Download
High-Users of Acute Care in a Teaching Hospital:
A Retrospective Chart Review and Survey of
Primary Care Physicians
Arpita Gantayet, MD, MASc, Pamela Mathura, MBA, Alexis Fong-Leboeuf, BMSc, Natalie McMurtry, MBA, Julie Zhang, MSc,
Finlay A. McAlister, MD, MSc, Narmin Kassam, MD, MHPE
About the Authors
Arpita Gantayet, MD, MASc, is a Resident in the Department of Internal Medicine, University of Alberta, Edmonton, Alberta, Canada
Pamela Mathura, MBA, is an Improvement Specialist for the Department Medicine, Alberta Health Services and a Clinical Lecturer
in the Department Medicine, University of Alberta, Edmonton, Alberta, Canada
Alexis Fong-Leboeuf, BMSc, is a Medical Student at the University of Alberta, Edmonton, Alberta, Canada.
Natalie McMurtry, MBA, is an Executive Director in Operations and Facility Development at Alberta Health Services, Edmonton,
Alberta, Canada
Julie Zhang, MSc. is a Performance Improvement Manager at the University of Alberta Hospital, Edmonton, Alberta, Canada
Finlay A. McAlister, MD, MSc is a Professor of General Internal Medicine and Alberta Health Services Chair in Cardiovascular Outcomes
Research, University of Alberta, Edmonton, Canada
Narmin Kassam, MD, MHPE is Professor of General Internal Medicine and Deputy Clinical Department Chair in the Department of
Medicine, University of Alberta, Edmonton, Canada
Corresponding Author: arpita.gantayet@gmail.com
Submitted: April 14, 2019. Accepted: October 18, 2019. Published: August 27, 2020. DOI: 10.22374/cjgim.v15i3.363
ABSTRACT
Purpose
To characterize high-users (HUs) of inpatient units, obtain insights from their primary care
physicians (PCPs) and identify factors that can be modified to reduce resource use.
Method
The study design included retrospective chart reviews of high-user patients and qualitative surveys
of their PCPs. HUs were defined as adults with 3 or more admissions to an index tertiary teaching
hospital in Edmonton as well as a cumulative length of stay (cLOS) greater than 30 days at any
hospital in the province of Alberta, between September 1, 2015 and September 30, 2016. The
charts of HUs were reviewed to assess demographics, admitting and consulting services, medical
profile, social profile, community supports, and scores on pre-existing risk-stratification tools
to identify patient factors that might be characteristic of HUs. Additionally, a survey comprising
12 multiple-choice and 8 short-answer questions was faxed to their PCPs to assess HU attitudes
and behaviors and collect recommendations to prevent high use of acute care.
Results
Of 125 HUs (median 62 years old, 5 admissions, cLOS 49 days, 14 emergency department (ED)
visits, 10 medications), 74% lived at home, 86% had a PCP, 56% received homecare pre-admission
and 34% had at least one critical care admission. HUs accounted for 2474 admissions or ED
Canadian Journal of General Internal Medicine
28 Volume 15, Issue 3, 2020
Gantayet et al.
CJGIM_3_2020_175414.indd 28CJGIM_3_2020_175414.indd 28 8/25/20 4:35 PM8/25/20 4:35 PM
visits (median 14, IQR 10-22) at all sites in the year studied; 41% of their 1605 ED visits and
21% of their 869 admissions were at other hospitals. Their most prevalent comorbidities were
hypertension, depression, and diabetes. 49 responses were received to 114 faxed surveys (43%
response rate). Only 14 of 49 responding PCPs suggested interventions to address ED revisits
and readmissions; PCPs most frequently cited living conditions and lack of social supports as
key causative factors.
Conclusions
We have characterized high-user patients and discussed PCP perspectives and strategies to
optimize their healthcare use.
Resume
Objet
Caractériser les grands utilisateurs (HU) des unités d’hospitalisation, obtenir des informations
de leurs médecins de soins primaires (PCP) et identifier les facteurs qui peuvent être modifiés
pour réduire lutilisation des ressources.
Méthode
La conception de létude comprenait des examens rétrospectifs de dossiers de patients très
utilisateurs et des enquêtes qualitatives sur leurs PPC. Les UH ont été définis comme des adultes
ayant été admis à trois reprises ou plus dans un hôpital universitaire tertiaire d’Edmonton et
dont la durée de séjour cumulée (DSC) est supérieure à 30 jours dans n’importe quel hôpital
de la province de lAlberta, entre le 1er septembre 2015 et le 30 septembre 2016. Les tableaux
des HU ont été examinés afin d’évaluer les données démographiques, les services d’admission
et de consultation, le profil médical, le profil social, les soutiens communautaires et les scores
des outils de stratification des risques préexistants afin d’identifier les facteurs des patients qui
pourraient être caractéristiques des HU. En outre, une enquête comprenant 12 questions à
choix multiple et 8 questions à réponse courte a été envoyée par fax à leurs PCP afin d’évaluer
les attitudes et les comportements des HU et de recueillir des recommandations pour prévenir
un recours élevé aux soins de courte durée.
Résultats
Sur 125 HU (âge médian 62 ans, 5 admissions, cLOS 49 jours, 14 visites aux urgences, 10
médicaments), 74 % vivaient à domicile, 86 % avaient un PCP, 56 % recevaient des soins à
domicile avant leur admission et 34 % avaient au moins une admission en soins intensifs. Les
HU ont représenté 2474 admissions ou visites aux urgences (médiane 14, IQR 10-22) dans tous
les sites au cours de lannée étudiée ; 41% de leurs 1605 visites aux urgences et 21% de leurs
869 admissions se sont faites dans dautres hôpitaux. Leurs comorbidités les plus fréquentes
étaient l’hypertension, la dépression et le diabète. 49 réponses ont été reçues pour 114 enquêtes
envoyées par fax (taux de réponse de 43 %). Seuls 14 des 49 PCP ayant répondu ont suggéré des
interventions pour remédier aux problèmes des visites aux urgences et des réadmissions; les PCP
ont le plus souvent cité les conditions de vie et le manque de soutien social comme principaux
facteurs de causalité.
