Search tips
Search criteria 


Logo of nihpaAbout Author manuscriptsSubmit a manuscriptHHS Public Access; Author Manuscript; Accepted for publication in peer reviewed journal;
Arch Intern Med. Author manuscript; available in PMC 2011 January 11.
Published in final edited form as:
PMCID: PMC2897009

Unexplained Variation across U.S. Nursing Homes in Antipsychotic Prescribing Rates



Serious safety concerns related to use of antipsychotics have not decreased the prescribing of these agents to nursing home (NH) residents. We assessed the extent to which resident clinical characteristics and institutional prescribing practice were associated with antipsychotic prescribing.


Antipsychotic prescribing was assessed for a nationwide, cross-sectional population of 16,586 newly-admitted NH residents in 2006. We computed facility-level antipsychotic rates based on the previous year’s (2005) prescribing patterns. Poisson regressions with generalized estimating equations were used to identify the likelihood of resident-level antipsychotic utilization in 2006 given 2005 facility-level prescribing pattern and NH resident indication for antipsychotic therapy (psychosis, dementia, and behavioral disturbance).


Over 30% of study residents received at least one antipsychotic medication in 2006, of whom 32% (n=1605) had no identified clinical indication for this therapy. Residents entering NHs with the highest facility-level antipsychotic rates were 1.34 times more likely to receive antipsychotics relative to those entering the lowest prescribing rate NHs, after adjusting for potential clinical indications (relative risk [RR]=1.34; 95% confidence interval [CI], 1.22–1.47). The elevated risk associated with facility-level prescribing rates was apparent for only NH residents with dementia but no psychosis (RR=1.37; 95%CI, 1.21 to 1.54), and residents without dementia or psychosis (RR=1.51; 95% CI: 1.23–1.86).


NH antipsychotic prescribing practice was independently associated with use of antipsychotics in NH residents. Future research is needed to determine why such a prescribing culture exists and whether it could result in adverse health consequences.


Nearly 1 in 3 nursing-home resident sin the United States received antipsychotic drugs in 2007[1], which is the highest reported level of use in more than a decade. Serious safety concerns related to these agents are increasing.[24] In 2005, the Food and Drug Administration (FDA) issued warnings of excess mortality associated with the use of these antipsychotic agents for behavioral symptoms in elderly patients with dementia.[2] Furthermore, a large NIH-sponsored clinical trial [5] recently concluded that the adverse side effects of atypical antipsychotics offset their advantages in older adults with Alzheimer's disease. [6]

Recently published data from Ontario indicate that there is wide variation in the proportion of NH residents on antipsychotics.[7] This prescribing variation may reflect differences in the patient case-mix and the prevalence of diagnoses of psychoses or dementia with severe behavior problems. Alternatively, antipsychotic use may also be driven by NHs’ facility-level antipsychotic prescribing rates.[7] Such prescribing patterns may be considered a “visible artifact of deeper cognitive processes shared by organizational members”,[8] and perhaps indicate an institutional prescribing culture. Although previous work supports the role of facility-level factors in resident-level prescribing in Canada [7], the extent to which this exists in the US in unclear.

The objective of the present study is to examine the association between facility-level antipsychotic rates and the use of antipsychotics among individual nursing home residents in the U. S. We hypothesize that residents who enter NHs with high facility-level antipsychotic prescribing rates would be more likely to receive antipsychotic medications, independent of their clinical indications for this treatment. Findings from this study will help inform policies to target NHs with high prescribing rates of antipsychotics.



