Heart & Lung 44 (2015) 408e415

Contents lists available at ScienceDirect

Heart & Lung

journal homepage: www.heartandlung.org

Care of Critically Ill Adults and Children

Quality of care and resource use among mechanically ventilated

patients before and after an intervention to assist nurse-nonvocal
patient communication
Mary Beth Happ, PhD, RN, FAAN a, b, c, *, Susan M. Sereika, PhD d, Martin P. Houze, MS d,
Jennifer B. Seaman, PhD, RN b, c, Judith A. Tate, PhD, RN a, Marci L. Nilsen, PhD, RN c,
Jennifer van Panhuis, BSN, RN e, Andrea Scuilli, MS g, Brooke Paull, MS, SLP-CCC g,
Elisabeth George, PhD, RN, CCRN g, Derek C. Angus, MD, MPH b, e,
Amber E. Barnato, MD, MPH, MS e, f
a

The Ohio State University College of Nursing, Columbus, OH, USA

The CRISMA Laboratory (Clinical Research, Investigation, and Systems Modeling of Acute Illness), Department of Critical Care Medicine, University of
Pittsburgh, Pittsburgh, PA, USA
c
Department of Acute and Tertiary Care, University of Pittsburgh School of Nursing, Pittsburgh, PA, USA
d
Department of Health and Community Systems, University of Pittsburgh School of Nursing, Pittsburgh, PA, USA
e
Department of Medicine, University of Pittsburgh School of Medicine, Pittsburgh, PA, USA
f
Department of Health Policy and Management, University of Pittsburgh Graduate School of Public Health, Pittsburgh, PA, USA
g
University of Pittsburgh Medical Center, Pittsburgh, PA, USA
b

a r t i c l e i n f o

a b s t r a c t

Article history:
Received 25 March 2015
Received in revised form
1 July 2015
Accepted 4 July 2015

Objectives: Implement and test unit-wide patient-nurse assisted communication strategies (SPEACS).
Background: SPEACS improved nurse-patient communication outcomes; effects on patient care quality
and resource use are unknown.
Methods: Prospective, randomized stepped-wedge pragmatic trial of 1440 adults ventilated
2 days and
awake for at least one shift in 6 ICUs at 2 teaching hospitals 2009e2011 with blinded retrospective
medical record abstraction.
Main results: 323/383 (84%) nurses completed training; their communication knowledge (p < .001) and
satisfaction and comfort (p < .001) increased. ICU days with physical restraint use (p .44), heavy
sedation (p .73), pain score documentation (p .97), presence of ICU-acquired pressure ulcers
(p .78), coma-free days (p .76), ventilator-free days (p .83), ICU length of stay (p .77), hospital
length of stay (p .22), and median costs (p .07) did not change.
Conclusions: SPEACS improved ICU nurses knowledge, satisfaction and comfort in communicating with
nonvocal MV patients but did not impact patient care quality or resource use.
2015 Elsevier Inc. All rights reserved.

Keywords:
Augmentative and alternative
communications systems
Intubation
Endotracheal
Nurses
Education
Quality of health care

Abbreviations: ICU: intensive care unit; SPEACS: Study of PatienteNurse Effectiveness with Assisted Communication Strategies; MV: mechanical ventilation; SLP:
speech language pathologist; EMR: electronic medical record; APACHE: Acute Physiology and Chronic Health Evaluation; NCS: Nurse Communication Survey; ITT: intention to treat.
Author contributions: Study concept and design e M.B.H., A.E.B., J.A.T., S.M.S., D.C.
A.; Acquisition of data e J.B.S., A.S., J.V.P., J.A.T.; Statistical analysis e M.P.H., S.M.S.;
Analysis and interpretation of the data e M.B.H., A.E.B., J.A.T., J.B.S., S.M.S., D.C.A., E.
G.; Drafting of the manuscript e A.E.B., M.B.H., S.M.S., J.A.T., M.F.N., M.P.H.; Critical
revisions to the manuscript for important intellectual content e all authors;
Obtained funding e M.B.H, A.E.B., J.A.T., S.M.S, E.G., B.B., D.C.A.; Study supervision e
M.B.H., A.E.B., J.A.T., S.M.S, E.G., B.B., D.C.A.
0147-9563/$ e see front matter 2015 Elsevier Inc. All rights reserved.
http://dx.doi.org/10.1016/j.hrtlng.2015.07.001

Source of support: Robert Wood Johnson Foundation Interdisciplinary Nursing

Quality Research Initiative grant #66633.
Potential conict of interest: The SPEACS-2 program is accessible online at http://
go.osu.edu/speacs2. Dr. Happ holds the Creative Commons copyright.
* Corresponding author. The Ohio State University, 378 Newton Hall, 1585 Neil
Ave, Columbus, OH 43210, USA. Tel.: 614 292 8336.
E-mail address: aeb2@pitt.edu (M.B. Happ).

M.B. Happ et al. / Heart & Lung 44 (2015) 408e415

Intubation for mechanical ventilation (MV) precludes the ability

to speak. Thus, intensive care unit (ICU) patients who are placed on
MV suddenly acquire a profound communication disability, which
can be a source of distress, frustration, anxiety, and agitation.1,2
Communication disability among nonvocal ICU patients reduces
the accuracy of pain and symptom assessment,3,4 predisposes patients to preventable adverse events,5 and may lead to increased
use of immobilizing treatments such as sedatives or physical
restraint.
These negative patient outcomes may be ameliorated by assistive communication strategies, such as tagged yes-no questions, communication boards, hearing ampliers, writing tools.
Yet nurses in the ICU lack training in assistive communication
strategies, readily available communication materials, and access
to communication experts (e.g., speech language pathologists).6
In a prior clinical trial of ICU nurse training in the use of simple assistive communication techniques, communication tools
provision, and speech pathologist support, the Study of Patiente
Nurse Effectiveness with Assisted Communication Strategies
(SPEACS),7 we demonstrated improved communication between
individual nurses and their nonvocal ICU patients.8,9 Secondary
analyses also suggested a positive relationship between
communication process and patient outcomes, such as pain
management and sedation level.10 The training format used in
SPEACS e a 4-h small group workshop e limited the feasibility of
dissemination.
The purpose of the current study was to translate the
multi-component SPEACS program into a disseminable format
(SPEACS-2: web-based ICU nurse training in assistive communication techniques, provision of low tech communication tools,
and expert consultation), then evaluate prospectively whether
unit-wide implementation of the SPEACS-2 could improve nurse
knowledge, satisfaction, and comfort in communicating with
nonvocal mechanically ventilated patients and thereby improve
patient-level quality of care and resource use.
Materials and methods
Design
We received University of Pittsburgh Institutional Review
Board approval for the study, including approval for a waiver of
informed consent for medical record review. We conducted a
randomized crossover cluster (stepped-wedge) quality improvement trial of unit-wide implementation of SPEACS-2 in 6 intensive care units (ICUs) at 2 teaching hospitals between August
2009 and July 2011. Data collectors were blind to the intervention
assignment.
Intervention
The original SPEACS training involved a 4-h course consisting of
interactive lecture with PowerPoint slides and video exemplars,
demonstration, role-play and instructor feedback on performance
of communication strategies.7,8 To facilitate unit-wide, scalable
dissemination, we modied SPEACS into SPEACS-2. The SPEACS-2
communication skills training intervention involved 1 h of online training, including video exemplars of communication techniques, for all bedside nurses (http://go.osu.edu/speacs2; see
Table 1). Experienced and novice clinicians and assistive communication experts pretested the on-line training version of SPEACS
for feasibility, acceptability, and content. The intervention also
included the provision of communication supplies (e.g., communication boards, notebooks, felt-tip pens, clipboards, hearing aid
batteries, etc.), and weekly bedside teaching rounds with a speech

