Patient related factors in frequent readmissions: the influence of condition, access to services and patient choice.

Abstract

People use emergency department services for a wide variety of health complaints, many of which could be handled outside hospitals. Many frequent readmissions are due to problems with chronic disease and are preventable. We postulated that patient related factors such as the type of condition, demographic factors, access to alternative services outside hospitals and patient preference for hospital or non-hospital services would influence readmissions for chronic disease. This study aimed to explore the link between frequent readmissions in chronic disease and these patient related factors.
A retrospective analysis was performed on emergency department data collected from a regional hospital in NSW Australia in 2008. Frequently readmitted patients were defined as those with three or more admissions in a year. Clinical, service usage and demographic patient characteristics were examined for their influence on readmissions using multivariate analysis.
The emergency department received about 20,000 presentations a year involving some 16,000 patients. Most patients (80%) presented only once. In 2008 one hundred and forty four patients were readmitted three or more times in a year. About 20% of all presentations resulted in an admission. Frequently readmitted patients were more likely to be older, have an urgent Triage classification, present with an unplanned returned visit and have a diagnosis of neurosis, chronic obstructive pulmonary disease, dyspnoea or chronic heart failure. The chronic ambulatory care sensitive conditions were strongly associated with frequent readmissions. Frequent readmissions were unrelated to gender, time, day or season of presentation or country of birth.
Multivariate analysis of routinely collected hospital data identified that the factors associated with frequent readmission include the type of condition, urgency, unplanned return visit and age. Interventions to improve patient uptake of chronic disease management services and improving the availability of alternative non-hospital services should reduce the readmission rate in chronic disease patients.

Full text

Page 1

RESEARCH ARTICLE Open Access
Patient related factors in frequent readmissions:
the influence of condition, access to services
and patient choice
Sue E Kirby*, Sarah M Dennis, Upali W Jayasinghe, Mark F Harris
Abstract
Background: People use emergency department services for a wide variety of health complaints, many of which
could be handled outside hospitals. Many frequent readmissions are due to problems with chronic disease and are
preventable. We postulated that patient related factors such as the type of condition, demographic factors, access
to alternative services outside hospitals and patient preference for hospital or non-hospital services would influence
readmissions for chronic disease. This study aimed to explore the link between frequent readmissions in chronic
disease and these patient related factors.
Methods: A retrospective analysis was performed on emergency department data collected from a regional
hospital in NSW Australia in 2008. Frequently readmitted patients were defined as those with three or more
admissions in a year. Clinical, service usage and demographic patient characteristics were examined for their
influence on readmissions using multivariate analysis.
Results: The emergency department received about 20,000 presentations a year involving some 16,000 patients.
Most patients (80%) presented only once. In 2008 one hundred and forty four patients were readmitted three or
more times in a year. About 20% of all presentations resulted in an admission. Frequently readmitted patients were
more likely to be older, have an urgent Triage classification, present with an unplanned returned visit and have a
diagnosis of neurosis, chronic obstructive pulmonary disease, dyspnoea or chronic heart failure. The chronic
ambulatory care sensitive conditions were strongly associated with frequent readmissions. Frequent readmissions
were unrelated to gender, time, day or season of presentation or country of birth.
Conclusions: Multivariate analysis of routinely collected hospital data identified that the factors associated with
frequent readmission include the type of condition, urgency, unplanned return visit and age. Interventions to
improve patient uptake of chronic disease management services and improving the availability of alternative non-
hospital services should reduce the readmission rate in chronic disease patients.
Background
Emergency departments aim to provide treatment for
more urgent and serious conditions. In Australia, the
role of emergency departments has been specified as
“prevention, diagnosis and management of acute and
urgent aspects of illness and injury affecting patients of
all age groups with a full spectrum of undifferentiated
physical and behavioural disorders”[1]. However, many
people who access emergency departments have
complex social needs as well as a clinical condition
requiring treatment [2-4]. Emergency departments
across the world are reporting serious overcrowding
resulting in lengthy waits which impair the quality of
care and patient outcomes [5-7]. Investigation of patient
profiles and the reasons behind the choice of emergency
department services by frequent attenders and fre-
quently readmitted patients is vital to guide the develop-
ment of policy and design of interventions to address
more appropriate patient management strategies and to
prevent overcrowding [8].
Overcrowding of emergency departments has spawned
a plethora of research on frequent emergency
* Correspondence: s.kirby@student.unsw.edu.au
Centre for Primary Health Care & Equity, School of Public Health &
Community Medicine, University of New South Wales, UNSW, Sydney 2052,
NSW, Australia
Kirby et al. BMC Health Services Research 2010, 10:216
http://www.biomedcentral.com/1472-6963/10/216
© 2010 Kirby et al; licensee BioMed Central Ltd. This is an Open Access article distributed under the terms of the Creative Commons
Attribution License (http://creativecommons.org/licenses/by/2.0), which permits unrestricted use, distribution, and reproduction in
any medium, provided the original work is properly cited.