This study used two merged data sources: 1) a nation-wide database of 2005–2006 NH prescription drug dispensing records; and 2) the 2006 MDS. Both datasets are previously described in detail. [9] The prescription data come from the pharmacy claims of over 2.5 million unique individuals living in approximately 16,000 NHs from across 48 states. Pharmacy claims include a variety of prescription drug plans (including private insurance, Medicaid, Medicare Part D) and those without insurance coverage for drugs. The drug dispensing data include all medications prescribed and administered to NH residents, including over-the-counter drugs and drugs administered on an as needed basis. Data elements include the National Drug Code, dispensing date, and the state where the NH was located. Linkable Minimum Data Set (MDS) records were available for about one-third of individuals with prescription data. The MDS is a federally-mandated health assessment tool used in U.S. NHs that captures over 300 items about a residents’ physical and cognitive functioning. Full assessments occur upon admission, significant change in clinical status, and annually. Most elements of the MDS demonstrate excellent to good reliability.[10, 11]

Study sample

The sampling frame for this study was 66,181 NH residents newly-admitted in 2006 who had at least one drug dispensing record. We excluded 46,610 short-stay residents, defined as individuals having fewer than three consecutive months of a NH stay, because previous research have shown these individuals differ from long-stay residents.[12, 13] Furthermore, we excluded residents living in small NHs with fewer than 5 residents (n=2,985). The final sample size was 16,586 residents admitted to 1,257 NHs.

NH facility-level antipsychotic prescribing rate and characteristics

To measure a NH facility-level antipsychotic prescribing rate, we adopted a method developed by Rochon et al. [7] For each NH with newly-admitted residents in 2006, we examined the records of prescription drugs dispensed for all resident sin that NH in the previous year (i.e. 2005). We used 2005 data for this calculation in order to establish independence of facility-level antipsychotic prescribing rate from the actual use of antipsychotic drugs for newly-admitted residents. This also established the temporal relation ship between the two variables. The facility-level antipsychotic prescribing rate was defined as the proportion of long-stay residents in the NH receiving at least one antipsychotic prescription in 2005. For 1,473 residents with stays in multiple NHs, we used their first NH stay for this analysis. Based on the distribution of antipsychotic prescribing rates, NHs were categorized into quintiles (Quintile 1 to Quintile 5, noted as Q1 to Q5 in the later text) of facility prescribing rates. As a sensitivity analysis, we recalculated the antipsychotic rates including short-stay residents and found that the quintile assignments remained substantially similar. (Data not shown but available upon request). NH characteristics included number of long-stay residents in the NH during 2005 and location as categorized by U.S. census region.

Resident characteristics

Resident characteristics were drawn from the first MDS admission assessment in 2006. Demographics included age, sex, marital status, and race/ethnicity. We calculated the MDS-Changes in Health, End-stage disease and Symptoms and Signs score (CHESS) to measure frailty of the resident. [14] The CHESS score, ranging from 0 (no frailty) to 5 (high frailty), is a strong predictor of mortality and health instability in NH residents. [14] Severity of behavioral problems was measured by the Behavioral Index, which is based on the frequency and number of behaviors including wandering, and being verbally or physically abusive and socially inappropriate. [1517] Ranging from 0 to 2, the Behavioral Index was categorized into normal/mild (0 or 1) and moderate/severe behavioral problems (2). Cognitive impairment was assessed by the Cognitive Performance Scale (CPS) and categorized as minimal (0–1), moderate (2–3), and severe (4–6). [18] The CPS has been shown to be highly correlated with the Mini-Mental State Exam (MMSE). [10, 18] Residents were classified as having dementia if there was a diagnosis of Alzheimer's disease (AD), dementia other than AD, or if they received a prescription for an acetyl cholinesterase inhibitor, an ergot alkaloid or non-competitive N-methyl d-aspartate receptor antagonist (i.e. memantine). Based on a method described by Oliveria et al. [19] we defined residents as having psychoses if they were diagnosed with schizophrenia, schizoaffective disorder, mood disorder with psychotic features, psychotic symptoms accounted for by a substance, or current major depressive episode with psychotic symptoms of hallucinations or delusions. As behavioral components were used to compute variables such as the Behavioral Index, CPS and dementia, it was possible that they were correlated. We performed Pearson correlation coefficient to examine the correlation among the three variables and found the highest correlation coefficient was lower than 0.45. Therefore, all three variables were included in the analysis.

