Nursing Home Transfers for Behavioral Concerns: Findings from the OPTIMISTIC demonstration project

Abstract

OBJECTIVES:

To characterize pre-transfer on-site nursing home (NH) management, transfer disposition, and hospital discharge diagnoses of long-stay residents transferred for behavioral concerns.

DESIGN:

Secondary data analysis of the Optimizing Patient Transfers, Impacting Medical Quality, Improving Symptoms: Transforming Institutional Care (OPTIMISTIC) project, in which clinical staff embedded in the NH setting conducted medical, transitional, and palliative care quality improvement initiatives and gathered data related to resident transfers to the emergency department/hospital setting. R software and Microsoft Excel were used to characterize a subset of transfers prompted by behavioral concerns.

SETTING:

Nursing homes in central Indiana (N = 19)

PARTICIPANTS:

Long-stay nursing home residents with behavioral concerns prompting transfer for acute emergency department/hospital evaluation (N = 355 transfers)

MEASUREMENTS:

Symptoms prompting transfer; resident demographics and baseline characteristics (MDS variables including scores for the Cognitive Function Scale, ADL Functional Status, behavioral symptoms directed towards others, and pre-existing psychiatric diagnoses); on-site management (e.g., medical evaluation in-person or by phone, testing, and interventions); avoidability rating; transfer disposition (inpatient versus emergency department only); and hospital discharge diagnoses.

RESULTS:

Over half of transfers, 56%, had a medical evaluation prior to transfer, and diagnostic testing was conducted prior to 31% of transfers. Following transfer, 80% were admitted. The most common hospital discharge diagnoses were dementia-related behaviors (27%) and altered mental status (27%), followed by a number of medical diagnoses.

CONCLUSION:

Most transfers for behavioral concerns merited hospital admission and medical discharge diagnoses were common. There remain significant opportunities to improve pre-transfer management of NH transfers for behavioral concerns.

INTRODUCTION

Nursing home (NH) residents are a medically complex and frail population with high rates of emergency department utilization^1,2 and hospital admission^3,4 for both acute and chronic conditions. Transfers that result in hospitalizations are associated with increased risk for delirium,⁵ functional decline,⁶ errors related to continuity of care, and higher financial costs.^7,8 Growing awareness of these risks has led to systems-level efforts to decrease hospitalizations and incentivize continuity of care in lower acuity settings. Despite these efforts, many NH transfers are considered to be potentially avoidable,^7,8 meaning they might be preventable or manageable in the NH setting.

To address this issue, the Optimizing Patient Transfers, Impacting Medical Quality, Improving Symptoms: Transforming Institutional Care (OPTIMISTIC) project was undertaken across nineteen nursing homes in central Indiana. As detailed in prior publications, OPTIMISTIC is a multicomponent clinical demonstration project funded by the Centers for Medicare and Medicaid Services (CMS) that utilizes proactive on-site medical, transitional, and palliative care interventions in an effort to reduce potentially avoidable hospitalizations among long-stay NH residents.^9,10 Notably, the first phase of the project yielded relative reductions of 19% in all-cause hospitalizations and 33% in potentially avoidable hospitalizations.¹¹

Prior OPTIMISTIC publications have explored avoidability of hospitalizations,^12,13 particularly those stemming from a defined set of conditions commonly identified as potentially avoidable reasons for hospital transfer: pneumonia, urinary tract infection, dehydration, congestive heart failure, skin infection, and chronic obstructive pulmonary disease or asthma.^14,15 This secondary data analysis explores NH transfers prompted by behavioral concerns, a common and clinically challenging situation in the NH setting.

Given the complexity of the typical NH resident, any acute behavioral change may have multiple relevant and potentially contributory factors, including exacerbations of medical or psychiatric conditions and/or delirium. Delirium is highly under-recognized and was found to have an incidence of 40% over the duration of a nursing home stay, even among a sample definition excluding those with moderate or more severe baseline cognitive impairment,¹⁶ making it an intriguing opportunity for investigation in the OPTIMISTIC data set.

