Characteristics of elderspeak communication in hospital dementia care: Findings from the Nurse Talk observational study

Abstract

Background:

Elderspeak communication is typically viewed as patronizing and infantilizing by older adults and can lead to resistive behaviors in persons living with dementia. Little is known about the presence of elderspeak communication in hospitals in the United States. Understanding this phenomenon in the hospital setting is needed in order to improve hospital dementia care.

Objectives:

The purpose of the Nurse Talk study was to (1) describe attributes of elderspeak use in hospital dementia care and to (2) determine what characteristics are associated with nursing staff use of elderspeak communication with hospitalized patients with dementia.

Design:

A cross-sectional observational study design was used to collect and analyze audio-recordings of nursing staff during care for hospitalized patients with dementia.

Setting:

Three hospital units in one Midwestern university hospital in the United States.

Participants:

A convenience sample of 53 staff nurses and nursing assistants that provided direct care to 16 patients with mild or more severe dementia recruited from October 2019 through mid-March 2020.

Methods:

Eighty-eight care encounters were audio-recorded and coded for elderspeak communication using the Iowa Coding of Elderspeak scheme to determine the frequency and characteristics of elderspeak communication. A linear mixed effects model was used to determine what characteristics were associated with elderspeak and the frequency of elderspeak use by nursing staff to hospitalized patients with dementia.

Results:

Over a quarter (28.7%) of all nursing staff speech directed towards patients with dementia constituted elderspeak and nearly all (96.6%) care encounters included some elderspeak. Particularly common attributes of elderspeak were minimizing words and mitigating expressions, childish terms and phrases, and collective pronoun substitution. A statistically significant interaction was identified between staff role and age (95% CI: −0.02, −0.00, $p=0.008$) in predicting the frequency of elderspeak use, indicating that elderspeak was used more often by older staff nurses, whereas the age of nursing assistants remained constant across elderspeak use. Statically significant effects for delirium and length of stay were also demonstrated. Elderspeak use was 12.5% higher with patients with delirium (95% CI: 0.02, 0.23, $p=0.025$) and increased 1.5% for each additional day the patient with dementia was hospitalized (95% CI: 0.00, 0.03, $p=0.035$).

Conclusions:

Elderspeak is present and pervasive in the acute care setting. Interventions targeted towards older staff nurses and nursing staff from hospital units that care for patients with delirium and longer lengths of stay are needed.

Tweetable Abstract:

This study identified that nursing staff are frequently using elderspeak (infantilizing speech) with hospitalized patients with dementia. @claireshaw_phd @IowaNursing

Introduction

Communication is a critical aspect of nursing that is more challenging in dementia care. Changes in language and cognition are hallmark manifestations of dementia and create barriers to effective communication (American Psychiatric Association, 2013; Ryan, Meredith, et al., 1995). Ineffective communication has been directly linked to the occurrence of responsive behaviors that can disrupt care, highlighting the need to understand the communication-behavior relationship as a basis for preventing responsive dementia behaviors (Roth et al., 2002; Williams et al., 2009). Strategies for successful dementia care communication are needed to support personhood and complete essential care tasks (Buron, 2008; Wilson et al., 2013). Informal and formal caregivers advocate for enhanced education and skills training in communication to prevent responsive behaviors that are a common and challenging part of dementia care (Feast et al., 2016; Page & Hope, 2013).

Elderspeak is a form of communication overaccommodation used with older adults that is evidenced by inappropriately juvenile lexical choices and/or exaggerated prosody; arises from implicit ageist stereotypes; carries goals of expressing care, exerting control, and/or facilitating comprehension; and may lead to negative self-perceptions in older adults and resistive behaviors in persons living with dementia (Shaw & Gordon, 2021). In long-term care, elderspeak comprises 22% to 58% of staff-resident interactions (Caporael, 1981; Herman & Williams, 2009). Elderspeak communication is more common in care encounters with persons living with dementia than with cognitively intact older adults and increases with dementia severity and functional decline (Herman & Williams, 2009; Schnabel et al., 2020; Williams, 2006; Williams et al., 2009). Despite the negative responses of elderspeak in persons living with dementia, its use by nursing staff is usually inspired by positive intentions of conveying comfort, encouraging cooperation, and/or enhancing comprehension (Grimme et al., 2015). Simplified speech may be considered beneficial in dementia care (Eggenberger et al., 2013); however, most features of elderspeak—such as a high pitch and over-exaggeration—actually diminish comprehension (Kemper & Harden, 1999; Shaw & Gordon, 2021). Elderspeak communication by care staff can also result in rejection of care by residents with dementia. Sequential analyses have demonstrated that elderspeak doubles the probability of rejection of care and that, when elderspeak communication is reduced, rejection of care by residents decreases (Williams et al., 2009; Williams et al., 2017).

The communication predicament of aging model proposes that elderspeak arises due to implicit ageism (Ryan, Hummert, et al., 1995). This model describes a negative feedback loop that occurs when a younger speaker views the older adult communication partner as incompetent and dependent based on recognition of old age cues. The speaker then overaccommodates by using infantilizing speech to the older adult. Forty years of research on elderspeak has confirmed that older adults generally find elderspeak patronizing and infantilizing (Shaw & Gordon, 2021). Using elderspeak thus represents a non-person-centered approach to care because it directly violates respecting and recognizing an individual’s personhood (Kitwood, 1990; Fazio et al., 2018). A later theory, the communication enhancement model, was developed to address the need for individualized assessment to determine and limit accommodations to specific communication deficits of each older adult (Ryan, Meredith, et al., 1995). The communication enhancement model re-directs the communication predicament of aging model cycle to circumvent patronizing communication with older adults. This is accomplished by assessing cues on an individual basis rather than on stereotypes of aging. Further, accommodations adapt as communication needs are continually assessed throughout the communicative exchange. This contrasting model thus uses a person-centered approach to communication by focusing on the individuality of a person’s needs.