409

Table 1
Components of the SPEACS-2 Intervention.
Intervention components
1. Six 10-min on-line educational modules involving narrated text slides and
video exemplars of communication assessment and techniques (60 min).
2. Reference manual, pocket reference cards, assessment e intervention
algorithm.
3. Communication cart in the ICU containing assistive communication tools and
materials.
4. Communication resource nurses (champions) e minimum of 2 per ICU.
5. Weekly teaching posters communication strategy of the week.
6. Weekly patient case conference with Speech Language Pathologist.

language pathologist (SLP) on the unit for a period of 25 weeks

(2 quarters). Unit poster displays reviewing a different component of the training program each week reinforced learning. We
randomized each ICU to a 3-month intervention period across 6
consecutive quarters (18 months) beginning on November 1, 2009,
February 1, 2010, May 1, 2010, August 1, 2010, November 1, 2010, or
February 1, 2011 (Fig. 1). As soon as we deployed the intervention in
a unit, we gave nurses access to the online training and encouraged
them to complete this self-directed training.
We measured delity to the intervention (delivery, receipt and
enactment) by tracking the number of eligible nurses who
completed the on-line course, attendance at weekly communication rounds conducted by the SLP, change in scores on a 10-item
pre- and post-test knowledge quiz, and use of communication
tools and strategies assessed by communication supply inventory
and randomly scheduled weekly observations for communication
tools in the room, patient communication support in accordance
with the assessment-intervention algorithm, and bedside use of a
written communication plan.
Study ICUs
The six specialty ICUs included in this study were nave to the
SPEACS and SPEACS-2 programs at the study onset. The units
included: NeuroTrauma, Transplant, Medical, Cardiovascular,
Trauma, Neurological. All study ICUs provided 1:2 to 1:1 nurse e
patient ratios and 12-h shift rotations. The Critical Care Medicine
service provided attending physician coverage for all 6 ICUs across
two different hospitals in the academic health system. The SLP role
was limited to dysphagia consultations and swallowing evaluations; consultations for communication support were rare.
Participants
Nurse sample
The eligible nurse sample included all permanent full and
part-time staff nurses assigned to the study unit at the time of
intervention deployment. Unit nurse managers and clinical nurse
specialists provided endorsement of the study, introduction and
access to the unit nursing staff. Managers recommended 2e3
individuals to serve as nurse champions. Nurse champions
received an additional in-person introduction to the program and
review of the communication cart by study staff. Nurse champions to then liaised with the research team and encouraged and
supported unit nurses to complete the training program and use
the communication tools.
Patient sample
We retrospectively identied all potentially eligible control
(pre-intervention) and intervention cases from consecutive

410

M.B. Happ et al. / Heart & Lung 44 (2015) 408e415

Neuro

Trauma

Unit

Medical

Cardiovascular

Transplant

Neuro Trauma

1
Included
n=30
Excluded
n=61
Included
n=30
Excluded
n=30
Included
n=30
Excluded
n=38
Included
n=30
Excluded
n=16
Included
n=30
Excluded
n=47
Included
n=34
Excluded
n=44

2
Included
n=30
Excluded
n=65
Included
n=30
Excluded
n=18
Included
n=30
Excluded
n=36
Included
n=30
Excluded
n=26
Included
n=30
Excluded
n=47
Included
n=32
Excluded
n=32

3
Included
n=30
Excluded
n=81
Included
n=30
Excluded
n=25
Included
n=30
Excluded
n=49
Included
n=30
Excluded
n=22
Included
n=30
Excluded
n=58
Included
n=23
Excluded
n=35

Quarter
4
5
Included
Included
n=30
n=30
Excluded Excluded
n=49
n=32
Included
Included
n=30
n=30
Excluded Excluded
n=19
n=19
Included
Included
n=30
n=30
Excluded Excluded
n=23
n=24
Included
Included
n=30
n=30
Excluded Excluded
n=15
n=15
Included
Included
n=30
n=30
Excluded Excluded
n=28
n=39
Included
Included
n=24
n=26
Excluded Excluded
n=28
n=24

6
Included
n=30
Excluded
n=56
Included
n=30
Excluded
n=20
Included
n=30
Excluded
n=41
Included
n=30
Excluded
n=23
Included
n=30
Excluded
n=43
Included
n=37
Excluded
n=21

7
Included
n=30
Excluded
n=38
Included
n=30
Excluded
n=21
Included
n=30
Excluded
n=30
Included
n=30
Excluded
n=21
Included
n=30
Excluded
n=44
Included
n=30
Excluded
n=36

8
Included
n=30
Excluded
n=51
Included
n=30
Excluded
n=18
Included
n=30
Excluded
n=44
Included
n=30
Excluded
n=19
Included
n=30
Excluded
n=45
Included
n=34
Excluded
n=31

Unshaded pre-intervention quarter; Light shading intervention deployment quarter; Dark shading post-intervention quarter. The total
number of randomly selected patients screened for abstraction eligibility in each quarter = included + excluded; the proportion of patients
meeting abstraction eligibility criteria = included/(included + excluded).
Fig. 1. [Box] Screening and Eligibility by Study Unit and Quarter. The unshaded section of the box represents pre-intervention quarters, the light-shaded section represents the
intervention deployment quarter, and dark shaded section represents post-intervention quarters. The total number of randomly selected patients screened for abstraction eligibility
in each quarter included excluded; the proportion of patients meeting abstraction eligibility criteria included/(included excluded).