Despite existing research on associated symptoms in paradigms such as neuropsychiatric symptoms^17–19 and behavioral and psychological symptoms of dementia,^20,21 there is limited research to guide the clinician faced with an undifferentiated behavioral concern in the general NH population. As such, the purpose of this analysis is to better understand pre-transfer on-site NH management, transfer disposition, and eventual hospital discharge diagnoses of those transferred for behavioral concerns. Behavioral concerns were defined broadly within the framework of previously collected OPTIMISTIC data, including any transfer presenting with behavioral, psychiatric (depressive affect, change in appetite, suicidal ideation), and/or cognitive (confusion or worsening cognitive function, change in mental status) symptoms, reflective of the wide range of presentations that clinicians face. This analysis seeks to provide context for the OPTIMISTIC objective of reducing potentially avoidable hospitalizations and inform clinical care. Given the persistence and ubiquity of behaviorally-related transfers, customized interventions beyond current measures focused on in-place management of medical conditions may be required, with characterization of current management and diagnoses as a necessary first step.

METHODS

Data Collection

In each of the participating OPTIMISTIC NHs, a registered nurse led multicomponent quality improvement efforts with clinical support from a nurse practitioner. Any time that a transfer occurred, an OPTIMISTIC registered nurse on site completed a transfer root cause analysis form. This form included documentation of the symptom(s) prompting transfer, any medical evaluation, testing, or interventions that occurred prior to transfer, and staff rating of whether the transfer was avoidable. Symptoms were elicited with three separate prompts, specifically primary, secondary, and other symptoms; with each prompt, the OPTIMISTIC nurse also had the option to select “Others (specify)” and write in a response.

Residents returning to the facility were seen by the OPTIMISTIC nurse practitioner (NP) for a post-discharge visit, and data were collected via completion of the NP Transition Visit form. This included data related to transfer disposition and hospital discharge diagnosis. Diagnoses were elicited with prompts for both primary and secondary diagnoses; as with pre-transfer symptoms, with each prompt, the OPTIMISTIC nurse practitioner also had the option to select “Others (specify)” and write in the response.

Minimum Data Set 3.0 (MDS)^22–24 data was collected quarterly for each nursing home resident, including an annual comprehensive assessment. MDS data for each transfer was taken from the assessment most closely preceding each transfer event. As such, the temporal relationship of the MDS data to the time of transfer could vary by as much as three months. For this reason, the Confusion Assessment Method results from the MDS were not utilized.

Participants

Participants were selected from OPTIMISTIC data during an 18-month period between January 1, 2015 and June 30, 2016. For the OPTIMISTIC study as a whole, eligible participants included long-stay nursing home residents, defined as those with facility stays >100 days or with no plan for discharge in the MDS.⁹ Residents and their families, per CMS guidelines, had the opportunity to opt-out of the project, and less than 1% of eligible residents opted out in the study timeframe.

Beyond the general OPTIMISTIC inclusion criteria, transferred residents were selected for inclusion into this secondary data analysis based on the presenting symptom(s) that prompted transfer. Residents were eligible for inclusion if they had one or more of the following qualifying symptoms at time of transfer: behavioral symptoms, confusion or worsening cognitive function, change in mental status, change in appetite, depressive affect, or suicidal ideation. This cluster of symptoms was selected in an effort to broadly define behavioral concerns, with particular attention to potential overlap of medical and psychiatric symptoms and/or the presence of delirium. Residents without these behavioral concerns were incorporated as a comparison group in a logistic regression model to characterize differences between these groups.

Measures

Variables of interest came from both OPTIMISTIC registered nurse and nurse practitioner data collection as well as the MDS. OPTIMISTIC variables included demographic data, on-site management undertaken such as type of evaluation (in-person or by phone), testing, and interventions, avoidability rating,¹² and transfer disposition (inpatient versus emergency department only). Avoidability rating was based on OPTIMISTIC registered nurse clinical judgment, guided by observations while embedded at the NH and/or review of medical records or other data collected by staff. MDS variables incorporated into this analysis included scores for the Cognitive Function Scale,²⁵ ADL Functional Status,²⁶ behavioral symptoms directed towards others, and pre-existing psychiatric and dementia diagnoses. ADL Functional Status was calculated based on summation of seven ADL scores (dressing, hygiene, toileting, locomotion, transfer, bed mobility, and eating) ranging from 0 (independent) to 4 (total dependence).

