Quality Measurement Considerations

The recommendations of this guideline are consistent with several existing Choosing Wisely recommendations. For example, the American Psychiatric Association (2015a) advises, “Don’t prescribe antipsychotic medications to patients for any indication without appropriate initial evaluation and appropriate ongoing monitoring” and “Don’t routinely use antipsychotics as first choice to treat behavioral and psychological symptoms of dementia.” The latter recommendation is echoed by the Choosing Wisely recommendation of the American Geriatrics Society (2015). In addition, two existing process measures relating to the use of antipsychotics in individuals with dementia have been endorsed by the National Quality Forum (NQF) (Pharmacy Quality Alliance 2014). For one of the measures (NQMC-9260), the denominator includes “patients 65 years and older with either a diagnosis of dementia and/or two or more prescription claims and greater than 60 days supply for a cholinesterase inhibitor or an N-methyl-D-aspartate receptor antagonist.” The numerator is defined by “the number of patients in the denominator who had at least one prescription AND greater than 30 days supply for any antipsychotic medication during the measurement period and do not have a diagnosis for schizophrenia, bipolar disorder, Huntington’s disease or Tourette’s syndrome.” The other measure (NQMC-9907) applies to long-stay nursing home residents with dementia who are age 18 years or older and examines the percentage of individuals who have been receiving an antipsychotic medication for 12 days or longer. Again, individuals with a diagnosis of schizophrenia, bipolar disorder, Huntington’s disease, or Tourette’s syndrome are excluded from the measure.

Available administrative data allow calculations of the rates of antipsychotic use in nursing homes (Partnership to Improve Dementia Care in Nursing Homes 2015) and other settings. Such data show significant regional and state-to-state variability; however, they have a number of confounds and do not provide details about the reasons these medications are being prescribed or the severity of symptoms exhibited by the patient. Thus, these data reflect antipsychotic use but, like the currently endorsed NQF measures, do not provide information about appropriate use of antipsychotic medications in individuals with dementia.

In terms of other recommendations, the typical practices of psychiatrists and other health professionals are unknown, but anecdotal observations suggest possible variability across healthcare settings and specialty practices. Such variability could indicate a need to strengthen clinician knowledge, improve training, or increase the time available to assess patients and document decision making. Variability could also indicate a need to address barriers to care such as geographic or socioeconomic differences in the availability of health professionals, skilled staff, specific medications, nonpharmacological interventions, or other care-related resources.

Measures could be developed that focus on the assessment of behavioral and psychological symptoms in individuals with dementia, including the type, frequency, severity, pattern, and timing of symptoms (statement 1), potentially modifiable contributors to symptoms (statement 2), and factors that may influence choices of treatment (statement 2). The use of a quantitative measure (statement 3) would be difficult to implement as a quality measure because available rating scales are primarily designed for research. Less formal approaches to quantitative measurement would be better suited to typical clinical settings. Nevertheless, quantitative measures (statement 3) could be one option of several approaches for documenting symptom type, frequency, severity, pattern, and timing (statement 1). Typically, measures of assessment or screening should be matched to a measure that evaluates follow-up treatment and can therefore affect patient outcomes. Given the weak evidence for efficacy of nonpharmacological and pharmacological treatments for agitation and psychosis in dementia, pairing of a treatment-specific measure may not be appropriate. However, these measures could be paired with a measure relating to the presence of a documented treatment plan (statement 4).

Several recommendations (statements 5, 6, and 7) relate to the decision making that should precede consideration of nonemergency antipsychotic treatment in an individual with dementia. In particular, such treatment should be used only “when symptoms are severe, are dangerous, and/or cause significant distress to the patient” (statement 5), after “reviewing the clinical response to nonpharmacological interventions” (statement 6), and after assessing “the potential risks and benefits from antipsychotic medication” (statement 7). Statement 7 also recommends that “the potential risks and benefits from antipsychotic medication be assessed by the clinician and discussed with the patient (if clinically feasible) as well as with the patient’s surrogate decision maker (if relevant) with input from family or others involved with the patient.” This could be incorporated into the above measure as a process focused internal quality improvement measure, or a family/surrogate-reported satisfaction measure could be developed with patient input obtained, when clinically appropriate. For such measures, the measure denominator would focus on patients who received nonemergency treatment with an antipsychotic medication. Several other recommendations (statements 10 and 12) are related to attempts at tapering and discontinuing antipsychotic medications. Since many patients with dementia exhibit both agitation and psychosis and clinical responses can be subtle, it would be difficult to develop distinct measures to address each of these recommendations. However, a composite measure could be used to determine whether an attempt to taper the antipsychotic had occurred within 4 months of treatment initiation. Statement 11 also focuses on decision making and discussion with the patient, surrogate decision maker, and family, in this case related to tapering of antipsychotic medication in a patient who had experienced a positive response to treatment. The latter inclusion criteria would make it difficult to use this statement as a quality measure.

