be unique to the individual and his or her past experiences, needs, desires, preferences, and values to provide comprehensive person-centered care that is aimed at alleviating distress, promoting comfort, and enhancing quality of life.

The plan of treatment should also be reassessed over time, with modifications made to address changes in the patient's cognitive status, symptom evolution, and treatment response. This may entail reassessing for contributing or mitigating factors as well as continuing effective behavioral interventions or environmental modifications, adding other approaches if symptoms are not well controlled, and discontinuing ineffective nonpharmacological approaches. Any prescribed medications should also be reviewed for their benefits and for evidence of adverse effects. For example, benzodiazepine use is common, despite minimal evidence of benefit (Defrancesco et al. 2015) and an association with an increased risk of falls (Woolcott et al. 2009), worsening of cognition (Defrancesco et al. 2015), and potentially increased mortality (Huybrechts et al. 2011).

Given the risks associated with antipsychotic medications, if nonemergent use of antipsychotic medication is being considered to address agitation or psychosis, it is important to review all aspects of the assessment and the treatment plan. The aims of such a review are to determine the frequency and severity of symptoms in a systematic fashion, identify consequences of agitation or psychosis (e.g., distress to the patient, danger to self or others), discover previously unrecognized contributors to agitation or psychosis, reassess the clinical response to nonpharmacological or pharmacological treatments, and decide whether different interventions might be indicated.

If agitation or psychosis results in significant negative consequences to the patient and to his or her quality of life, the potential for benefits of an antipsychotic medication should be weighed against the potential for harmful effects (see section “Potential Benefits and Harms” earlier in this guideline). This is particularly important given the modest benefits and demonstrated risks of antipsychotic treatment in clinical trials and in less rigorous observational and cohort studies. In emergent situations, when there is risk of harm to the patient or others, acute treatment may need to proceed to allow the immediate crisis to be stabilized. However, in other contexts, discussing potential benefits and harms with the patient’s family or other surrogate decision makers and eliciting their concerns, values, and preferences are essential in helping them arrive at an informed decision about treatment that will be person-centered and focused on overall quality of life. Patients may also be able to appreciate these factors and offer input on their current and future treatment preferences depending on their level of cognitive impairment (O’Rourke et al. 2015). Open-ended questioning and discussion will likely be helpful in identifying potential benefits and side effects of treatment that are most important to the person living with dementia. For example, individuals may be particularly concerned about effects of the medication on their remaining capabilities in terms of cognition and communication. On the other hand, calming effects of medication may be viewed as particularly helpful if they ease distressing anxiety or suspiciousness or alleviate aggressive episodes, allowing individuals to remain safely in their homes. If medication calms the individual for even a few hours, it can facilitate attendance at an adult day program, giving them pleasure through program activities and granting a caregiver a few hours of respite. In all settings of care, such preferences of patients, family, and other caregivers should be respected, documented, and reviewed in ongoing discussions as part of the treatment planning process.

The subtype of dementia is another important factor to establish before the potential benefits and risks of antipsychotic treatment are considered (Chare et al. 2014; Mrak and Griffin 2007; Pressman and Miller 2014). For example, in individuals with Lewy body dementia and Parkinson’s disease dementia, the risks of extrapyramidal side effects of antipsychotic medication and the potential for cognitive worsening will be significantly greater than in individuals with other types of dementia (Aarsland et al. 2005; Stinton et al. 2015) and in some instances have been reported to include irreversible cognitive decompensation or death. Although clozapine and quetiapine may be better tolerated than the other antipsychotic medications in these patients, the evidence for efficacy of these agents in treating psychosis is minimal (Stinton et al. 2015). Consequently, it may be better to avoid antipsychotic treatment for the visual hallucinations that are common among individuals with Lewy body dementia and for the psychotic symptoms with Parkinson’s disease and dopamine agonist therapy. Individuals with frontotemporal lobar degeneration may also have a heightened sensitivity to antipsychotic medication (Pijnenburg et al. 2003). Even in individuals with a diagnosis of Alzheimer’s disease, pathological evidence of Lewy body disease may be present (Mrak and Griffin 2007) and warrants review of diagnosis before antipsychotic medications are prescribed.