admissions to the 6 study ICUs using administrative (e.g., billing)

data. Potentially eligible admissions met three criteria: (1) age 18
years and older; (2) rst ICU admission during the hospital stay;
and (3) MV of at least 2 days duration. From this list of potentially
eligible admissions, we randomly screened electronic medical record (EMR) charts to assess eligibility until we reached 30 patients
per unit per quarter (3 month period) across 8 quarters, yielding
the pre-specied sample of 240 patients per unit (total n 1440;
Fig. 1).
Denitively eligible admissions were those conrmed by EMR
review to involve: (1) the rst ICU admission during the hospital
stay; (2) invasive MV via endotracheal tube or tracheostomy for 2
or more calendar days (e.g., non-invasive MV or invasive MV
for < 2 days excluded); and (3) being awake, alert, and responsive to verbal communication from clinicians. We operationalized
awake, alert and responsive as being awake for at least one 12h nursing shift while receiving MV dened as responding to and/
or following commands; nursing note description as alert,
arousable, anxious, or awake; or a score of 6 (obeys verbal
commands) for the Best Motor Response on the Glasgow Coma
Scale,12 and/or a score of 1e3 on the Modied Ramsay Scale,11
Riker13 Sedation Agitation Score
3 or responsive to verbal
communication from clinicians via head nods, gestures, or other
nonvocal method. Only patients meeting these basic communication criteria could be served by the communication skills and
tools included in the intervention. We required patients to be
intubated, ventilated for 2 or more days to exclude brief intubations in which patients were extubated shortly after awakening and to allow for the communication intervention to effect

quality of care outcomes. Full details of our subject screening

results are published elsewhere.14

Evaluation
Nurse outcomes
We assessed several process measures to evaluate the delity
of the intervention.15 Specically, we assessed training completion (intervention delivery) by dates of unit nurses completion
of the web-based training and post-test, with an a priori target
of 85% completion during the quarter of intervention deployment. We assessed nurse knowledge acquisition (intervention
receipt) by comparing nurses performance on a 10-item preand post-training test. Expert clinicians and nurse educators
vetted the tests after viewing all six training modules. We
recorded communication supply use during weekly restocking,
attendance at SLP bedside teaching rounds, and adherence
to training principles (intervention enactment) via bedside
observation.
We also assessed nurse satisfaction and comfort with communication with nonvocal patients using a 16-item Nurse Communication Survey (NCS) administered immediately before and 3
months after nurse training.16 This was adapted from a previously
validated 12-item NCS with 4 novel items drawn from our focus
group analysis from the original SPEACS study.9 The revised 16item NCS survey revealed 6 factors or subscales explaining 66% of
the item variance. We conned the outcome analysis to subscales
with internal consistency above a .70.

M.B. Happ et al. / Heart & Lung 44 (2015) 408e415

Patient outcomes
We selected quality of care measures proximately related to the
hypothesized mechanism of action of the communication skills
training, informed by video-analysis8 and focus group interviews
with participants in the original SPEACS trial.9 Specically, we
hypothesized that improved communication with nonvocal but
awake mechanically-ventilated ICU patients would: 1) increase
effectiveness of patient-nurse communication and thereby increase the frequency of pain score assessment and documentation,
and 2) decrease the frequency of patient frustration and agitation
and thereby decrease physical restraint and/or heavy sedation to
prevent device disruption and resultant ICU-acquired pressure
ulcers. We based these hypotheses on research linking communication difculty to feelings of anxiety and agitation during
MV,1,2,17 survey research indicating that patients perceived that
their frustration with communication would have been signicantly lower if communication tools had been offered during
MV,18 and our previous nding that seriously ill mechanically
ventilated patients communicated more often during periods of
non-restraint.19
We operationalized and measured these outcomes in the EMR
as follows: physical restraint (proportion of mechanically ventilated ICU days with one or more upper extremity restraint at any
point during the 24 h interval), heavy sedation (proportion of
mechanically ventilated ICU days in which the patient did not meet
awake criteria for 8 out of 12 h AM or PM as dened by Modied
Ramsay score 1e3, Glasgow Coma motor score of 6, nursing note
documentation of being alert, awake, arousable, responsive, or
communicative), coma-free days (number of days out of 28 in
which patients were assessed as awake or not in heavy sedation
state for at least 8/12 h of both AM and PM nursing shifts),20 pain
documentation (proportion of mechanically ventilated ICU days
with pain score documented, given documented presence of any
pain), any ICU-acquired pressure ulcer grade II or greater, and unplanned endotracheal or tracheal tube extubation (see Online
Supplement for greater detail).
Secondary patient outcomes included resource use. We
hypothesized that reductions in restraint and heavy sedation would
reduce length of MV, ICU and hospital stay, and costs. Specically,
we measured ventilator-free days,21 ICU and hospital length of stay,
and cost-adjusted charges.
Trained staff, blinded to intervention assignment, abstracted
clinical data (physical restraint, heavy sedation, coma-free days,
pain documentation, ICU-acquired pressure ulcers, unplanned
extubation, ventilator days) from the electronic medical record
(PowerChart, Cerner Corporation, Kansas City, MO) during the
incident ICU admission, for up to 28 days, using a standardized
abstraction instrument. From among the 1440 abstracted cases, 108
were randomly selected for co-abstraction by a single rater (10% for
quarters 1e4; 5% for quarters 5e8) with inter-rater reliability of

0.8e1.0 (near-perfect agreement)22 for all measures except the
presence of pressure ulcers (k 0.69). We report details of the
development and validation of the abstraction instrument elsewhere.23 We collected administrative data (ICU length of stay,
hospital length of stay, hospital charges) from billing records
(Medical Archival System, Pittsburgh, PA).
Statistical analyses
The study was appropriately powered and an adequate patient
sample was obtained to detect a small to medium effect size.
Specically, 30 patients per unit per quarter provided at least 80%
power to detect unit by quarter interactions as small as f 0.136
[small to medium effect size based on Cohen (1980)] using a
two-sided F-test from a two-way general linear model at a