Page 2

department users. Byrne et al [2] reported that frequent
emergency department users were more likely to be
males from low socioeconomic backgrounds with severe
psychosocial problems who have a high morbidity and
mortality. There have been similar findings from
researchers in the UK [9-12], Canada [12,13], Sweden
[14-16], USA [17-25], Italy [26], Taiwan [27,28] and
Australia [29,30].
Frequently readmissions, however, have received less
research attention. The Patients At Risk of Readmission
(PARR) tool, used extensively in the UK for identifica-
tion of patients for case management to prevent read-
missions, [31,32] indicates that age, sex, ethnicity,
number of previous admissions, and clinical condition
are associated with readmissions. Howell et al [33] in a
2009 study in Australia identified age, co-morbidities,
economic disadvantage, number of previous admissions
as risk factors for frequent readmissions through a sta-
tistical algorithm derived from inpatient data. The Aus-
tralian model was only moderately successful because of
the relatively high number of false negatives.
Andersen’s model of health service utilisation [34-36]
provides a theoretical framework for thinking about why
patients are frequently readmitted. This model postu-
lates that health service utilisation is dependent on a
range of factors such as the environment, population
characteristics and health behaviours resulting in out-
comes such as perceived health status, evaluated health
status and consumer satisfaction. The environment
includes available services and access issues. Population
characteristics include demographic details, age, gender,
cultural background, needs and resources. Health beha-
viours are influenced by personal attitudes and beliefs.
Identifying the clinical, service utilisation details and
demographic patient characteristics available from hos-
pital data can identify risk factors for frequent readmis-
sions and thereby help to develop strategies to divert
these patients away from hospitals where clinically pos-
sible. The direct application of this model to emergency
department utilisation summarised by Padgett et al
[37,38], who suggests that only about 15% of emergency
visits in the USA are for life threatening reasons and
that poor mental health, anxiety about health combined
with symptoms and injury influence people to decide to
use the emergency department.
How important is access to primary care in readmis-
sions? Hospitalisations for ambulatory care sensitive
(ACS) conditions is used an outcome indicator for
access to primary care [39]. It is acknowledged that
treatment of ACS conditions outside hospitals, including
chronic disease, reduces hospitalisations [40]. It would
be expected that ACS conditions would contribute to
frequent readmissions in areas with limited access to
primary care services.
The problem that a small number of emergency
department users account for a disproportionate use of
scarce resources has been reported since the 1980s
[41,42] and continues to plague the health system [43].
An Australian survey of emergency departments indi-
cated that patients awaiting admission are a major con-
tributing factor in emergency department overcrowding
[44]. Two Australian States, NSW and Victoria, report
diversion of chronic disease patients to chronic disease
management services, including those provided by
primary care, can reduce readmission rates [25,45].
We know little about the reasons behind the choice of
emergency department services over other alternative
medical services in the community. However, studies in
NSW Australia have reported that patients came to the
emergency department because they thought their con-
dition warranted urgent services, with access to a doctor
and tests or x-rays done in the same place but there was
a mismatch between the views of clinicians and patients
[46,47]. In a qualitative study performed in the USA
[48], patients reported that they were unable to obtain
an appointment with a primary care provider; were
referred by the staff to be evaluated in the emergency
department; and it took less of their time to be seen in
the emergency department than it did to contact their
primary care provider, only to then be told to go to the
emergency department.
Fulde and Duffy[3] argued that patients who present
frequently to emergency departments are a vulnerable
and marginalised group perceived as not looking after
themselves. Padgett et al [38] suggest that the emer-
gency department is the “least appropriate setting for
treating multiple co-morbidities”. These authors also
reported that people who feel powerless and isolated
tend to be high emergency department users.
Hong et al [49] suggest that we need a radical para-
digm shift to legitimise the role of emergency depart-
ments to include primary care services for those people
on the lower end of the socioeconomic scale who tend
not to have a primary care provider in the community.
The counter opinion would be to restrict the role of
emergency departments and to increase the availability
of alternative primary care services and to thereby
reduce emergency department overcrowding.
This research analysed emergency department data
routinely collected by hospitals in the state of NSW,
Australia, to identify patient related factors associated
with three or more admissions a year and with admis-
sion for ACS conditions. The analysis was designed to
answer questions about the impact on frequent readmis-
sions of three sets of variables: demographic, clinical
and arrival times and dates. The answers provide a basis
for exploring the reasons underlying frequent readmis-
sions. The research sets the scene for further
Kirby et al. BMC Health Services Research 2010, 10:216
http://www.biomedcentral.com/1472-6963/10/216
Page 2 of 8

