Psychoactive Medication Therapy and Delirium Screening in Skilled Nursing Facilities

Abstract

Background:

A positive delirium screen at skilled-nursing facility (SNF) admission can trigger a simultaneous diagnosis of Alzheimer’s Disease or related dementia (AD/ADRD) and lead to psychoactive medication treatment despite a lack of evidence supporting use.

Methods:

This was a nationwide historical cohort study of 849,086 Medicare enrollees from 2011–2013 who were admitted to the SNF from a hospital without history of dementia. Delirium was determined through positive Confusion Assessment Method screen and incident AD/ADRD through active diagnosis or claims. Cox proportional hazard models predicted risk of receiving one of three psychoactive medications (i.e., antipsychotics, benzodiazepines, antiepileptics) within 7 days of SNF admission and within the entire SNF stay.

Results:

Of 849,086 newly-admitted SNF patients (62.6% female, mean age 78), 6.1% had delirium (of which 35.4% received an incident diagnosis of AD/ADRD); 12.6% received antipsychotics, 30.4% benzodiazepines, and 5.8% antiepileptics. Within 7 days of admission, patients with delirium and incident dementia were more likely to receive an antipsychotic (relative risk [RR] 3.09; 95% confidence interval [CI] 2.99 to 3.20), or a benzodiazepine (RR 1.23; 95% CI 1.19 to 1.27) than patients without either condition. By the end of the SNF stay, patients with both delirium and incident dementia were more likely to receive an antipsychotic (RR 3.04; 95% CI 2.95 to 3.14) and benzodiazepine (RR 1.32; 95% CI 1.29 to 1.36) than patients without either condition.

Conclusion:

In this historical cohort, a positive delirium screen was associated with a higher risk of receiving psychoactive medication within 7 days of SNF admission, particularly in patients with an incident AD/ADRD diagnosis. Future research should examine strategies to reduce inappropriate psychoactive medication prescribing in older adults admitted with delirium to SNFs.

INTRODUCTION

Patients entering skilled nursing facilities (SNF) after hospitalization often have delirium, an acute disorder of attention and cognition associated with poor clinical outcomes.¹ Delirium screening is required at SNF admission. SNF patients who screen positive for delirium at admission have a nearly 3-fold increased risk for receiving a new Alzheimer’s Disease and Alzheimer’s Disease related Dementia (AD/ADRD) diagnosis.¹ Despite the lack of high-quality evidence supporting the use of medication-based strategies to treat either delirium or dementia, psychoactive medication use is highly prevalent in long-stay populations but little has been published about prescribing rates in the first days after a SNF admission.²

The American Geriatrics Society Beers Criteria© do not recommend prescribing psychoactive medications to older adults with delirium or dementia due to their lack of proven benefit and concerns about medication-related cognitive decline and death.³ Psychoactive agents, particularly benzodiazepines, may exacerbate delirium.⁴ Antiepileptics have insufficient evidence demonstrating effectiveness in treating dementia-associated agitation and significant safety concerns.⁵ Clinical practice guidelines for delirium and dementia recommend nonpharmacological interventions first and foremost.^6–8

Data are limited on the psychoactive prescribing that occurs in patients who screen positive for delirium at SNF admission, especially in the context of simultaneously diagnosing AD/ADRD. The purpose of this study was to describe the exposure to psychoactive medications in this group using a sample of newly-admitted SNF patients.

METHODS

Data

This study used 2011–2013 Medicare data: 1) SNF Minimum Data Set (MDS) 3.0; 2) Medicare enrollment files; 3) Medicare Parts A, B claims; 4) all-payer daily dispensing prescription database from approximately half of US SNFs, and 5) SNF Certification and Survey Provider Enhanced Report (CASPER). Details of these data are described in previous work, but the dispensing prescription data are gathered through access agreements between NHs and the largest US long-term care pharmacy provider (Omnicare Inc.).¹

Study population

Our study sample included Medicare enrollees who entered a SNF facility from a hospital and stayed in the facility for at least 7 days (n=1,402,850), ensuring complete data capture. Individuals without prescribing data were excluded (n=334,697) (Supplementary Figure S1.). We conducted sensitivity analyses excluding people with CMS-defined exclusionary conditions for antipsychotics and FDA indications for antiepileptics (Supplementary Table S2).