The key antecedent to elderspeak is the recognition of being an older adult as described in the models above, yet little else is known about what attributes of the speaker, recipient, and environment lead to elderspeak communication use, especially in the hospital setting (Shaw & Gordon, 2021). Prior research has demonstrated that adults who appear more frail and/or cognitively impaired are more likely to be the recipients of elderspeak (Shaw & Gordon, 2021). Research has varied on whether gender of the person living with dementia elicits a difference in elderspeak frequency. Some research suggests that females are the more likely recipient (Cavallaro et al., 2016; Sachweh, 1998), while other research has found no difference in the receipt of elderspeak due to gender (Cockrell, 2020; O’Connor & Rigby, 1996). Elderspeak use does not seem to depend on the gender of the care staff (Shaw & Gordon, 2021).

Most research on elderspeak to date has been conducted in long-term care settings and little is known about elderspeak in acute inpatient care settings, particularly in the United States (Shaw & Gordon, 2021). There has been one recent study which evaluated the use of elderspeak in German hospitals and demonstrated that elderspeak and controlling communication were used more by nursing staff when caring for older adults with greater functional declines (Schnabel et al., 2019; Schnabel et al., 2020), indicating that nursing staff do adopt negative communication practices with certain populations. Other observational research in acute care also suggests that communication failures are present, and interventions to both improve dyadic interactions and reduce negative outcomes such as rejection of care are needed (Allwood et al., 2017; Backhouse et al., 2020; Honda et al., 2016). The purposes of the current study, entitled the Nurse Talk study, are to (1) describe the attributes of elderspeak in hospital dementia care and (2) determine what characteristics of the nursing staff, the patient with dementia, and the environment are associated with elderspeak communication by nursing staff. It is hypothesized that (1) elderspeak will be present and pervasive in hospital dementia care and that (2) nursing staff will increase use of elderspeak when caring for patients with more advanced dementia.

Methods

The Nurse Talk study used a cross-sectional approach and convenience sampling (supplemental methods). Nursing staff and patients with dementia were recruited from three units in a single university hospital in the Midwestern United States. Point-of-care encounters between nursing staff and patients with dementia were audio-recorded to collect samples of communication between October 2019 and March 2020. All study procedures were approved by the research ethics board.

Sample

Staff nurses and nursing assistants were recruited from three study units—Family Medicine, Internal Medicine, and Neurology. These hospital units were selected because they care for a high proportion of patients with dementia with acute illnesses. Nursing staff were included if they were 18 years or older, English speaking, and provided direct care to patients with dementia on a participating unit.

Patients living with dementia were included if they were admitted to a participating unit and had an ICD-9 or ICD-10 dementia diagnosis (Ernecoff et al., 2018), a report of rejection of care since admission (noted in the medical record or reported by staff), and no excluding characteristics, as described below. Persons living with dementia were excluded if they were not fluent in English; were staged as having less than a mild severity of dementia on the Functional Assessment Staging (FAST) instrument (Reisberg et al., 1985; Sclan & Reisberg, 1992); or had one of the following diagnoses which may impact the outcome of rejection of care through different neurobiological mechanisms: mild cognitive impairment, Huntington’s disease, alcohol-related dementia, schizophrenia, traumatic brain injury, a brain tumor, or developmental delay. Additional findings on the relationship between elderspeak and rejection of care are reported elsewhere (Shaw et al., In Press).

Recruitment

Nursing staff were notified of their unit’s involvement in a communication study through emails, presentations at staff meetings, and individual introductions with the research team. Elderspeak communication was not mentioned during recruitment, data collection, or in consent documents in order to reduce bias during the recordings and observations. To ensure that there was no deception, nursing staff were informed that the study goal was to investigate the impact of communication on responsive behaviors by persons living with dementia.

Recruitment took place in two phases. In Phase 1, nursing staff were recruited. The investigator met with available nursing staff individually to introduce the study and establish trust. Staff understood that they may or may not be recorded with a patient with dementia. In Phase 2, patients with dementia were recruited along with additional nursing staff. Hospitalized patients with dementia were identified through a custom screening report in the electronic medical record, which searched the census of all patients on the participating units for ICD-9 or ICD-10 dementia diagnoses. If the patient met inclusion criteria after discussion with their primary nurse, the legally authorized representative of the patient was approached for consent and the patient was approached for assent. The legal representative was approached by the researcher in-person or via telephone after permission was received by the primary nurse. The legal representative then provided consent for the patient. The study was also discussed with the patient and verbal assent was received prior to any study procedures taking place. The study included minimal risk to the patients with dementia as it involved audio-recording care encounters that occurred inherently as a part of their hospital stay. Verbal assent was repeated prior to each recording for both the patients with dementia and nursing staff. During this study phase additional nursing staff caring for consented patients with dementia were invited to participate if they had not been previously approached. A $25 honorarium was provided to participating patients with dementia.

Data Collection

Data about the participants were collected by self-report, electronic medical record extraction, and observation. The staff self-reported on nursing related measures. The patient with dementia and their legal representative provided demographic and descriptive self-report measures. Observational study data was collected through observation by the principal investigator (first author) and co-occurring audio-recording of the care encounter. A care encounter was operationalized as a care interaction between the patient with dementia and the nursing staff such as activities of daily living, assessment, medication administration, and procedural care. Non-care task activities were not included (e.g., care updates) and care encounters were excluded if they involved non-consenting staff or family members.

Nursing staff and patients with dementia were audio-recorded on multiple occasions with multiple communication partners using an iPod touch with an attached high-quality microphone to capture nursing staff communication. When two or three nursing staff were present in the same encounter, the communication of both were examined separately. After each nursing staff member had been recorded and observed with the same patient with dementia for 25 minutes, no further observations of that dyad were collected.

Measures

Elderspeak Communication

The Iowa Coding of Elderspeak (ICodE) scheme was used to determine the frequency and characteristics of elderspeak. The ICodE contains five mutually exclusive categories of communication states during staff-patient care encounters: elderspeak communication to patients with dementia, neutral communication (i.e., staff-to-patient with dementia communication without elderspeak), staff to staff communication, patient speech, and silence. The elderspeak category includes three non-mutually exclusive major categories (i.e., semantics, discourse, and prosody) and 11 non-mutually exclusive subcategories. The semantics category identifies infantilizing word choices such as childish terms, diminutives, collective pronouns, and short words or phrases subcategories. The discourse category includes types of speech that reflect an imbalance between the communicative goals of conveying care, respect, and control. It contains subcategories of directive/imperative phrases, exaggerated praise, tag questions, minimizing words or mitigating expressions, reflective phrases, interruptions, and laughing at or otherwise belittling. The prosody category identifies altered prosodic, or intonational, patterns such as a raised pitch, a sing-song pattern of intonation, excessive changes in pitch range or volume, or over-articulation. Prosodic modifications are not broken down into subcategories. Additional information on the development, reliability, and validity of the ICodE will be reported elsewhere.