411

signicance level of .05, adjusting for clustering effect by unit. To

examine nurse outcomes at post-intervention, we calculated that a
sample size averaging 50 nurses per unit (300 total), allowing for
nonparticipation and attrition, achieves provided about 80% power
to detect a medium-sized effect of d 0.523 between the group
means assuming an intra-unit correlation of at most .010 using a
two-sided Wald test (or more conservatively t-test) at a signicance
level of .05. Although an adequate sample of nurses participated
overall, we did not achieve 50 nurses per unit with data at both pre
and post-intervention periods.
We calculated the proportion of regular staff unit nurses in each
study ICU who completed the pre-test, training, and post-test
during the quarter of intervention deployment. In some units,
most of the nurses completed the training expeditiously, and so
more of the days during the intervention period were staffed by
nurses who had completed the modules already. In other units,
nurses delayed completion, and therefore, based upon the date of
each nurses post-test, we calculated the proportion of trained
nurse days among nurses who completed SPEACS-2 training by
unit to estimate the median exposure during the quarter of intervention deployment. This number could theoretically range from
0 if all the nurses who completed SPEACS-2 training completed the
post-test on the last day of the quarter (akin to the pre-intervention
control period) to 1 if all the trained nurses completed the post-test
on the rst day of the quarter (akin to the post-intervention period).
We used a paired sample Wilcoxon rank sum test of the difference
between pre- and post-training knowledge and attitude scores
of nurses. We used a Wilcoxon rank-sign test to compare nurses
pre- and post-training knowledge and attitude scores for the total
sample.
We conducted analyses of intervention effect on patient outcomes using intention to treat (ITT). That is, we considered patients
admitted to a study ICU in the intervention and post-intervention
period ICU exposed to SPEACS-2, even if the patients nurse
failed to complete the training or completed the training but did
not deploy any of the skills or tools from SPEACS-2 in patient care.
We summarized demographic and clinical characteristics
(age, sex, race, ethnicity, ICU admission diagnosis, severity of illness
upon ICU admission using the Acute Physiology and Chronic Health
Evaluation III score to measure illness severity24 calculated
from clinical variables, and pre-hospital functional status) of the
randomly selected and abstracted patient sample and compared
the characteristics of those admitted during pre-intervention and
intervention quarters using t-test, chi-square, and Fishers exact
tests as appropriate.
We used mixed-effects linear regression to estimate and test the
effect of the intervention on normally-distributed outcomes
(physical restraint, heavy sedation, pain score documentation, and
ventilator-free days) and mixed-effects logistic regression to assess
effect on binary outcomes (ICU-acquired pressure ulcer). Mixedeffects linear regression was employed to assess the effect on
continuous outcomes (ICU length of stay, hospital length of stay,
and hospital costs) after Winsorizing (e.g., top-coding) outliers
at the 90th percentile. We treated intervention assignment and
quarter as xed effects and unit as a random effect, with and
without adjustment for covariates selected a priori based upon
their inuence on the outcome of interest (age, sex, APACHE score).
We treated the intervention assignment during the quarter of
intervention deployment as a fraction between 0 and 1 based upon
the median trained nurse days during the quarter.25
In sensitivity analyses, we explored alternate specications used
in the stepped-wedge literature, including treating the step
involving intervention deployment as control (0) or intervention
(1).26 Finally, we conducted post-hoc analyses stratied by change
in knowledge score after the training intervention.

412

M.B. Happ et al. / Heart & Lung 44 (2015) 408e415

neurological ICU) accrued more time in the post-intervention

deployment intervention period and therefore contribute more of
their patients with a neurologic admission diagnosis (Table 2).

Results
Nurse characteristics
A total of 323 of 383 eligible nurses completed the on-line
course. These study nurses were predominantly women (77%),
43% were bachelors-prepared, 5.5% were masters-prepared, and
17% were critical care certied.
Patient characteristics
From a sample of 5476 patients identied as potentially eligible
using administrative claims, we screened 3087 patient charts to
identify 1440 eligible patients for full chart abstraction (Fig. 1).
There were no statistically signicant differences in demographic
or clinical characteristics age, sex, race, APACHE III, or pre-hospital
functional status among abstracted patients admitted during
control and intervention quarters (Table 2). The distribution of
primary admission diagnoses classied by organ system differed
signicantly among control and intervention quarters due to the
order of crossover of subspecialty study ICUs. Subspecialty ICUs
that were randomized to cross-over earlier in the study period (e.g.,

Table 2
Demographic and clinical characteristics of abstracted eligible patients, by intervention period.
Pre-intervention
(N 626)
Age, mean (SD), y
59.97 (17.65)
Male sex, n (%)
326 (52.1)
b
Race (N 1435), n (%)
White
560 (89.7)
Black
57 (9.1)
Asian
2 (0.3)
Other
5 (0.8)
APACHE III score, mean (SD)
67.40 (27.8)
Primary admission diagnosis
classied by organ system, n (%)
Cardiovasculard
93 (14.9)
Respiratory
153 (24.4)
Neurological
54 (8.6)
Gastrointestinal
42 (6.7)
Trauma
105 (16.8)
Renal or liver
19 (3.0)
Transplant
23 (3.7)
Sepsis
31 (5.0)
Post-surgical
101 (16.1)
Othere
5 (0.8)
Pre-hospital functional dependence, n (%)b
Eating (N 1036)
25 (4.0)
Grooming (N 1033)
25 (4.0)
Bathing (N 1037)
30 (4.8)
Dressing (N 1031)
27 (4.3)
Transfers (N 1022)
34 (5.4)
Toileting (N 1221)
33 (5.3)

Intervention
groupa
(N 814)

p-value

62.06 (16.26)
428 (52.6)

0.07
0.85
0.33c

731
75
4
1
64.87

(90.1)
(9.2)
(0.5)
(0.1)
(27.41)

141
177
181
41
82
6
5
31
137
13

(17.3)
(21.7)
(22.2)
(5.0)
(10.1)
(0.7)
(0.6)
(3.8)
(16.8)
(1.6)

20
26
30
29
28
42

(2.5)
(3.2)
(3.7)
(3.6)
(3.4)
(5.2)

Intervention delity (delivery, receipt, and enactment)

Overall, 323/383 (84%) nurses completed the training during
their units intervention deployment quarter (range: 77e93%) with
a median fraction of trained nurse days during the deployment
quarter of 46% (range: 11e65%) (Table 3). Among the 273/323 (85%)
nurses who completed both the SPEACS-2 course pre- and posttests, average test scores increased from 49% to 55% correct
(p < .001).
We restocked an average of 10.8 items per unit each week,
equivalent to 0.5 items/ICU bed (64.8 items/124 ICU beds) during
communication cart inventory. A majority (94%) of scheduled
weekly SLP rounds were completed during a 25-week period (two
quarters) for each ICU with a total of 24e40 (mean 30.3; 156/383
(41%)) distinct nurses participating per ICU. We conducted 225
intervention delity monitoring observations during the intervention quarter and 257 observations during the quarter immediately following intervention deployment. We did not see any
decrease (i.e., decay) in the adherence to SPEACS-2 program components between the intervention quarter and the quarter immediately following intervention deployment (see Fig. 2).
Nurse comfort and satisfaction with communication
Among the 264/323 (82%) nurses who completed the perceptions of communication survey, the overall mean item score
increased from 3.21 to 3.43 (p < .001). These changes were principally driven by improvements in items related to comfort and
satisfaction with communication.
Quality of care