Conclusions
Nous avons caractérisé les patients grands utilisateurs et discuté des perspectives et des stratégies
de la PCP pour optimiser leur utilisation des soins de santé.
Canadian Journal of General Internal Medicine
Volume 15, Issue 3, 2020 29
Original Article
CJGIM_3_2020_175414.indd 29CJGIM_3_2020_175414.indd 29 8/25/20 4:35 PM8/25/20 4:35 PM
As life expectancy increases and healthcare needs become more
complex, it is becoming increasingly important to improve
efficiency in healthcare delivery.
1,2
It has been observed in various
settings that a small proportion of the population accounts
for disproportionate use of healthcare resources; 5% of the
population accounted for 64% of total health care spending in
Ontario
3
and 66% in Alberta.
4
The biggest driver of these costs
are inpatient (IP) admissions.
1,3
Although multiple risk scores
have been developed to predict single readmissions after hospital
discharge,
5-7
there are few risk-stratification tools to predict
which individuals will become high system users.
The objective of this study was to use a patient-focused
approach to gain insights into high-users (HUs) and potential
approaches to optimizing their care. The goal was to identify and
characterize current HUs, obtain insights from their primary care
physicians (PCPs) on their healthcare behaviors and attitudes,
discover patient and system factors that predispose them to
frequent readmissions, and to suggest strategies to intervene
against modifiable factors.
METHODS
This study was performed at a large tertiary care teaching hospital,
the University of Alberta Hospital (UAH), Edmonton between
Sep 1, 2015 and Sep 30, 2016. The study design included chart
reviews of high-user patients and qualitative surveys of their
PCPs. As per the Canadian Institute for Health Information
(CIHI) definition, HUs were any adult patients with three or
more admissions at our index hospital and cumulative length of
stay (cLOS) greater than 30 days at any hospital in the province
of Alberta, during that year. Patients who met the HU definition
were included even if they died in the hospital or during the
study period.
High user data was obtained from the Alberta Health
Services (AHS) medical site administrative office using the Data
Integration Monitoring and Reporting (DIMR) unit. We used
DIMR to collect all data on visits to ED or acute care hospitals
anywhere in the province of Alberta – this allowed us to track
resource use by HUs regardless of where else they received care
in the province. The descriptive variables derived from the UAH
local database and DIMR were organized into the following
categories; patient gender, age at last admit, postal code, date of
last admission between Sep 1/15 - Sep 30/16, number of UAH
admissions in study period, number of non-UAH admissions,
number of UAH ED visits, number of non-UAH ED visits,
cumulative LOS (days) between Sep 1/15 - Sep 30/16 at UAH
only and cumulative LOS (days) between Sep 1/15 - Sep 30/16 at
all hospitals in Alberta. DIMR also provided data on whether the
identified HUs had any admissions in the prior two years if the
HU patient was deceased at the time of analysis, and the number
of ambulance arrivals. We obtained the most common admitting
diagnosis list from UAH under the hospital’s International
Statistical Classification of Disease (ICD) codes. HU data was
derived for all specialties except Obstetrics/Gynecology, which
is not available at the UAH site, and Pediatrics, as this patient
population was not the focus of the study. Admitting service
was categorized by specialties; General Internal Medicine
(GIM)/Cardiology/Critical care/Gastroenterology/Hematology/
Nephrology/Family Medicine/Surgical specialties including ENT/
Psychiatry/Geriatrics. Quantitative variables were loaded on a
‘Dashboard’ database created by the Performance Improvement
Manager at UAH and analysis was performed by the first author
using its’ filter applications. The data was documented as number
counts and ranges of minimum to maximum, where applicable.
Excel worksheets were then used by the first author to calculate
proportions, percentages as well as medians with an interquartile
range from the 25
th
to 75
th
percentile.
We obtained ethics approval for chart reviews and PCP
surveys, with a waiver of informed patient consent, from the
University of Alberta Research Ethics Board (Pro00073914).
Physician (PCP) informed consent was obtained by faxing a consent
sheet along with the survey. Chart reviews were standardized
by the first author using a list of definitions and categorization
protocols and were performed by the first and third authors in
this study. Only charts for the last admission in the study period
were reviewed for each HU, as comprehensive data collection and
analysis was not feasible given the large number of admissions
for all HUs combined. Descriptive variables derived from the
chart review were categorized as living facility at the time of the
last admission, time lived in that facility, prior or current home
care, duration of home care, types of home supports, types of
inter-professional care supports in the community, independent
for all activities of daily living (ADLs) and iADLs (in the form of
yes, no or unknown), number of regularly prescribed, scheduled
medications at the last admission, list of discharge medications
after the last admission in the study period, the cumulative
number of specialties involved and goals of care documented in
the chart at the time of the last admission in the study period.
Comorbidities listed in the chart under ‘Past Medical History’
were categorized into Neurologic, Cardiovascular, Pulmonary
Gastrointestinal, Hematologic, Nephrology, Oncology, Infectious
Diseases, Rheumatology, Endocrinology, Psychiatry, Geriatrics,
Transplant and Other. These comorbidities were used to calculate
the Charlson Comorbidities score.
To determine if existing scores are predictive of these high-
users, we calculated LACE and FAM-FACE-SG scores for each
HU based on their last admission during the study period.
The LACE score is a 19-point score which assigns points for
length of stay, acuity of admission, comorbidities and number
Canadian Journal of General Internal Medicine
30 Volume 15, Issue 3, 2020
Gantayet et al.
CJGIM_3_2020_175414.indd 30CJGIM_3_2020_175414.indd 30 8/25/20 4:35 PM8/25/20 4:35 PM
of ED visits and is considered high risk for readmission if
≥ 10.
5
The FAM-FACE-SG score was modified to include age,
number of admissions, frequency of ED use (≥3 ED visits in
6 months before index admission), antidepressant use in past
year, Charlson Comorbidity score,
9
ESRD on dialysis and lasix
dose (intravenous 40 mg and above during last admission), but
no need for financial assistance or subsidized ward stay, with a
score ≥14 considered high risk.
8
PCP affiliation was determined by the third author by checking
if the HU had a physician name listed in their medical record.