Analysis

Though symptoms prompting transfer were recorded in the root cause analysis transfer form as primary, secondary, or other, because staff were not given specific direction on how to prioritize multiple presenting symptoms, the three tiers were not distinguished in the data analysis. Similarly, primary and secondary hospital discharge diagnoses were treated equivalently. For residents meeting inclusion criteria, open-ended responses in cases where the symptom selected was “Other (please specify)” were individually reviewed and, if appropriate, re-categorized as a symptom from the standard list. Similarly, open-ended responses for primary and/or secondary discharge diagnoses were individually reviewed and, if appropriate, re-categorized (for example, encephalopathy was re-categorized as altered mental status since this was the term on the NP Transition Visit form), and a separate category was created for psychiatric diagnoses or suicidal ideation. Likewise, open-ended responses for pre-transfer testing and interventions were reviewed and re-categorized as appropriate.

The data analysis was performed using R software.²⁷ In order to facilitate re-categorization of open-ended responses for symptoms and diagnoses and more easily analyze co-occurring symptoms and associated diagnoses, an R software output file was generated for further analysis in Microsoft Excel. The data were analyzed at the transfer level rather than the participant level given that the research questions centered on the characterization of NH resident transfers. Individual residents may have more than one transfer.

A logistic regression model was utilized in two analyses. The first examined transfer characteristics associated with disposition of hospital admission versus emergency department evaluation only. The second examined characteristics associated with transfer for behavioral concerns, versus the previously-excluded sample transferred for non-behavioral concerns. For each of these analyses, a generalized estimating equations (GEE) logistic model was used to account for the potential correlations among transfers by the same resident, including variables of age, sex, race, ADL rating, Cognitive Function Scale score, on-site evaluation, on-site testing, on-site intervention, avoidability rating, MDS diagnostic categorizations including anxiety, mood disorder, and psychotic disorder, and eventual discharge diagnosis; the first model also incorporated presenting symptoms as either behavioral only or both behavioral and non-behavioral. Temporal correlations were accounted for by an autoregressive (AR1) working correlation matrix in the GEE model. All associations were measured with adjusted odds ratios; 95% confidence intervals and p values were provided. Variables with p values less than 0.05 were considered as statistically significant. Analyses were conducted using R software²⁷ and the geepack package.²⁸

RESULTS

Of 1502 transfers during the study period, 355 transfers met secondary data analysis inclusion criteria, composed of 279 distinct residents. Characteristics of residents involved in transfers are provided in Table 1, with 80% transferred once during the study period, 15% transferred twice, and the remaining 5% transferred between three and five times. Demographic data reported below is based upon transfer-level data. Transferred residents were 76 years of age on average, 60% were female, and 74% were white. While 36% had moderate or severe baseline cognitive impairment on the Cognitive Function Scale, 53% had a dementia diagnosis. Based on MDS assessment data most closely preceding the transfer event, baseline behavioral symptoms directed towards others in the past week were most often verbal (7%), followed by physical (4%) and other (2%). Rates of baseline psychiatric diagnoses included 58% with depression, 33% with anxiety, and 17% with schizophrenia or other psychotic disorder.

Table 1:

Characteristics of Transferred Participants^*

Demographic Data (N = 355 transfers)	Mean (SD) or N (%)
Age Upon Transfer (N = 344)	76.1 (12.4)
Sex (N = 344)
Male	138 (40%)
Female	206 (60%)
Race (N = 344)
White	253 (74%)
Black	61 (18%)
Hispanic	10 (3%)
Other	20 (6%)
Married (N = 308)	71 (23%)
Transfers during study period (resident-level data, N = 279)	1.3 (0.6)
One	223 (80%)
Two	41 (15%)
Three	12 (4%)
Four	1 (0.4%)
Five	2 (1%)
Baseline Cognitive and Behavioral Status	Mean (SD) or N (%)
MDS Cognitive Function Scale Score (N = 342)
Cognitively Intact	122 (36%)
Mildly Impaired	96 (28%)
Moderately Impaired	118 (35%)
Severely Impaired	6 (2%)
MDS ADL Functional Status Score (N = 342)	18.9 (3.6)
MDS Behavioral Symptoms Directed Towards Others (N = 343)
Physical
Behavior Occurred 1–3 Days of Past Week	13 (4%)
Verbal
Behavior Occurred 1–3 Days of Past Week	22 (6%)
Behavior Occurred 4–6 Days of Past Week	1 (0.3%)
Other
Behavior Occurred 1–3 Days of Past Week	5 (1%)
Behavior Occurred 4–6 Days of Past Week	2 (1%)
MDS Selected Active Diagnoses (N = 344)	N (%)
Anxiety	115 (33%)
Depression (Other Than Bipolar)	199 (58%)
Dementia (Alzheimer’s or non-Alzheimer’s)	184 (53%)
Manic Depression (Bipolar Disease)	19 (6%)
Psychotic Disorder (Other than Schizophrenia)	24 (7%)
Schizophrenia	35 (10%)

^*Mean (SD) and percentages were calculated only on available data. Missing data were excluded from these calculations.