Statistical analyses

Descriptive statistics were performed to examine the distribution of the facility and individual characteristics among the facility-level antipsychotic quintiles and between individuals with and without antipsychotic therapy. Chi square tests were used to compare proportions.

Risk ratios of antipsychotic therapy for calendar year 2006 were estimated using Poisson regression and generalized estimating equations. This modeling approach does not require a rare-disease assumption, provides valid confidence intervals using robust estimation,[20] and adjusts for clustering of residents within a facility. Unadjusted (Model 1) and adjusted (Models 2–4) models were conducted in the overall population and in mutually exclusive clinical subgroups indicated for antipsychotic therapy: 1) residents with psychosis, 2) residents with dementia and no psychosis, 3) residents without psychosis or dementia. The first two groups were considered potential clinical indication groups while the last as non-indication group. Adjusted models included successive sets of additional covariates: Model 2-facility characteristics (facility size and region);Model 3-resident characteristics, including demographics (age, gender, marital status and race/ethnicity) and general health status; and Model 4-antipsychotic indications (Behavioral Index, indicator of dementia, and indicator of psychosis). Note that Model 4 was defined individually for each clinical subgroup. We used such a stepwise approach in order to: 1) separate the effect of NH characteristics from resident characteristics and 2) separate the effect of antipsychotic indications from non-antipsychotic indications.

STATA 10.0 (STATA Corp, Texas) was used to conduct all statistical analyses and p value < 0.05 was considered statistically significant.


We indentified 16,586 long-stay NH residents who were newly-admitted to 1,257 NHs in 2006. The facility-level antipsychotic prescribing rates of these NHs in the preceding year ranged from 0%–24.4% in Q1 to 43.8%–100% in Q5. About 30% (5,005) of all residents received at least one antipsychotic medication in 2006. Residents with psychosis (n=972) had the highest level of use with 77% using at least one antipsychotic, followed by residents with dementia and no psychosis (43%, n=2,694) and then residents without dementia or psychosis (17%, n=9,426). (Figure 1) About 32% (1,605/5005) of antipsychotics were dispensed to residents without any clinical indication.

Figure 1
Use of antipsychotics in 2006

Table 1 describes the study NHs by their antipsychotic prescribing rates quintiles in 2005. Comparing Q5 and Q1 NHs, we found that a higher proportion of Q5 NHs were more likely to be located in the South and have fewer than 250 residents. A higher proportion ofQ1 NHs were located in the Mid-West, and had more than 250 residents.

Table 1
Characteristics of nursing homes by facility-level antipsychotic prescribing rate

Table 2 describes the study population by quintiles of facility-level antipsychotic prescribing rates. Compared to residents in Q5 NHs, those in Q1 NHs tended to be older (age > 75: 75% vs. 59%, p<0.001), female (69.4% vs. 60.9%, p<0.001), and white (85% vs. 71.9%, p<0.001). More residents in Q1 NHs were frail (CHESS 3–5: 24.7% vs. 14.0%, p<0.001) and had a higher CPS score (CPS 0–1: 39.2% vs. 31.4%, p<0.001) than residents in Q5 NHs. More residents in Q5 NHs had moderate or severe behavioral problems (Behavioral Index moderate/severe: 23.5%vs. 12.6%, p<0.001), dementia (52.3% vs. 41.4% p<0.001) and psychosis (10.3% vs. 4.0% p<0.001), as compared to those in Q1 NHs.

Table 2
Facility-level antipsychotic prescribing rate quintile by resident characteristics

Table 3 shows the distribution individual characteristics between antipsychotic users and non-users. Residents who were prescribed antipsychotic medications were younger (age ≥ 65: 13.8% vs. 10.1%; 66–75: 20.1 vs. 18.7, p<0.001), male (37.8% vs. 33.9%, p<0.001) and less frail (CHESS score of 0: 25.1% vs. 17.4%, p<0.001) as compared to those who were not on antipsychotics. Antipsychotics tended to be given to residents with moderate and severe behavioral problems (32.3% vs. 8.5%, p<0.001), dementia (68.7% vs. 36.5%, p<0.001) and psychosis (15.1% vs. 1.9%, p<0.001).