If yes, the PCP’s medical specialty and contact information was
obtained from the College of Physicians and Surgeons of Alberta
(CPSA) website; some PCPs could not be contacted as their
contacts were not available on the CPSA website or they were
deceased. For the remainder, we first contacted the PCP offices
by phone and clarified if the patient had ever been affiliated
with the PCP. If yes, we additionally asked for permission to
send a survey and determined the accurate fax number or email
address. The first round of surveys was sent by fax or email
between February 18, 2018 - March 2, 2018. If a response was
not received within two weeks, a reminder call was given and
surveys were re-faxed. All responses received by May 30, 2018,
were included in the analysis.
The survey consisted of 12 multiple-choice questions and
8 short answer questions. A sample survey can be seen in Table
S2 in the Supporting Information section. All analysis of PCP
surveys was performed by the first author. For the multiple-choice
questions, when multiple options were selected, all options were
included in the count. Unfortunately, not all PCPs responded to
every question, therefore each question has a different number
of responses. We calculated the percentage of each response in
excel, using a common denominator, as a proportion of the total
number of PCPs who responded to the survey.
For the eight short answer questions, the PCP’s handwritten
responses were first typed into excel. For the question ‘What are
the 4 most common medical conditions the patient visits you
for?’, responses were categorized by the first author as broad
organ systems and reported based on the highest frequency of
reason for visits to the PCP. The response to ‘What 4 specialists
are most commonly enlisted in this patients care?’ and ‘What 4
outpatient clinics is this patient followed most often in?’ could
not be analyzed as most PCPs provided a physician name rather
than a specialty or clinic. The responses to ‘What 4 community
supports are enlisted in this patients care most often?’ were
diverse and could not be categorized. Qualitative responses to
‘What interventions, if any, could prevent this patients ED revisits
and inpatient admissions?’ and ‘Is there anything else you would
like to share?’ were categorized under the headings ‘PCP has
suggestions’ versus ‘PCP has no suggestions for intervention. PCP
suggestions with similar themes were then compiled under broad
categories (see Table 3). Qualitative responses to the question
‘What patient factors, if any, predispose this patient to high-cost
healthcare use?’ were categorized as ‘medical issues only’ and
additional contributing social factors. The ‘Select-All-That-Apply’
question ‘Can you identify any ‘Social Determinants of Health’ that
influence this patients high-cost use? provided options; income
and income distribution, education, unemployment, and job
security, employment and working conditions, early childhood
development, food insecurity, housing, social exclusion, social
safety network, health services, aboriginal status, gender, race,
disability, living situation, social supports and other. The number
of selections for each factor were calculated and reported as a
percentage of the total number of social factors selected.
Results
High-User Demographics and Social Profile
A total of 125 HUs were identified at UAH between Sep 1, 2015
and Sep 30, 2016. The total number of patients admitted to
UAH during this study period was 23,643. Thus, the 125 HUs
accounted for 0.5% of all patients admitted over that year. The
125 HUs (median 62 years old) had 688 admissions to UAH
between Sep 2015-2016 accounting for 7716 acute hospital days.
During the same time the UAH had a total of approximately
13,400 medical discharges and 11,800 surgical discharges. Thus,
the HUs accounted for 688 of 25,200 (2.7%) of all admissions at
UAH that year. For the HUs, UAH hospitalizations were only
the tip of the iceberg as they also had 1605 additional ED visits
(41% at hospitals other than the UAH) and 181 admissions
(21% of their total) to other non-UAH hospitals elsewhere in
the province. The HUs arrived by ambulance 662 times (41% of
their UAH visits) with median 4 (IQR 2-7) ambulance arrivals
in the year studied. Of these 125 HUs for 2015/16, 76 (61%)
had also been admitted at least once in Sep 2014/2015 and 58
(46%) between Sep 2013/2014, as obtained from DIMR data.
Table 1 summarizes the characteristics of these HUs. Almost
half of HUs (46%) were younger than 60 years and 70 (56%)
were receiving home care. 62 (50%) were dependent for at least
some ADLs as per documentation in patient charts, however,
these were rarely broken down into specifics. Types of home
supports included 19 HUs with personal care (15% of HUs), 14
with medication assist (11%), 14 with home care/support (11%),
12 with wound care (10%), 11 with bath assist (9%) and 11 with
respiratory services (9%). Based solely on the last admission of
the study period, the median number of discharge medications
was 10 (IQR 7-14), median Charlson Comorbidity score was 2
(IQR 1-3), median LACE score was 14 (IQR 12-16), and median
FAM-FACE-SG score (excluding financial components) was 32
(IQR 25-38) (Table 1).
Canadian Journal of General Internal Medicine
Volume 15, Issue 3, 2020 31
Original Article
CJGIM_3_2020_175414.indd 31CJGIM_3_2020_175414.indd 31 8/25/20 4:35 PM8/25/20 4:35 PM
High-User Medical Profile
Most (93%) HUs (116) had at least one admission to a medical
specialty and 51 (41%) had at least one admission to a surgical
specialty in the 13 months studied. Table S1 in the Appendix
summarizes rates of admission and consultation amongst services.
Of the HU hospitalizations, 28% (185 admissions) were on GIM
wards, 16% (116) Pulmonary, 15% (100) Hematology, and 15%
(98) Gastroenterology. The median number of services involved
at the last admission (either as admitting or consulting service)
was 2 (IQR 2, 4) up to a maximum of 10. The most consulted
specialties were Surgical Specialties with 46 consults (16% of all
consults), GIM with 39 consults (13%), Gastroenterology 34 (12%),
Pulmonary 28 (9%), Infectious Diseases 26 (9%), Psychiatry 19
(6%) and Hematology 18 (6%) consults. About one third (34%)
of HUs (42) were admitted to critical care at some point over
the study year at least once, with a total of 66 ICU admissions
and 397 ICU days for these patients with a median of 2.6 days
(IQR 0.8-11.1) per patient in the study year. Figure 1 displays
the most commonly listed comorbidities under ‘Past Medical
History’ during chart reviews. These listed comorbidities were
categorized by grouping under predefined categories. Figure S1
in the Appendix displays the most common admitting diagnosis
under ICD codes. As of the last admission, goals of care were
documented under mandatory admission documentation as full
Table 1. Demographics and Social Characteristics of High-Users (HUs).