De-identified audio recordings were transcribed verbatim by a professional transcription service and then coded with the ICodE scheme. Audacity Version 2.3.3 software (https://www.audacityteam.org) and Microsoft Excel were utilized. The length of each utterance was determined to the tenth of a second and coded as belonging to one of the five mutually exclusive communication states and each appropriate elderspeak category when appropriate. Coding was completed by two coders who maintained interrater and intra-rater reliability at ICC > 90% after extensive training.

Observation-Level Measures

Observation-level and environmental characteristics that changed or had the possibility to change with each observation included care activity, familiarity with the caregiver, familiarity with environment, level of stimulation, sedation level, time of day, delirium status, and pain severity. The care activity referred to the task or tasks being performed during the care encounter, which were categorized post hoc into activities of daily living and health-care activities (i.e., medication administration, vital signs, assessment, and needle sticks). Familiarity with the caregiver was operationalized as the number of previous shifts the nursing staff had caring for the patient and familiarity with the environment was operationalized as how many days the patient had been in the hospital at the time of the care encounter. Level of stimulation was represented by the number nursing staff providing care at one time. Sedation level was operationalized as the presences of a pharmacologically sedating medication (i.e., opioids, benzodiazepines, antipsychotics, and first-generation antihistamines).

Delirium status was measured with Confusion Assessment Method (CAM) short form (Wei et al., 2008). Delirium is detected if acute onset and fluctuating course and inattention, and disorganized thinking or altered level of consciousness are present. The CAM has been demonstrated to be a reliable and valid tool for non-physician raters, such as nurses or trained laypersons (Jones et al., 2019). Pain severity was measured with the Pain Assessment in Advanced Dementia Scale (PAINAD). The PAINAD is a five-item scale, with each item (breathing, vocalization, facial expression, body language, and consolability) scored from 0 to 2 for (Warden et al., 2003). All categories are summed for a pain severity rating of 0–10 in which 0 represents no pain and 10 represents severe pain. A PAINAD score of two or greater indicates that pain is present.

Persons-Living-with-Dementia-Level Measures

Characteristics of the patients with dementia that were consistent across observations included hospital unit, demographic factors (age, gender, race, ethnicity, and education level); and factors about the dementia and other health conditions (admitting diagnosis, type and severity of dementia, severity of comorbidities). Comorbidity severity was measured with the Cumulative Illness Rating Scale (CIRS), a 14-item scale, in which each item represents a separate body system (Salvi et al., 2008). Each body system is rated on a five-point scale for severity (0=no problem; 4=extremely severe problem) and summed for a total score of 0–56 in which 0 represents no comorbidities and 56 represents severe multi-system failure. Dementia severity was measured with the FAST, which has seven categories with 16 items representing function in dementia (Reisberg, 2007; Reisberg et al., 1985). A score of 0 represents no cognitive deficits, and a score of 8 represents severe dementia. The FAST was scored using the patient’s function prior to hospitalization.

Nursing-Staff-Level Measures

Staff participant characteristics included their confidence in dementia care, knowledge of dementia, nursing role, hospital unit, years as a healthcare worker, age, gender, race, ethnicity, primary language, and highest level of completed education. All characteristics were collected using self-report including confidence in dementia which was measured with the Confidence in Dementia (CODE) scale and knowledge of dementia which was measured with the Knowledge in Dementia (KIDE) scale. The CODE scale has nine items scored on a five-point Likert scale in which 1=not confident and 5=very confident (Elvish et al., 2014; Elvish et al., 2018). The items are summed for a total score of 9–45 in which 9=not confident and 45=extremely confident in providing care to persons living with dementia. The KIDE scale has 16 statements that to which the rater responds ‘agree’ or ‘disagree’. Endorsed items (agree responses) receive a score of 1, and all items are summed for a total score of 0–16, wherein 0=no items endorsed and 16=all items endorsed (Elvish et al., 2014; Elvish et al., 2018).

Analysis

The use of elderspeak was calculated in two ways: (1) as the proportion of time spent using elderspeak relative to all communication states (i.e., time of elderspeak ÷ observation time); and (2) as the proportion of time spent using elderspeak relative to all nursing staff speech directed towards the patient (i.e., time of elderspeak ÷ [time of elderspeak + time of neutral speech]). The duration of elderspeak was calculated at both the observation level (how much elderspeak was communicated to the patient by all nursing staff present in the care encounter) and the individual nursing staff level (how much elderspeak was enacted by each nursing staff member). Descriptive results are presented for each attribute are presented at the observation level to identify the frequency of each attribute occurring in care encounters.

A linear mixed effects model was used to determine what characteristics are associated with elderspeak use by nursing staff during dementia hospital care. The outcome of interest was the proportion of elderspeak in nursing staff speech. Although the proportion of elderspeak is bounded between zero and one, the conditional residuals were found to be roughly normally distributed, indicating that the linear model is adequate. A random intercept for nurse was included to account for the within-subject correlation arising from repeated measurements. The model was fit with the MIXED procedure in SAS 9.4, using the Kenward-Roger procedure to estimate the degrees of freedom. Estimation via maximum likelihood was used for model selection, and the final reported parameter values were estimated using restricted maximum likelihood. Forward selection was used with the Bayesian Information Criterion (BIC) to ensure model fit. The BIC was selected because it balances the trade-off between goodness of fit and model parsimony to aid in selecting a model which best describes the relationships between the independent variables and the outcome of interest. The forward selection procedure evaluates the addition of each potential variable individually, then adds the variable which results in the best fit to the model. This is repeated until the inclusion of any additional variable does not improve model fit. Additional information on model building is provided in the results section below.

