- Email: [email protected]
Preventable Hospitalizations Among Nursing Home Residents With Dementia and Behavioral Health Disorders
JAMDA 20 (2019) 1280e1286
JAMDA journal homepage: www.jamda.com
Original Study
Preventable Hospitalizations Among Nursing Home Residents With Dementia and Behavioral Health Disorders Helena Temkin-Greener PhD a, *, Xi Cen MS a, Michael J. Hasselberg MS, PhD b, Yue Li PhD a a b
Department of Public Health Sciences, University of Rochester School of Medicine and Dentistry, Rochester, NY Department of Psychiatry, University of Rochester School of Medicine and Dentistry, Rochester, NY
a b s t r a c t Keywords: Avoidable hospitalizations dementia behavioral disorders nursing homes
Objective: Nursing home (NH) residents with Alzheimer’s disease/related dementias (ADRD) and/or behavioral health disorders (BHD) are at high risk of hospitalizations, many of which are potentially avoidable. Empirical evidence regarding potentially avoidable hospitalizations (PAHs) among these residents is quite sparse and mixed. The objectives of this study were to (1) examine the risk of PAH among residents with ADRD only, BHD only, ADRD and BHD compared to residents with neither and (2) identify associations between individual- and facility-level factors and PAH in these subgroups. Design: Retrospective, CY2014-2015. Setting and Participants: Long-term residents age 65þ (N ¼ 807,630) residing in 15,234 NHs. Methods: We employed the Minimum Data Set, MedPAR, Medicare beneficiary summary, and Nursing Home Compare. Hospitalization risk was the outcome of interest. Individual-level covariates were used to adjust for health conditions. Facility-level covariates and state dummies were included. Multinomial logistic regression models were fit to estimate the risk of PAH and nonepotentially avoidable hospitalizations (N-PAH). Results: Compared to residents without ADRD or BHD, those with ADRD had at least a 10% lower relative risk ratio (RRR) of N-PAH and a significantly lower risk of PAH, at 16% (P < .0001). Residents with BHD only had a statistically higher, but clinically very modest (RRR ¼ 1.03) risk of N-PAH, with no difference in the risk of PAH. Focusing on specific BHD conditions, we found no difference in N-PAH or PAH among residents with depression, lower PAH risk among those with schizophrenia/psychosis (RRR ¼ 0.92), and an increased risk of both N-PAH (RRR ¼ 1.15) and PAH (RRR ¼ 1.09) among residents with bipolar disorders. Conclusions and Relevance: We observed a lower risk of PAH and N-PAH among residents with ADRD, with the risk for residents with BHD varying by condition. Substantial variations in PAH and N-PAH were evident across states. Future research is needed to identify state-level modifiable factors that explain these variations. Ó 2019 AMDA e The Society for Post-Acute and Long-Term Care Medicine.
Research suggests that 60% to 90% of nursing home (NH) residents have a diagnosis of Alzheimer’s or related dementias (ADRD) and/or behavioral health disorders (BHD; eg, schizophrenia, bipolar disorder, depression, and substance abuse).1e3 Residents with these conditions
This work was supported by the Donaghue Medical Research Foundation’s 2016 Another Look Program. The authors declare no conflicts of interest. * Address correspondence to Helena Temkin-Greener, PhD, Department of Public Health Sciences, University of Rochester School of Medicine and Dentistry, 265 Crittenden Blvd, CU 420644, Rochester, NY 14642. E-mail address: [email protected] (H. TemkinGreener). https://doi.org/10.1016/j.jamda.2019.03.006 1525-8610/Ó 2019 AMDA e The Society for Post-Acute and Long-Term Care Medicine.
experience disparities in accessing high-quality facilities,4 and within a NH, availability of effective mental health treatments is often lacking.5,6 Faced with serious physical health problems as well, these residents are at high risk of hospitalizations, which tend to exacerbate iatrogenic illness, functional decline, and mortality.7 Approximately one-fourth of NH residents are hospitalized annually,8 and at least half of these hospitalizations are considered to be potentially avoidable, that is, either preventable or manageable within NHs.7,9,10 Policy makers have focused considerable attention on factors influencing the potentially avoidable hospitalizations (PAHs) among NH residents in an effort to improve the quality of care and reduce Medicare costs. The most recent such effort introduced by the Centers for Medicare &
H. Temkin-Greener et al. / JAMDA 20 (2019) 1280e1286
Medicaid Services (CMS) in 2012 included multiyear interventions but produced mixed results, with NHs in 2 of 7 participating states showing consistent 2-year declines in PAH.11 For residents with ADRD and/or BHD, NHs may have difficulty comanaging behavioral and physical comorbidities, because on-site capacity to provide and/or manage mental health services is often limited.12 NH nurses and social workers have reported that when residents with ADRD and other behavioral health needs “act out,” they are more likely to be hospitalized.13 But empirical evidence on PAHs among NH residents with ADRD and/or BHD is quite sparse and mixed. One study found no statistically significant difference in the occurrence of PAHs among residents with (23.3%) and without (22.0%) dementia.14 In the last year of life, although the rates of PAH were considerably higher, there was still no difference between the 2 subpopulations. A Massachusetts study of Medicare/Medicaid NH residents with dementia found lower PAH rates among those with dementia compared with other residents, but only in facilities with overall lower rates of hospitalizations.15 A Florida study focusing on time to first PAH among residents with dementia and mental disorders found that having dementia, major psychotic, bipolar, depressive, or substance abuse disorder increased the risk of PAH compared to residents without these conditions. Motivated by this sparse and conflicting literature, our objectives were to (1) examine and compare the risk of PAH among all NH residents with ADRD only, BHD only, ADRD and BHD, and neither and (2) identify associations between individual- and facility-level characteristics and PAH in these subgroups. Methods Study Design and Data Sources This was a retrospective study using nationwide assessment and claims data for CY2014-2015. We used 100% Minimum Data Set (MDS) to identify long-term residents. The MDS contains clinical assessments for all residents at admission, quarterly, annually, when there is a significant change in health status, and at discharge. Assessments include information on residents’ sociodemographics, diagnoses,
1281
treatments, behaviors, and functional and cognitive impairments. The MDS data, including the behavioral health items, have been found to be of high research quality.16e18 We used the Medicare beneficiary summary file to identify and exclude the Medicare-Advantage plan enrollees for whom hospitalization data in the Medicare claims may not be accurate. The Medicare Provider and Analysis Review (MedPAR) file was used to track inpatient admissions and identify those that were potentially avoidable. We used the CMS Nursing Home Compare data to obtain information on facility characteristics. Medicare beneficiary and provider identifiers were used to merge these databases. The study protocol was reviewed and approved by the Institutional Research and Subjects Review Board. Study Population Based on the MDS for the second quarter of CY2014, we identified residents with a quarterly, annual, or a significant change in status assessments, and those with a re-entry record for a prior long-term stay, as being long-stay residents. After excluding the Medicare Advantage enrollees and beneficiaries younger than 65 years of age, the analytical file included 807,630 long-stay residents in 15,234 homes. For this population, we tracked inpatient admissions from the second quarter of 2014 through the second quarter of 2015. Outcome Variable The outcome of interest was the risk of PAH. We used the definition of PAH developed by the RTI International10 and most recently employed in the evaluation of the CMS Initiative to Reduce Potentially Avoidable Hospitalizations among Nursing Facility Residents.11 We focused on 17 conditions (Table 1) deemed potentially preventable and/or manageable (https://www.cms.gov/ResearchStatistics-Data-and-Systems/Statistics-Trends-and-Reports/Reports/ downloads/costdriverstask2.pdf). We modeled preventable and manageable conditions separately and also combined as a single outcome. The results for separate conditions were not measurably different from the model in which they were combined, and because
Table 1 Distribution of Potentially Avoidable Hospitalizations by Condition Diagnostic Condition
Potentially Avoidable Hospitalizations Total Number
Lower respiratory: pneumonia bronchitis Urinary tract infection Falls and trauma Congestive heart failure Dehydration, acute renal failure, hypokalemia, hyponatremia Chronic obstructive pulmonary disease, asthma, chronic bronchitis Cellulitis Diarrhea, gastroenteritis with nausea and vomiting Seizures Hypertension, hypotension Skin ulcers including pressure ulcers Hyper- and hypoglycemia: diabetes mellitus with ketoacidosis or hyperosmolar coma Anemia Constipation or fecal impaction obstipation Weight loss, nutritional deficiencies, adult failure to thrive Psychosis, severe agitation, organic brain syndrome Altered mental status/acute confusion/delirium
Percent
Preventable/Manageable
25,508 23,784 20,849 19,330 17,985
18.6 17.3 15.2 14.1 13.1
N/Y Y/Y Y/N Y/Y Y/Y
7868
5.7
Y/Y
6340 3663
4.6 2.7
N/Y N/Y
2888 2009 1836 1669
2.1 1.5 1.3 1.2
Y/Y Y/Y Y/Y Y/Y
1343 679 585
1.0 0.5 0.4
N/Y Y/Y Y/N
557
0.4
N/Y
396 137,289
0.3 100.0
Y/N
1282
H. Temkin-Greener et al. / JAMDA 20 (2019) 1280e1286
(ABS ¼ 0), mild (ABS ¼ 1-2), moderate (ABS ¼ 3-5), and severe (ABS ¼ 6-12).1,21 Mood distress was measured using PHQ-9 assessing both symptom presence and frequency (scale 0-27), coded as none (<¼4), mild (5-9), moderate (10-14), moderate to severe (15-19), and severe (20-27). Number of comorbid chronic conditions (heart failure, chronic obstructive pulmonary disease, diabetes, stroke/transient ischemic attack, ischemic heart disease) was included as a count variable (0-5). We controlled for sociodemographic factors such as age, gender, race (non-Hispanic white, non-Hispanic black, Hispanic, non-Hispanic Asian, other), marital status, and dual Medicare/Medicaid eligibility (yes/no). To examine associations with facility-level characteristics, we included NH size (beds), occupancy rate, percentage of Medicare and Medicaid residents, and skilled care mix (ratio of registered nurses to licensed practical nurses and certified nurse assistants). Ownership was dichotomized as nonprofit or for-profit, and chain or not. We controlled for facility quality using CMS 5-star metrics based on overall and staffing quality, categorizing NHs as 1-, 2- to 4-, or 5-star.