High-user Variable Total number
Range
(min-max)
Median
(Interquartile
Range
25
th
-75
th
) Subgroups
Number of
HUs (% of
125 HUs)
Total admissions and ED visits between Sep
2015-16
2474 5 - 127 14 (10,22)
Cumulative UAH LOS (days) 7716 13 - 219 49 (35,73)
UAH admissions 688 3 - 11 5 (4,6)
Non-UAH admissions 181 0 - 13 1 (0,2)
UAH ED visits 941 1 - 41 6 (4,8)
Non-UAH ED visits 664 0 - 89 1 (0,3)
Number of ambulance arrivals 662 1 - 57 4 (2,7)
Age (years) 22 - 95 62 (46,72)
Number of discharge meds 0 - 30 10 (7,14)
Charlson Comorbidity score
9
0 - 15 2 (1,3)
LACE Score
5
(High risk ≥10) 4 - 19 14 (12,16)
FAM-FACE-SG score
8
excluding Medifund &
subsidized stay (High risk ≥14)
10 - 50 32 (25,38)
Sex Male 66 (53)
Female 59 (47)
Living facility at time of last admission
Home 92 (74)
LTC 12 (10)
Supportive Living 11 (9)
No Fixed Address 9 (7)
Prior or Current Home Care
<6 months 25 (20)
≥6 months 45 (36)
None 55 (44)
ED = Emergency Department; LOS = length of stay; UAH = University of Alberta Hospital
Canadian Journal of General Internal Medicine
32 Volume 15, Issue 3, 2020
Gantayet et al.
CJGIM_3_2020_175414.indd 32CJGIM_3_2020_175414.indd 32 8/25/20 4:35 PM8/25/20 4:35 PM
resuscitation for 46% (58 HUs), resuscitation without CPR for
4% (5), ICU support without CPR and intubation for 7% (9),
medical management but no resuscitation or ICU support in
27% (34), comfort measures only in 11% (14) and unknown/
presumed full resuscitation in 4% (5). At the time of analysis in
February 2018, 54 (43%) of the 125 HUs were deceased.
Family Physician Affiliation and Survey Responses
In our cohort, 96% (120) of 125 HUs listed in their medical
record that they had a PCP; 108 (86%) were Family Physicians,
5 were General Internists, 3 were Cystic Fibrosis specialists,
and 4 others listed a psychiatrist, medical oncologist, palliative
care physician, and nurse practitioner. During surveys, 6 PCPs
could not be contacted as their contacts were not available on
the CPSA website or they were deceased. We received responses
to 49 of the faxed 114 surveys (43% response rate); 4 declined
participation and 9 responded stating that the patient is not
affiliated with them or their clinic. Table 2 summarizes the
PCP responses to the first 11 multiple-choice questions in the
survey including duration of affiliation, frequency of visits,
patient behaviors and attitudes, goals of care discussions, and
discharge summaries. In cases where multiple options were
selected, all options were included. Some PCPs did not respond
to select questions. Question 12 (What is the patients attitude
to you?) is not reported in Table 2 as the PCP understanding
of the question was variable and not accurately interpretable.
In the short answer section, when PCPs were asked about
factors that predispose the patient to high use, prominent responses
were limited resources and lack of social/home supports, low level
of education/illiteracy, patient and family anxiety, personality
traits and poor capacity. 16 PCPs responded that medical issues
were the root of the problem while 24 PCPs believed that social
factors additionally contributed, including lack of social supports
which was suggested 15 times (17% of 82 suggested social
factors), disability in 11 HUs (13% of all social factors), living
situation in 9 (11%), and income and income distribution in 9
(10%). Pulmonary and gastrointestinal conditions were the most
common reasons for HU visits to the PCP office. Other general
Figure 1. Most prevalent comorbidities amongst high-users (HUs).
Canadian Journal of General Internal Medicine
Volume 15, Issue 3, 2020 33
Original Article
CJGIM_3_2020_175414.indd 33CJGIM_3_2020_175414.indd 33 8/25/20 4:35 PM8/25/20 4:35 PM
Table 2. Primary Care Physician (PCP) Responses to the First 11 Multiple-
Choice Questions on the Survey
Multiple-Choice Questions
# of PCP
responses (%
responses out
of 49)
1. Are you still the Primary Care Provider for this
patient?
49 (100)
a. Yes 28 (57)
b. No, the patient transferred to another
provider
3 (6)
c. No, the patient never followed up 2 (4)
d. No, the patient is now deceased 13 (27)
e. No, for other reasons 3 (6)
2. If yes, for how long have you been this
patient’s GP?
44 (90)
a. < 1 year 5 (10)
b. 1-5 years 18 (37)
c. 6-10 years 8 (16)
d. 10-20 years 8 (16)
e. >20 years 4 (8)
f. Not applicable 1 (2)
No response provided 5 (10)
3. How many times has the patient visited your
clinic in the last year?
47 (96)
a. 0 16 (33)
b. 1-5 14 (29)
c. 6-10 5 (10)
d. 10-15 4 (8)
e. >15 2 (4)
Note: Home visits/nursing home/LTC 6 (12)
No response provided 2 (4)
4. Does this patient visit you within 2 weeks of a
hospital visit?
46 (94)
a. Always 7 (14)
b. Mostly 11 (22)
c. Occasionally 12 (24)
d. Rarely 6 (12)
e. Never 7 (14)
No response provided 3 (6)
Multiple-Choice Questions
# of PCP
responses (%
responses out
of 49)
5. Do you consistently receive discharge
summaries on the patient after hospital visits?
46 (94)
a. Always 18 (37)
b. Mostly 16 (33)
c. Occasionally 5 (10)
d. Rarely 3 (6)
e. Never 1 (2)
No response provided 3 (6)
6. Does the patient miss appointments with
you?
46 (94)
a. Never 20 (41)
b. Rarely 12 (24)
c. Occasionally 5 (10)
d. Frequently 5 (10)
e. Always 1 (2)
No response provided 3 (6)
7. Do you think this patient is managing chronic
diseases well?