Results

Sample

Nursing Staff

The final sample included 101 nursing staff, of which 53 were recorded and observed. A total of 160 nursing staff were approached across the three study units in Phase 1 and Phase 2, of which 36.9% declined participation (Figure S1). The primary reason that nursing staff provided for declining was not being interested in participating in research (49.2%), followed by finding the study procedures too burdensome (32.2%), and concerns for privacy related to the audio recordings and observations (18.6%). Of the 101 nursing staff who consented, slightly over half (52.5%, $\mathrm{n}=53$) were ultimately audio-recorded and observed while caring for a consenting patient with dementia. No demographic differences were identified between the nursing staff who were audio-recorded and the nursing staff that did not have the opportunity to be audio-recorded $(\mathrm{n}=48)$ because they were not assigned to care for a consented patient with dementia during the study period (Table S1). The CONSORT diagram and demographic characteristics of the nursing staff, patients, and observations can be found in the Supplemental materials.

The recorded nursing staff sample $(\mathrm{n}=53)$ consisted of 27 staff nurses (50.9%) and 26 nursing assistants (49.1%). Most of the sample were female (84.9%), White (81.1%), and not Hispanic or Latinx (92.5%). The nursing staff were generally young ($M=29.7$ years, $\mathrm{S}\mathrm{D}=10.9$), had less than five years of healthcare experience (62.3%), and were relatively new to their hospital unit ($M=2.2$ years, $\mathrm{S}\mathrm{D}=3.3$). There were two differences between the staff nurses and nursing assistants. Staff nurses had a significantly higher level of education than the nursing assistants $(\mathrm{p}<.001)$ and staff nurses scored significantly higher $(\mathrm{p}<.001)$ on the KIDE scale averaging a score of 90.0% compared to the average nursing assistant score of 74.3% indicating that the staff nurses have a greater knowledge of dementia. All other characteristics between the staff nurses and nursing assistants were not significantly different at the $\mathrm{p}=.05$ level (Tables S2–S3).

Patients with Dementia

The final sample included 16 patients with dementia. A total of 73 persons living with dementia were admitted to the three hospital units during the study period. Approximately half (49.3%) did not meet inclusion criteria because they did not exhibit rejection of care (58.3%), had less than mild dementia (25.0%), had a translator present indicating lack of English fluency (8.3%), or had an excluding diagnosis (8.3%). Of the 37 eligible patients with dementia, over half (56.8%) did not enroll because contact with their legal representative could not be made before discharge (57.1%), they were not interested in participating in research (38.1%), or they died prior to consent (4.8%). The 16 remaining patients with dementia were subsequently observed in at least one care encounter.

No differences in gender, race or ethnicity, primary dementia diagnosis, or age were identified between the enrolled patients with dementia $(\mathrm{n}=16)$, the eligible patients with dementia who did not enroll $(\mathrm{n}=20)$, and the ineligible patients with dementia $(\mathrm{n}=36)$. A p-value of .051 was computed for gender, reflecting that more males were enrolled in the final sample than in the excluded sample (Table S4).

The final 16 patients with dementia represented a diversity of dementias (Tables S5), including Alzheimer’s disease ($\mathrm{n}=5$, 31.3%), unspecified dementia ($\mathrm{n}=6$, 37.5%), and other dementias ($\mathrm{n}=5$, 31.3%). The participants were mostly staged as having moderately severe dementia ($\mathrm{n}=12$, 75.0%). The sample was fairly split between males ($\mathrm{n}=9$, 56.3%) and females ($\mathrm{n}=7$, 43.8%), and between community-dwellers ($\mathrm{n}=7$, 43.8%) and persons coming from long-term care ($\mathrm{n}=9$, 56.3%). The patients with dementia were admitted for a variety of medical complaints, including general medical conditions ($\mathrm{n}=8$, 49.5%), traumatic falls ($\mathrm{n}=4$, 25.0%), and neurologic impairments ($\mathrm{n}=4$, 25.0%). The final sample was primarily non-Hispanic White ($\mathrm{n}=15$, 93.8%).

Observations

The final sample included 88 observations between the 16 patients with dementia and 53 nursing staff (Table S6). Observations of care encounters were completed from October 2019 through mid-March 2020, when data collection abruptly ceased due to the COVID-19 pandemic. A total of 91 care encounters were observed and audio-recorded; however, three of these encounters were excluded from the final analysis because a non-consenting person was included. The remaining description of the observations focuses on the 88 care encounters that were analyzed.

The 88 care encounters totaled 10 hours 47 minutes and 10 seconds of observation time. Each patient was observed an average of 5.5 times ($\mathrm{S}\mathrm{D}=2.6,\mathrm{R}\mathrm{a}\mathrm{n}\mathrm{g}\mathrm{e}=1-10$) with an average of 3.8 different nursing staff members ($\mathrm{S}\mathrm{D}=1.8,\mathrm{R}\mathrm{a}\mathrm{n}\mathrm{g}\mathrm{e}=1-7$) across all observations, totaling an average of 40 minutes and 26 seconds ($\mathrm{S}\mathrm{D}=29$ minutes and 43 seconds) per patient. Each nursing staff was recorded an average of 2.4 times ($\mathrm{S}\mathrm{D}=1.2,\mathrm{R}\mathrm{a}\mathrm{n}\mathrm{g}\mathrm{e}=1-6$) with an average of 1.1 patients ($\mathrm{S}\mathrm{D}=0.3,\mathrm{R}\mathrm{a}\mathrm{n}\mathrm{g}\mathrm{e}=1-2$), totaling an average of 20 minutes and 47 seconds ($\mathrm{S}\mathrm{D}=16$ minutes and 8 seconds). Most of the observations ($\mathrm{n}=53$, 60.2%) were dyadic in which there was one staff and one patient, followed by triadic in which there were two staff and one patient ($\mathrm{n}=31$, 35.2%), and the remaining were tetradic in which there were three staff and one patient ($\mathrm{n}=4$, 4.6%).