Other benefits and risks of treatment will relate to the individual characteristics and circumstances of the patient. For example, individuals with preexisting diabetes have an increased risk of hospitalization for hyperglycemia with antipsychotic initiation (Lipscombe et al. 2009), whereas those with preexisting problems with gait may be at an increased risk for falls if they develop extrapyramidal side effects. Lowering of blood pressure and development of orthostasis can also contribute to falls, particularly in combination with use of other medications or dehydration. Other co-occurring conditions such as cerebrovascular disease or cardiac disease may also influence the risk of side effects from antipsychotic medications. On the other hand, if agitation or psychosis is severe and distressing to the patient and can be reduced by judicious treatment with an antipsychotic, some individuals may experience an enhanced quality of life (Beerens et al. 2013) and be able to remain in the community for longer periods of time because of reductions in caregiver burden (Mohamed et al., 2012). When behavioral and psychological symptoms are associated with dangerous behaviors to the individual or to others, treatment with an antipsychotic medication may also be appropriate and can reduce risk.

After a risk-benefit assessment and discussion with the family or other surrogate decision makers, if antipsychotic treatment is clinically indicated on a nonemergent basis, it is important to begin the medication at a low dose. Typical starting doses for frail or older patients will be one-third to one-half the starting dose used to treat psychosis in younger individuals or the smallest size of tablet that is available. Doses should be titrated gradually to the lowest dose associated with clinical response. Factors such as drug-drug interactions, medication half-life, and renal and hepatic function should be taken into consideration during titration of medications to avoid dose adjustments that are too rapid. Because of variations in the metabolism of antipsychotic medications and variations in the time needed to reach steady-state medication levels, it is not possible to predict the time needed to reach an adequate dose of medication for an individual patient. However, doses used in clinical trials in patients with dementia can serve as a guide to the typical dose of medication required with each agent.

As dose titration proceeds and at all points in the course of treatment with an antipsychotic, the clinician will want to assess the patient and obtain information from caregivers about response to treatment, possible medication side effects, and adherence. As described above, use of quantitative measures can be helpful in tracking longitudinal response. Poor adherence may be due to factors such as cost, difficulties with swallowing, resistance to taking medication, or intolerable side effects. If side effects are observed or reported, the nature, frequency, and severity of these side effects will determine whether the risks and benefits of treatment favor ongoing treatment, an attempt at tapering, or immediate discontinuation of the medication. Monitoring for tolerability is also important so that sedation, extrapyramidal effects, gait disturbance, cognitive impairing effects, and other side effects can be minimized. Specific recommendations about the timing of laboratory monitoring have not been developed for individuals with dementia who are being treated with antipsychotic medication; however, in individuals with schizophrenia, it has been suggested that an Abnormal Involuntary Movement Scale (AIMS) be done at least every 6 months in geriatric patients (American Psychiatric Association 2004). Monitoring blood pressure, weight, body mass index (BMI), waist circumference, fasting glucose, fasting lipid profile, and personal/family history have been suggested at baseline for individuals receiving antipsychotic medication, with additional personal/family history and waist circumference annually, blood pressure and fasting plasma glucose at 12 weeks and annually, lipid profile at 12 weeks and every 5 years, and weight with calculation of BMI monthly for 3 months, then quarterly (American Diabetes Association et al. 2004). Hemoglobin A_1C monitoring may be substituted for a fasting glucose level (American Diabetes Association 2015).

If a partial response to antipsychotic treatment occurs, further dose titration may be indicated depending on whether side effects are present and on the relative balance of benefits and harms for the patient. When patients are being treated for psychotic symptoms, relief of distress or associated agitation may occur even though hallucinations or delusions persist. In such circumstances, further dose adjustments may not be necessary and would add to the potential for side effects. If there is no clinically significant response within 4 weeks of reaching a typical therapeutic dose of medication, the medication should be tapered and stopped to avoid potential harms of medication treatment without any offsetting benefit. If severe, dangerous, or significantly distressing symptoms persist, a trial of a different antipsychotic medication may be considered after reevaluation for contributing factors to the patient’s symptoms, additional review of the risks and benefits of treatment, and discussion with the patient and surrogate decision maker, with input from family and other involved individuals.