Table 3
Resident-level use of antipsychotics by resident characteristics

The association between resident use of antipsychotics and facility-level prescribing rates in the full sample is shown in Table 4. Residents in Q5 NHs had nearly double the relative risk (RR)of receiving antipsychotics (Model 1: RR=1.94, 95% Confidence Interval [CI]: 1.75–2.15) compared to residents in Q1 NHs. Adjusting for NH characteristics did not change the magnitude of the association (Q5 vs. Q1 Model 2: RR=1.91, 95%CI: 1.71–2.13). Adjusting for demographics and health status, reduced the RR (Q5 vs. Q1 Model 3: RR=1.62, 95% CI: 1.46–1.80). The RR was further reduced after controlling for potential indication of antipsychotics; however, compared with residents in Q1 NHs, those in Q5 NHs still had higher risk of being prescribed antipsychotics (Model 4, RR=1.34, 95% CI: 1.22–1.47).

Table 4
Adjusted relative risks of resident-level use of antipsychotics by facility-level antipsychotic prescribing rate, according to resident clinical subgroups **

Table 4 also shows that antipsychotic use across the quintiles of facility-level prescribing rates varied by clinical subgroups. Among the residents with psychosis, antipsychotic use did not vary significantly across quintile sin the full model. After adjusting for all covariate sets (Model 4 for psychosis), the RR forQ5 to Q1 NHs was 1.10 (95% CI, 0.96–1.25).

However, facility-level prescribing quintile did predict use of antipsychotics for the two other clinical subgroups. For residents with dementia and no psychosis, those residing in Q5 NHs were more likely to be prescribed antipsychotics (Model1: RR=1.62; 95% CI: 1.43–1.78) relative to those in Q1 NHs, and the magnitude and significance of RR changed little after adjusting for facility characteristics (Model 2: RR=1.56, 95% CI 1.38–1.77), and then for demographics and health status (Model 3: RR=1.47, 95% CI 1.29–1.66). After adjusting for Behavioral Index, the RR was still significant (Model 4: RR=1.37, 95% CI 1.21–1.54).

Among the residents without psychosis or dementia, facility-level prescribing quintile was significantly associated with use of antipsychotics (Q5 vs. Q1 Model 1: RR=1.76, 95% CI: 1.43–2.17). After full adjustment the association remained statistically significant (Q5 vs. Q1 Model 5: RR=1.51; 95% CI: 1.23–1.86).


This study provides evidence of a facility-level variation in the prescribing of antipsychotics in U.S. NHs. We found that the likelihood of a newly-admitted NH resident to receive an antipsychotic was strongly and independently related to the facility-level antipsychotic prescribing rate, even after adjustment for clinical and socio demographic characteristics. Residents newly admitted to NHs with the highest prescribing rates were 1.34 times more likely to receive an antipsychotic medication relative to those in the NHs with the lowest prescribing rates. The influence of the facility-level prescribing rate was most apparent in residents without psychosis, who have the weakest indication for antipsychotic use.

Another important finding in this study is the high use of antipsychotics in NHs in the period after the 2005 FDA mortality warnings for antipsychotic agents. Our finding that 30% of newly-admitted NH residents received antipsychotic medication sin 2006 is corroborated by other sources,[3] including a sample of 8 states in 2006 reporting antipsychotic prevalence of 27.6% among NH residents[21] and a sample from Canada.[7]

The high use of antipsychotics may reflect a growing proportion of NH residents diagnosed with psychoses.[21] However, it was shown that residents diagnosed with schizophrenia, bipolar disorder, or aggressive behavioral symptoms of dementia accounted for only a small proportion of antipsychotic use. [21] In addition, we found that 17% of residents who had no clinical indication for antipsychotic therapy (no psychoses and no dementia) received antipsychotic medications. Ad hoc analyses to isolate the role of behavior on the use of antipsychotics showed that the risk of receiving antipsychotics steadily increased with higher facility-level prescribing rates, but only for residents with dementia and normal/mild behavior problems. In contrast, this association was not evident for residents with dementia and moderate/severe behavior problems (data not shown but available upon request). This suggests that managing behavioral problems plays an important role in facility-level decisions about antipsychotic prescribing.