of the significant overlap between preventable and manageable conditions, we only present the results on PAH as a single outcome. Independent Variables The key independent variables of interest were presence of (1) BHD only, (2) ADRD only, and (3) ADRD with comorbid BHDs. The diagnoses constituting BHD were schizophrenia or psychosis, bipolar disorder, depression or anxiety, personality disorder, and substance abuse. Based on prior literature,4,19 these conditions were hierarchically assigned in this order. Conditions were identified from the MDS section I (active diagnoses) and included both prepopulated fields and a search for specific ICD-9 codes. All conditions were identified from the last available assessment in the second quarter of 2014 (baseline). Conditions defined as PAH are based solely on diagnoses and do not reflect illness severity. Therefore, we also included variables reflecting the residents’ clinical conditions and comorbidities at baseline. Residents’ functional status was measured by the number of impairments in activities of daily living (ADL) based on the score ranging from 0 to 28 (higher score signifying greater impairment). We assessed cognitive function using a 0 to 3 scale.20 Aggressive behaviors were measured using 4 items (MDS section E; Agressive Behavior Scale [ABS]) reflecting both the prevalence and the frequency of these behaviors in the last week on a scale from 0 to 12, and coded as none
Statistical Analyses Descriptive analyses were conducted across resident groups characterized by the presence/absence of ADRD and/or BHD. We compared these groups on sociodemographic, functional, clinical, and
Table 2 Descriptive Characteristics of the Study Population at Baseline: By Dementia (ADRD) and Behavioral Health Disorders (BHD) Characteristics Hospitalizations Residents with at least 1 hospitalization Residents with at least 1 PAH Potentially avoidable hospitalizations Sociodemographic factors Age, y, mean SD Male gender Married Race/ethnicity*** Non-Hispanic white Non-Hispanic black Hispanic Non-Hispanic Asian Others Function and diagnostic factors Number of chronic conditions (0-5) Activities of daily living (0-28) Mood distress (PHQ-9)*** Severe Moderate to severe Moderate Mild None Cognitive Function Score (CFS)*** Severe impairment Moderate impairment Mild impairment Intact Behavior health disorder (BHD) Schizophrenia/psychosis Bipolar disorder Depression/anxiety Personality Substance abuse Aggressive Behavior Scale score*** Severe Moderate Mild None
ADRD Only [n ¼ 135,610 (16.79%)]
BHD Only [n ¼ 206,694 (25.59%)]
ADRD and BHD [n ¼ 350,963 (43.46%)]
No ADRD or BHD [n ¼ 114,363 (14.16%)]
27.26*** 10.99*** 47.69***
37.90*** 15.97*** 52.06***
28.59*** 12.08*** 50.02***
36.41 15.45 51.76
86 8.9*** 28.9*** 19.9***
74 14.0*** 36.0*** 16.0***
83 10.0*** 27.4*** 18.9***
80 13.8 38.8 17.5
73.1 16.3 5.3 3.0 2.4
80.7 12.0 3.8 1.0 2.5
80.9 10.9 5.0 1.1 2.1
72.2 17.2 4.8 2.8 3.0
1.8 0.9*** 19.1 6.3***
1.2 1*** 16.2 6.9***
1.9 1*** 18.6 6.1***
11 17.1 6.7
0.2 1.0 4.4 12.0 81.4
0.3 1.5 5.7 14.8 77.0
0.4 1.7 6.1 15.8 75.5
0.1 0.7 3.0 10.1 84.4
24.2 46.7 18.3 10.3 d d d d d d
4.5 17.9 23.7 53.2
18.4 46.8 20.2 14.1
24.3 5.0 70.2 0.2 0.4
31.1 3.4 65.2 0.1 0.3
8.9 18.9 22.8 48.8 d d d d d d
1.1 4.2 11.1 83.3
0.8 4.0 11.9 82.9
2.0 7.2 16.2 74.4
0.3 1.8 6.7 90.1
SD, standard deviation. ***P value < .0001, compared to “No ADRD or BHD.” Values are percentages unless otherwise noted.
H. Temkin-Greener et al. / JAMDA 20 (2019) 1280e1286
1283
Table 3 Long-Stay Nursing Home Residents With Dementia (ADRD) and/or Behavioral Health Disorders (BHD): Adjusted RRRs for Nonavoidable and Potentially Avoidable Hospitalizations Diagnostic Group*
Sample Size (No. of Residents)
ADRD only Unadjusted Adjustedy BHD only Unadjusted Adjustedy ADRD and BHD Unadjusted Adjustedy
135,610
Nonavoidable Hospitalizations, RRR (95% CI)
Potentially Avoidable Hospitalizations, RRR (95% CI)
P Value for the Comparison of RRRs
0.68 (0.66, 0.69) 0.90 (0.87, 0.92)
0.62 (0.61, 0.64) 0.84 (0.81, 0.85)
<.0001 <.0001
1.07 (1.05, 1.09) 1.03 (1.01, 1.05)
1.06 (1.03, 1.08) 0.98 (0.96, 1.01)
.3491 .0013
0.70 (0.69, 0.71) 0.88 (0.87, 0.90)
0.70 (0.68, 0.71) 0.85 (0.83, 0.87)
.5702 .0022
206,694
350,963
Note. Bold values are statistically significant P < .05. *Reference group ¼ neither ADRD nor BHD; groups are mutually exclusive. y RRRs are adjusted for individual- and facility-level covariates shown in Supplementary Table 1.
diagnostic factors. Chi-square and t tests were employed for statistical inference as appropriate. We fit multinomial logistic regression models to estimate the risk of PAH and nonepotentially avoidable hospitalizations (N-PAH) for ADRD only, BHD only, and both, while controlling for resident and facility-level covariates and for state dummies. Separately, we also modeled the risk of PAH and N-PAH for residents with behavioral health diagnoses of schizophrenia/psychosis, bipolar disorder, depression/anxiety, personality disorder, and substance abuse. All analyses included residents without ADRD and BHD as the reference group. The state dummies were included in all models to control for state-level variations in NH payment rates, minimum staffing requirements, and other quality of care regulations. We conducted a sensitivity analysis to assess the impact of residents’ deaths. We compared the analysis in which residents who died during the follow-up period were excluded to the analysis based on the entire sample. Because the findings from these analyses were not measurably different, we presented only the results based on the entire sample. Analyses were performed using SAS, version 9.4 (SAS Institute, Inc, Cary, NC). Results For long-stay residents, we identified 439,822 hospitalizations occurring during the study period, with 31.86% having at least 1 hospital stay. Just more than half (50.7%) of all hospitalizations were considered potentially avoidable, and 13.37% of the residents experienced at least 1 PAH. Five conditions (pneumonia, urinary tract infection, falls and trauma, congestive heart failure, and dehydration) were responsible for more than 78% of PAHs (Table 1). Residents with ADRD (with or without BHD) were less likely to have had any hospitalizations (27.26%) compared to residents with neither condition (36.41%) or with BHD only (37.90%). The same was true for the percentage of residents with at least 1 PAH (Table 2). Overall, the
proportion of PAHs was also lowest among residents with ADRD only (47.69%) compared to those with neither ADRD nor BHD (51.76%). Residents with ADRD were older, included more women, had more comorbid conditions, and had higher ADL and cognitive impairments than residents with BHD only and those with neither condition. Unadjusted and risk-adjusted results from the multinomial logistic regressions are summarized in Table 3, with all comparisons made to residents with neither ADRD nor BHD (reference group). For residents with ADRD only, the adjusted relative risk ratio (RRR) of N-PAH was 10% lower and of PAH 16% lower, with the 2 RRRs being statistically significantly different from each other (P < .0001). Findings for the ADRD and BHD group were very similar, with RRRs of 0.88 and 0.85 for N-PAH and PAH, respectively. Residents with BHD only had a statistically higher, but clinically very modest, adjusted risk of N-PAH (RRR ¼ 1.03), whereas their risk of PAH was not different from that of the reference group. We further examined the risk of N-PAH and PAH focusing on residents with specific BHD diagnoses (Table 4). The risk of N-PAH among residents with depression/anxiety, the largest diagnostic group, was not statistically different from the reference group, and the risk of PAH, though statistically significantly different, was clinically negligible (RRR ¼ 1.02). For residents with serious mental illness, the risk of hospitalizations varied by diagnosis. Residents with schizophrenia/ psychosis had the same risk of N-PAH as residents in the reference group, but their risk of PAH was 8% lower (RRR ¼ 0.92). On the other hand, residents with bipolar disorder had increased risk of both NPAH (RRR ¼ 1.15) and PAH (RRR ¼ 1.09) compared to the reference group. Among residents with personality disorder and substance abuse, both accounting for a very small proportion of the residents, the risks of N-PAH and PAH were substantially lower than in the reference group. The adjusted risk of PAH varied substantially across states. In Figure 1, we depicted both the risk of N-PAH and PAH by state. Alaska, which has one of the lowest rates of hospitalizations among NH residents8 was the comparison (reference) state. For the most part, the 2
Table 4 Long-Stay NH Residents With Behavioral Health Disorders (BHD): Adjusted RRRs* for Nonavoidable and Potentially Avoidable Hospitalizations Diagnostic Groupy
Sample Size (No. of Residents)
Nonavoidable Hospitalizations, RRR* (95% CI)
Potentially Avoidable Hospitalizations, RRR* (95% CI)
P Value for the Comparison of RRRs
Schizophrenia/psychosis Bipolar disorder Depression/anxiety Personality disorder Substance abuse
159,388 21,984 373,736 782 1767
1.01 1.15 1.00 0.93 0.68
0.92 1.09 1.02 0.74 0.55
<.0001 .0613 .0278 .1314 .0692
(0.99, (1.10, (0.99, (0.76, (0.59,
1.03) 1.19) 1.02) 1.15) 0.77)
Note. Bold values are statistically significant P < .05. CI, confidence interval. *RRRs are adjusted for the individual- and facility-level covariates shown in Supplementary Table S1. y Reference group ¼ neither ADRD nor BHD.