The most common symptom at time of transfer was change in mental status (N=148, 42%), followed by behavioral symptoms (N=122, 34%), with additional symptoms detailed in Table 2. For transfers studied, 46% (N=166) had only one presenting symptom recorded. However, the average number of recorded symptoms per transfer was 2.1, with a range from 1 to 12.

Table 2:

List of Symptoms Prompting Transfer

Symptoms Ranked by Frequency	%	N
Change in mental status	42%	148
Behavioral symptoms	34%	122
Confusion or worsening cognitive function	28%	101
Fever	9%	32
Shortness of breath/high respiratory rate	9%	31
Tachycardia	7%	26
Infection	7%	26
Vomiting	5%	18
Malaise	5%	17
Change in appetite	5%	17
Hypotension	5%	16
Hypertension	5%	16
Unresponsiveness	4%	15
Critical lab value	4%	15
Pain	4%	14
Hypoxia	4%	14
Falls with injury	4%	14
New neurologic weakness	3%	12
Suicidal ideation	3%	11
Falls without obvious injury	3%	11
Seizure	2%	8
Cough	2%	7
Urinary symptoms or incontinence	2%	6
Fluid imbalance	2%	6
Abdominal pain	1%	5
Limb swelling	1%	4
Depressive affect	1%	4
Diarrhea	1%	2
Chest pain	1%	2
Suspected soft tissue injury	0.3%	1
Loss of consciousness	0.3%	1
Head trauma	0.3%	1
Dizziness	0.3%	1
Bradycardia	0.3%	1
Bloody stool	0.3%	1
Bleeding (other than GI)	0.3%	1
Anemia	0.3%	1
Other (please specify)	4%	15

Symptoms from inclusion criteria are listed in bold.

In terms of on-site management prior to transfer (Table 3), over half of transfers (N=195, 56%) had a medical evaluation prior to transfer, either an in-person evaluation (N=110, 32%) or by phone only (N=84, 24%). Diagnostic testing was conducted on 31% (N=101) of transfers, including blood tests (N=61, 19%), urinalysis and/or culture (N=36, 11%), x-ray (N=32, 10%), ECG (N=2, 1%), venous Doppler (N=1, 0.3%), and other testing (N=20, 6%), with other testing or assessments described by OPTIMISTIC registered nurses including checking vital signs, glucose level, or neurological exam. Pre-transfer intervention occurred in 49% (N=163) of transfers and included medication adjustments (N=78, 23%), oxygen supplementation (N=40, 12%), IV or subcutaneous fluids (N=10, 3%), palliative care (N=1, 0.3%), and other interventions (N=75, 23%) such as redirection, non-pharmacologic intervention, one-to-one or other increased monitoring, or breathing treatments. Most transfers were rated as definitely not avoidable (N=102, 29%) or probably not avoidable (N=159, 46%), with the remainder evaluated as probably avoidable (N=68, 19%) or definitely avoidable (N=20, 6%).

Table 3:

Management Prior to Transfer^*

Management Technique	N (%)
Medical Evaluation (N = 346 transfers)	195 (56%)
In Person	110 (32%)
Phone Only	84 (24%)
No Evaluation	152 (44%)
Testing (N = 325 transfers)	101 (31%)
Blood Tests	61 (19%)
EKG	2 (1%)
Urinalysis and/or Culture	36 (11%)
Venous Doppler	1 (0.3%)
X-Ray	32 (10%)
Other (open-ended responses)	20 (6%)
Vital Signs	11 (3%)
Glucose Check(s)	5 (2%)
Neurological Checks	3 (1%)
Psychiatric Evaluation	2 (1%)
Stool Studies	2 (1%)
Interventions (N = 332 transfers)	163 (49%)
New/Changed Medication	78 (23%)
IV or Subcutaneous Fluids	10 (3%)
Oxygen	40 (12%)
Palliative Care	1 (0.3%)
Other (open-ended responses)	75 (23%)
Redirection	16 (5%)
1:1 Supervision	14 (4%)
Non-Pharmacologic Intervention	11 (3%)
Other Increased Monitoring	7 (2%)
Nebulizer or Other Breathing Intervention	6 (2%)
Neurological Checks	6 (2%)
Vital Signs	4 (1%)
Glucose Checks	4 (1%)
Glucagon	2 (1%)
Other	6 (2%)
Inappropriate– Non-Intervention Response	9 (3%)

^*Percentages were calculated only on available data. Missing data were excluded from these calculations.

Most resident transfers associated with behavioral concerns resulted in admission to a hospital for inpatient treatment (N=280, 79%). The others were evaluated in the emergency department and not admitted (N=72, 20%), admitted for observation (N=2, 1%), or admitted with uncertain status (N=1, 0.3%). Of transfers, 91% (N=322) returned to their original facility at the conclusion of their post-transfer management, while the remainder died in the hospital (N=22, 6%) or transferred to another setting (N=11, 3%).

The most common discharge diagnoses (Table 4 and Figure 1) were dementia-related behaviors (N=96, 27%), and altered mental status (N=95, 27%), with these two diagnoses often overlapping (N=23). The next most common diagnoses were pneumonia (N=50, 14%), UTI without sepsis (N=47, 13%), and diabetes-related complications (N=34, 10%). A significant number of subjects were missing discharge diagnosis data (N=63, 18%).

Table 4:

List of Hospital Discharge Diagnoses

Discharge Diagnosis (N = 355 transfers)	%	N
Dementia-related behaviors	27%	96
Altered mental status	27%	95
Pneumonia	14%	50
UTI without sepsis	13%	47
Diabetes-related complications	10%	34
COPD exacerbation	8%	27
Fall with or without injury	7%	24
Psychiatric diagnosis or suicidal ideation	7%	24
Heart failure	6%	23
Sepsis	6%	20
Urosepsis	5%	18
Seizures	5%	17
Stroke or TIA	4%	15
Acute renal failure	4%	15
Hypertensive urgencies and related problems	4%	14
Uncontrolled pain	3%	12
Electrolyte abnormalities	3%	11
Dehydration unspecified reason	2%	7
Cellulitis, abscesses and other related problems	1%	5
Constipation/impaction	1%	5
C Diff and related problems	1%	4
Ulcers/wounds	1%	4
Hip Fracture	1%	3
MI	1%	2
Fracture other than hip	1%	2
Laceration	0%	1
Others (specify)	30%	106
Missing	18%	63

Figure 1.

In the OPTIMISTIC study, 80% of nursing home transfers for behavioral concerns were admitted to the hospital, with a wide range of eventual discharge diagnoses including many medical diagnoses.

In examining behavioral transfers versus non-behavioral transfers (N=1102), behaviorally-related transfers were associated in a statistically significant manner with greater likelihood of eventual discharge diagnosis of dementia-related behavior or altered mental status, and lower likelihood of eventual medical or other non-psychiatric discharge diagnosis. In examining transfer disposition (hospital admission versus emergency department only), statistically significant predictors of hospital admission included male sex, greater ADL limitations, on-site intervention prior to transfer, and assessment of the transfer as not avoidable by the OPTIMISTIC nurse at the time of transfer. Logistic regression GEE model results are detailed in the supplemental data section (Supplementary Tables S1 and S2).

DISCUSSION

OPTIMISTIC is a high-impact and innovative program with demonstrated quality improvement in the NH setting. This is the first OPTIMISTIC publication to characterize on-site management of behavioral concerns prior to transfer, and it extends the limited literature base related to behavioral symptoms as potential signals of medical illness.²⁹ It also expands the scope of OPTIMISTIC study findings into the realm of behavioral health and the interface of medicine and psychiatry. Multiple findings from these analyses merit further discussion, including baseline sample characteristics, lower than expected rates of on-site management, high rates of hospital admission, and frequency of medical conditions in hospital discharge diagnoses.