Even when benefit is apparent, patients’ symptoms and need for an antipsychotic medication may change. Consequently, in an effort to reduce the potential harms of treatment, an attempt should be made to taper the antipsychotic medication within 4 months of treatment initiation. However, earlier attempts at tapering the medication may also be warranted given the ongoing risk of harms with continued treatment.

In the same way that clinical and patient-specific circumstances will require clinical judgment in the decision to initiate treatment with an antipsychotic, the clinician will need to weigh multiple factors in a decision to attempt a taper of medication. Discussion with the patient, surrogate decision maker, family, or others involved with the patient is also important. The aim of such a discussion is to elicit their preferences and concerns as well as to review the initial goals, observed benefits, and side effects of antipsychotic treatment; potential risks of continued exposure to antipsychotics; and past experience with antipsychotic medication trials and tapering attempts. The duration of treatment before an attempt at tapering may depend on the chronicity of the symptom prior to treatment initiation and on the severity and degree of dangerousness of the target symptoms. If the initial reasons for antipsychotic medication treatment are unclear after information is obtained from treating health professionals, medical records, family members, or other sources of collateral information, an earlier attempt at tapering may be warranted. When symptoms have been long-standing or associated with significant physical risks, more caution will be needed in efforts at medication tapering. Similarly, if symptoms have recurred with previous tapering attempts, it may be appropriate to continue treatment without an attempt at tapering. In addition, this recommendation is not intended to apply to individuals with a preexisting psychotic disorder such as schizophrenia for whom ongoing antipsychotic treatment may be necessary. As with decisions about initiating antipsychotic treatment, it is essential to obtain input from patients, family, and other caregivers on an ongoing basis and review their preferences, values, and concerns about continued treatment or tapering in a person-centered fashion.

When a medication taper is attempted, close monitoring will be needed to note signs of recurrent symptoms, with monthly symptom assessments recommended during the taper and for at least 4 months after medication discontinuation. The nature of such assessment may vary and can include face-to-face assessments, telephone contact, or other approaches to following symptoms and behaviors. Again, it can be helpful to use quantitative measures or other structured approaches. If breakthrough symptoms are noted with tapering, this suggests that the benefit of the medication may outweigh the potential risks of continued treatment, that other contributing factors may need to be addressed, or that other nonpharmacological or pharmacological interventions may be indicated.

If an antipsychotic medication is being initiated, a number of factors warrant consideration when a specific agent is being selected. For example, patients, surrogate decision makers, or family members may express a preference for a specific medication or note concerns about specific side effects (e.g., weight gain, diabetes, sedation, additional cognitive impairment). Such preferences should be considered in concert with the other factors noted below. Barriers to choice of specific medications are also common and typically involve regulatory stipulations, cost considerations, formulary coverage, or preauthorization requirements.

The potential side effects of specific medications are also important considerations. In studies using administrative databases that have examined a wide range of antipsychotics, the risk of mortality with an FGA in individuals with dementia was generally greater than the risk with an SGA. Head-to-head comparison data from randomized trials are limited, and the bulk of the available evidence on FGAs relates to haloperidol. Thus, because of the greater risk of harms with haloperidol treatment reported in clinical trials and cohort studies, this medication is not recommended as a first-line agent for nonemergent use in individuals with dementia. On the basis of the available data on harms, it may be preferable to avoid use of other FGAs as well. In emergent situations or in the context of delirium, use of haloperidol may still be appropriate, given its availability in an intravenous and short-acting intramuscular formulation and its relatively rapid onset of action relative to other parenteral antipsychotic medications. However, if longer-term treatment is indicated, a different agent should be chosen as a first-line medication.