(0.90, (1.04, (1.01, (0.56, (0.46,
0.94) 1.14) 1.04) 0.96) 0.67)
1284
H. Temkin-Greener et al. / JAMDA 20 (2019) 1280e1286
Fig. 1. Relative risk ratios for potentially avoidable and nonavoidable hospitalizations by state (Alaska ¼ reference group).
measures appear to have been correlated. For example, residents in Louisiana experienced the highest risk of PAH (RR ¼ 1.38) and high risk of N-PAH (RRR ¼ 1.19), whereas residents in Hawaii had the lowest risk of PAH (RRR ¼ 0.30) and low risk of N-PAH (RRR ¼ 0.60). After adjusting for individual-level covariates, facility characteristics did not seem to be particularly importantly associated with N-PAH or PAH, except for profit status. Residents in for-profit facilities had incrementally higher risk of N-PAH (14.5%) and of PAH (13.5%) than residents in not-for-profits (Supplementary Table 1). Discussion This is the first study to explore, in a national population of longstay NH residents, differences in the risk of PAH by ADRD and BHD status. We found 50% of hospitalizations among these residents to be
potentially avoidable. Although this is consistent with prior studies, which based on data from a decade ago reported PAH ranging from 47% to 60%,10,22 we were somewhat surprised to find similar rates with substantially newer data, particularly as CMS has recently reported a 31% decline in PAHs in US NHs between 2010 and 2015.23 Although our list of potentially avoidable diagnoses is more extensive than that employed by CMS, with a decline of this magnitude, we would have expected to report lower average rates of PAH. With regard to our population of interest, there have been only a few prior, older, and largely not comparable studies.14,15,24 Although Becker, in a study of Florida NHs, reported higher risk of hospitalizations for ambulatory sensitive conditions for residents with dementia, schizophrenia, and major depression,24 we found substantially lower relative risk of PAH for residents with dementia and schizophrenia. For residents with depression, our findings indicated only a marginally
H. Temkin-Greener et al. / JAMDA 20 (2019) 1280e1286
increased risk of PAH. However, unlike Becker, we included all diagnoses of depression and did not specifically focus on major depression, so this may account for the differences in our findings. Other differences, such as the scope of study population (1 state vs national) and the span of time, may also account for these differences. Lower rates of PAH are considered to be an indicator of higher NH quality. We found that residents with ADRD have 10% to 12% lower risk of N-PAH as well as 15% lower risk of PAH. It is not clear, however, at what point such lower risk of hospitalizations no longer reflects better quality. We observed that residents with higher cognitive impairment, which is frequent among residents with ADRD, had further incrementally reduced the risk of PAH (and N-PAH). If residents with ADRD have a substantially lower risk of hospitalizations because as a result of an impaired ability to effectively communicate with NH staff they are less likely to be sent to the hospital for further evaluation (even for N-PAH), this may not necessarily be a positive finding. Although we are unable to fully explicate this issue, further research in this regard is urgently needed as new payment models, such as the CMS Initiative to Reduce Avoidable Hospitalizations among Nursing Facility Residents or future programs to include PAH as a metric in value-based purchasing, are already in place or poised to be implemented. Although these programs are designed to combat unnecessary hospitalizations, they may inadvertently have negative consequences, particularly for the most at risk residents such as those with ADRD. Recent findings from the OPTIMISTIC demonstration, part of the CMS initiative to reduce PAHs,25 showed that only 25% of the acute transfers of residents with potentially avoidable diagnoses were actually deemed avoidable by project registered nurses.26 The primary reason, offered by the OPTIMISTIC nurses, for transferring patients with potentially avoidable conditions was lack of NH resources to safely manage residents in place. Other reasons were poor communication about preferences and absence of advance directives. Studies have shown that residents with dementia are more likely to have completed advance directives at admission, and more likely to change from more to less aggressive end-of-life treatment preferences after admission than residents with other conditions.27 Thus, it is possible that a medical provider has a higher threshold when making a decision to hospitalize residents with ADRD who were more likely to have made their treatment preferences clear, compared to residents without ADRD who have not yet made such preferences clear. For residents with BHD only, we found essentially no difference in PAH compared to those without ADRD/BHD. This is largely driven by residents with depression/anxiety who account for almost one-half of the total population of long-stayers. Other residents with BHD exhibited different risks depending on primary diagnosis. For individuals with schizophrenia, having a stable place of residence where basic care needs are met, and frequently present comorbid metabolic disorders are attended to, may in itself promote lower risk of PAH. Residents with bipolar disorders exhibited increased risk of both NPAH and PAH. It is possible that this increase is an unintended consequence of the National Partnership to Improve Dementia Care launched by CMS in 2012. The Partnership implemented public reporting of NHlevel antipsychotic use with an intent to reduce prescribing. In regulating the use of antipsychotic medications, which are FDA approved to treat bipolar disorders, CMS identified only schizophrenia, Huntington’s disease, and Tourette’s syndrome as approved diagnostic clusters. This over regulation, combined with known lack of access to psychiatric expertise in this care setting, may have resulted in patients with bipolar disorders having their medications stopped or reduced, thus increasing the risk of both N-PAH and PAH. Finally, we found variation across NHs and states in hospitalization risk. Consistent with prior literature, the risks of both N-PAHs and PAHs were significantly higher in for-profit NHs.28,29 Variations across
1285
states were substantial, but states in which residents had an overall high risk of N-PAH had a similarly high risk of PAH, with the reverse being true. Explicating the reasons for these variations is beyond the scope of this study. However, future research is needed to determine if important state policy factors, such as NH payment rates and quality regulations, help to reduce PAHs and N-PAHs. We acknowledge possible limitations. Although the MDS is an excellent source for identifying ADRD and BHD, these are rarely principal admitting conditions and may be underdiagnosed, resulting in underestimating the difference in the relative risk of hospitalizations among residents with and without these conditions. We were unable to directly control for the severity of ADRD or BHD. However, we controlled for many individual-level risk factors associated with disease severity, especially cognitive impairment, and this probably sufficiently compensated for this limitation. Finally, defining which specific admissions are preventable based on diagnostic codes alone is not very precise.7,26,30 However, employing conditions specifically identified as either preventable or manageable for NH residents10 allowed us to level the playing field and offer findings that can be compared to other existing/future studies. Conclusions and Implications In conclusion, we found lower risk of PAH and N-PAH among residents with ADRD but not among those with BHD only. Today, NHs are subject to many national and state initiatives focused on quality improvements (eg, reductions in readmissions) and cost-efficiencies (eg, value-based purchasing) directed largely at residents in skilled care facilities, but also increasingly impacting long stayers (eg, reductions in antipsychotics). Although these programs are undoubtedly intended to improve overall care, they may have negative and unintended consequences for individuals with ADRD and BHD. Future studies are needed to better understand how such initiatives impact these most vulnerable of residents. References 1. Cen X, Li Y, Hasselberg MJ, et al. Aggressive behaviors among nursing home residents: Association with dementia and behavioral health disorders. J Am Med Dir Assoc 2018;19:1104e1109.e4. 2. Boccuti C, Casillas G, Neuman T. Reading the stars: Nursing home quality star ratings, nationally and by state. Medicare. San Francisco, CA: Henry J. Kaiser Family Foundation. Accessed March 30, 2019. 3. Molinari V, Hyer K, Branch L, et al. Mental health treatment in nursing homes. Marquette Elder’s Advisor 2011;13. article 5. Available at: http://scholarship. law.marquette.edu/elders/vol13/iss1/5. Accessed March 30, 2019. 4. Temkin-Greener H, Campbell L, Cai X, et al. Are post-acute patients with behavioral health disorders admitted to lower-quality nursing homes? Am J Geriatr Psychiatry 2018;26:643e654. 5. Bartels SJ, Miles KM, Dums AR, Levine KJ. Are nursing homes appropriate for older adults with severe mental illness? Conflicting consumer and clinician views and implications for the Olmstead decision. J Am Geriatr Soc 2003;51:1571e1579. 6. Grabowski DC, Aschbrenner KA, Rome VF, Bartels SJ. Quality of mental health care for nursing home residents: A literature review. Med Care Res Rev 2010; 67:627e656. 7. Ouslander JG, Lamb G, Perloe M, et al. Potentially avoidable hospitalizations of nursing home residents: Frequency, causes, and costs. J Am Geriatr Soc 2010; 58:627e635. 8. US Department of Health and Human Services. Office of Inspector General. Medicare Nursing Home Resident Hospitalization Rates Merit Additional Monitoring. Washington, DC: Department of Health and Human Services, Office of Inspector General; 2013. 9. Saliba D, Raynard K, Buchanan J, et al. Appropriateness of the decision to transfer nursing facility residents to the hospital. J Am Geriatr Soc 2000;48: 154e163. 10. Walsh EG, Wiener JM, Haber S, et al. Potentially avoidable hospitalizations of dually eligible Medicare and Medicaid beneficiaries from nursing facility and home- and community-based services waiver programs. J Am Geriatr Soc 2012;60:821e829. 11. Ingber M, Feng Z, Khatutsky G, et al. Initiative to reduce avoidable hospitalizations among nursing facility residents shows promising results. Health Aff 2017;36:441e450.