Page 3

examination of the reasons why people with chronic
conditions are repeatedly admitted to hospital rather
than seeking chronic disease management services.
Methods
Study design and site
In this study, we analysed 2008 data from the Emer-
gency Department Information System collected at a
regional hospital. The University of Wollongong/South
Eastern Sydney Illawarra Health Service Medical
Human Research Ethics Committee approved the
research study (approval HE07/271). All patient data
were de-identified. The study was carried out in a one
hundred and fifty bed regional hospital in south eastern
Australia which is part of a network with a major teach-
ing and referral hospital. The hospital, funded by the
State and Commonwealth Government, is located in a
coastal regional urban area with a feeder population of
around 100,000.
Data Analysis
All statistical analyses were performed using SPSS statis-
tical software (Version18; SPSS, Chicago, Illinois, USA).
Two-sided P values of less than 0.05 were considered
statistically significant.
Predictor variables
The original variables were: patient identifier number,
age, gender, date and time of arrival, country of birth,
ICD 9 diagnosis, level of urgency based on the Austra-
lian Triage Classification [50], type of visit and mode of
separation (treated in emergency or admitted). Type of
visit is a local code referring to the type of emergency
presentation: normal, planned return and unplanned
return within 28 days of the original presentation. Each
patient presentation was accorded one primary ICD 9
diagnosis based on the major reason for the presenta-
tion. Date and time of arrival variables were converted
into hour, day and season of arrival.
The data was analysed either from the perspective of
presentations in row-for-each-presentation format or
from the perspective of patients in a row-for-each-
patient format with nested data for each patient. The
primary research question to uncover the patient char-
acteristics associated with frequent readmissions
required the row-for-each-patient format. However, pre-
liminary univariate analysis was performed on the row-
for-each-presentation format on the variables for which
there were different or potentially different, values for
each presentation: hour, day and season of arrival;
urgency; unplanned return visit and diagnosis. Univari-
ate analysis was also performed with the variables which
were constant for each presentation (age, gender and
country of birth).
Primary outcome variable
The primary outcome variable was a binary variable of
the number of admissions set with a cut off point at
three or more admissions denoting frequent readmis-
sions. Presenting patients judged to have conditions
warranting admissions may be either admitted to the
study hospital or to another hospital in the network if
the bed supply is inadequate. All admissions, including
those for which patients were placed in a bed in another
hospital, were included in the sample. The number of
admissions variable was obtained by separating the pre-
sentations for patients admitted, converting to the row-
for-each-patient format and merging the admissions
variable in to the main file.
Regression models and manipulation of predictor
variables
Two logistic regression models for multivariate analysis
were constructed with the binary dependent variable.
The independent variables were transformed by count-
ing the number of occurrences in the nested variables
for each patient and dividing by the total number of
presentations for that patient resulting in a variable of
the proportion of occurrences and used in the multivari-
ate analysis. As the values in the proportion variables
ranged from zero to one, they were treated as continu-
ous covariates. The independent diagnosis variables
included the proportion of the most common ICD 9
diagnosis codes for presentations resulting in frequent
readmissions: neurosis, chronic heart failure, chronic
obstructive pulmonary disease (COPD), dyspnoeas and
chest pain.
In the second logistic regression model was created to
determine the impact of diagnoses which are deemed
suitable for non-hospital services. Ambulatory care sen-
sitive conditions are defined as those which hospitalisa-
tion is considered potentially avoidable through
preventive care and early disease management, usually
delivered in an ambulatory setting, such as primary
health care (for example by general practitioners or
community health centres). For the second logistic
regression model, the independent continuous propor-
tion-of-diagnoses variables based on the ACS groupings:
rapid onset conditions and chronic conditions [40]. were
substituted for the continuous ICD 9 diagnoses propor-
tion variables in the first model.
Logistic regression was considered the best option for
analysis of the possible predictor variables because the
dependent variable demonstrated Poisson distribution,
with the variance higher than the mean, rendering linear
regression models inappropriate [51,52]. However, there
has been debate in the literature about the impact of
intra-patient variability in emergency department pre-
sentations on the choice of statistical analysis. This has
Kirby et al. BMC Health Services Research 2010, 10:216
http://www.biomedcentral.com/1472-6963/10/216
Page 3 of 8