Study measures

Delirium:

Delirium during the SNF stay was determined by a positive screen for delirium using the Confusion Assessment Method (CAM) diagnostic algorithm. The CAM algorithm requires: acute onset or fluctuating symptoms AND inattention with either disorganized thinking OR altered level of consciousness to establish delirium presence. In the SNF setting, the CAM is conducted as part of the required MDS assessment by a registered nurse or other staff who assess the signs and symptoms of delirium from direct observation of the patient, communications with family members, and a review of the medical record. Research shows trained SNF clinical nursing staff can accurately detect these symptoms and score the CAM, although sensitivity is poor.^{9, 10} A history of delirium was determined by international classification of disease version 9 (ICD-9) diagnosis of delirium in the Medicare claims prior to SNF admission. Because delirium diagnoses are frequently underreported, we included a broad range of delirium-related codes to maximize sensitivity.

AD/ADRD:

We used our previous methods of looking at all Medicare claims for up to two years before the SNF admission^11–14 to determine the presence of AD/ADRD: 1) one ICD-9 diagnosis in the hospital claims, 2) two diagnoses in the outpatient claims, or 3) an active diagnosis in MDS within the first 30 days of the SNF stay (Supplementary Table S1). We defined incident AD/ADRD as occurring after the SNF admission. We excluded patients with prevalent AD/ADRD (n=218,527) defined as occurring before the SNF admission or use of cholinesterase inhibitor within 7 days of SNF admission.

The study sample was categorized into four mutually-exclusive groups based on a positive CAM and new AD/ADRD diagnosis within the first 30 days of the SNF admission: 1) no delirium and no AD/ADRD; 2) no delirium and AD/ADRD; 3) delirium and no AD/ADRD; and 4) delirium and AD/ADRD. Our previous research shows that nearly all positive delirium screens and new AD/ADRD diagnoses occur within the first 30 days of admission.¹

Psychoactive prescribing:

We examined the use of three drug classes commonly prescribed to patients with delirium: antipsychotics, benzodiazepines, and antiepileptics. We excluded both cholinesterase inhibitors and antidepressants from our analysis given they are rarely used to treat delirium but are often prescribed in the SNF to treat AD/ADRD and depression. Each drug was identified through generic medication names (Supplementary Table S3) and national drug codes. For each drug class, we calculated the proportion of persons receiving each medication over the proportion of persons observed. We also distinguished the timing of the dispensed medication as occurring within first 7 days after admission, or by the end of SNF stay. Use within the first 7 SNF days was the primary time endpoint for our analyses. Although patients may already be taking these at the time of SNF admission (e.g., prescribed in hospital), we did not consider pre-SNF use in our analyses.

Covariates:

Demographics and health status diagnoses were extracted from the Medicare enrollment and the first MDS admission assessment. Facility characteristics were extracted from CASPER.¹⁵

Statistical analysis

Descriptive statistics were calculated for patient and facility characteristics as well as the prescribing rates of each study drug class. All multivariable analyses were risk adjusted using the 2012 Diagnostic Cost Group Hierarchical Condition Category (DCG/HCC).¹⁶

We conducted multivariable cox proportional hazard and logistic regression models with generalized estimation equations to account for the nested structure of residents within SNFs using SAS 9.4 (Cary, NC). This study was approved by the Northeastern University institutional review board.

RESULTS

At baseline, 849,086 newly-admitted SNF individuals without prevalent AD/ADRD were included in the analysis. Delirium was recognized in 6.1% (n=51,844). Among patients with a positive CAM delirium screen, 35.4% received a new diagnosis of AD/ADRD (n=18,361) within 30 days of SNF admission. This rate of a new AD/ADRD diagnosis was nearly 4 times higher than in patients without delirium (9.2%, n=73,634). At admission, 10.1% were taking an antipsychotic, 24.7% a benzodiazepine, and 5.2% an antiepileptic.