1286
H. Temkin-Greener et al. / JAMDA 20 (2019) 1280e1286
12. Grabowski DC, Aschbrenner KA, Feng Z, Mor V. Mental illness in nursing homes: Variations across states. Health Aff 2009;28:689e700. 13. Perry M, Cummings J, Jacobson G, et al. To Hospitalize or Not to Hospitalize? Medical Care for Long-Term Care Facility Residents. Health Costs. San Francisco, CA: Henry J. Kaiser Family Foundation; 2010. 14. Feng Z, Coots LA, Kaganova Y, Wiener JM. Hospital and ED use among Medicare beneficiaries with dementia varies by setting and proximity to death. Health Aff 2014;33:683e690. 15. Carter MW, Porell FW. Vulnerable populations at risk of potentially avoidable hospitalizations: The case of nursing home residents with Alzheimer’s disease. Am J Alzheimers Dis Other Demen 2005;20:349e358. 16. Hawes C, Morris JN, Phillips CD, et al. Reliability estimates for the minimum data set for nursing home resident assessment and care screening (MDS). Gerontologist 1995;35:172e178. 17. Mor V, Angelelli J, Jones R, et al. Inter-rater reliability of nursing home quality indicators in the U.S. BMC Health Serv Res 2003;3:20. 18. Saliba D, Jones M, Streim J, et al. Overview of significant changes in the minimum data set for nursing homes version 3.0. J Am Med Dir Assoc 2012;13: 595e601. 19. Fullerton CA, McGuire TG, Feng Z, et al. Trends in mental health admissions to nursing homes, 1999-2005. Psychiatr Serv 2009;60:965e971. 20. Thomas K, Dosa D, Wysocki A, Mor V. The Minimum Data Set 3.0 Cognitive Function Scale. Med Care 2017;55:e68ee72. 21. Perlman CM, Hirdes JP. The Aggressive Behavior Scale: A new scale to measure aggression based on the minimum data set. J Am Geriatr Soc 2008;56: 2298e2303. 22. Spector WD, Limcangco R, Williams C, et al. Potentially avoidable hospitalizations for elderly long-stay residents in nursing homes. Med Care 2013;0:1e9.
23. Brennan N, Engelhardt T. Data brief: Sharp reductions in avoidable hospitalizations among long-term care facility residents. CMS website. 2017. Available at: https://www.cms.gov/blog/data-brief-sharp-reduction-avoidable-hospitali zations-among-long-term-care-facility-residents. Accessed March 30, 2019. 24. Becker MA, Boaz TL, Andel R, et al. Predictors of preventable nursing home hospitalizations: The role of mental disorders and dementia. Am J Geriatr Psychiatry 2010;18:475e482. 25. Centers for Medicare and Medicaid Services (CMS). Initative to reduce avoidable hospitalizations among nursing facility residents. 2019. Available at: https://www.cms.gov/Medicare-Medicaid-Coordination/Medica re-and-Medicaid-Coordination/Medicare-Medicaid-Coordination-Office/Initi ativetoReduceAvoidableHospitalizations/AvoidableHospitalizationsamong NursingFacilityResidents.html. Accessed March 30, 2019. 26. Unroe KT, Hickman SE, Carnahan JL, et al. Investigating the avoidability of hospitalizations of long stay nursing home residents: Opportunities for improvement. Innov Aging 2018;2:1e9. 27. Mukamel DB, Ladd H, Temkin-Greener H. Stability of cardiopulmonary resuscitation and do-not-resuscitate orders among long-term nursing home residents. Med Care 2013;51:666e672. 28. Konetzka RT, Spector W, Shaffer T. Effects of nursing home ownership type and resident payer source on hospitalization for suspected pneumonia. Med Care 2004;42:1001e1008. 29. Hirth RA, Grabowski DC, Feng Z, et al. Effect of nursing home ownership on hospitalization of long-stay residents: An instrumental variables approach. Int J Health Care Finance Econ 2014;14:1e18. 30. Grabowski DC, O’Malley AJ, Barhydt NR. The costs and potential savings associated with nursing home hospitalizations. Health Aff 2007;26: 1753e1761.
1286.e1
H. Temkin-Greener et al. / JAMDA 20 (2019) 1280e1286
Appendix Supplementary Table 1 Individual- and Facility-Level Factors and Their RRRs for Nonavoidable and Potentially Avoidable Hospitalizations for Long-Stay Residents Variable
Individual resident characteristics Age, y Gender ¼ male Marital status ¼ married Race (white ¼ reference) Black Hispanic Asian Other Aggressive behavior (reference ¼ none) Mild Moderate Severe ADL score (0-28) Mood distress (PHQ-9) (reference ¼ none) Mild Moderate Moderate-severe Severe Cognitive function score (reference ¼ none) Mild Moderate Severe Number of chronic conditions Medicare/Medicaid eligible Nursing home characteristics Bed size % Medicaid % Medicare Occupancy rate RN/nursing staff ratio Chain For-profit Overall star rating (reference ¼ 2-4) Star ¼ 1 Star ¼ 5 Note. Bold values are statistically significant P < .05. CI, confidence interval; RN, registered nurse; RRR, relative risk ratio.