45 (92)
a. Yes 20 (41)
b. Somewhat 4 (8)
c. No 8 (16)
d. Unsure 7 (14)
No response provided 4 (8)
8. Is the patient compliant with health
management recommendations?
45 (92)
a. Yes 31 (63)
b. Somewhat 5 (10)
c. No 5 (10)
d. Unsure 3 (6)
No response provided 4 (8)
Table 2. Primary Care Physician (PCP) Responses to the First 11 Multiple-
Choice Questions on the Survey (continued)
(continued )
(continued )
Canadian Journal of General Internal Medicine
34 Volume 15, Issue 3, 2020
Gantayet et al.
CJGIM_3_2020_175414.indd 34CJGIM_3_2020_175414.indd 34 8/25/20 4:35 PM8/25/20 4:35 PM
surprising that HUs have multiple chronic conditions, high
medication use, longitudinal patterns of high-use, and lower
socioeconomic status, we were surprised that most of the high-
users we identified had a regular PCP and lived at home. It is
also noteworthy that PCP perspectives on HUs generally reveal
positive health behaviors and attitudes, with 65% of responding
PCPs stating the patient rarely or never missed appointments.
41% of PCPs believed the HU was managing his/her chronic
diseases well, and only 10% thought the HUs were non-compliant
with prescribed therapies (Table 2). Third, the greatest number
of admissions of HUs were to GIM wards and GIM was also the
most frequently consulted service when patients were admitted to
Table 3. Compilation of Primary Care Physician (PCP) Recommendations for
Improvement
Targetable
Strategies
(Number of
PCPs who
made the
suggestion)
Summary of PCP recommendations
Higher Level of
Care (5)
move to supportive living
psychiatric-oriented facility
direct readmission to hospice
greater medical capacity at patient living
facilities
Better community
supports (7)
medication supports and better home care
supports for individuals with cognitive issues
early childhood development, aboriginal status
supports, family support, income support
Better access to
medical resources
(5)
home visits by RN/NPs, PCPs and/or specialists
24/7 access to PCPs
quicker access to urgent surgeries
patient’s local pharmacy to stock infrequently
used meds
Targeting patient
behaviors (8)
health education to better understanding
primary condition
mental health resources for personality
disorders, anxiety, and opioid dependency
encourage patients towards proactive clinic
visits and better communication with their
physicians
Administrative
interventions (4)
detailed patient information sheets for clinic
visits and hospital discharges
ensuring PCPs receive discharge summaries at
time of discharge
efforts to improve and enhance communication
between specialists and primary care
Multiple-Choice Questions
# of PCP
responses (%
responses out
of 49)
9. Have you discussed Goals of Care with this
patient or family?
44 (90)
a. Yes 22 (45)
b. Partially 2 (4)
c. No 14 (29)
d. Cannot recall 2 (4)
No response provided 5 (10)
10. Where does this patient live? 43 (88)
a. Home alone 8 (16)
b. Home with family/friends 26 (53)
c. Assisted living facility 3 (6)
d. Long-term care facility 3 (6)
e. Homeless shelter 0 (0)
f. Other 3 (6)
No response provided 6 (12)
11. What is the patient’s attitude to his/her
medical condition?
44 (90)
a. Aware and compliant 31 (63)
b. Aware but non-compliant 6 (12)
c. Unaware/indifferent 1 (2)
d. Lacks capacity 4 (8)
No response provided 5 (10)
*A total of 49 PCPs responded to surveys; some questions were skipped by some PCPs.
Table 2. Primary Care Physician (PCP) Responses to the First 11 Multiple-
Choice Questions on the Survey (continued)
issues that were noted by their PCP but not picked up during
chart reviews included insomnia, smoking cessation, pressure
ulcers, fractures, fertility, prescription renewal, and poor mobility.
When asked about recommendations for improvement, 14 PCPs
provided one or more suggestions; these have been grouped into
categories by the authors and presented in Table 3.
Discussion
There are 4 key findings in our work that deserve emphasis. First,
amongst HUs, 41% of ED visits and 21% of readmissions were
at other hospitals despite these patients having well-established
links to staff at the main study hospital. Second, while it is not
Canadian Journal of General Internal Medicine
Volume 15, Issue 3, 2020 35
Original Article
CJGIM_3_2020_175414.indd 35CJGIM_3_2020_175414.indd 35 8/25/20 4:35 PM8/25/20 4:35 PM
a surgical or subspecialty service. Fourth, published readmission
prediction scores have high predictive value in this cohort: at
the time of their last admission in the study period, 88% of HUs
had a high-risk LACE score (≥10) whereas 99% had a high-risk
FAM-FACE-SG score ≥14.
The high rate of readmission and ED visits to the non-
index hospital is concerning a lack of continuity is associated
with higher mortality in other studies.
10,11
For example, Staples
etal., 2014 found that 18% of patients who were re-admitted
to an alternative hospital in their study had a 3% higher 30-day
mortality rate.
10
While ambulance staff decide which hospital
to take a patient to based on ED census numbers, consideration
should be given to prioritize HUs so that they are preferentially
taken to the hospital (and physicians, nurses, system) most
familiar with them.
Our characterization of HUs is consistent with prior studies
in that we found that hypertension, depression, diabetes, and
obstructive lung disease were the most common comorbidities.
12
We found that only 10% of our HUs lived in LTC, which is
consistent with a report that 14% of HUs in another Canadian
province (Ontario) lived in LTC.
13
Moreover, our HUs tend to
have a longitudinal pattern of increased use, consistent with an
Ontario study that reported that 45% of adult HUs persisted
as HUs in the subsequent year.
3
Current and past healthcare
utilization are strongly associated with HUs indicating a need
for targeted interventions even for those patients with ongoing
continuity of care.
14
Although we identified HUs using a standard, CIHI-endorsed
definition and were able to track all of their hospitalizations and
ED visits anywhere in the province, there are some limitations
to our study. For one, there are multiple other definitions of
HUs in the literature based not on total LOS in one year, as used
by the CIHI definition, but instead based on healthcare costs,
3
population cost percentiles,
15
number of readmissions regardless
of LOS
8
and use of other healthcare resources. We recognize
that CIHI-defined HUs do not necessarily reflect those with the
highest hospital costs.