Transfers related to behavioral concerns in this analysis were, on average, older than the typical OPTIMISTIC resident in prior publications. They also had a high rate of baseline psychiatric diagnosis; for example, MDS data reported that over half the sample had depression and 10% had schizophrenia. This is striking though not unexpected given that the sample was defined to investigate those with symptoms that are potentially psychiatric in origin. That said, the most common symptoms in this sample (change in mental status, behavioral symptoms, and confusion or worsening cognitive functioning) are certainly not specific to psychiatric etiologies.

In these findings, some aspects of pre-transfer management are less robust than might be expected, particularly in the context of an initiative to reduce avoidable hospitalizations. For example, only 32% of transfers had an in-person evaluation by a treating physician, physician assistant, or nurse practitioner prior to transfer, and only 31% had diagnostic testing prior to transfer. Although urinary tract infection is prominent in the differential for unexplained behavioral change in older adults, only 11% of subjects received a urinalysis and/or culture prior to transfer; 18% received discharge diagnoses of urosepsis or urinary tract infection without sepsis. Prior analyses of OPTIMISTIC data also found that project RNs cited lack of use of available NH resources and problems with communication as areas of quality improvement in potentially avoidable transfer events.⁹

Though these findings suggest opportunities to expand on-site management, disposition data nonetheless suggests that many transfers were sufficiently high acuity for escalation to hospital evaluation, with 80% of transfers admitted to the hospital under either observation or inpatient status (Figure 1). This is consistent with OPTIMISTIC nurse assessment of avoidability at the time of transfer; most transfers were not felt to be avoidable, and assessment as non-avoidable was associated with inpatient admission. However, this admission rate is notably higher than what was seen in several past studies, in which about one third to one half of emergency department residents were admitted,¹ though these studies included community-dwelling older adults. One study limited to those age 75 or older found that 65% of those presenting to the ED were admitted,³⁰ closer to the admission rate seen in this analysis. It is also important to note that other studies included a broader range of presenting symptoms than this behaviorally-focused analysis. This difference in sample population as well as varying living situations and associated differences in complexity and functional status may explain some of the disparity in admission rates, but the higher observed admission rates may also reflect the effectiveness of OPTIMISTIC. Demonstrated reductions in hospitalization rates among participating NHs suggest that OPTIMISTIC is having a meaningful impact on clinical practice and transfer decision-making in the NH setting. Furthermore, in this analysis, on-site intervention prior to transfer was positively associated with eventual inpatient admission. Stated another way, among those for whom on-site intervention was both attempted and insufficient, emergency department management alone was less likely to suffice. This reinforces that OPTIMISITC efforts to reduce potentially avoidable hospitalizations may be positively influencing the appropriateness of transfers and increasing the acuity of transfers that do occur.

Although medical discharge diagnoses were more strongly associated with non-behavioral transfers, hospital discharge diagnoses nonetheless highlight the overlap between medical and psychiatric conditions among NH residents transferred for broadly-defined behavioral concerns. For this sample selected based on qualifying symptoms including altered mental status, psychiatric symptoms, and general behavioral symptoms, the most common discharge diagnoses were dementia-related behaviors (27%) and altered mental status (27%). However, the next six most common discharge diagnoses were common medical conditions: pneumonia (14%), UTI without sepsis (13%), diabetes-related complications (10%), COPD exacerbation (8%), and heart failure (6%), as well as 7% with falls with or without injury. Incidence of each of these individual diagnoses was greater than or equal to that of aggregated psychiatric hospital discharge diagnoses (7%). Importantly, among the fifteen transfers for depressive affect or suicidal ideation, more than half (N=8) had medically-based discharge diagnoses such as pneumonia, UTI without sepsis, and acute renal failure. These results underscore the importance of appropriate medical workup for psychiatric chief complaints.

Notably, delirium was not listed as a discharge diagnosis for any transfer, despite the fact that more than a quarter had a discharge diagnosis of altered mental status. It is likely that some or many of the altered mental status diagnoses (N=95) may represent delirium, particularly as the diagnosis co-occurred with a number of medical conditions including diabetes-related complications (N=16), UTI without sepsis (N=14), pneumonia (N=13), heart failure (N=10), sepsis (N=9), and COPD exacerbation (N=9). Delirium could also be contributory in cases diagnosed as dementia-related behaviors if the behaviors represented a deviation from baseline.