Nonavoidable Hospitalizations
Potentially Avoidable Hospitalizations
RRR (95% CI)
RRR (95% CI)
0.986 (0.98, 0.99) 1.261 (1.24, 1.28) 1.075 (1.06, 1.09)
0.997 (0.99, 0.99) 1.001 (0.99, 1.02) 1.104 (1.08, 1.13)
1.279 1.303 1.261 1.088
(1.25, (1.26, (1.19, (1.04,
1.31) 1.35) 1.34) 1.14)
1.174 1.273 1.047 1.102
(1.14, (1.21, (0.97, (1.04,
1.20) 1.33) 1.13) 1.16)
1.040 1.081 1.147 1.016
(1.02, (1.05, (1.08, (1.01,
1.06) 1.12) 1.22) 1.02)
1.047 1.071 1.096 0.998
(1.02, (1.03, (1.02, (0.99,
1.07) 1.11) 1.18) 0.99)
1.006 1.012 0.959 0.903
(0.99, (0.98, (0.90, (0.79,
1.03) 1.04) 1.02) 1.03)
1.051 1.041 0.933 0.868
(1.03, (1.00, (0.87, (0.76,
1.07) 1.08) 0.99) 1.00)
0.885 0.734 0.570 1.194 0.794
(0.87, (0.72, (0.55, (1.19, (0.69,
0.90) 0.75) 0.59) 1.20) 0.71)
0.835 0.669 0.446 1.332 0.814
(0.82, (0.65, (0.43, (1.32, (0.80,
0.85) 0.68) 0.46) 1.34) 0.83)
1.000 1.005 1.012 1.000 0.994 0.989 1.145
(1.00, (1.00, (1.01, (0.99, (0.99, (0.97, (1.12,
1.00) 1.01) 1.01) 1.00) 0.99) 1.01) 1.17)
1.000 1.005 1.015 0.999 0.992 0.975 1.135
(1.00, (1.04, (1.01, (1.00, (0.99, (0.95, (1.10,
1.00) 1.06) 1.02) 1.00) 0.99) 1.00) 1.17)
1.005 (0.97, 1.04) 0.986 (0.94, 1.03)
0.974 (0.94, 1.01) 0.961 (0.91, 1.01)
JAMDA journal homepage: www.jamda.com
Original Study
Preventable Hospitalizations Among Nursing Home Residents With Dementia and Behavioral Health Disorders Helena Temkin-Greener PhD a, *, Xi Cen MS a, Michael J. Hasselberg MS, PhD b, Yue Li PhD a a b
Department of Public Health Sciences, University of Rochester School of Medicine and Dentistry, Rochester, NY Department of Psychiatry, University of Rochester School of Medicine and Dentistry, Rochester, NY
a b s t r a c t Keywords: Avoidable hospitalizations dementia behavioral disorders nursing homes
Objective: Nursing home (NH) residents with Alzheimer’s disease/related dementias (ADRD) and/or behavioral health disorders (BHD) are at high risk of hospitalizations, many of which are potentially avoidable. Empirical evidence regarding potentially avoidable hospitalizations (PAHs) among these residents is quite sparse and mixed. The objectives of this study were to (1) examine the risk of PAH among residents with ADRD only, BHD only, ADRD and BHD compared to residents with neither and (2) identify associations between individual- and facility-level factors and PAH in these subgroups. Design: Retrospective, CY2014-2015. Setting and Participants: Long-term residents age 65þ (N ¼ 807,630) residing in 15,234 NHs. Methods: We employed the Minimum Data Set, MedPAR, Medicare beneficiary summary, and Nursing Home Compare. Hospitalization risk was the outcome of interest. Individual-level covariates were used to adjust for health conditions. Facility-level covariates and state dummies were included. Multinomial logistic regression models were fit to estimate the risk of PAH and nonepotentially avoidable hospitalizations (N-PAH). Results: Compared to residents without ADRD or BHD, those with ADRD had at least a 10% lower relative risk ratio (RRR) of N-PAH and a significantly lower risk of PAH, at 16% (P < .0001). Residents with BHD only had a statistically higher, but clinically very modest (RRR ¼ 1.03) risk of N-PAH, with no difference in the risk of PAH. Focusing on specific BHD conditions, we found no difference in N-PAH or PAH among residents with depression, lower PAH risk among those with schizophrenia/psychosis (RRR ¼ 0.92), and an increased risk of both N-PAH (RRR ¼ 1.15) and PAH (RRR ¼ 1.09) among residents with bipolar disorders. Conclusions and Relevance: We observed a lower risk of PAH and N-PAH among residents with ADRD, with the risk for residents with BHD varying by condition. Substantial variations in PAH and N-PAH were evident across states. Future research is needed to identify state-level modifiable factors that explain these variations. Ó 2019 AMDA e The Society for Post-Acute and Long-Term Care Medicine.
Research suggests that 60% to 90% of nursing home (NH) residents have a diagnosis of Alzheimer’s or related dementias (ADRD) and/or behavioral health disorders (BHD; eg, schizophrenia, bipolar disorder, depression, and substance abuse).1e3 Residents with these conditions
This work was supported by the Donaghue Medical Research Foundation’s 2016 Another Look Program. The authors declare no conflicts of interest. * Address correspondence to Helena Temkin-Greener, PhD, Department of Public Health Sciences, University of Rochester School of Medicine and Dentistry, 265 Crittenden Blvd, CU 420644, Rochester, NY 14642. E-mail address: [email protected] (H. TemkinGreener). https://doi.org/10.1016/j.jamda.2019.03.006 1525-8610/Ó 2019 AMDA e The Society for Post-Acute and Long-Term Care Medicine.
experience disparities in accessing high-quality facilities,4 and within a NH, availability of effective mental health treatments is often lacking.5,6 Faced with serious physical health problems as well, these residents are at high risk of hospitalizations, which tend to exacerbate iatrogenic illness, functional decline, and mortality.7 Approximately one-fourth of NH residents are hospitalized annually,8 and at least half of these hospitalizations are considered to be potentially avoidable, that is, either preventable or manageable within NHs.7,9,10 Policy makers have focused considerable attention on factors influencing the potentially avoidable hospitalizations (PAHs) among NH residents in an effort to improve the quality of care and reduce Medicare costs. The most recent such effort introduced by the Centers for Medicare &
H. Temkin-Greener et al. / JAMDA 20 (2019) 1280e1286
Medicaid Services (CMS) in 2012 included multiyear interventions but produced mixed results, with NHs in 2 of 7 participating states showing consistent 2-year declines in PAH.11 For residents with ADRD and/or BHD, NHs may have difficulty comanaging behavioral and physical comorbidities, because on-site capacity to provide and/or manage mental health services is often limited.12 NH nurses and social workers have reported that when residents with ADRD and other behavioral health needs “act out,” they are more likely to be hospitalized.13 But empirical evidence on PAHs among NH residents with ADRD and/or BHD is quite sparse and mixed. One study found no statistically significant difference in the occurrence of PAHs among residents with (23.3%) and without (22.0%) dementia.14 In the last year of life, although the rates of PAH were considerably higher, there was still no difference between the 2 subpopulations. A Massachusetts study of Medicare/Medicaid NH residents with dementia found lower PAH rates among those with dementia compared with other residents, but only in facilities with overall lower rates of hospitalizations.15 A Florida study focusing on time to first PAH among residents with dementia and mental disorders found that having dementia, major psychotic, bipolar, depressive, or substance abuse disorder increased the risk of PAH compared to residents without these conditions. Motivated by this sparse and conflicting literature, our objectives were to (1) examine and compare the risk of PAH among all NH residents with ADRD only, BHD only, ADRD and BHD, and neither and (2) identify associations between individual- and facility-level characteristics and PAH in these subgroups. Methods Study Design and Data Sources This was a retrospective study using nationwide assessment and claims data for CY2014-2015. We used 100% Minimum Data Set (MDS) to identify long-term residents. The MDS contains clinical assessments for all residents at admission, quarterly, annually, when there is a significant change in health status, and at discharge. Assessments include information on residents’ sociodemographics, diagnoses,
1281
treatments, behaviors, and functional and cognitive impairments. The MDS data, including the behavioral health items, have been found to be of high research quality.16e18 We used the Medicare beneficiary summary file to identify and exclude the Medicare-Advantage plan enrollees for whom hospitalization data in the Medicare claims may not be accurate. The Medicare Provider and Analysis Review (MedPAR) file was used to track inpatient admissions and identify those that were potentially avoidable. We used the CMS Nursing Home Compare data to obtain information on facility characteristics. Medicare beneficiary and provider identifiers were used to merge these databases. The study protocol was reviewed and approved by the Institutional Research and Subjects Review Board. Study Population Based on the MDS for the second quarter of CY2014, we identified residents with a quarterly, annual, or a significant change in status assessments, and those with a re-entry record for a prior long-term stay, as being long-stay residents. After excluding the Medicare Advantage enrollees and beneficiaries younger than 65 years of age, the analytical file included 807,630 long-stay residents in 15,234 homes. For this population, we tracked inpatient admissions from the second quarter of 2014 through the second quarter of 2015. Outcome Variable The outcome of interest was the risk of PAH. We used the definition of PAH developed by the RTI International10 and most recently employed in the evaluation of the CMS Initiative to Reduce Potentially Avoidable Hospitalizations among Nursing Facility Residents.11 We focused on 17 conditions (Table 1) deemed potentially preventable and/or manageable (https://www.cms.gov/ResearchStatistics-Data-and-Systems/Statistics-Trends-and-Reports/Reports/ downloads/costdriverstask2.pdf). We modeled preventable and manageable conditions separately and also combined as a single outcome. The results for separate conditions were not measurably different from the model in which they were combined, and because
Table 1 Distribution of Potentially Avoidable Hospitalizations by Condition Diagnostic Condition
Potentially Avoidable Hospitalizations Total Number
Lower respiratory: pneumonia bronchitis Urinary tract infection Falls and trauma Congestive heart failure Dehydration, acute renal failure, hypokalemia, hyponatremia Chronic obstructive pulmonary disease, asthma, chronic bronchitis Cellulitis Diarrhea, gastroenteritis with nausea and vomiting Seizures Hypertension, hypotension Skin ulcers including pressure ulcers Hyper- and hypoglycemia: diabetes mellitus with ketoacidosis or hyperosmolar coma Anemia Constipation or fecal impaction obstipation Weight loss, nutritional deficiencies, adult failure to thrive Psychosis, severe agitation, organic brain syndrome Altered mental status/acute confusion/delirium
Percent
Preventable/Manageable
25,508 23,784 20,849 19,330 17,985
18.6 17.3 15.2 14.1 13.1
N/Y Y/Y Y/N Y/Y Y/Y
7868
5.7
Y/Y
6340 3663
4.6 2.7
N/Y N/Y
2888 2009 1836 1669
2.1 1.5 1.3 1.2
Y/Y Y/Y Y/Y Y/Y
1343 679 585
1.0 0.5 0.4
N/Y Y/Y Y/N
557
0.4
N/Y
396 137,289
0.3 100.0
Y/N
1282
H. Temkin-Greener et al. / JAMDA 20 (2019) 1280e1286
(ABS ¼ 0), mild (ABS ¼ 1-2), moderate (ABS ¼ 3-5), and severe (ABS ¼ 6-12).1,21 Mood distress was measured using PHQ-9 assessing both symptom presence and frequency (scale 0-27), coded as none (<¼4), mild (5-9), moderate (10-14), moderate to severe (15-19), and severe (20-27). Number of comorbid chronic conditions (heart failure, chronic obstructive pulmonary disease, diabetes, stroke/transient ischemic attack, ischemic heart disease) was included as a count variable (0-5). We controlled for sociodemographic factors such as age, gender, race (non-Hispanic white, non-Hispanic black, Hispanic, non-Hispanic Asian, other), marital status, and dual Medicare/Medicaid eligibility (yes/no). To examine associations with facility-level characteristics, we included NH size (beds), occupancy rate, percentage of Medicare and Medicaid residents, and skilled care mix (ratio of registered nurses to licensed practical nurses and certified nurse assistants). Ownership was dichotomized as nonprofit or for-profit, and chain or not. We controlled for facility quality using CMS 5-star metrics based on overall and staffing quality, categorizing NHs as 1-, 2- to 4-, or 5-star.