16
A second limitation is the possibility of
response bias in our PCP surveys. PCPs had an overall favorable
view of HUs. The 43% response rate may represent physicians
who are more diligent and actively involved in patient care.
Another limitation of our PCP surveys is that these were not done
prospectively soon after a readmission occurred and therefore
did not provide PCP feedback in real-time. The surveys did
not ask PCPs to rate the ‘preventability’ of that readmission or
identify factors/potential interventions that could have prevented
that specific readmission. Instead, the survey only asked in a
general sense about the factors contributing to high system use
or potential interventions to reduce unnecessary use. There is
also a selection bias in characterizing HUs as this study was
conducted in a tertiary care teaching hospital and is not reflective
of a community hospital. For example, the high readmission rates
for hematology and gastroenterology were likely due to the high
number of leukemia patients and the hepatology/liver transplant
service at this centre. It is thus important to take the healthcare
centre into account when generalizing admission patterns.
There is also a survivor bias in characterizing HUs as they had
to have at least three admissions and we did not include those
who died during their first or second hospitalization within the
study period. Additionally, while the CIHI definition includes
all patient admissions in a year, we applied it only to a single
hospital thereby missing out on hospitalizations at other centers
and thus under-capturing HU patients.
In our surveys, only 14 (29%) out of 49 PCPs provided
suggestions for improvement. Our finding that most PCPs
believed that HU readmissions were not preventable is consistent
with prior studies in this area. For example, a systematic review
of 34 studies found that the median proportion of readmissions
deemed avoidable was 27%, ranging from 5 to 79% between
studies
5
and in another similar survey of PCPs of high user
patients, 58% of PCPs believed that no interventions could
prevent readmissions.
17
Our study thus re-enforces the notion
of ‘unavoidability’ of high user readmissions. On the other hand,
a literature review suggested that 12 to 75% of readmissions
might be preventable by patient education and pre- and post-
discharge care
18
and a Kaiser Permanente case series found that
47% of 30-day readmissions might be preventable by targeting
transition and follow-up care planning, medication management
and advance care planning.
19
Thus, the healthcare arena is
gradually adopting a focus on transitions of care with a renewed
emphasis on patient-centered care. Ultimately, even though HU
readmissions are generally perceived as unavoidable, we must
continue to identify modifiable system factors to intervene on.
In our surveys, a prominent recommendation by 5 PCPs was
for higher levels of care and community supports including LTC,
supportive living, psychiatric-oriented facilities, and hospice, as
well as increased depth of medical care at these facilities. One PCP
stated, “need more capacity to manage episodes in her facility
as they will call the ambulance quite readily. There was also a
suggestion for “efforts to improve and enhance communication
between specialists and primary care” as only 34 (74%) PCPs stated
they mostly-always receive discharge summaries from hospital
admissions before HU follow-up visits. PCPs also notably suggested
increased mental health resources for patients with personality
disorders, depression, anxiety, and opioid dependency; we found
that depression and anxiety together accounted for 6.3% (63) of
all listed comorbidities (see Figure 1), psychiatry accounted for
6% of inpatient consults amongst HUs (Table S1) and 39% (49)
of HUs were on an antidepressant over the past year according
Canadian Journal of General Internal Medicine
36 Volume 15, Issue 3, 2020
Gantayet et al.
CJGIM_3_2020_175414.indd 36CJGIM_3_2020_175414.indd 36 8/25/20 4:35 PM8/25/20 4:35 PM
to chart reviews. Thus, an important intervention would be to
connect identified HUs with mental health resources and ensure
they have appropriate community supports.
A popular recommendation by PCPs was for home visits
whether by nurses, PCPs or even specialists; however, only 6 (13%)
of 46 PCPs stated that they do home/facility visits (ranging from
weekly to 1-4 times per year) (see Table 2). In the US, there has
been a push towards the multidisciplinary team-based Patient-
Centered Medical Home (PCMH) which provides a shift in focus
from acute rescue to proactive maintenance of chronic diseases.
20
The ‘Virtual Ward’ is another model which uses a predictive
tool to identify HUs at risk of frequent readmissions and then
provides intensive multidisciplinary ward-type care to these high
user patients in a community setting.
8, 20–26
A Toronto-based trial
by Dhalla et al. 2014 focused only on patients with a high LACE
score and found no reduction in readmission rates with intensive
post-discharge care transitions.
27
However, another Community
Virtual Ward model in Ireland designed to support older patients
with complex social and health care needs found a reduction in
ED presentations and unplanned hospital admissions using a
proactive integrated multidisciplinary approach.
24
Thus, future models should continue to incorporate a broader
definition of HUs, as characterized in our study, and should focus
on targeting social factors and preventative medicine in addition
to care transitions. Predictive tools that use electronic medical
records to extract information on patient characteristics are now
being developed and would further improve readmission predictive
analysis.
28
Hospital physicians can also play a role by ensuring
timely and effective communication with PCPs and connecting
high user patients with adequate mental health resources and
community supports. Potential next steps for hospital quality
improvement projects is to use predictive models to firstly
identify high-users and then engage identified HUs in system-
based interventions to study their effectiveness in improving
readmission outcomes. A future direction for our group is to
develop a ‘Virtual Hospital’ for our HU patient population and
investigate its effectiveness in reducing readmissions and ED visits.
Conclusion
In conclusion, our analysis of HUs identified characteristics that
are consistent with prior studies and recognized pre-existing
models. We found that most of the HUs we identified had a
regular PCP and lived at home with home care and family
supports. While HU patients were seen in all specialties, the
greatest number were admitted to GIM wards and GIM was the
most frequently consulted service when patients were admitted
to a surgical or subspecialty service. While PCP suggestions do
not necessarily correlate with specific HU characteristics, they do
provide a wealth of frontline experience that we can build on to
develop practical interventions. In keeping with prior literature,
PCP perspectives in this study reinforce the notion that many
high user readmissions are unavoidable and guide us to look
away from individual patient characteristics and instead to look
at system-based preventative and transitional care factors for
targetable interventions. Future studies must focus on predicting
HUs early on and studying the effect of suggested system-based
interventions on preventing their readmissions.