0.08
<0.01

Overall, 49.2% of mechanically ventilated ICU days involved

upper extremity physical restraint and 34.7% involved heavy
sedation (corresponding to 20.8 coma-free days) and, among those
with pain documented, 26.2% involved documentation of the pain
score. 133/1440 (9.2%) patients developed ICU-acquired pressure
Table 3
Nurse training receipt, knowledge, and attitude scores.
Measure

0.10
0.42
0.30
0.47
0.07
0.93

SD e standard deviation; APACHE III e Acute Physiology Age and Chronic Health
Evaluation (APACHE) III score.
a
Includes 180 patients abstracted from a quarter of intervention deployment. In
our base-case analysis, we treated the unit as being partially exposed to the
intervention during the quarter in which the nurses received training. We assigned
a value between 0 and 1 corresponding to the median proportion of trained nurse
days in the unit. Units in which the nurses completed earlier in the quarter have
numbers closer to 1 than units in which the nurses completed training later in the
quarter (see Table 3).
b
The proportion reported is among patients with non-missing data; we report
the number of subjects with non-missing data in parenthesis.
c
p-value obtained from Fishers exact test.
d
Includes Cardio/thoracic/vascular surgery, cardiomyopathy, myocardial infarction, and arrhythmia.
e
Includes hematology/oncology.

Overall

Eligible nurses, n
383
Receipt
Completed training during quarter,a n (%)
323 (84)
Trained nurse days in quarter,a median proportion (IQR)
0.46 (0.52)
Nurse knowledge about strategies for communication with nonvocal patients
Nurses with both pre- and post-test data, n
273
Pre-test knowledge: % correct,b mean (SD)
49.0 (14.2)
Post-test knowledge: % correct,b mean (SD)
55.4 (14.4)
Change: % correct,b mean (SD)
6.3 (17.4)
p-value
<0.001
Nurse attitudes toward communication with nonvocal patients
Nurses with attitude data, n
264
Pre-training attitude score,c mean (SD)
3.21 (0.42)
Post-training attitude score,c mean (SD)
3.43 (0.42)
Change: mean item response, (SD)
0.22 (0.44)
p-value
<0.001
IQR e interquartile range (75th percentile value minus 25th percentile value); SD e
standard deviation.
a
Quarter during which the intervention was deployed.
b
Knowledge outcomes restricted to the 273/323 (84.5%) trained nurses who
competed both the pre- and post-test knowledge assessment tests.
c
Mean item response for the 16-item survey; each item had a possible range of
1e5 indicating strength of agreement.

Percentage Adherent to Protocol

M.B. Happ et al. / Heart & Lung 44 (2015) 408e415

Intervention Fidelity Monitoring (IFM) during and postintervention quarters

100%
90%
80%
70%
60%
50%
40%
30%
20%
10%
0%

p=.107

p=.122
p=.141

p=.268

Communication wall
cards/careplans

Algorithm use

Communication at all, Communication tools

with or without
in room
algorithm

Bedside Intervention Fidelity Items

IFM Intervention Quarter

IFM Post-Intervention Quarter

Fig. 2. [Bar chart] Intervention Fidelity Monitoring (IFM). Direct observations during
the intervention implementation quarter (N 225) and during the rst postintervention implementation quarter (N 257) demonstrated no decreases in the
adherence to SPEACS-2 program components (written communication care plans, algorithm use, communication at all e with or without algorithm, and assistive and
augmentative communication tools in the patients room) over time.

413

Resource use
Overall, the sample had an average of 19.2 ventilator free days, a
median ICU length of stay of 9.0 days, a median hospital length of
stay of 15.0 days, and a median cost of $48,774. There were no
statistically signicant differences between intervention and control quarters in ventilator-free days (.15, p .83), ICU length of stay
(.20, p .77), hospital length of stay (1.22, p .22), and median
costs ($6,380, p .07) (Table 4).
These ndings were not qualitatively changed in analyses
adjusting for age, sex, race, admission APACHE III score, and
neurologic disorder as the admitting diagnosis (Table 4), or in
sensitivity analyses treating the intervention deployment quarter
as a control period (0) or a full intervention period (1), instead of a
fraction corresponding to the median proportion of trained nurse
days (data not shown). In post-hoc analyses stratied by units with
greater versus lesser nurse knowledge change, the intervention
was associated with an increase of $14,258 in per patient costs in
the 3 study units with less than 5% absolute increase in post-test
knowledge (p .03).
Discussion

ulcers grade II or greater. There were no statistically signicant

differences between intervention and control quarters in proportion of ICU days with physical restraint (2.44, p .44), heavy
sedation (1.08, p .73), or pain score documentation (.11, p .97),
in ICU-acquired pressure ulcers (.11, p .78), or coma-free days
(.32, p .76), or (Table 4).
These ndings were not qualitatively changed in analyses
adjusting for age, sex, race, admission APACHE III score, and
neurologic disorder as the admitting diagnosis (Table 3), or in
sensitivity analyses treating the intervention deployment quarter
as a control period (0) or a full intervention period (1), instead of a
fraction corresponding to the median proportion of trained nurse
days, or in post-hoc analyses stratied by greater versus lesser
nurse knowledge change (see Supplemental Appendix).

In this study conducted in 2 hospitals in a single academic

health system, the SPEACS-2 program implemented sequentially
across six heterogeneous specialty ICUs improved nurse knowledge of and attitudes toward assistive communication strategies
very modestly, but had no impact on selected nursing care
quality indicators (i.e., physical restraint use, heavy sedation, pain
rating adherence, and ICU-acquired pressure) or utilization outcomes (i.e., ventilator-free days, length of ICU stay, length of
hospital stay, cost).
There are several possible explanations for the lack of impact of
SPEACS-2 on nursing quality indicators and utilization outcomes.
The rst is that the communication skills training was not
received. Although we met the 85% training target in all but

Table 4
Nursing care quality and utilization outcomes, by intervention period.
Control
(N 626)

Interventiona
(N 814)

Unadjustedb intervention
effect b [95% CI]
p-value

Adjustedc intervention
effect b [95% CI]
p-value

47.9 (36.0)

50.1 (36.5)

% ICU days with heavy sedation

37.9 (31.3)

32.3 (29.7)

Coma-free days, d

20.3 (10.2)

21.1 (9.9)

% ICU days with pain score documented

29.0 (30.1)