It is possible that differing nomenclature or diagnostic ambiguity explain some diagnoses of non-delirium altered mental status. For example, terminology in discharge documentation may reflect billing requirements rather than the nuances of clinician understanding. OPTIMISTIC staff were not instructed to explicitly assess for and record delirium as a diagnosis, as the study was not designed to focus specifically on delirium. Despite these caveats, the complete absence of delirium diagnoses is not consistent with clinical experience and is likely reflective of the well-documented under-recognition of delirium. In fact, Han et al. found that almost 40% of nursing home residents presenting to the emergency department have delirium.³¹ It is critical that providers consider and recognize delirium as a specific entity to limit its serious sequelae including increased mortality. Delirium is often preventable and evidence-based non-pharmacologic management techniques exist,³² but it must be included in the differential diagnosis for research evidence to translate to clinical benefit.

A number of limitations should be noted. Most prominently, while this secondary data analysis is possible because of the broader OPTIMISTIC study, this also limits its generalizability. Findings are drawn from a multicomponent quality improvement effort with a demonstrated impact on hospitalization rates in central Indiana NHs. As such, usual on-site management is likely different than in non-OPTIMISTIC NHs. Transfers investigated in this analysis may also represent a particularly challenging subset of the typical spectrum of cases that might prompt consideration of transfer in a usual NH.

Another limitation stems from the data collection process. OPTIMISTIC was designed to focus on medical conditions targeted as potentially avoidable with proactive on-site management. It was not specifically focused on mental health concerns or delirium. Thus the lists of presenting symptoms and discharge diagnoses were not tailored specifically to investigation of behaviorally-related or mental health symptoms. For example, neither delirium nor any psychiatric diagnoses were included in the standard list of discharge diagnoses utilized for data collection at the post-discharge visit. OPTIMISTIC staff did have the option to list alternative open-ended responses, but this extra step may mean that some mental health or delirium diagnoses were not captured. Furthermore, for purposes of this analysis, several of the listed discharge diagnoses were vague, not fully clarifying the clinical situation, as with the frequently-selected “altered mental status.” Further complicating interpretation, a number of open-ended diagnosis entries were symptoms without elaboration as to cause. While the hospital setting has more resources than the NH for diagnostic work-up, the nonspecific nature of these discharge diagnoses may reflect real-world emphasis on ruling out emergent etiologies as opposed to clarifying the nuances of the case.

Despite these limitations, this secondary data analysis offers a clinically-relevant characterization of residents transferred for behavioral concerns. It underscores the frequent co-occurrence of medical and behavioral or psychiatric symptoms, and eventual discharge diagnoses support behavioral symptoms as potential indicators of medical illness. Rates of pre-transfer management suggest opportunities for improvement, though in light of the high rates of inpatient admission, it is unclear whether escalated on-site management beyond the OPTIMISTIC quality improvement efforts would be sufficient to avert these transfers.

Since implemented OPTIMISTIC efforts have already been shown to successfully reduce hospitalizations among long-stay NH residents, further quality improvement efforts could focus on targeted interventions for common but challenging issues in NHs, with acute behavioral concerns as a clear area of opportunity. Proactive on-site management specifically tailored to psychiatric and behavioral changes, including preventive environmental measures, behavioral de-escalation techniques, and evaluation of potential medical contributors, could improve resident quality of life and further limit preventable transfers and hospitalizations.

Supplementary Material

supinfo

Supplemental Table S1. Among nursing home transfers for behavioral concerns, statistically significant predictors of hospital admission included male sex, greater ADL limitations, on-site intervention prior to transfer, and assessment of the transfer as not avoidable by the OPTIMISTIC nurse at the time of transfer. Note: * indicates p value < 0.05. Age and ADL Functional Status are continuous variables. All other variables are categorical and unless noted, were coded as yes vs. no for the condition stated.

Supplemental Table S2. Nursing home transfers for behavioral concerns, relative to non-behaviorally-related transfers, were statistically significantly associated with greater likelihood of eventual discharge diagnosis of dementia-related behavior or altered mental status, and lower likelihood of eventual medical or other non-psychiatric discharge diagnosis. Note: * indicates p value < 0.05. Age and ADL Functional Status are continuous variables. All other variables are categorical and unless noted, were coded as yes vs. no for the condition stated.