The care encounters mostly occurred in the morning (64.8%) and were focused on ADL care (57.9%) compared to non-ADL care (42.0%). Observations took place from the first to the 13^th day of hospitalization ($M=5.6,\mathrm{S}\mathrm{D}=3.1$). The patients were rated as being in pain in almost half of the observations (49.9%), with an average pain score of 2.6 $(\mathrm{S}\mathrm{D}=2.9)$. Delirium was identified in 11 (12.5%) of the observations. Sedative medication had been given in the previous six hours in nearly one third (30.7%) of the observations.

Use and Characteristics of Elderspeak Communication

Elderspeak was identified in 85 of the 88 all care encounters (96.6%). Table 1 and Figure 1 examine the durations of all communication states in each observation demonstrating that the most time was spent in silence, followed by neutral communication by nursing staff, and then patient speech. On average, elderspeak comprised 11.7% $(\mathrm{S}\mathrm{D}=10.4\mathrm{\%})$ of the observation time and 28.7% $(\mathrm{S}\mathrm{D}=21.0\mathrm{\%})$ of the nursing staff communication. In other words, over one-quarter of the spoken communication by nursing staff directed towards patients with dementia consisted of elderspeak. Per observation, elderspeak ranged from 0.0% of the time to a maximum of 80.5% of the time in nursing staff communication and over half ($\mathrm{n}=51$, 57.9%) of the observations contained nurse communication with 20% or greater elderspeak. Similar patterns were present when examining elderspeak use at the nurse level. On average, individual nursing staff enacted elderspeak in 30.4% $(\mathrm{S}\mathrm{D}=17.7\mathrm{\%})$ of their spoken communication to patients with dementia. This was highly variable, with some nursing staff never enacting elderspeak and other nursing staff using elderspeak in up to 77.6% of their spoken communication to patients with dementia.

Table 1.

Count and Average Percentage of Time of Communication States (N=88)

	Count of Recordings State is Present		Average % of Time Across All Observations (N=88)
	n	%	Mean %	SD %	Min %	Max %
Elderspeak	85	96.6	11.7	10.4	0.0	57.4
Neutral	88	100.0	31.2	17.6	5.1	78.0
Silence	84	95.5	39.6	25.1	0.0	92.1
Staff-to-Staff	41	46.6	5.0	7.9	0.0	45.6
Patient	77	87.5	12.4	13.7	0.0	54.2

Figure 1.

Average Percentage of Time in Each Communication States $(\mathrm{N}=88)$

Each elderspeak major category was represented in most observations (Table 2 and Figure S2). Overall, semantic attributes were present in over two thirds of observations (69.3%). Over one-third of observations contained childish terms or phrases (44.3%), such as “I know, you’re kind of a roly poly today, aren’t ya?”. Collective pronoun substitution was also used in more than a third of the observations (42.0%); for example: “Okay, do we wanna stay on the bed pan a little longer?”. Diminutives were used in over a quarter (29.6%) of observations and included inappropriate terms of endearment like “pumpkin”, “sweetie”, “dear”, “honey”, “bud”, and “buddy”. Short words or phrases was less frequently identified than other semantic attributes. Prosodic alterations were present in nearly three quarters of observations (73.9%).

Table 2.

Frequency of Elderspeak by Attributes by Observation (N=88) with Examples

Attribute	Definition	N	%	Example
Semantics		61	69.3
Childish term/phrase	Expressions commonly used with children or in childhood.	39	44.3	• I’m gonna get your armpit—no stinky armpits today
				• Are your feetsies hot still?
				• I know, you’re kind of like a roly poly today, aren’t ya?
Collective	Plural forms of pronoun substituted for the individual, singular form.	37	42.0	• Well, why are we taking off our clothes?
				• Okay, do we wanna stay on the bed pan a little longer?
				• Oh, that’s not something we wanna do
Diminutive	Terms of endearment or pet names inappropriate to a healthcare provider-patient relationship.	26	29.6	• All right, pumpkin, open up that there mouth.
				• Oh, I’m sorry, sweetie.
				• Be strong, buddy.
Short words/ phrase	Utterances that are extremely brief and feature monosyllabic words as if talking to young child.	10	11.4	• Lay back. Lay back. Just lay back. Lay back. There ya’ go. Yeah, just lay back. Lay back.
				• Hi, Bob. Hi, Bob. Iťs okay, Bob.
				• Just relax. You’re okay. Just relax.
Discourse		83	94.3
Minimizing word/ Mitigating expression	Speech modifications (words, expressions, utterances) that minimize or diminish. Only when the action applies to the patient.	77	87.5	• It’ll be really quick, just a little pinch, iťll be okay.
				• I know, that’s kinda cold, I know.
				• Gonna feel a little poke in your belly, okay?
Laugh at/ belittle patient	Staff laughter at the patient not with the patient or expressions/utterances that belittle/ignore patient.	32	36.4	• Tryin’ to escape, Joe? [Laughs]
				• Donť you get up, though. [Chuckles]
				• Leťs put the . . . put your leash on.
Exaggerated praise	Comments praising patients are overdone, repeated, or false.	26	29.6	• Good job, nice work, awesome—you did good.
				• I’m so proud of you.
				• Repeat of “good job”/“fabulous”/“magnificent” for same task
Tag question	A question is asked while providing the desired answer.	25	28.4	• That was easy, wasn’t it?
				• We’re having a fun day, aren’t we?
				• That wasnť too bad, was it?
Reflective	Phrase that requests a demanding action from the patient on behalf of someone else.	24	27.3	• Can you stay still for me?
				• Let go of that for me.
				• Would you leave that on for me?
Directive/ Imperative	Tone of communication is bossy and indicates that the speaker is dominant and expects the recipient to do what they are told.	8	9.1	• No, you canť.
				• No, Betty, hold still.
				• No, no, no, no, no, no!
Interrupting	Staff interrupts patient’s communication.	1	1.1
Prosodic		65	73.9

Elderspeak attributes at the discourse level were present in nearly all observations (94.3%). Nearly all observations (87.5%) contained minimizing words and mitigating expressions like, “It’ll be really quick, just a little pinch, it’ll be okay,” in which the nursing staff minimized the discomfort of receiving an injection. Nurses laughed at or belittled the patient in over one-third (36.4%) of the observations. A notable example of the patient being laughed at was, “Are you gettin’ cold, Frank? [chuckles]”. An example of a patient being belittled was, “Let’s put the…put your leash on,” when talking about placing oxygen tubing on the patient for ambulation. Exaggerated praise, tag questions, and reflectives were each identified in over a quarter of the observations. A notable example of a tag question included, “That wasn’t too bad, was it?” wherein the nursing staff controlled the patient’s response of determining that the care task was a positive experience. An example of a reflective utterance was, “Would you leave that on for me?” wherein the patient was asked to leave telemetry monitoring on to appease the nursing staff rather than for their own health. Directives and interrupting were less frequently identified attributes of discourse.