of the significant overlap between preventable and manageable conditions, we only present the results on PAH as a single outcome. Independent Variables The key independent variables of interest were presence of (1) BHD only, (2) ADRD only, and (3) ADRD with comorbid BHDs. The diagnoses constituting BHD were schizophrenia or psychosis, bipolar disorder, depression or anxiety, personality disorder, and substance abuse. Based on prior literature,4,19 these conditions were hierarchically assigned in this order. Conditions were identified from the MDS section I (active diagnoses) and included both prepopulated fields and a search for specific ICD-9 codes. All conditions were identified from the last available assessment in the second quarter of 2014 (baseline). Conditions defined as PAH are based solely on diagnoses and do not reflect illness severity. Therefore, we also included variables reflecting the residents’ clinical conditions and comorbidities at baseline. Residents’ functional status was measured by the number of impairments in activities of daily living (ADL) based on the score ranging from 0 to 28 (higher score signifying greater impairment). We assessed cognitive function using a 0 to 3 scale.20 Aggressive behaviors were measured using 4 items (MDS section E; Agressive Behavior Scale [ABS]) reflecting both the prevalence and the frequency of these behaviors in the last week on a scale from 0 to 12, and coded as none
Statistical Analyses Descriptive analyses were conducted across resident groups characterized by the presence/absence of ADRD and/or BHD. We compared these groups on sociodemographic, functional, clinical, and
Table 2 Descriptive Characteristics of the Study Population at Baseline: By Dementia (ADRD) and Behavioral Health Disorders (BHD) Characteristics Hospitalizations Residents with at least 1 hospitalization Residents with at least 1 PAH Potentially avoidable hospitalizations Sociodemographic factors Age, y, mean SD Male gender Married Race/ethnicity*** Non-Hispanic white Non-Hispanic black Hispanic Non-Hispanic Asian Others Function and diagnostic factors Number of chronic conditions (0-5) Activities of daily living (0-28) Mood distress (PHQ-9)*** Severe Moderate to severe Moderate Mild None Cognitive Function Score (CFS)*** Severe impairment Moderate impairment Mild impairment Intact Behavior health disorder (BHD) Schizophrenia/psychosis Bipolar disorder Depression/anxiety Personality Substance abuse Aggressive Behavior Scale score*** Severe Moderate Mild None
ADRD Only [n ¼ 135,610 (16.79%)]
BHD Only [n ¼ 206,694 (25.59%)]
ADRD and BHD [n ¼ 350,963 (43.46%)]
No ADRD or BHD [n ¼ 114,363 (14.16%)]
27.26*** 10.99*** 47.69***
37.90*** 15.97*** 52.06***
28.59*** 12.08*** 50.02***
36.41 15.45 51.76
86 8.9*** 28.9*** 19.9***
74 14.0*** 36.0*** 16.0***
83 10.0*** 27.4*** 18.9***
80 13.8 38.8 17.5
73.1 16.3 5.3 3.0 2.4
80.7 12.0 3.8 1.0 2.5
80.9 10.9 5.0 1.1 2.1
72.2 17.2 4.8 2.8 3.0
1.8 0.9*** 19.1 6.3***
1.2 1*** 16.2 6.9***
1.9 1*** 18.6 6.1***
11 17.1 6.7
0.2 1.0 4.4 12.0 81.4
0.3 1.5 5.7 14.8 77.0
0.4 1.7 6.1 15.8 75.5
0.1 0.7 3.0 10.1 84.4
24.2 46.7 18.3 10.3 d d d d d d
4.5 17.9 23.7 53.2
18.4 46.8 20.2 14.1
24.3 5.0 70.2 0.2 0.4
31.1 3.4 65.2 0.1 0.3
8.9 18.9 22.8 48.8 d d d d d d
1.1 4.2 11.1 83.3
0.8 4.0 11.9 82.9
2.0 7.2 16.2 74.4
0.3 1.8 6.7 90.1
SD, standard deviation. ***P value < .0001, compared to “No ADRD or BHD.” Values are percentages unless otherwise noted.
H. Temkin-Greener et al. / JAMDA 20 (2019) 1280e1286
1283
Table 3 Long-Stay Nursing Home Residents With Dementia (ADRD) and/or Behavioral Health Disorders (BHD): Adjusted RRRs for Nonavoidable and Potentially Avoidable Hospitalizations Diagnostic Group*
Sample Size (No. of Residents)
ADRD only Unadjusted Adjustedy BHD only Unadjusted Adjustedy ADRD and BHD Unadjusted Adjustedy
135,610
Nonavoidable Hospitalizations, RRR (95% CI)
Potentially Avoidable Hospitalizations, RRR (95% CI)
P Value for the Comparison of RRRs
0.68 (0.66, 0.69) 0.90 (0.87, 0.92)
0.62 (0.61, 0.64) 0.84 (0.81, 0.85)
<.0001 <.0001
1.07 (1.05, 1.09) 1.03 (1.01, 1.05)
1.06 (1.03, 1.08) 0.98 (0.96, 1.01)
.3491 .0013
0.70 (0.69, 0.71) 0.88 (0.87, 0.90)
0.70 (0.68, 0.71) 0.85 (0.83, 0.87)
.5702 .0022
206,694
350,963
Note. Bold values are statistically significant P < .05. *Reference group ¼ neither ADRD nor BHD; groups are mutually exclusive. y RRRs are adjusted for individual- and facility-level covariates shown in Supplementary Table 1.
diagnostic factors. Chi-square and t tests were employed for statistical inference as appropriate. We fit multinomial logistic regression models to estimate the risk of PAH and nonepotentially avoidable hospitalizations (N-PAH) for ADRD only, BHD only, and both, while controlling for resident and facility-level covariates and for state dummies. Separately, we also modeled the risk of PAH and N-PAH for residents with behavioral health diagnoses of schizophrenia/psychosis, bipolar disorder, depression/anxiety, personality disorder, and substance abuse. All analyses included residents without ADRD and BHD as the reference group. The state dummies were included in all models to control for state-level variations in NH payment rates, minimum staffing requirements, and other quality of care regulations. We conducted a sensitivity analysis to assess the impact of residents’ deaths. We compared the analysis in which residents who died during the follow-up period were excluded to the analysis based on the entire sample. Because the findings from these analyses were not measurably different, we presented only the results based on the entire sample. Analyses were performed using SAS, version 9.4 (SAS Institute, Inc, Cary, NC). Results For long-stay residents, we identified 439,822 hospitalizations occurring during the study period, with 31.86% having at least 1 hospital stay. Just more than half (50.7%) of all hospitalizations were considered potentially avoidable, and 13.37% of the residents experienced at least 1 PAH. Five conditions (pneumonia, urinary tract infection, falls and trauma, congestive heart failure, and dehydration) were responsible for more than 78% of PAHs (Table 1). Residents with ADRD (with or without BHD) were less likely to have had any hospitalizations (27.26%) compared to residents with neither condition (36.41%) or with BHD only (37.90%). The same was true for the percentage of residents with at least 1 PAH (Table 2). Overall, the
proportion of PAHs was also lowest among residents with ADRD only (47.69%) compared to those with neither ADRD nor BHD (51.76%). Residents with ADRD were older, included more women, had more comorbid conditions, and had higher ADL and cognitive impairments than residents with BHD only and those with neither condition. Unadjusted and risk-adjusted results from the multinomial logistic regressions are summarized in Table 3, with all comparisons made to residents with neither ADRD nor BHD (reference group). For residents with ADRD only, the adjusted relative risk ratio (RRR) of N-PAH was 10% lower and of PAH 16% lower, with the 2 RRRs being statistically significantly different from each other (P < .0001). Findings for the ADRD and BHD group were very similar, with RRRs of 0.88 and 0.85 for N-PAH and PAH, respectively. Residents with BHD only had a statistically higher, but clinically very modest, adjusted risk of N-PAH (RRR ¼ 1.03), whereas their risk of PAH was not different from that of the reference group. We further examined the risk of N-PAH and PAH focusing on residents with specific BHD diagnoses (Table 4). The risk of N-PAH among residents with depression/anxiety, the largest diagnostic group, was not statistically different from the reference group, and the risk of PAH, though statistically significantly different, was clinically negligible (RRR ¼ 1.02). For residents with serious mental illness, the risk of hospitalizations varied by diagnosis. Residents with schizophrenia/ psychosis had the same risk of N-PAH as residents in the reference group, but their risk of PAH was 8% lower (RRR ¼ 0.92). On the other hand, residents with bipolar disorder had increased risk of both NPAH (RRR ¼ 1.15) and PAH (RRR ¼ 1.09) compared to the reference group. Among residents with personality disorder and substance abuse, both accounting for a very small proportion of the residents, the risks of N-PAH and PAH were substantially lower than in the reference group. The adjusted risk of PAH varied substantially across states. In Figure 1, we depicted both the risk of N-PAH and PAH by state. Alaska, which has one of the lowest rates of hospitalizations among NH residents8 was the comparison (reference) state. For the most part, the 2
Table 4 Long-Stay NH Residents With Behavioral Health Disorders (BHD): Adjusted RRRs* for Nonavoidable and Potentially Avoidable Hospitalizations Diagnostic Groupy
Sample Size (No. of Residents)
Nonavoidable Hospitalizations, RRR* (95% CI)
Potentially Avoidable Hospitalizations, RRR* (95% CI)
P Value for the Comparison of RRRs
Schizophrenia/psychosis Bipolar disorder Depression/anxiety Personality disorder Substance abuse
159,388 21,984 373,736 782 1767
1.01 1.15 1.00 0.93 0.68
0.92 1.09 1.02 0.74 0.55
<.0001 .0613 .0278 .1314 .0692
(0.99, (1.10, (0.99, (0.76, (0.59,
1.03) 1.19) 1.02) 1.15) 0.77)
Note. Bold values are statistically significant P < .05. CI, confidence interval. *RRRs are adjusted for the individual- and facility-level covariates shown in Supplementary Table S1. y Reference group ¼ neither ADRD nor BHD.