It may also be possible to develop a measure that assesses the use of haloperidol in individuals with dementia (statement 14). However, such a measure would require documenting whether or not the patient was experiencing delirium, whether or not the use of antipsychotic was on an emergency basis, and whether or not a different antipsychotic medication had been tried and stopped (e.g., due to side effects or lack of efficacy).

Other statements would be difficult or inappropriate to develop into quality measures because of the lack of a discrete and measurable numerator and denominator (statements 8, 9, and 13). Since long-acting injectable antipsychotic medications would be expected to constitute a small fraction of prescribed antipsychotic medications, the impact of a quality measure based on statement 15 is likely to be limited.

For all of these recommendations, there are important practical barriers to the derivation and utility of quality measures. For example, to assess a clinician’s performance of a clinical process, a measure must clearly define the applicable patient group (i.e., the denominator) and the process that is measured (i.e., the numerator). Furthermore, the clinician’s performance of the process must be readily ascertained from chart review or administrative data. When quality measures relate to patient assessment, clinical judgment must be used to determine the factors that merit emphasis in the evaluation of an individual patient. Clinical judgment is also needed to determine the clinical response to nonpharmacological interventions, weigh the potential benefits and harms of antipsychotic treatment, and decide on the appropriate timing of attempts to taper antipsychotic medication.

Additional barriers relate to a lack of standardization in how findings are documented. Information in medical records may be lacking or incomplete; more often it does not fully align with the specific requirements of a particular performance measure. Many clinicians appropriately use free text prose to describe symptoms, response to treatment, discussions with family, plans of treatment, and other aspects of care and clinical decision making. Reviewing these free text records for measurement purposes would be impractical, and it would be inappropriate to hold clinicians accountable to such measures, without significant increases in electronic medical record use and advances in natural language processing. The presence or absence of scoring from a relevant measurement tool could be included as one of several approaches to fulfill a measure that relates to symptom assessment. Another approach could be to measure only for the presence or absence of text in relevant free text fields of an electronic medical record. This approach would allow for maximum flexibility in how clinicians document findings of their assessments; however, a liability of this approach is that it would have limited utility to address variability in how clinicians assess patients with dementia and document treatment planning and clinical decision making. Such an approach could also lead to documentation burden and overuse of standardized language that does not accurately reflect what has occurred in practice. On the other hand, if multiple discrete fields are used to capture information on a paper or electronic record form, oversimplification is a possible unintended consequence of measurement. For example, implementation of a measure relating to haloperidol use (statement 14) would minimally require that a clinician’s medical record capture yes or no answers about current delirium, emergent need for treatment, and prior antipsychotic trials. Not all electronic medical records may do this without costly modifications, and even if they do, information may not be captured in an easily retrievable and reportable format. In addition, crucial clinical information might be lost through this type of documentation (e.g., information on responses or side effects from prior antipsychotic trials).

As a result of these practical barriers, it may be difficult to derive meaningful performance measures from these recommendations. Consequently, quality improvement activities including performance measures derived from these guidelines should yield improvements in quality of care to justify any clinician burden (e.g., documentation burden). Possible unintended consequences of any derived measures would also need to be addressed in testing of a fully specified measure.

In addition to the possible use of these guidelines to develop formal quality measures, these guideline statements can also be used to promote quality care in other ways. For example, quality of care might be improved through educational activities or through electronic clinical decision support. With appropriate controls for case-mix and comorbidities, organizations could examine the effects of the recommendations on overall outcomes (e.g., proportion of individuals with significant behavioral and psychological symptoms of dementia, proportion of individuals experiencing adverse effects of antipsychotic medication, rates of transition from community to nursing care settings). Quality improvement initiatives could then be developed to improve these outcomes.