Factors Contributing to the Frequency of Elderspeak Communication

The relationship between elderspeak use and potential explanatory variables were evaluated at the patient with dementia $(\mathrm{n}=16)$, nursing staff $(\mathrm{n}=53)$, or observation $(\mathrm{n}=88)$ level. The outcome of interest, use of elderspeak, was observed for each speaking nursing staff per observation $(\mathrm{n}=126)$ in which there were 53 nursing staff in dyadic observations $(\mathrm{n}=53)$, 31 nursing staff in triadic observations $(\mathrm{n}=61)$, and four tetradic observations $(\mathrm{n}=12)$. One nursing staff did not speak in a triadic observation making the total speaking staff per observation 126.

Clinically relevant variables were evaluated with F-tests to determine whether a significant amount of variation in elderspeak was explained by each potential variable and the variables were selected in the final model based on the p-values of the F-test (Tables S7–S8). Statistically significant variables at the 10% level were evaluated as fixed effects of in the model. This included the age and role of the nursing staff; the age, gender, and education of the patients with dementia; number of staff present during the observation, number of days caring for the patient with dementia, length of stay, presence of pain, and presence of delirium. Three clinically relevant interactions were tested: staff role * staff age, number of staff * delirium, and number of staff * pain presence. Of these three interactions, staff role * staff age was the only one which was statistically significant at the 10% level. Forward selection was initiated based on this interaction, that is, the interaction was entered first and additional variables were then included using the BIC.

The final model included the staff role * staff age interaction, delirium status of the patient with dementia, and length of staff of the patient with dementia. For interpretability, staff age was centered so the intercept occurred at age 18 (i.e., the minimum staff age in our sample). The significant interaction between staff role and staff age ($\widehat{\beta }=-0.013,95\mathrm{\%}\mathrm{C}\mathrm{I}=-0.023,-0.004,p=.008$) indicates that the effect of age on elderspeak is stronger for staff nurses than nursing assistants (Table 3 and Figure 2). For staff nurses, a one-year increase in age was associated with a 1.3% increase in the proportion of elderspeak used ($\hat{\beta }=0.013,95\mathrm{\%}\mathrm{C}\mathrm{I}=0.005,0.021,p=.002$). By contrast, age differences did not affect elderspeak use in nursing assistants. This is illustrated by the magnitude of the interaction effect, which is the same as the age effect for registered nurses. The main effect of staff role indicates that there is not a significant difference in elderspeak use between nursing assistants and staff nurses at age 18 ($\hat{\beta }=0.111,95\mathrm{\%}\mathrm{C}\mathrm{I}=-0.43,0.265,p=.153$). A significant effect of delirium was also demonstrated, with elderspeak use an average of 12.5% higher when the patient with dementia screened positive for delirium ($\hat{\beta }=0.125,95\mathrm{\%}\mathrm{C}\mathrm{I}=0.016,0.235,p=.025$). Finally, nursing staff used an average of 1.5% more elderspeak for each additional day a patient with dementia was hospitalized ($\hat{\beta }=0.015,95\mathrm{\%}\mathrm{C}\mathrm{I}=0.001,0.035,p=.035$).

Table 3.

Impact of Nursing Staff, Patient with Dementia, and Observational Characteristics on Elderspeak Use by Nursing Staff

Variable	Estimate	SE	DF	t	95% CI	p
Intercept	0.060	0.080	50	0.76	−0.100, 0.220	.453
Role (Ref: Nurse)
Nursing Assistant	0.111	0.076	46	1.45	−0.043, 0.265	.153
Age of staff by year	0.013	0.004	44	3.39	0.005, 0.021	.002
Role*Age of Staff
Nursing Assistant	−0.013	0.005	45	−2.79	−0.023, −0.004	.008
Delirium (Ref: No)
Yes	0.125	0.055	60	2.30	0.016, 0.235	.025
Length of stay in days	0.015	0.007	65	2.15	0.001, 0.035	.035

Figure 2.

Interaction Between Nursing Staff Age and Role for Percent of Elderspeak in Nurse Speech by Patients With and Without Delirium

The variables not evaluated in the final model also represent important findings on the relationship between elderspeak communication and nursing staff, patient, and environmental characteristics. The following variables were found to have no relationship with elderspeak communication: nursing staff characteristics of gender $(p=.937)$, race/ethnicity $(p=.714)$, years of experience as healthcare worker $(p=.899)$, confidence in dementia care $(p=.508)$, and knowledge of dementia $(p=.630)$; patient with dementia characteristics of admitting diagnosis $(p=.235)$, dementia severity $(p=.972)$, and comorbidity severity $(p=.964)$; and observational characteristics of time of day $(p=.400)$ and care activity $(p=.250)$.

Discussion

The Nurse Talk study identified that elderspeak communication was present and pervasive in this sample of acute care nursing staff. On average, over one quarter (28.7%) of all nursing staff speech directed towards patients with dementia constituted elderspeak and nearly all (96.6%) care encounters included some elderspeak. Some notable examples of elderspeak from the Nurse Talk study that sounded nearly identical to infant-directed speech were: “are your feetsies hot still?”; “you’re bein’ such a good girl”; and “can we keep that on?”