(0.90, (1.04, (1.01, (0.56, (0.46,
0.94) 1.14) 1.04) 0.96) 0.67)
1284
H. Temkin-Greener et al. / JAMDA 20 (2019) 1280e1286
Fig. 1. Relative risk ratios for potentially avoidable and nonavoidable hospitalizations by state (Alaska ¼ reference group).
measures appear to have been correlated. For example, residents in Louisiana experienced the highest risk of PAH (RR ¼ 1.38) and high risk of N-PAH (RRR ¼ 1.19), whereas residents in Hawaii had the lowest risk of PAH (RRR ¼ 0.30) and low risk of N-PAH (RRR ¼ 0.60). After adjusting for individual-level covariates, facility characteristics did not seem to be particularly importantly associated with N-PAH or PAH, except for profit status. Residents in for-profit facilities had incrementally higher risk of N-PAH (14.5%) and of PAH (13.5%) than residents in not-for-profits (Supplementary Table 1). Discussion This is the first study to explore, in a national population of longstay NH residents, differences in the risk of PAH by ADRD and BHD status. We found 50% of hospitalizations among these residents to be
potentially avoidable. Although this is consistent with prior studies, which based on data from a decade ago reported PAH ranging from 47% to 60%,10,22 we were somewhat surprised to find similar rates with substantially newer data, particularly as CMS has recently reported a 31% decline in PAHs in US NHs between 2010 and 2015.23 Although our list of potentially avoidable diagnoses is more extensive than that employed by CMS, with a decline of this magnitude, we would have expected to report lower average rates of PAH. With regard to our population of interest, there have been only a few prior, older, and largely not comparable studies.14,15,24 Although Becker, in a study of Florida NHs, reported higher risk of hospitalizations for ambulatory sensitive conditions for residents with dementia, schizophrenia, and major depression,24 we found substantially lower relative risk of PAH for residents with dementia and schizophrenia. For residents with depression, our findings indicated only a marginally
H. Temkin-Greener et al. / JAMDA 20 (2019) 1280e1286
increased risk of PAH. However, unlike Becker, we included all diagnoses of depression and did not specifically focus on major depression, so this may account for the differences in our findings. Other differences, such as the scope of study population (1 state vs national) and the span of time, may also account for these differences. Lower rates of PAH are considered to be an indicator of higher NH quality. We found that residents with ADRD have 10% to 12% lower risk of N-PAH as well as 15% lower risk of PAH. It is not clear, however, at what point such lower risk of hospitalizations no longer reflects better quality. We observed that residents with higher cognitive impairment, which is frequent among residents with ADRD, had further incrementally reduced the risk of PAH (and N-PAH). If residents with ADRD have a substantially lower risk of hospitalizations because as a result of an impaired ability to effectively communicate with NH staff they are less likely to be sent to the hospital for further evaluation (even for N-PAH), this may not necessarily be a positive finding. Although we are unable to fully explicate this issue, further research in this regard is urgently needed as new payment models, such as the CMS Initiative to Reduce Avoidable Hospitalizations among Nursing Facility Residents or future programs to include PAH as a metric in value-based purchasing, are already in place or poised to be implemented. Although these programs are designed to combat unnecessary hospitalizations, they may inadvertently have negative consequences, particularly for the most at risk residents such as those with ADRD. Recent findings from the OPTIMISTIC demonstration, part of the CMS initiative to reduce PAHs,25 showed that only 25% of the acute transfers of residents with potentially avoidable diagnoses were actually deemed avoidable by project registered nurses.26 The primary reason, offered by the OPTIMISTIC nurses, for transferring patients with potentially avoidable conditions was lack of NH resources to safely manage residents in place. Other reasons were poor communication about preferences and absence of advance directives. Studies have shown that residents with dementia are more likely to have completed advance directives at admission, and more likely to change from more to less aggressive end-of-life treatment preferences after admission than residents with other conditions.27 Thus, it is possible that a medical provider has a higher threshold when making a decision to hospitalize residents with ADRD who were more likely to have made their treatment preferences clear, compared to residents without ADRD who have not yet made such preferences clear. For residents with BHD only, we found essentially no difference in PAH compared to those without ADRD/BHD. This is largely driven by residents with depression/anxiety who account for almost one-half of the total population of long-stayers. Other residents with BHD exhibited different risks depending on primary diagnosis. For individuals with schizophrenia, having a stable place of residence where basic care needs are met, and frequently present comorbid metabolic disorders are attended to, may in itself promote lower risk of PAH. Residents with bipolar disorders exhibited increased risk of both NPAH and PAH. It is possible that this increase is an unintended consequence of the National Partnership to Improve Dementia Care launched by CMS in 2012. The Partnership implemented public reporting of NHlevel antipsychotic use with an intent to reduce prescribing. In regulating the use of antipsychotic medications, which are FDA approved to treat bipolar disorders, CMS identified only schizophrenia, Huntington’s disease, and Tourette’s syndrome as approved diagnostic clusters. This over regulation, combined with known lack of access to psychiatric expertise in this care setting, may have resulted in patients with bipolar disorders having their medications stopped or reduced, thus increasing the risk of both N-PAH and PAH. Finally, we found variation across NHs and states in hospitalization risk. Consistent with prior literature, the risks of both N-PAHs and PAHs were significantly higher in for-profit NHs.28,29 Variations across
1285
states were substantial, but states in which residents had an overall high risk of N-PAH had a similarly high risk of PAH, with the reverse being true. Explicating the reasons for these variations is beyond the scope of this study. However, future research is needed to determine if important state policy factors, such as NH payment rates and quality regulations, help to reduce PAHs and N-PAHs. We acknowledge possible limitations. Although the MDS is an excellent source for identifying ADRD and BHD, these are rarely principal admitting conditions and may be underdiagnosed, resulting in underestimating the difference in the relative risk of hospitalizations among residents with and without these conditions. We were unable to directly control for the severity of ADRD or BHD. However, we controlled for many individual-level risk factors associated with disease severity, especially cognitive impairment, and this probably sufficiently compensated for this limitation. Finally, defining which specific admissions are preventable based on diagnostic codes alone is not very precise.7,26,30 However, employing conditions specifically identified as either preventable or manageable for NH residents10 allowed us to level the playing field and offer findings that can be compared to other existing/future studies. Conclusions and Implications In conclusion, we found lower risk of PAH and N-PAH among residents with ADRD but not among those with BHD only. Today, NHs are subject to many national and state initiatives focused on quality improvements (eg, reductions in readmissions) and cost-efficiencies (eg, value-based purchasing) directed largely at residents in skilled care facilities, but also increasingly impacting long stayers (eg, reductions in antipsychotics). Although these programs are undoubtedly intended to improve overall care, they may have negative and unintended consequences for individuals with ADRD and BHD. Future studies are needed to better understand how such initiatives impact these most vulnerable of residents. References 1. Cen X, Li Y, Hasselberg MJ, et al. Aggressive behaviors among nursing home residents: Association with dementia and behavioral health disorders. J Am Med Dir Assoc 2018;19:1104e1109.e4. 2. Boccuti C, Casillas G, Neuman T. Reading the stars: Nursing home quality star ratings, nationally and by state. Medicare. San Francisco, CA: Henry J. Kaiser Family Foundation. Accessed March 30, 2019. 3. Molinari V, Hyer K, Branch L, et al. Mental health treatment in nursing homes. Marquette Elder’s Advisor 2011;13. article 5. Available at: http://scholarship. law.marquette.edu/elders/vol13/iss1/5. Accessed March 30, 2019. 4. Temkin-Greener H, Campbell L, Cai X, et al. Are post-acute patients with behavioral health disorders admitted to lower-quality nursing homes? Am J Geriatr Psychiatry 2018;26:643e654. 5. Bartels SJ, Miles KM, Dums AR, Levine KJ. Are nursing homes appropriate for older adults with severe mental illness? Conflicting consumer and clinician views and implications for the Olmstead decision. J Am Geriatr Soc 2003;51:1571e1579. 6. Grabowski DC, Aschbrenner KA, Rome VF, Bartels SJ. Quality of mental health care for nursing home residents: A literature review. Med Care Res Rev 2010; 67:627e656. 7. Ouslander JG, Lamb G, Perloe M, et al. Potentially avoidable hospitalizations of nursing home residents: Frequency, causes, and costs. J Am Geriatr Soc 2010; 58:627e635. 8. US Department of Health and Human Services. Office of Inspector General. Medicare Nursing Home Resident Hospitalization Rates Merit Additional Monitoring. Washington, DC: Department of Health and Human Services, Office of Inspector General; 2013. 9. Saliba D, Raynard K, Buchanan J, et al. Appropriateness of the decision to transfer nursing facility residents to the hospital. J Am Geriatr Soc 2000;48: 154e163. 10. Walsh EG, Wiener JM, Haber S, et al. Potentially avoidable hospitalizations of dually eligible Medicare and Medicaid beneficiaries from nursing facility and home- and community-based services waiver programs. J Am Geriatr Soc 2012;60:821e829. 11. Ingber M, Feng Z, Khatutsky G, et al. Initiative to reduce avoidable hospitalizations among nursing facility residents shows promising results. Health Aff 2017;36:441e450.
