CoChat

A Protocol-Driven, Bedside Digital Conversational Agent to Support Nurse Teams and Mitigate Risks of Hospitalization in Older Adults: Case Control Pre-Post Study

Open access

Published version of a preprint

The DOI matches its Crossref record; 43 of the 43 references it cites were found in a registry or an index; 22 of the 52 sources its claims cite were checked for support.

Abstract

BACKGROUND: Hospitalized older adults often experience isolation and disorientation while receiving care, placing them at risk for many inpatient complications, including loneliness, depression, delirium, and falls. Embodied conversational agents (ECAs) are technological entities that can interact with people through spoken conversation. Some ECAs are also relational agents, which build and maintain socioemotional relationships with people across multiple interactions. This study utilized a novel form of relational ECA, provided by Care Coach (care.coach, inc): an animated animal avatar on a tablet device, monitored and controlled by live health advocates. The ECA implemented algorithm-based clinical protocols for hospitalized older adults, such as reorienting patients to mitigate delirium risk, eliciting toileting needs to prevent falls, and engaging patients in social interaction to facilitate social engagement. Previous pilot studies of the Care Coach avatar have demonstrated the ECA's usability and efficacy in home-dwelling older adults. Further study among hospitalized older adults in a larger experimental trial is needed to demonstrate its effectiveness. OBJECTIVE: The aim of the study was to examine the effect of a human-in-the-loop, protocol-driven relational ECA on loneliness, depression, delirium, and falls among diverse hospitalized older adults. METHODS: This was a clinical trial of 95 adults over the age of 65 years, hospitalized at an inner-city community hospital. Intervention participants received an avatar for the duration of their hospital stay; participants on a control unit received a daily 15-min visit from a nursing student. Measures of loneliness (3-item University of California, Los Angeles Loneliness Scale), depression (15-item Geriatric Depression Scale), and delirium (confusion assessment method) were administered upon study enrollment and before discharge. RESULTS: Participants who received the avatar during hospitalization had lower frequency of delirium at discharge (P<.001), reported fewer symptoms of loneliness (P=.01), and experienced fewer falls than control participants. There were no significant differences in self-reported depressive symptoms. CONCLUSIONS: The study findings validate the use of human-in-the-loop, relational ECAs among diverse hospitalized older adults.

Overview

CoChat Summary

## TL;DR

This case-control quasi-experimental pre-post study tested a human-in-the-loop, protocol-driven tablet avatar (Care Coach) in 95 hospitalized adults over 65 and found lower delirium and loneliness at discharge versus a control unit. The intervention did not improve depressive symptoms, and the paper reports fewer falls on the avatar units.

## Key findings

  • In the intervention group, delirium presence decreased from enrollment to discharge, with McNemar tests showing P<.001; the control group showed no change (P=.25).
  • Participants with avatars had a decrease in loneliness on the UCLA Loneliness Scale compared with controls (P=.01).
  • There was no statistically significant difference in depressive symptoms between participants with and without avatars.
  • Falls data were reported as unit-level quarterly rates: falls were reduced by 33% on one avatar unit with partial data, reduced by 82% on the avatar unit with complete data, and increased by 86% on the control unit.

## Methods

  • Design: Case control quasi-experimental pre-post clinical study in 3 medical-surgical units of a 600+ bed community hospital in New York City; 1 control unit and 2 intervention units.
  • Sample/data: 95 adults over age 65 (intervention n=41, control n=54); inclusion required consent or proxy consent, and patients who did not speak English were excluded.
  • Analysis: Descriptive tests used Student t tests and chi-square; McNemar tests assessed within-group delirium change; ANCOVA/general linear models compared loneliness and depression between groups; falls were examined descriptively using quarterly unit-level quality improvement data.

## Limitations

  • The paper states the study was limited by convenience sampling and possible referral bias by nurse managers.
  • The authors note they could not control for differences in unit care model and participant exposure to care across units.
  • Only English-speaking patients were enrolled, which excluded many potential participants in a diverse setting.
  • Falls data were only available quarterly by unit, and slower recruitment on the second intervention unit led to partial falls data; additionally, individual-level comorbidity data were unavailable, limiting analysis of patient-specific effects.