Among the SGAs, the choice of a specific medication involves consideration of a number of factors. As described in the sections “Potential Benefits and Harms” earlier in this guideline and “Review of Supporting Research Evidence” in Appendix A, data from randomized placebo-controlled trials suggest efficacy for risperidone in treating psychosis and for risperidone, olanzapine, and aripiprazole in treating agitation. There was insufficient information from trials of quetiapine to determine whether it was efficacious in treating either agitation or psychosis, and it appeared to be no better than placebo in treating behavioral or psychological symptoms of dementia overall. In terms of potential risks, the pooled data from randomized trials indicate a greater risk of mortality with use of an SGA relative to placebo but do not show significant differences in mortality between placebo and individual antipsychotic medications. However, the total number of deaths in each study is small. When pooled placebo-controlled RCT data are considered along with data from larger observational cohort studies and research using administrative databases, the evidence suggests that there may be differences in risk between individual antipsychotic agents, but confidence intervals are overlapping and effects are dose dependent. In addition, the number of individuals who had received aripiprazole was very small relative to the number who had received risperidone or olanzapine. There is no information about the benefits or harms of asenapine, brexpiprazole, cariprazine, clozapine, iloperidone, lurasidone, paliperidone, or ziprasidone in individuals with dementia. The lack of head-to-head comparison data among antipsychotic medications on efficacy and on harms makes it difficult to designate a specific antipsychotic as being most appropriate to use as a first-line agent in treating agitation or psychosis in individuals with dementia.

As with all medication-related decisions, choice of a medication will also depend on factors such as the patient’s prior responses to a specific agent; co-occurring medical conditions; the pharmacokinetic properties of the medication, such as absorption and half-life; and the potential for drug-drug interactions and additive side effects with other medications that the patient is already taking. Some antipsychotic medications have active metabolites of the parent drug that may be relevant in medication selection. For example, norquetiapine has significantly greater anticholinergic side effects than quetiapine; interactions of other medications with quetiapine’s primary metabolic pathway (i.e., cytochrome P450 3A4) can also worsen anticholinergic effects. The side effect profile of a medication is another important factor in selecting a specific agent. In addition to the potential risk of serious adverse events such as mortality or stroke, commonly relevant side effects include sedation, hypotension, cardiac effects (including QTc interval prolongation), extrapyramidal effects, akathisia, falls, dysphagia with associated risk of aspiration pneumonia, effects on seizure threshold, and metabolic effects (including weight gain, diabetes, dyslipidemia, and metabolic syndrome). Anticholinergic effects of antipsychotic medications can worsen cognition or narrow angle glaucoma as well as contribute to urinary retention and constipation. The frequency of these adverse effects will vary depending on the antipsychotic medication that is chosen.

Features that individuals in the expert survey noted may influence their prescribing of specific medications include the long half-life, potential for drug-drug interactions, and partial agonist mechanism of action and rates of akathisia with aripiprazole; greater likelihood of extrapyramidal effects and hyperprolactinemia with risperidone; anticholinergic effects, sedation, metabolic effects, and weight gain with olanzapine; and QTc prolongation and changes in absorption with food for ziprasidone. For individuals with Lewy body dementia or Parkinson’s disease dementia, quetiapine and clozapine were noted as the most appropriate medications because of the risk of worsened motor symptoms with the other antipsychotic agents.

The available formulations of the antipsychotic may also play a role in the medication selection process. For example, for patients who have difficulty swallowing pills, it would be preferable to choose a medication that is available as a rapidly dissolving tablet or oral concentrate formulation. If an intramuscular formulation of antipsychotic is indicated for short-term use in individuals who are unable to take oral medications or in emergent situations, care should be taken to use a short-acting parenteral preparation.

The long-acting injectable decanoate formulation of haloperidol and other long-acting injectable formulations of antipsychotic medications are likely to carry a greater risk of side effects in individuals with dementia. However, individuals with a chronic psychotic disorder, such as schizophrenia, may benefit from treatment with a long-acting injectable antipsychotic medication if they have a history of poor adherence and have tolerated oral formulations of medication. In other selected circumstances, a low dose of a long-acting injectable antipsychotic may aid adherence and minimize struggles over the taking of oral medications. Individuals with a preexisting chronic psychotic illness may also have adherence enhanced by administering long-acting medication. Nevertheless, if used, caution is needed to assure that oral medication is well tolerated before shifting to a long-acting injectable. Furthermore, care must still be taken in dosing of long-acting intramuscular formulations because of aging-related changes in medication pharmacokinetics, changes in body composition, and impairments in renal or hepatic function.