1286
H. Temkin-Greener et al. / JAMDA 20 (2019) 1280e1286
12. Grabowski DC, Aschbrenner KA, Feng Z, Mor V. Mental illness in nursing homes: Variations across states. Health Aff 2009;28:689e700. 13. Perry M, Cummings J, Jacobson G, et al. To Hospitalize or Not to Hospitalize? Medical Care for Long-Term Care Facility Residents. Health Costs. San Francisco, CA: Henry J. Kaiser Family Foundation; 2010. 14. Feng Z, Coots LA, Kaganova Y, Wiener JM. Hospital and ED use among Medicare beneficiaries with dementia varies by setting and proximity to death. Health Aff 2014;33:683e690. 15. Carter MW, Porell FW. Vulnerable populations at risk of potentially avoidable hospitalizations: The case of nursing home residents with Alzheimer’s disease. Am J Alzheimers Dis Other Demen 2005;20:349e358. 16. Hawes C, Morris JN, Phillips CD, et al. Reliability estimates for the minimum data set for nursing home resident assessment and care screening (MDS). Gerontologist 1995;35:172e178. 17. Mor V, Angelelli J, Jones R, et al. Inter-rater reliability of nursing home quality indicators in the U.S. BMC Health Serv Res 2003;3:20. 18. Saliba D, Jones M, Streim J, et al. Overview of significant changes in the minimum data set for nursing homes version 3.0. J Am Med Dir Assoc 2012;13: 595e601. 19. Fullerton CA, McGuire TG, Feng Z, et al. Trends in mental health admissions to nursing homes, 1999-2005. Psychiatr Serv 2009;60:965e971. 20. Thomas K, Dosa D, Wysocki A, Mor V. The Minimum Data Set 3.0 Cognitive Function Scale. Med Care 2017;55:e68ee72. 21. Perlman CM, Hirdes JP. The Aggressive Behavior Scale: A new scale to measure aggression based on the minimum data set. J Am Geriatr Soc 2008;56: 2298e2303. 22. Spector WD, Limcangco R, Williams C, et al. Potentially avoidable hospitalizations for elderly long-stay residents in nursing homes. Med Care 2013;0:1e9.
23. Brennan N, Engelhardt T. Data brief: Sharp reductions in avoidable hospitalizations among long-term care facility residents. CMS website. 2017. Available at: https://www.cms.gov/blog/data-brief-sharp-reduction-avoidable-hospitali zations-among-long-term-care-facility-residents. Accessed March 30, 2019. 24. Becker MA, Boaz TL, Andel R, et al. Predictors of preventable nursing home hospitalizations: The role of mental disorders and dementia. Am J Geriatr Psychiatry 2010;18:475e482. 25. Centers for Medicare and Medicaid Services (CMS). Initative to reduce avoidable hospitalizations among nursing facility residents. 2019. Available at: https://www.cms.gov/Medicare-Medicaid-Coordination/Medica re-and-Medicaid-Coordination/Medicare-Medicaid-Coordination-Office/Initi ativetoReduceAvoidableHospitalizations/AvoidableHospitalizationsamong NursingFacilityResidents.html. Accessed March 30, 2019. 26. Unroe KT, Hickman SE, Carnahan JL, et al. Investigating the avoidability of hospitalizations of long stay nursing home residents: Opportunities for improvement. Innov Aging 2018;2:1e9. 27. Mukamel DB, Ladd H, Temkin-Greener H. Stability of cardiopulmonary resuscitation and do-not-resuscitate orders among long-term nursing home residents. Med Care 2013;51:666e672. 28. Konetzka RT, Spector W, Shaffer T. Effects of nursing home ownership type and resident payer source on hospitalization for suspected pneumonia. Med Care 2004;42:1001e1008. 29. Hirth RA, Grabowski DC, Feng Z, et al. Effect of nursing home ownership on hospitalization of long-stay residents: An instrumental variables approach. Int J Health Care Finance Econ 2014;14:1e18. 30. Grabowski DC, O’Malley AJ, Barhydt NR. The costs and potential savings associated with nursing home hospitalizations. Health Aff 2007;26: 1753e1761.
1286.e1
H. Temkin-Greener et al. / JAMDA 20 (2019) 1280e1286
Appendix Supplementary Table 1 Individual- and Facility-Level Factors and Their RRRs for Nonavoidable and Potentially Avoidable Hospitalizations for Long-Stay Residents Variable
Individual resident characteristics Age, y Gender ¼ male Marital status ¼ married Race (white ¼ reference) Black Hispanic Asian Other Aggressive behavior (reference ¼ none) Mild Moderate Severe ADL score (0-28) Mood distress (PHQ-9) (reference ¼ none) Mild Moderate Moderate-severe Severe Cognitive function score (reference ¼ none) Mild Moderate Severe Number of chronic conditions Medicare/Medicaid eligible Nursing home characteristics Bed size % Medicaid % Medicare Occupancy rate RN/nursing staff ratio Chain For-profit Overall star rating (reference ¼ 2-4) Star ¼ 1 Star ¼ 5 Note. Bold values are statistically significant P < .05. CI, confidence interval; RN, registered nurse; RRR, relative risk ratio.
Nonavoidable Hospitalizations
Potentially Avoidable Hospitalizations
RRR (95% CI)
RRR (95% CI)
0.986 (0.98, 0.99) 1.261 (1.24, 1.28) 1.075 (1.06, 1.09)
0.997 (0.99, 0.99) 1.001 (0.99, 1.02) 1.104 (1.08, 1.13)
1.279 1.303 1.261 1.088
(1.25, (1.26, (1.19, (1.04,
1.31) 1.35) 1.34) 1.14)
1.174 1.273 1.047 1.102
(1.14, (1.21, (0.97, (1.04,
1.20) 1.33) 1.13) 1.16)
1.040 1.081 1.147 1.016
(1.02, (1.05, (1.08, (1.01,
1.06) 1.12) 1.22) 1.02)
1.047 1.071 1.096 0.998
(1.02, (1.03, (1.02, (0.99,
1.07) 1.11) 1.18) 0.99)
1.006 1.012 0.959 0.903
(0.99, (0.98, (0.90, (0.79,
1.03) 1.04) 1.02) 1.03)
1.051 1.041 0.933 0.868
(1.03, (1.00, (0.87, (0.76,
1.07) 1.08) 0.99) 1.00)
0.885 0.734 0.570 1.194 0.794
(0.87, (0.72, (0.55, (1.19, (0.69,
0.90) 0.75) 0.59) 1.20) 0.71)
0.835 0.669 0.446 1.332 0.814
(0.82, (0.65, (0.43, (1.32, (0.80,
0.85) 0.68) 0.46) 1.34) 0.83)
1.000 1.005 1.012 1.000 0.994 0.989 1.145
(1.00, (1.00, (1.01, (0.99, (0.99, (0.97, (1.12,
1.00) 1.01) 1.01) 1.00) 0.99) 1.01) 1.17)
1.000 1.005 1.015 0.999 0.992 0.975 1.135
(1.00, (1.04, (1.01, (1.00, (0.99, (0.95, (1.10,
1.00) 1.06) 1.02) 1.00) 0.99) 1.00) 1.17)
1.005 (0.97, 1.04) 0.986 (0.94, 1.03)
0.974 (0.94, 1.01) 0.961 (0.91, 1.01)