Page 4

led to questioning of the appropriateness of logistic
regression and the proposition that negative binomial
regression should be used [53,54]. The results of the
logistic regression were therefore checked against those
of negative binomial regression. A backward elimination
approach was adopted in the interest of achieving a par-
simonious model [55,56]. The categorical variables, male
gender and born in Australia both met the criterion of a
minimum of ten cases in each category.
Results
There were 21,956 presentations in 2008, of which 20%
were admitted. Results of the univariate analysis on
patient data in Table 1 indicate that there were signifi-
cant differences between the general patient population
and the frequent readmissions group in age, country of
birth and gender. Table 2, presentation data, comparing
the frequencies of the independent proportion variables
in the general presenting population and presentations
resulting in frequently readmissions, shows there were
significant differences in all of the five most common
diagnosis codes amongst frequently readmitted presenta-
tions and the “other” diagnoses, in urgent presentations,
unplanned return visits and the chronic and “other”
ambulatory care sensitive variables.
Of the seasonal variables, only in summer was there a
significant difference between the two groups. There
were no differences observed in time or day of arrival.
Direct logistic regression uncovered the impact of the
independent variables on frequent readmissions. Both
models were initially set up with the independent vari-
ables gender, age, country of birth, and the proportion
variables unplanned return visit, urgency, season, week-
end, after hours and the diagnosis. Subsequently, step-
wise backward elimination of the variable with the
highest p value was performed to establish the most
parsimonious model for the independent variables. The
first model including the “proportion” variables for the
five most common ICD 9 diagnoses was statistically sig-
nificant with a c2value of 278.3 (7 degrees of freedom,
N = 15806, p < 0.01) indicating the model was able to
distinguish between frequently readmitted and patients
not frequently readmitted based on the definition of fre-
quent admissions as three or more. As a whole, the ICD
9 diagnoses model explained up to 17.1% (R2Nagelkerk
value) of the variance [57,58] in the dependent variable.
As Table 3 reveals, seven of the independent variables,
age, proportion of unplanned return visits, urgent, diag-
noses of neurosis, chronic heart failure, COPD and dys-
pnoeas, made a statistically significant (p < 0.05) positive
contribution to the model.
In the second model, a similar logistic regression ana-
lysis was performed substituting three variables of the
proportion of preventable, rapid onset and chronic ACS
conditions for the diagnosis variables in the first model.
The other independent variables were similar to those
used for model one. Overall, the ACS model was statis-
tically significant with a c2value of 192.8 (4 degrees of
freedom, N = 15806, p < 0.01) indicating that the model
could distinguish between frequently readmitted patients
and non-frequently readmitted patients. As a whole, the
ACS model explained up to 12.3% (Nagelkerk R2value
[57]) of the variance of the dependent variable. In model
two, four of the independent variables age, proportion of
unplanned return visits, urgency and proportion of ACS
chronic conditions made a statistically significant (p <
0.05) positive contribution to the model (see Table 4
The results of negative binomial regression models con-
firmed the logistic regression models.
Discussion
The majority of presentations to the emergency depart-
ment were by patients who had a single presentation in
the year. The patients under investigation in this study
were the one hundred and forty four frequently read-
mitted patients who had three or more admissions in a
year. The univariate analysis showed significant differ-
ences in all the independent variables except the seasons
of autumn, winter and spring and time and day of arri-
val. However, the multivariate analysis identified age,
chronic conditions of neurosis, COPD, dyspnoeas and
chronic heart failure, ACS chronic conditions, urgency
and unplanned return visit as being associated with fre-
quent readmissions. The study highlights the advantage
of the multivariate analysis using the patient rather than
presentation perspective to separate factors important in
frequent readmissions.
Using the framework of Andersen’s model of health
service utilisation, the results can be considered in
terms of patient demographics, clinical significance,
Table 1 Univariate analysis of patient characteristics comparing the general patient population with frequently
readmitted patients 2008
All patients (n = 15,806)Patients admitted = >3 times (n = 144) 1% of all patients
48% p < 0.051
66 (21, 2-96) p < 0.05
27% p < 0.051
Males
Mean age (SD, range)
Born outside Australia
52%
39 (26, 1-106)
16%
1Chi square results
Kirby et al. BMC Health Services Research 2010, 10:216
http://www.biomedcentral.com/1472-6963/10/216
Page 4 of 8