24.0 (27.2)

Patients With one or more pressure ulcers

62 {9.9}

71 {8.7}

2.44, [10.27; 4.50]

0.44
1.08, [5.09; 7.25]
0.73
0.32, [1.73; 2.36]
0.76
0.11, [6.03; 5.80]
0.97
0.11, [0.90; 0.67]
0.78

2.44, [9.80; 4.92]

0.52
0.56, [5.36; 6.48]
0.85
0.96, [0.88; 2.81]
0.31
1.44, [5.97; 3.10]
0.53
0.17, [0.98; 0.64]
0.68

19.1 (7.5)

19.3 (7.3)

ICU length of stay, d

10.0 [12.0]

9.0 [11.0]

Hospital length of stay, dd

15.0 [15.0]

15.0 [13.0]

Cost, $d

45,876 [47,195]

50,458 [53,305]

0.15, [1.22; 1.51]

0.83
0.20, [1.18; 1.59]
0.77
1.22, [0.66; 3.09]
0.20
6,380, [579; 13,339]
0.07

0.37, [0.88; 1.61]

0.55
0.08, [1.28.; 1.13]
0.90
0.89, [0.89; 2.68]
0.33
5,797, [936; 12,529]
0.09

Outcome

Nursing care quality measures, mean (SD); N {%}

% ICU days with upper extremity physical restraint

Utilization outcomes, mean (SD); median [IQR]

Ventilator-free days, d
d

SD e standard deviation; IQR e interquartile range; ICU e intensive care unit.

a
Includes quarter of intervention deployment, during which the intervention is treated as a fraction corresponding to the median trained nurse days (see Table 2).
b
Mixed effect regression (standard errors adjusted for clustering by unit).
c
Mixed effect regression with additional adjustment for patient age, sex, race, admission APACHE III, and neurological disorder as admitting diagnosis.
d
Cost-adjusted charges calculated from hospital administrative claims data for charges, deated by the hospital-specic cost-to-charge ratio. Outliers winsorized
(top-coded) at the 90th percentile.

414

M.B. Happ et al. / Heart & Lung 44 (2015) 408e415

one unit, there were only modest improvements in knowledge

observed and a relatively high rate of incorrect answers on the
post-test among nurses who completed SPEACS-2 training. The
pre- and post-test was a 10-item knowledge quiz. The observed
difference of 6-percentage points between the pre- and post-test
corresponds to answering 0.6 (<1) additional item correctly. This
test was pretested with clinicians but did not undergo psychometric assessment and test performance cannot be mapped onto
bedside nursing behaviors.
These small, though statistically signicant, knowledge gains
raise the concern that the training intervention was ineffective.
The on-line SPEACS-2 was a condensed version of the face-to-face
communication skills training course developed and tested in the
original SPEACS study,7,8 revised with input from those nurse
participants9 and vetted with clinicians and communication disorders specialists. We did not, however, pretest the online version
to replicate SPEACS study ndings. We chose to deliver this content via an online platform to maximize accessibility (24 h a day, 7
days a week) and consistency in content delivery. Nurses, particularly those who accessed the training program from bedside
computers, may have been distracted or inattentive to the training
program content and video exemplars. Although we sought to
reinforce learning with posters featuring a different assessment or
communication strategy each week and weekly communication
rounds with the Speech Language specialist, enactment was
suboptimal, and overall, delity to nurse implementation of the
intervention in terms of real practice change was low. A recent online training in AAC techniques developed by speech language
pathologists at the University of Iowa Hospitals demonstrated
even less nurse participation.27 Communication skills transfer
may require more intensive role modeling, skills practice,
boosters, and feedback.28e30
Second, it is possible that, despite some knowledge gain, the
communication skills training program did not change communication behaviors and achieve practice change. We had no predetermined expectations of adherence. Critical care provider
adherence to known clinical practice guidelines varies widely as
evidenced by reported adherence to care bundles in the Surviving
Sepsis Campaign (36%),31 trauma patient transfer guidelines
(<50%),32 endotracheal suctioning guidelines (51%),33 and enteral
nutrition guidelines (>80%).34 Although we identied nurse
champions in each ICU to facilitate nurse engagement in the study
and obtained approval and support of the health systems chief
nursing ofcer, we conceptualized the intervention as an intervention to change knowledge and attitudes. Even demonstrably
acquired skills may not generalize to actual practice if existing
communication practices are deeply embedded. Attitudinal barriers reported by previous nurse focus groups, such as prioritization
of technological caregiving and life sustaining treatments over
communication and psychosocial needs and preferences for patients who are sedated and quiet,9 may require a broader
organizational initiative to change communication practice,
recognition for good performance, and more prolonged implementation of 12e24 months.35 For example, a recent study of
physician simulation-based communication skills training to
improve end-of-life communication between physicians/nurse
practitioners and seriously ill patients did not change practice or
patient and family perceptions of encounters with the health care
provider.36
Third, the study results may be constrained by limitations in
the outcomes measurement, which depended on the accuracy
and completeness of EMR documentation and assumed sensitivity of these particular measures to the impact of changes
in communication processes between nurses and nonvocal patients in ICU. This is the rst study to attempt longitudinal

quantication of nursing care quality indicators in ICU to measure the effect of communication. The intervention demonstrated
modest improvements in nurse knowledge, and their comfort
and satisfaction with communication with nonvocal, intubated
ICU patients. Greater improvements in nurses knowledge
about communication using AAC tools and techniques may have
been achieved with formal educational follow-up or boosters
(e.g., webinar, simulation, role play, communication coaching).
Because we have no information on minimal clinically important
differences in scales measuring satisfaction and comfort with
communication, we are careful to not over-interpret the signicance. Although the gains were modest, the fact that the program
was able to change attitudes, particularly comfort with and
satisfaction in communicating with impaired patients is an
important step toward practice change. Program enactment
measures showed gradual increases over time from the deployment quarter to the post-implementation quarter, most notably,
in the use of AAC tools and techniques beyond simple yes/no
appropriate to the patients level of communication function, and
placement of AAC tools in the rooms of eligible mechanically
ventilated patients. Although program adherence and adoption
may not have reached levels necessary to show meaningful
change in quality indicators or patient outcomes, these achievements suggest that the program can produce measurable practice
change and that longer and more robust implementation may yield
better results.
Fourth, we powered the study to nd a small to moderate effect
size. If the effect size is considerably smaller, the null nding would
reect type II error.
Post-hoc sensitivity analysis indicating that units with low
gains in nurse knowledge experienced a substantive increase
in per patient costs in the post-intervention quarters are
difcult to interpret because other utilization outcomes that
typically track with costs, such as length of stay and ventilatorfree days, did differ between units with low and high gains in
nurse knowledge, and may reect Type I error (See Online
Supplement).
The stepped wedge design was chosen because individual level
randomization was not possible and to avoid threats to external
validity in a classic cluster randomized trial, particularly with
respect to heterogeneity of ICUs,26 and because the intervention
showed initial evidence of effectiveness in improving communication process measures.8 However, the protracted stepped
implementation extends trial duration while limiting the time
available for intervention implementation as well as the opportunity for follow-up measurement.37
Finally, more granular data collection on nurse dose and
implementation may have provided a more robust test of practice
change and the impact of that change on patient well-being. One
approach may be to quantify the proportion of days that the patients nurse(s) were SPEACS-2 trained. Nurses could be observed
for actual use of SPEACS-2 techniques and tools during patient
communication.
In conclusion, unit-wide implementation of a program of
nurse training in assistive communication skills, low-tech
communication tools and clinical resources (SPEACS-2) achieved
modest improvements in ICU nurses knowledge, satisfaction and
comfort in communicating with nonvocal mechanically ventilated patients but did not signicantly impact on nursing care
quality or clinical outcomes. Future studies should include longer
intervention periods, pragmatic clinical trial designs38 that
employ iterative intervention testing responsive to the ICU
setting, more granular intervention measurement, validation of
outcome measures, and consideration of patient and family
psychological outcomes.