Table 1 provides characteristics for each of the four groups. The delirium and AD/ADRD group, compared to the other 3 groups, was the oldest (mean age 84.5), the most functionally limited (mean ADL 10.6 [ADL range 0–16; higher values indicate more functional limitation]),¹⁷ and had the highest rates of prior delirium (13.7%). This group also had a much higher rate of severe or moderate cognitive impairment (68.9%) compared to the no delirium and AD/ADRD (38.3%), delirium and no AD/ADRD (44.3%), and no delirium and no AD/ADRD (7.8%) groups.

Table 1.

Characteristics of newly-admitted SNF residents by delirium and dementia

	No Delirium		Delirium
Characteristics	No AD/ADRD Diagnosis	AD/ADRD Diagnosis	No AD/ADRD Diagnosis	AD/ADRD Diagnosis
No. of Residents	723,608	73,634	33,483	18,361
Age, mean (SD)	77.6 (11.10)	84.1 (8.80)	77.9 (12.30)	84.5 (8.90)
Female, n (%)	452652 (62.55)	45858 (62.28)	18737 (55.96)	11178 (60.88)
Non-White, n (%)	117317 (16.21)	11815 (16.05)	5506 (16.44)	2445 (13.32)
ADL Score, mean (SD)	7.8 (4.05)	9.27 (4.19)	9.97 (4.36)	10.59 (4.15)
Depression, n (%)	211924 (29.29)	26706 (36.27)	11134 (33.25)	6541 (35.62)
History of Delirium, n (%)	38058 (5.26)	8351 (11.34)	4001 (11.95)	2522 (13.74)
Medicaid Eligible, n (%)	190320 (26.30)	22589 (30.68)	11697 (34.93)	6206 (33.80)
US Census Region, n (%)
Midwest	196271 (27.12)	17829 (24.21)	10670 (31.87)	5221 (28.44)
Northeast	165125 (22.82)	16795 (22.81)	6555 (19.58)	4025 (21.92)
South	231799 (32.03)	24643 (33.47)	10251 (30.62)	5706 (31.08)
West	121687 (16.82)	13163 (17.88)	5641 (16.85)	3138 (17.09)
CFS Score, n (%)
Severe Impairment	27798 (3.84)	9452 (12.84)	6835 (20.41)	5091 (27.73)
Moderate Impairment	28619 (3.96)	18709 (25.41)	7993 (23.87)	7565 (41.2)
Mild Impairment	109854 (15.18)	22508 (30.57)	10159 (30.34)	4104 (22.35)
Cognitively Intact	557229 (77.01)	22937 (31.15)	8491 (25.36)	1593 (8.68)
Risk Adjustment Score, mean (SD)	1.59 (0.33)	1.50 (0.30)	1.60 (0.36)	1.47 (0.29)
SNF Size, No. of residents, n (%)
<50	62092 (8.58)	5979 (8.12)	2774 (8.28)	1501 (8.17)
50–100	242293 (33.48)	24324 (33.03)	13375 (39.95)	6824 (37.17)
101–250	388320 (53.66)	39550 (53.71)	16305 (48.70)	9265 (50.46)
>250	22178 (3.06)	2577 (3.50)	663 (1.98)	500 (2.72)
Chain ownership, n (%)	480852 (66.45)	48140 (65.38)	24105 (71.99)	12987 (70.73)
For-profit status, n (%)
For Profit	567753 (78.46)	57705 (78.37)	27731 (82.82)	14731 (80.23)
Non-Profit	135021 (18.66)	13408 (18.21)	4784 (14.29)	2933 (15.97)
Government	12109 (1.67)	1317 (1.79)	602 (1.80)	426 (2.32)

Abbreviations: Delirium= positive Confusion Assessment Method, AD/ADRD= incident Alzheimer’s disease and Alzheimer’s Disease related dementias, ADL=activities of daily living, SNF=Skilled nursing facility, CFS=Cognitive Functioning scale, SNF=skilled-nursing facility

Figure 1 highlights the proportion of patients receiving psychoactive medication for each medication class across the four groups and by their duration of SNF stay. Antipsychotics were the psychoactive medication class most strongly associated with delirium and AD/ADRD. Within 7 days of SNF admission, 28.0% of the delirium and AD/ADRD group received an antipsychotic compared 19.8% in the no delirium and AD/ADRD group, 17.8% in the delirium and no AD/ADRD group, and 8.3% in the no delirium and no AD/ADRD group. By SNF discharge, 37.8% of the delirium and AD/ADRD group received an antipsychotic; use in the other three groups ranged from 10.3% to 24.6%.