Funding/Support
Alberta Health Services Quality Innovation Fund.
Ethical Approval
Research Ethics Board at the University of Alberta Hospital
(Pro00073914).
Other Disclosures and Disclaimers
None.
Previous Presentations
None.
Acknowledgements
The authors wish to thank Alberta Health Services, Strategic
Clinical Improvement Committee for their valuable assistance and
access to resources. We thank the Data and Health Information
Resources for compiling data on HUs. We thank Xing Sun for
contacting the primary care physician offices. The authors thank
the PCPs who responded to our surveys.
References
1. Guilcher SJT, Bronskill SE, Guan J, et al. Who are the high-users? A
method for person-centred attribution of health care spending. PLoS ONE
[Electronic Resource]. 2016;11(3).
2. Lown BA, McIntosh S, Gaines ME, et al. Integrating compassionate,
collaborative care (the “triple C”) into health professional education to
advance the triple aim of health care. Acad Med 2016;91(3):310–16.
3. Wodchis WP, Austin PC, Henry DA. A 3-year study of high-users of health
care. CMAJ 2016;188(3):182–88.
4. Canadian Foundation for Healthcare Improvement. [Internet]. Available at:
https://www.cfhi-fcass.ca. Revised February 5, 2019. Accessed February 7,
2019.
5. Van Walraven C, Bennett C, Jennings A, et al. Proportion of hospital
readmissions deemed avoidable: A systematic review. CMAJ 2011;183(7):402.
6. Bardsley M, Lewis G. Reflections from the NHS in England. Healthc Papers
2014;14(2):26–30.
7. Billings J, Dixon J, Mijanovich T, et al. Case finding for patients at risk of
readmission to hospital: Development of algorithm to identify high risk
patients. BMJ 2006;333(7563):327–30.
8. Low LL, Liu N, Lee KH, et al. FAM-FACE-SG: A score for risk stratification
of frequent hospital admitters. BMC Med Inform Decis Mak 2017;17:35.
9. Charlson ME, Pompei P, Ales KL, et al. A new method of classifying
prognostic comorbidity in longitudinal studies: Development and validation.
J Chronic Dis 1987;40(5):373–83.
Canadian Journal of General Internal Medicine
Volume 15, Issue 3, 2020 37
Original Article
CJGIM_3_2020_175414.indd 37CJGIM_3_2020_175414.indd 37 8/25/20 4:35 PM8/25/20 4:35 PM
10. Staples JA, Thiruchelvam D, Redelmeier DA. Site of hospital readmission
and mortality: A population-based retrospective cohort study. CMAJ Open
2014;2(2):77.
11. McAlister FA, Youngson E, Kaul P. Patients with heart failure readmitted to
the original hospital have better outcomes than those readmitted elsewhere. J
Am Heart Assoc 2017;6(5).
12. Chechulin Y, Nazerian A, Rais S, et al. Predicting patients with high risk of
becoming high-cost healthcare users in Ontario (Canada). Healthc Policy
2014;9(3):68–79.
13. Rosella LC, Fitzpatrick T, Wodchis WP, et al. High-cost health care users
in Ontario, Canada: Demographics, socio-economic, and health status
characteristics. BMC Health Serv Res 2014;14:532.
14. Condelius A, Hallberg IR, Jakobsson U. Hospital and outpatient clinic
utilization among older people in the 3-5 years following the initiation
of continuing care: A longitudinal cohort study. BMC Health Serv Res
2011;11:136.
15. Rais S, Nazerian A, Ardal S, et al. High-users of Ontarios healthcare services.
Healthc Policy 2013;9(1):44–51.
16. Nguyen OK, Tang N, Hillman JM, et al. What’s cost got to do with it?
Association between hospital costs and frequency of admissions among
“high-users” of hospital care. J Hosp Med (Online). 2013;8(12):665–71.
17. Gantayet A, Ang M, Cao X, et al. Persistent and non-persistent high-users
of acute care resources: A deeper dive into the patient and system factors.
Healthc Q 2017; 20(2):31–4.
18. Benbassat J, Taragin M. Hospital readmissions as a measure of
quality of health care: Advantages and limitations. Arch Intern Med
2000;160(8):1074–81.
19. Feigenbaum P, Neuwirth E, Trowbridge L, et al. Factors contributing to all-
cause 30-day readmissions: A structured case series across 18 hospitals. Med
Care 2012;50(7):599–605.
20. Ortiz G, Fromer L. Patient-centered medical home in chronic obstructive
pulmonary disease. J Multidiscip Healthc 2011;4:357–65.
21. Lewis G, Vaithianathan R, Wright L, et al. Integrating care for high-risk
patients in England using the virtual ward model: Lessons in the process of
care integration from three case sites. Int J Integr Care 2013;13:e046.
22. Lewis G, Bardsley M, Vaithianathan R, et al. Do ‘virtual wards’ reduce rates
of unplanned hospital admissions, and at what cost? A research protocol
using propensity matched controls. Int J Care Pathw [Electronic Resource]
2011;11:e079.
23. Leung DYP, Lee DT, Lee IFK, et al. The effect of a virtual ward program on
emergency services utilization and quality of life in frail elderly patients after
discharge: A pilot study. Clin Interv Aging 2015;10:413–420.
24. Lewis C, Moore Z, Doyle F, et al. A community virtual ward model to
support older persons with complex health care and social care needs. Clin
Interv Aging 2017;12:985–993.
25. Jones J, Carroll A. Hospital admission avoidance through the introduction of
a virtual ward. Br J Community Nurs 2014;19(7):330–334.
26. Pacho C, Domingo M, Nunez R, et al. Early postdischarge STOP-HF-clinic
reduces 30-day readmissions in old and frail patients with heart failure. Rev
Esp Cardiol 2017;70(8):631–38.
27. Dhalla IA, O’Brien T, Morra D, et al. Effect of a postdischarge virtual ward
on readmission or death for high-risk patients: A randomized clinical trial.
JAMA 2014;312(13):1305–12.