Our study suggests that facility-level factors such as organizational culture may playa role in medication prescribing, and is consistent with previous studies suggesting such influence in the use of antipsychotics in Canada [7] and feeding tubes for NH residents with dementia[22]. There has been a growing interest in the role of organizational culture about medication prescribing in nursing homes. [8, 23] Organizational culture is a broad concept that encompasses the shared values, beliefs and assumptions of a group or members within a group, such as a NH and the NH’s clinicians and staff.[24] The perceptions shared by individuals working within a NH may exhibit itself as a facility-level preference for certain therapeutic modalities. Organizational culture may be particularly important in the use of antipsychotics in NHs since prescribing decisions often occur in NHs without direct contact between the prescriber and resident.

The study was subject to limitations. First, this is a cross-sectional study, thus we are not able to draw conclusions about casual relationships. Second, our results may not be generalizable to all Medicare enrollees in NHs as the data are from a single long term care pharmacy. However, a comparison of the geographic residence of our study sample to that of the nursing home residents in the December 2006 CMS OSCAR Data survey shows a similar distribution (Northeast: 24% vs. 23%; Midwest 36% vs. 29%; South 28% vs. 34%, and West 11% vs. 14%).[25] Third, we have excluded NHs with fewer than 5 residents in 2005 because their antipsychotic rates were unstable due to the small number of residents. We further excluded short-stay residents because of their distinct characteristics from long-stay residents. Limiting our study sample therefore prevents us from extending the interpretation of our findings to smaller facilities. Fourth, the prevalence of psychoses in our sample was lower compared to that found in another study using medical records.[19] Thus, we may have underestimated the prevalence of psychoses in this sample. Finally, due to data limitation, we may not have measured potentially important facility-level factors, such as staff-to-resident ratios, which have been previously linked to prescribing patterns in NHs.[26]


Safety concerns continue to persist in the use of antipsychotic medications in NH residents whose benefits from these agents are unclear. It appears that the risky prescribing of antipsychotics in residents with an uncertain indication seems to be a practice norm in some NHs. This study provides evidence that antipsychotic prescribing varies by NHs, independent of residents’ clinical characteristics, and NHs antipsychotic prescribing culture may be an important component to explain such variation. Future research is needed to determine why such a prescribing culture exists and whether there a read verse health consequences as a result of our observed facility-level antipsychotic prescribing rate. This study may also inform future policies to target NHs with high antipsychotic prescribing rates in order to improve quality of care for NH residents.