Figure 1a-c.

Proportion of Newly-Admitted SNF Residents Receiving Psychoactive Medications by Delirium Screen and Duration of Stay

Abbreviations: SNF=skilled nursing facility, AD/ADRD= Alzheimer’s disease and Alzheimer’s Disease related dementias;

*=use of cholinesterase inhibitors in first 7 days is an exclusion criterion.

During the first 7 days of SNF admission, benzodiazepines were widely prescribed across all study groups ranging from 23.7% in the no delirium and AD/ADRD group to 30.2% in the delirium and no AD/ADRD group. Until SNF discharge benzodiazepines increased particularly in the delirium and AD/ADRD (42.9%) and no delirium and AD/ADRD (41.3%) groups. Antiepileptics were the least commonly used during the first 7 SNF days (range 4.2%−6.6%) or by discharge (range 4.8%−7.5%).

After adjustment for patient and SNF characteristics, the risks of receiving psychoactive medications in the first 7 days of the SNF admission is shown in Figure 2a. In the delirium and AD/ADRD group, the relative prescribing risk was greatest for antipsychotics (relative risk [RR] 3.04; 95% confidence interval [CI]: 2.95–3.14), followed by benzodiazepines (RR 1.23; 95% CI: 1.19,1.27); no increased risk was found for antiepileptics (Supplemental Table S4a–f). The risk for antipsychotic or benzodiazepine initiation in the first 7 SNF days in the no delirium and AD/ADRD and delirium and no AD/ADRD groups was also elevated relative to the no delirium and no AD/ADRD group, but less than in the delirium and AD/ADRD group.

Figure 2a-b.

Relative risk of Psychoactive Prescribing by Alzheimer’s disease and related dementia (AD/ADRD) diagnosis and Confusion Assessment Method (CAM) delirium screen and Time in SNF

Abbreviations: SNF=Skilled Nursing Facility; CI=confidence interval; HR=hazard ratio; See supplementary table S4a–f for full models.

Model: Cox proportional hazards regression adjusted for age, sex, race, activities of daily living, risk adjustment score, cognitive functioning scale, history of delirium, Medicaid status, facility size, chain ownership, for-profit status, and missing facility values.

By SNF discharge, patients in the delirium and AD/ADRD group were significantly more likely to be administered an antipsychotic (RR 3.04; 95% CI: 2.95, 3.14) or benzodiazepine (RR 1.32; 95% CI: 1.32 (1.29, 1.36), relative to the no delirium and no AD/ADRD group (Figure 2b). Antiepileptic use was not elevated in this group.

DISCUSSION

Using linked Medicare claims, MDS, and all-payer daily dispensing prescription database on nearly 1 million SNF admissions, we report delirium is independently associated with increases in psychoactive medication prescribing within 7 days of SNF admission. Patients with both delirium and newly diagnosed dementia were over 3 times more likely to receive antipsychotics and 23% more likely to receive benzodiazepines within 7 days of SNF admission, relative to patients without delirium. These rates climb until SNF discharge.

Our study extends the existing literature¹⁸ to document the consequences of clinical recognition of delirium on psychoactive medication prescribing. The pattern observed in this historical cohort suggests that delirium can lead to a cascade of an AD/ADRD diagnosis and potentially inappropriate psychoactive medication therapy. This cascade is inconsistent with the guidance of the fifth edition Diagnostic Statistical Manual of Mental Disorders (DSM-5), which states dementia should not be diagnosed in the face of active delirium¹⁹ and with clinical practice guidelines, which recommend primarily nonpharmacological interventions for delirium.^{7, 8} However, advances in understanding in the field and policy changes may have led to improvements in SNFs since these data were collected. Our analysis provides a baseline understanding.

A strength of this study is the use of the all-payer daily dispensing prescription database. Our dataset uniquely captures all prescriptions dispensed in the facility; data on all medications dispensed during the early days of SNF stay are rare. Medicare Part D does not contain data on prescriptions bundled into the per diem rates of the facility, or prescriptions not covered by the plan, notably benzodiazepines. Our finding of rates up to 40% in our SNF populations are among the highest recorded.