28. Zolbanin HM, Delen, D. Processing electronic medical records to improve
predictive analytics outcomes for hospital readmissions. Decis Support Syst
2018;112:98–110.
Appendix
Supplementary Table S1. Rates of Admission and Consultation
Amongst Medical Services
Service
% of HU
admissions
(# of
admissions)
Ratio of #
admissions
to # HUs
accounting for
admissions
over the year
% of
specialties
consulted
on
inpatient
HUs (# of
consults)
General
Internal
Medicine
28 (185) 3.1 (185/59) 13 (39)
Surgery 17 (116) 2.3 (116/51) 16 (46)
Pulmonary 16 (104) 4.2 (104/25) 9 (28)
Hematology 15 (100) 5.3 (100/19) 6 (18)
Gastroenterology
15 (98) 3.3 (98/30) 12 (34)
Fam Med 6 (42) 1.8 (42/24) 3 (8)
Nephrology 3 (17) 3.4 (17/5) 2 (7)
Cardiology 1 (5) 1.0 (5/5) 2 (6)
Direct
admission to
ICU/CCU
0 (3) 1.5 (2/3) Not available
Geriatrics 0 (2) 1.0 (2/2) 0 (1)
Psychiatry 0 (1) 1.0 (1/1) 6 (19)
Neurology 0 (0) N/A 2 (5)
Infectious
Diseases
N/A N/A 9 (26)
Palliative care N/A N/A 5 (14)
Transplant N/A N/A 4 (12)
Pain service N/A N/A 2 (5)
Endocrinology N/A N/A 1 (2)
Rheumatology N/A N/A 0 (1)
Canadian Journal of General Internal Medicine
38 Volume 15, Issue 3, 2020
Gantayet et al.
CJGIM_3_2020_175414.indd 38CJGIM_3_2020_175414.indd 38 8/25/20 4:35 PM8/25/20 4:35 PM
Supplementary Table S2. Survey Faxed to the Primary Care Physicians (PCPs) of High-Users (HUs)
Dear Dr. ----------,
Mr/Ms. --------- is a high-cost user of the University of Alberta Hospital acute care services and has documented you as his/her family physician.
Can you confirm that Mr/Ms. ---------- has ever been your patient?
 NO, I have never seen this patient.
Thank you. You have completed this survey.
Please fax the questionnaire back to us at ---------------.
 YES. Then please answer the subsequent questions as best as you can.
1. Are you still the primary care provider for this patient?
a. Yes
b. No, the patient transferred to another provider
c. No, the patient never followed up
d. No, the patient is now deceased
e. No, for other reasons
2. If yes, for how long have you been this patient’s GP?
a. < 1 year
b. 1-5 years
c. 6-10 years
d. 10-20 years
e. >20 years
f. Not applicable
3. How many times has the patient visited your clinic in the last year?
a. 0
b. 1-5
c. 6-10
d. 10-15
e. >15
Canadian Journal of General Internal Medicine
Volume 15, Issue 3, 2020 39
Original Article
CJGIM_3_2020_175414.indd 39CJGIM_3_2020_175414.indd 39 8/25/20 4:35 PM8/25/20 4:35 PM
4. Does this patient visit you within 2 weeks of a hospital visit?
a. Always
b. Mostly
c. Occasionally
d. Rarely
e. Never
5. Do you consistently receive discharge summaries on the patient after hospital visits?
a. Always
b. Mostly
c. Occasionally
d. Rarely
e. Never
6. Does the patient miss appointments with you?
a. Never
b. Rarely
c. Occasionally
d. Frequently
e. Always
7. Do you think this patient is managing his/her chronic diseases well?
a. Yes
b. Somewhat
c. No
d. Unsure
e. Not applicable
8. Is the patient compliant with health management recommendations?
a. Yes
b. Somewhat
c. No
d. Unsure
e. Not applicable
Canadian Journal of General Internal Medicine
40 Volume 15, Issue 3, 2020
Gantayet et al.
CJGIM_3_2020_175414.indd 40CJGIM_3_2020_175414.indd 40 8/25/20 4:35 PM8/25/20 4:35 PM
9. Have you discussed Goals of Care with this patient or his/her family?
a. Yes
b. Partially
c. No
d. Cannot recall
e. Not applicable
10. Where does this patient live?
a. Home alone
b. Home with family/friends
c. Assisted living facility
d. Long-term care facility
e. Homeless shelter
f. Other
11. What is the patient’s attitude to his/her medical condition?
a. Aware and compliant
b. Aware but non-compliant
c. Unaware/indifferent
d. Lacks capacity
e. Other
12. What is the patient’s attitude to you?
a. Very good
b. Good
c. Neutral
d. Poor
e. Very poor
13. What top 4 medical conditions does the patient visit you for?
14. What top 4 specialists are enlisted in this patient’s care?
15. What top 4 outpatient clinics is this patient followed in?
16. What top 4 community supports are enlisted in this patient’s care?
17. What interventions, if any, could prevent this patient’s Emergency Department revisits and inpatient readmissions?
18. What patient factors, if any, predispose this patient to high-cost healthcare use?
Canadian Journal of General Internal Medicine
Volume 15, Issue 3, 2020 41
Original Article
CJGIM_3_2020_175414.indd 41CJGIM_3_2020_175414.indd 41 8/25/20 4:35 PM8/25/20 4:35 PM
19. Can you identify any ‘Social Determinants of Health’ that influence this patient’s high-cost use? Select all that apply.
• Income and Income Distribution
• Education
• Unemployment and Job Security
• Employment and Working Conditions
• Early Childhood Development
• Food Insecurity
• Housing
• Social Exclusion
• Social Safety Network
• Health Services
• Aboriginal Status
• Gender
• Race
• Disability
• Living situation
• Social supports
• Other:
20. Is there anything else you would like to share?
Figure S1. Admitting diagnosis based on International Statistical Classification of Disease (ICD) codes as a
percentage of all diagnoses.
Canadian Journal of General Internal Medicine
42 Volume 15, Issue 3, 2020
Gantayet et al.
CJGIM_3_2020_175414.indd 42CJGIM_3_2020_175414.indd 42 8/25/20 4:35 PM8/25/20 4:35 PM