Model-written summary. Check it against the paper before relying on it.

Integrity signals

Registry

Crossref · 10.2196/13440

Cites

43 of 43 references verified0 citation errors · 0 not checkable

Statistics

Nothing to recompute.

Figures

Figure 1 The Care Coach avatar system design (top). An avatar encourages a patient to take his medications (bottom).
Figure 1 The Care Coach avatar system design (top). An avatar encourages a patient to take his medications (bottom).CC BY · PubMed Central
Figure 2 An example of a simple fall prevention protocol.
Figure 2 An example of a simple fall prevention protocol.CC BY · PubMed Central

Verification

Claims the paper makes

15 Valid · 1 Not valid · 16 Not checked · 38 Own results and interpretations

22 more claims

Create a free account to see all 70 claims and every source verdict.

See all claims — free account
  • Probably not a substantial claim. 31
  • It may state two separate claims as one. 22

Create a free account to see these claims.

Claims other papers make about this one

Each is checked against this paper’s own text: does it say what the citing paper says it does?

Citations

Cites

Research works printed but not linked: 3

  1. Adams, J, 2013, Nurs Econ, 31, 83
  2. Demiris, G, 2016, AMIA Annu Symp Proc, 2016, 496
  3. Singh, I, 2016, Gerontol Geriatr Res, 2, 1015

Cited by

See all 65

Metadata

Authors
Bott, N., Wexler, S., Drury, L., Pollak, C., Wang, V., Scher, K., & Narducci, S.
Affiliations
  • Stanford Medicine (US)
  • Pace University (US)
  • Jamaica Hospital (US)
Corresponding author
Bott, Nicholas
Venue
Journal of Medical Internet Research
Published
2019-10-17
Volume
21
Issue
10
Pages
e13440
ISSN
1438-8871 (ISSN-L)
OpenAlex
W2969720013
Cited by
61
References
43
Citations by year
2026: 7 · 2025: 7 · 2024: 15 · 2023: 9 · 2022: 12 · 2021: 9 · 2020: 2
FWCI
7.22
Type
journal article
Language
en
Open access
Licence
cc-by
Locations
  • Journal of Medical Internet Research (publishedVersion, cc-by)
  • Europe PMC (PubMed Central) (publishedVersion, cc-by)
  • PubMed Central (publishedVersion, cc-by)
  • PubMed (publishedVersion)
Funders
  • New York City Economic Development Corporation
  • National Institute of Nursing Research (R44NR017842)
Topics
  • Social Sciences › Psychology › Social Psychology › Social Robot Interaction and HRI
  • Health Sciences › Medicine › Geriatrics and Gerontology › Technology Use by Older Adults
  • Social Sciences › Business, Management and Accounting › Marketing › AI in Service Interactions
Keywords
embodied conversational agents, relational agents, hospitalized older adults, loneliness, depression, delirium, falls, human-in-the-loop, clinical protocols, nursing team, social isolation, social engagement, human-robot interaction
MeSH terms
Aged, Aged, 80 and over, Communication, Female, Hospitalization, Humans, Interpersonal Relations, Male, Nurses, Pilot Projects, Case-Control Studies, Telemedicine

About these checks

Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.

The claim verdicts this report counts are experimental: their accuracy is not yet measured.

A language model drew the claims from the paper on 2026-09-29, under version 3 of the claim definition (model deepseek-v4.1-flash, prompt extract-v3).

A second model labelled each claim on 2026-09-29 (model jev-1.13-20260917, definition revision gold-guideline-v0.2).

Cited sources are checked with a language model’s help (gpt-6-sol).

Statistics and arithmetic are recomputed, not judged.

Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.

What counts as a claim ›How we measure confidence ›How we check this ›

Are you an author, or do you see a problem with this page? Request a review

Data sources

Bibliographic data: Semantic Scholar, OpenAlex, Crossref.

Contains information from the Semantic Scholar Academic Graph Datasets, made available under the ODC Attribution License (ODC-BY 1.0).

Semantic Scholar’s data is described in: Kinney, R., et al. (2023). The Semantic Scholar Open Data Platform. arXiv:2301.10140.