1. Lagnado L. Nursing Homes Struggle To Kick Drug Habit. 2007. December 2007]; Available from:
2. FDA. Public Health Advisory: Death with antipsychotics in elderly. Rockville, MD: 2005.
3. Briesacher BA, et al. The quality of antipsychotic drug prescribing in nursing homes. Arch Intern Med. 2005;165(11):1280–1285. [PubMed]
4. Schneider LS, Dagerman KS, Insel P. Risk of death with atypical antipsychotic drug treatment for dementia: meta-analysis of randomized placebo-controlled trials. JAMA. 2005;294(15):1934–1943. [PubMed]
5. Schneider LS, et al. Clinical Antipsychotic Trials of Intervention Effectiveness (CATIE): Alzheimer's disease trial. Schizophr Bull. 2003;29(1):57–72. [PubMed]
6. Schneider LS, et al. Effectiveness of atypical antipsychotic drugs in patients with Alzheimer's disease. N Engl J Med. 2006;355(15):1525–1538. [PubMed]
7. Rochon PA, et al. Variation in nursing home antipsychotic prescribing rates. Arch Intern Med. 2007;167(7):676–683. [PubMed]
8. Hughes CM, et al. Does organisational culture influence prescribing in care homes for older people? A new direction for research. Drugs Aging. 2007;24(2):81–93. [PubMed]
9. Briesacher BA, et al. Nursing Home Residents and Enrollment in Medicare Part D. Journal of the American Geriatrics Society. accepted April 2009. [PMC free article] [PubMed]
10. Hawes C, et al. Reliability estimates for the Minimum Data Set for nursing home resident assessment and care screening (MDS) Gerontologist. 1995;35(2):172–178. [PubMed]
11. Snowden M, et al. Validity and responsiveness of the Minimum Data Set. J Am Geriatr Soc. 1999;47(8):1000–1004. [PubMed]
12. Doshi JA, Shaffer T, Briesacher BA. National estimates of medication use in nursing homes: findings from the 1997 medicare current beneficiary survey and the 1996 medical expenditure survey. J Am Geriatr Soc. 2005;53(3):438–443. [PubMed]
13. Keeler EB, Kane RL, Solomon DH. Short- and long-term residents of nursing homes. Med Care. 1981;19(3):363–370. [PubMed]
14. Hirdes JP, Frijters DH, Teare GF. The MDS-CHESS scale: a new measure to predict mortality in institutionalized older people. J Am Geriatr Soc. 2003;51(1):96–100. [PubMed]
15. Liperoti R, et al. Conventional and atypical antipsychotics and the risk of hospitalization for ventricular arrhythmias or cardiac arrest. Arch Intern Med. 2005;165(6):696–701. [PubMed]
16. Liperoti R, et al. The use of atypical antipsychotics in nursing homes. J Clin Psychiatry. 2003;64(9):1106–1112. [PubMed]
17. Snowden M, Sato K, Roy-Byrne P. Assessment and treatment of nursing home residents with depression or behavioral symptoms associated with dementia: a review of the literature. J Am Geriatr Soc. 2003;51(9):1305–1317. [PubMed]
18. Morris JN, et al. MDS Cognitive Performance Scale. J Gerontol. 1994;49(4):M174–M182. [PubMed]
19. Oliveria SA, et al. Adverse events among nursing home residents with Alzheimer's disease and psychosis. Pharmacoepidemiol Drug Saf. 2006;15(11):763–774. [PubMed]
20. Greenland S. Model-based estimation of relative risks and other epidemiologic measures in studies of common outcomes and in case-control studies. Am J Epidemiol. 2004;160(4):301–305. [PubMed]
21. Crystal S, et al. Broadened Use Of Atypical Antipsychotics: Safety, Effectiveness, And Policy Challenges. Health Aff (Millwood) 2009 [PMC free article] [PubMed]
22. Mitchell SL, et al. Clinical and organizational factors associated with feeding tube use among nursing home residents with advanced cognitive impairment. JAMA. 2003;290(1):73–80. [PubMed]
23. Gruneir A, Lapane KL. It is time to assess the role of organizational culture in nursing home prescribing patterns. Arch Intern Med. 2008;168(2):238–239. author reply 239. [PubMed]
24. Davies HT, Nutley SM, Mannion R. Organisational culture and quality of health care. Qual Health Care. 2000;9(2):111–119. [PMC free article] [PubMed]
25. Nursing Facility Total, Average and Median Number of Patients Per Facility and ADL Dependence in CMS OSCAR Data Current Surveys. Washington, DC: American Health Care Association; 2006. H.S.R.a.E Group, Editor.
26. Lau DT, et al. Potentially inappropriate medication prescriptions among elderly nursing home residents: their scope and associated resident and facility characteristics. Health Serv Res. 2004;39(5):1257–1276. [PMC free article] [PubMed]