M.B. Happ et al. / Heart & Lung 44 (2015) 408e415

Acknowledgments
Our thanks to research mentorship students, Rebecca Nock,
Anna Evans, Jin Lee, Ian Joel, Tiffany Behringer, Rachel Orler Reid,
Joe Ciampoli, Joe Kaye, Lauren Mancuso, Cassandra Delp, Hannah
Park, Alexandra Dreyzin, who worked on this project and to Melissa
Saul, MS and Elan Cohen, MS for database and programming
assistance.
References
1. Khalaila R, Zbidat W, Anwar K, Bayya A, Linton DM, Sviri S. Communication
difculties and psychoemotional distress in patients receiving mechanical
ventilation. Am J Crit Care. 2011;20:470e479.
2. Rotondi AJ, Chelluri L, Sirio C, et al. Patients recollections of stressful experiences while receiving prolonged mechanical ventilation in an intensive care
unit. Crit Care Med. 2002;30:746e752.
3. Happ MB, Sereika S, Garrett K, et al. Nurse-patient communication interactions
in the ICU. Am J Crit Care. 2011;20:e28ee40.
4. Puntillo K, Neuhaus J, Arai S, et al. Challenge of assessing symptoms in seriously
ill intensive care unit patients: can proxy reporters help? Crit Care Med.
2012;40:2760e2767.
5. Bartlett G, Blais R, Tamblyn R, Clermont RJ, MacGibbon B. Impact of patient
communication problems on the risk of preventable adverse events in acute
care settings. CMAJ. 2008;178:1555e1562.
6. Magnus VS, Turkington L. Communication interaction in ICUePatient and staff
experiences and perceptions [Erratum appears in Intensive Crit Care Nurs.
2008 Aug;24(4):264]. Intensive Crit Care Nurs. 2006;22:167e180.
7. Happ MB, Sereika S, Garrett K, Tate J. Use of the quasi-experimental sequential
cohort design in the Study of PatienteNurse Effectiveness with Assisted
Communication Strategies (SPEACS). Contemp Clin Trials. 2008;29:801e808.
8. Happ MB, Garrett KL, Tate JA, et al. Effect of a multi-level intervention on
nurse-patient communication in the intensive care unit: results of the SPEACS
trial. Heart Lung. 2014;43:89e98.
9. Radtke JV, Tate JA, Happ MB. Nurses perceptions of communication training in
the ICU. Intensive Crit Care Nurs. 2012;28:16e25.
10. Nilsen ML, Happ MB, Sereika SM, et al. Nurse and patient interaction behaviors
effects on nursing care quality for mechanically ventilated, older adults in the
ICU. Res Gerontol Nurs. 2014 MayeJun;7(3):113e125.
11. Ramsay M, Savege T, Simpson BR, Goodwin R. Controlled sedation with
alphaxalone-alphadolone. Br Med J. 1974;2:656e659.
12. Teasdale G, Jennett B. Assessment of coma and impaired consciousness. A
practical scale. Lancet. 1974;2:81e84.
13. Riker RR, Picard JT, Fraser GL. Prospective evaluation of the Sedation-Agitation
Scale for adult critically ill patients. Crit Care Med. 1999;27:1325e1329.
14. Happ MB, Seaman JB, Nilsen ML, et al. The number of mechanically ventilated ICU
patients meeting communication criteria. Heart Lung. 2015 JaneFeb;44(1):45e
49. http://dx.doi.org/10.1016/j.hrtlng.2014.08.010. pii: S0147e9563(14)00330-6.
15. Bellg AJ, Borrelli B, Resnick B, et al. Enhancing treatment delity in health
behavior change studies: best practices and recommendations from the NIH
Behavior Change Consortium. Health Psychol. 2004;23:443e451.
16. Sereika S, Tate J, Happ M, Garrett JM. Nurses Attitutdes and Perceptions About
Communications With Nonspeaking ICU Patients. Washington DC: Council for
the Advancement of Nursing Science, 2008 State of the Science Congress;
September, 2008.
17. Bergbom-Engberg I, Haljamae H. Assessment of patients experience of discomforts during respirator therapy. Crit Care Med. 1989;17:1068e1072.