Despite elderspeak being present in nearly all the recorded care encounters, not all attributes were equally pervasive. Short words and phrases, directives and imperatives, and interrupting were rarely coded. Minimizing words and mitigating expressions were frequently coded, and because of their high frequency further exploration is needed to determine which patterns of minimizers are typical features of communication as opposed to patronizing communication of elderspeak. After minimizers, childish terms and phrases, collectives, and laughing at or belittling were the next most frequently enacted attributes. Both childish terms and collectives have been frequently measured in elderspeak research; however, there has been little research on nursing staff laughing at and belittling patients which requires further exploration. There has only been one other study that has evaluated the presence of elderspeak in the hospital environment (Schnabel et al., 2020). This recent study in a German hospital coded “likely harmful” and “hybrid features” of elderspeak used by staff nurses with both cognitively intact and cognitively impaired older adults. The likely harmful features included tag questions, diminutives, babyish terms, exaggerated praise, and collectives; and the hybrid features included focused on the syntactic aspects of elderspeak such as mean length of utterance and words per minute. Overall, this study reported higher frequencies of elderspeak attributes than the Nurse Talk study found. Tag questions were identified in 97% of their care encounters compared to 28% in the Nurse Talk study. Similarly, collective pronouns occurred in 70% of their care encounters compared to the current 42%, and babyish terms occurred in 57% of their care encounters compared to the current 44%. The only feature that was identified with greater frequency in the Nurse Talk study was diminutives, which occurred in 30% of the care encounters compared to 16% in the German study. More information is needed on the German study’s coding procedures to determine if these dissimilarities arise from linguistic differences between German and English, cultural differences in attitudes towards older adults, or differences in the operationalization of elderspeak attributes in the two studies.

The proportion of elderspeak in care encounters in US nursing homes has ranged from 36% to 58% (Herman & Williams, 2009; Williams et al., 2017), which is also more than the 29% identified in the Nurse Talk study. Again, it is unclear if this is due to the coding procedures used or to differences in the purposes or culture of communication in nursing homes compared to the hospital setting. For example, communication by nursing staff in hospital settings may focus more on the acute health care needs of the patients, whereas staff in long-term care settings may engage in more social communication. It may also be relevant that nursing staff members are more familiar with their patients in long-term care settings since, in the current study, elderspeak increased with length of stay. Future research should take the current coding scheme and procedures and apply them to the videos of nursing home care encounters for a direct comparison of the rates and patterns of elderspeak communication in the nursing home versus the hospital setting.

Role and age of nursing staff, presence of delirium, and length of stay were demonstrated to be important to elderspeak use by nursing staff in the Nurse Talk study. The interaction between nursing staff role and age is a novel finding. To better understand this interaction, the following example on differing elderspeak use between nursing assistants and staff nurses was quantified at two ages. For 25-year-olds, there is not a significant difference in elderspeak use between nursing assistants and staff nurses ($\hat{\beta }=0.019,95\mathrm{\%}\mathrm{C}\mathrm{I}=-0.090,0.128,\mathrm{p}=0.730$). However, for 50-year-olds, nursing assistants use elderspeak in 31.1% less of their communication, on average ($\hat{\beta }=-0.311,95\mathrm{\%}\mathrm{C}\mathrm{I}=-0.517,-0.104,\mathrm{p}=.004$). Prior research conducted in Germany concluded that middle-aged caregivers used the most elderspeak in nursing homes (Sachweh, 1998) and more recently that age of nursing staff did not impact the amount of controlling communication used in hospital care (Schnabel et al., 2019). The author of the prior study suggested that elderspeak patterns may reflect the type of speech these caregivers have used with their own children, assuming that younger caregivers were less likely to have children and therefore less likely to use elderspeak. Our results do not necessarily support this hypothesis, as it is unclear why elderspeak communication would increase with age in nurses but not nursing assistants. We speculate that these findings reflect the recent emphasis on person-centered communication that is increasingly integrated into health-care curricula such as those in nursing schools. The contradictory findings on the impact of staff nurse age on elderspeak use in German versus American hospitals warrants further investigation.

Increased length of stay was also identified as a contributing factor to increasing elderspeak use in hospital dementia care. Our model estimates that, after one week in the hospital, a patient with dementia would receive 10.5% more elderspeak from nursing staff than at admission, after controlling for staff role, age and delirium status of the patient with dementia. Although not in the final model, a trend for a similar relationship was seen based on familiarity with nursing staff. Elderspeak communication increased based on the number of days the staff member directly cared for the patient with dementia $(p=.083)$. The combination of these findings indicates that familiarity with both the environment and the care provider may contribute to elderspeak use in which increasing familiarity leads to more elderspeak.

More elderspeak was used with patients who had screened positive for delirium. Identifying and managing delirium in hospitalized patients with dementia continues to be an important priority in hospital dementia care because patients with dementia are at double the risk of developing delirium compared to cognitively intact older adults, and persons living with dementia with delirium are at increased risk of negative outcomes like mortality (Hshieh et al., 2020). The findings in the Nurse Talk study suggest that nursing staff are communicating differently to patients with delirium than those without which has significant ramifications in trying to reduce resistiveness to care in patients with delirium.

Non-significant findings from the Nurse Talk study are also important findings for understanding what factors elicit elderspeak communication. Neither dementia severity or comorbidity severity of the person with dementia contributed to more elderspeak use despite prior research indicating the elderspeak is more likely with frailer (Schnabel et al., 2020) and cognitively impaired older adults (Shaw & Gordon, 2021). What is unique about the Nurse Talk study is that all participants had at least mild dementia and were actively receiving hospital care (i.e., in a frail state) indicating that at a certain level of cognitive impairment, elderspeak may not be more likely to be used by nursing staff, or conversely that severe conditions like delirium may overpower any impact of cognitive impairment or frailty. Consistent with prior research, the Nurse Talk study found that elderspeak communication is used equally by both male- and female-identifying nursing staff (Schnabel et al., 2020; Shaw & Gordon, 2021).

The high rates of elderspeak identified in the Nurse Talk study provide evidence that education on person-centered communication is needed in hospital dementia care. However, developing interventions on elderspeak communication is particularly complex because elderspeak is typically used by well-intentioned nursing staff attempting to convey care to older adults rather than the patronization it really conveys (Grimme et al., 2015; Hummert & Shaner, 1994; Lombardi et al., 2014). This may be one reason why not all educational interventions on elderspeak reduction have exhibited positive results. For example, an intervention targeted toward occupational therapists in the hospital setting exacerbated ageist views of older adults (Alden & Toth-Cohen, 2015). Higher self-ratings of psychogeriatric knowledge by nursing staff have also been associated with more controlling elderspeak-like emotional tone when speaking to hospitalized older adults (Schnabel et al., 2019). By contrast, in the Nurse Talk study, elderspeak communication was not impacted by the nursing staff members’ confidence in dementia care (CODE) or knowledge of dementia (KIDE). These findings indicate that the relationship between education or knowledge and the use of elderspeak communication is complex.