Our finding that most psychoactive prescribing occurs within 7 days of admission suggests that some portion of the medication orders originated in hospital. This pattern of prescribing suggests some missed opportunities for deprescribing upon hospital discharge and SNF admission. However our estimates may not reflect CMS more recent initiatives to reduce antipsychotic prescribing.²⁰ It is critical to recognize and treat underlying conditions known to cause delirium and modify risks before administering psychoactive medications.²¹All the medications prescribed in this study have serious safety concerns including death³ and no rigorous evidence to support benefit in this population.

Our study has limitations. First, the lack of complete clinical data makes the delirium and dementia diagnoses impossible to confirm. Second, indications for psychoactive medications prescribed are not available although sensitivity analyses were conducted on potential indications (Supplementary Tables S5a–S6b). Our assessment of antipsychotic use applying the CMS exclusion criteria found receipt rates of 5.9–23.2% at 7 days and 7.8–32.9% by end of SNF stay. The adjusted models with exclusions are consistent with our main models and do not change our findings. Third, our method to exclude individuals with prevalent dementia may have missed up to 10% of cases; some individuals may have had an undiagnosed dementia before being admitted to the SNF.²² Fourth, our data span 2011–2013, and may be outdated and do not reflect more recent efforts to reduce unnecessary use of antipsychotics. However, at least one recent study has found evidence that declining rates of psychoactive medications in SNFs has reversed and are climbing again.²³ Albeit, not all studies have found this increase. Fifth, our exclusion criteria resulted in omitting 58% of our sample, mainly due to lack of prescription data.

Conclusion

Among older adults without evidence of prior dementia, a positive CAM screen for delirium within 7 days of SNF admission was significantly associated with higher rates of prescribing psychoactive medications. This risk of receiving antipsychotics, and benzodiazepines was highest after a positive CAM delirium screen, and among patients who also received an incident AD/ADRD diagnosis. These findings suggest potentially inappropriate prescribing among older adults admitted to SNFs with delirium, particularly in those also diagnosed with AD/ADRD. If confirmed in future studies, our findings suggest the need for policy and educational interventions to improve delirium assessment and psychoactive medication prescribing in SNF admissions.

Supplementary Material

Appendix

Figure S1. Consort diagram

Table S1. Definitions of measures for dementia, delirium, and dispensed psychoactive medications.

Table S2. Sensitivity Analyses excluding patients with approved indications for each Rx

Table S3. Generic Names of Medications Assessed

Table S4. Main analysis multivariate models; a. Hazard ratios for benzodiazapines-within 7 days of NH entry; b. Hazard ratios for benzodiazapines - EVER DURING STAY; c. Hazard ratios for antipsychotics WITHIN 7 DAYS OF NH ENTRY; d. Hazard ratios for antipsychotics - ANYTIME DURING STAY; e. Hazard ratios for Anti-epileptic - WITHIN 7 DAYS OF NH ENTRY; f. Hazard ratios for Anti-epileptic Rx ANYTIME DURING STAY

Table S5a. Sensitivity analysis: Hazard ratios for antiepileptic use within 7 days of NH entry excluding prevalent cases of seizure disorders N=798,534

Table S5b. Sensitivity analysis: Hazard ratios for antiepileptic use anytime during the NH stay excluding prevalent cases of seizure disorders N=798,534

Table S6a. Sensitivity analysis: Hazard ratios for antipsychotics within 7 days of NH entry, exclude cases of schizophrenia, tourette’s, and huntington, N= 837,833

Table S6b. Sensitivity analysis: Hazard ratios for antipsychotics anytime during the NH stay, exclude cases of schizophrenia, tourette’s, and huntington, N= 837,833

Key Points.

• Baseline data on a historical cohort that shows psychoactive therapy is high among newly-admitted patients in SNFs who screen positive for delirium and receive a new diagnosis of dementia.

• Most psychoactive therapy occurs within the first 7 days of the SNF stay, while delirium may still be unresolved.

Why Does this Paper Matter?

The full scope of psychoactive therapy in SNFs and a positive delirium screen has not been known previously.