415

18. Patak L, Gawlinski A, Fung NI, Doering L, Berg J, Henneman EA. Communication
boards in critical care: patients views. Appl Nurs Res. 2006;19:182e190.
19. Happ MB, Tuite P, Dobbin K, DiVirgilio-Thomas D, Kitutu J. Communication
ability, method, and content among nonspeaking nonsurviving patients treated
with mechanical ventilation in the intensive care unit. Am J Crit Care. 2004;13:
210e218. quiz 9e20.
20. Pandharipande P, Bun BT, Herr DL, et al. Effect of sedation with dexmedetomidine vs lorazepam on acute brain dysfunction in mechanically ventilated
patients: the MENDS randomized controlled trial. J Am Med Assoc. 2007;298:
2644e2653.
21. Rubenfeld GD, Angus DC, Pinsky MR, et al. Outcomes research in critical care:
results of the American Thoracic Society Critical Care Assembly workshop on
outcomes research. Am J Respir Crit Care Med. 1999;160:358e367.
22. Landis JR, Koch GG. The measurement of observer agreement for categorical
data. Biometrics. 1977;33:159e174.
23. Seaman JB, Happ MB, Sereika S, Barnato AE. Development and testing of a data
collection instrument for quality of care data. Washington, DC: Council for the
Advancement of Nursing Science, 2012 State of the Science Congress;
September, 2012.
24. Knaus WA, Wagner DP, Draper EA, et al. The APACHE III prognostic system.
Risk prediction of hospital mortality for critically ill hospitalized adults. Chest.
1991;100:1619e1636.
25. Hughes JP. Personal Communication; 2013.
26. Hussey M, Hughes JP. Design and analysis of stepped wedge cluster randomized trials. Contemp Clin Trials. 2007;28:182e191.
27. Downey DA. The Effectiveness of AAC Training Protocols for Acute Care Nurses: A
Randomized Controlled Trial of an Instructional On-line Medium for Clinical Skills
Teaching. Doctoral Dissertation. University of Iowa, http://ir.uiowa.edu/etd/
4616; 2014.
28. Bandura A. Self-efcacy toward a unifying theory of behavioral change. Psychol
Rev. 1977;84:191e215.
29. Glanz K, Rimber B. Theory at a Glance: A Guide for Health Promotion Practice.
Bethesda, MD: National Cancer Institute; 1997.
30. Kaufman DM. Applying educational theory in practice. BMJ. 2003;326:
213e216.
31. Levy MM, Dellinger RP, Townsend SR, et al. The Surviving Sepsis Campaign:
results of an international guideline-based performance improvement program targeting severe sepsis. Intensive Care Med. 2010;36:222e231.
32. Mohan D, Barnato AE, Angus DC, Rosengart MR. Determinants of compliance
with transfer guidelines for trauma patients: a retrospective analysis of CT
scans acquired prior to transfer to a Level I Trauma Center. Ann Surg. 2010;251:
946e951.
33. Frota OP, Loureiro MDR, Ferreira AM. Open system endotracheal suctioning:
practices of intensive care nursing professionals. Esc Anna Nery. 2014;18:296e
302.
34. Quenot JP, Plantefeve G, Baudel JL, et al. Bedside adherence to clinical practice
guidelines for enteral nutrition in critically ill patients receiving mechanical
ventilation: a prospective, multi-centre, observational study. Crit Care.
2010;14:R37.
35. Tiller MG, wilson DS, Resnick B, Shever LL. Dissemination and Implementation:
INQRIs potential impact. Med Care. 2013;51:S41eS46.
36. Curtis JR, Back AL, Ford DW, et al. Effect of communication skills training
for residents and nurse practitioners on quality of communication with patients with serious illness: a randomized trial. J Am Med Assoc. 2013;310:
2271e2281.
37. Kotz D, Spigt M, Arts I, Crutzen R, Viechtbauer W. The stepped wedge design
does not inherently have more power than a cluster randomized controlled
trial. J Clin Epidemiol. 2013;66:1059e1060.
38. Riley W, Glassgow R, Etheredge L, Abernathy A. Rapid, responsive, relevant
(R3) research: a call for a rapid learning health research enterprise. Clin Transl
Med. 2013;2.

415.e1

M.B. Happ et al. / Heart & Lung 44 (2015) 408e415

Appendix Figure A1

Patients reported on the initial

eligibility list:
n = 5476

Pre-Intervention
n=2,322

Intervention
n=3,154

EMR reviewed for eligibility:

n=1340

EMR reviewed for eligibility:

n=1747

Excluded:
n=714

Excluded:
n=933

Abstracted:
n=626

Abstracted:
n=814

Fig. 1. [CONSORT diagram]. We retrospectively identied a representative sample of

control (pre-intervention) and intervention cases from consecutive admissions to the 6
study ICUs using administrative (e.g., billing) data. Potentially eligible admissions met
three criteria: (1) age 18 and older; (2) rst ICU admission during the hospital stay; and
(3) MV of at least 2 days duration. From this list of potentially eligible admissions, we
randomly screened charts to assess eligibility until we reached 30 patients per unit per
quarter, yielding the pre-specied sample of 1440.

Appendix Table A1
Quality indicator denitions
Quality indicator

Denition

Restraint use

All restraint devices used within the 24 h interval being evaluated including: soft extremity restraints (specify number of limbs restrained), vests,
waist belts, full side-rails, mitts, and enclosure beds.
Evidence of heavy sedation at any point during the 24 h interval as measured by:
 Modied Ramsay score 4e6 or
 Riker score of 1e2 or
 Nursing note description of unresponsiveness to verbal or tactile stimulation, or being comatose or anesthetized.

Heavy sedation

Pain

ICU acquired
pressure ulcers

Awake for 8 out of 12 h for 12:00 ame11:59 am or 12:00 pme11:59 pm as dened by:
 Modied Ramsay score 1e3.
 GCS motor score of 6.
 Nursing note documentation of being alert, awake, arousable, responsive, or communicative.
Presence of any pain during the 24 h period being evaluated (Y/N).
Highest pain score on a scale of 1e10 (including half scores) for the 24 h interval.
Any use of the descriptor unable to communicate in the pain assessment documentation during the 24 h interval (Y/N).
Any pressure ulcer, stage II or greater, occurring during the index ICU stay that was not documented on admission [cumulative for ICU stay].

M.B. Happ et al. / Heart & Lung 44 (2015) 408e415

415.e2

Appendix Table A2
Post hoc analysis of outcome variables by unit nurse knowledge gain
Low knowledge gain units

Nursing care quality measures

% ICU days with upper extremity physical restraint
% days heavy sedation
Coma-free days, d
% ICU days with pain score documented
Patients With one or more pressure ulcers
Utilization outcomes
Ventilator-free days, d
Study ICU LOS (days)
Hosp LOS (days)
Total ICU LOS (days)
Cost ($)

High knowledge gain units

Intervention effect

Statistic

p value

2.63
8.32
1.22
3.36
0.66

t 0.38
t 1.43
t 0.63
t 0.65
Wald ChiSq 1.21

0.71
0.15
0.53
0.52
0.27

6.64
7.12
2.07
2.98
0.07

0.23
0.51
0.52
0.45
14,259

t
t
t
t
t

0.87
0.69
0.77
0.74
0.03

0.3
1.09
2.94
1.17
5913

0.17
0.4
0.29
0.33
2.15

Intervention effect

Statistic

p value

t 0.87
t 1.18
t 1.21
t 0.54
Wald ChiSq 0.01

0.38
0.24
0.23
0.60
0.93

t
t
t
t
t

0.23
0.84
1.77
0.88
0.86

0.82
0.40
0.08
0.38
0.39