An evidence-based approach to education about elderspeak is needed to change attitudes and improve care. One such educational intervention exists. The CHAnging Talk (CHAT) intervention has demonstrated reductions in both elderspeak communication by nursing home staff and the consequent resistiveness to care and antipsychotic use by nursing home residents with dementia (Shaw et al., 2018; Williams et al., 2017). The Nurse Talk study provides evidence of the need to adapt and disseminate the CHAT intervention to the hospital setting.

Limitations

The current sample of 16 persons living with dementia was small and limited by a lack of diversity. The current sample was mostly men (56.3%), whereas national estimates indicate that over 60% of hospitalized persons living with dementia are female (Anderson et al., 2020). This may have been because our sample was limited to patients who had demonstrated resistiveness to care which is a behavior more likely to occur in males with dementia (Ishii et al., 2010). This exclusion criteria may have also led to selection bias by limiting the study to patients with rejection of care which may therefore alter the occurrence of elderspeak. The one hospital participating in this study does have the Nurses Improving Care for Health Elders (NIHCE) certification. Although individual nursing staff training through NICHE was not collected, nursing staff knowledge of dementia was collected with the KIDE and there were no differences in elderspeak use based on KIDE score.

The current sample of persons living with dementia also varied from national trends on race and ethnicity (Anderson et al., 2020). Nationally, an estimated 72.4% of hospitalized patients are non-Hispanic White. The current sample only included only one participant who did not identify as non-Hispanic White. The study site did not reflect the national diversity during the study period with 86.3% of patients with dementia being non-Hispanic White in the study’s three participating units. However, the average age of the Nurse Talk sample (81.6 years, $\mathrm{S}\mathrm{D}=6.6$) was which was consistent with the national average for hospitalized patients with dementia (82.6 years, $\mathrm{S}\mathrm{D}=6.7$) (Anderson et al., 2020).

The sample of nursing staff closely aligned with national trends of nurse characteristics (Smiley et al., 2018). The current sample had a similar makeup of gender (90.9% female nationally, 84.9% in the current study), race (80.8% White nationally, 81.1% in the current study), and ethnicity (94.7% non-Hispanic/Latinx nationally, 92.5% in the current study). There are no national estimates for the CODE and KIDE scores; however, the scales have been used with nursing staff in UK and German hospitals. In these studies, hospital staff reported similar estimates of confidence in dementia care (CODE), ranging from 31.4 $(\mathrm{S}\mathrm{D}=4.6)$ to 35.3 $(\mathrm{S}\mathrm{D}=7.6)$, compared to our sample score of 33.9 $(\mathrm{S}\mathrm{D}=4.7)$ (Gehr et al., 2021; O’Brien et al., 2018; Parveen et al., 2021). Similarly, estimates of knowledge in dementia (KIDE) in hospital staff has ranged from 12.0 $(\mathrm{S}\mathrm{D}=2.2)$ to 13.8 $(\mathrm{S}\mathrm{D}=1.9)$ compared to our sample of 13.4 $(\mathrm{S}\mathrm{D}=2.1)$ (Parveen et al., 2021; Schneider et al., 2020).

A further limitation of this study is that the presence of the researcher during observations may have altered the naturally occurring communication patterns between the nursing staff and the patients with dementia. The researcher attempted to alleviate such a Hawthorne effect by spending time on the study units and becoming familiar with the nursing staff, and by ensuring that nursing staff were blind to the study’s focus on elderspeak. Additionally, non-verbal attributes of elderspeak were unable to be explored in the Nurse Talk study and have remained relatively unexplored in elderspeak research. However, non-verbal communication remains an important component to understanding patronization or infantilization during intergenerational encounters. Future research should identify non-verbal behaviors that coincide with verbal elderspeak attributes using video recordings of dyadic care encounters.

Conclusion

Much of the previous research on elderspeak with older adults has focused on general intergenerational encounters with either experimental or survey designs (Shaw & Gordon, 2021). The Nurse Talk study collected recordings of actual care encounters between nursing staff and hospitalized patients with dementia to understand actual patterns of communication in hospital dementia care. Elderspeak was present and pervasive in this sample with over a quarter of speech directed to patients with dementia consisting of elderspeak. Particularly common attributes of elderspeak were minimizing words and mitigating expressions, childish terms and phrases, and collective pronoun substitution. Elderspeak communication was more common with staff nurses than nursing assistants as their age increased and with patients with dementia diagnosed with delirium than those without delirium.

The Nurse Talk study also confirms previous finding that elderspeak is highly individualized with some nursing staff rarely enacting its attributes and others doing so repeatedly (Herman & Williams, 2009; Williams et al., 2017). These findings thus support the need for educational interventions targeted towards older staff nurses and nursing staff from hospital units that care for patients with delirium and longer lengths of stay. Future research will adapt and evaluate the CHAT intervention for implementation in acute care with the goal of increasing person-centered communication by nursing staff to hospitalized patients with dementia.

Contribution of the Paper.

What is already known:

• Elderspeak communication sounds like baby talk to older adults and arises from implicit ageism.

• Elderspeak is often used by formal caregivers during health care encounters in nursing home settings and can lead to rejection of care by residents with dementia.

• The frequency of elderspeak use by nursing staff in hospital dementia care and responses of hospitalized dementia patients to elderspeak have not been explored in the United States.

What this paper adds:

• Elderspeak communication was identified in 96.6% of hospital care encounters between nursing staff and patients with dementia.

• The frequency of elderspeak is affected by characteristics of the nursing staff (role, age), the status of the patient (delirium), and the environment (length of stay).

• Nursing staff gender, confidence, knowledge about dementia care and patient comorbidities and type and severity of dementia were not associated with the frequency of elderspeak communication.