Page 5

access to services and individual patient preferences.
The finding that age significantly influenced frequent
readmissions is in line with other studies [31-33] These
studies also reported that other demographic factors,
sex and ethnicity, are associated with readmissions.
However, there was no discernible sex difference, nor
was there a difference between Australian born and
those born outside Australia, in our study.
The type of condition is important in readmissions.
Our model looked specifically at the most common con-
ditions in presentations by patients admitted to a
hospital bed. The diagnoses of neurosis, COPD, dys-
pnoeas and chronic heart failure, all serious conditions,
were associated with frequent readmissions. Chest pain
in our model was not associated with frequent readmis-
sions. Although the regression models did not include a
variable for severity of illness per se, other variables
point to severity. The triage level specifically indicates
urgency and more urgent conditions are generally more
severe.
Unplanned return visits within 28 days of the previous
visit for the same condition are reported by all public
Table 2 Univariate analysis independent variables in the general presenting population compared to presentations
resulting in frequent readmissions 2008
Variable groupVariable General presenting
population N = 21956
ICD9 diagnostic code2
Dyspnoeas (ICD 78609)
2.3%
Neurotic disorder (ICD 3079)
3.4%
Chest pain (ICD 78650)
4.4%
Chronic obstructive pulmonary
disease (ICD 496)
Congestive heart failure (ICD
4280)
All other ICD codes
89.0%
Triage categoryUrgent (Triage 1 & 2)
5.9%
Visit typeUnplanned return visit
4.3%
Season of arrival Autumn
24.9%
Winter
24.3%
Spring
25.0%
Summer
25.8%
Day of arrival Weekend
31.2%
Time of arrivalAfter hours
49.0%
Ambulatory Care Sensitive
Conditions4
ACS chronic
1.3%
ACS other
94.8%
Presentations resulting in frequently
readmissions N = 227
15.9% p < 0.013
10.6% p < 0.01
9.7% p < 0.01
7% p < 0.01 0.7%
0.3% 2.6% p < 0.01
54.2% p < 0.01
15.4% p < 0.01
8.8% p < 0.01
30.4% NS
28.6% NS
28.6% NS
12.3% p < 0.01
28.2% NS
44.9% NS
4.8% NS
ACS rapid onset
3.9%
7.5% P < 0.01
87.7% P < 0.01
2There were 1004 other ICD 9 codes with at least one presentation.
3Chi square results.
4ACS preventable was excluded because there was only one presentation in the general presenting population and none in the presentations resulting in
frequent readmissions group.
Table 3 Odds ratio and confidence limits for the variables having a significant impact on frequent readmissions in
model one five most common ICD 9 diagnoses, results from logistic regression backwards elimination
Variable P value Odds ratio95% Confidence Interval for odds ratio
LowerLower
Age (years)
Proportion of urgent (Triage categories 1 and 2)
Proportion of neuroses (ICD 9 code 3079)
Proportion of congestive heart failure (ICD 9 code 4280)
Proportion of chronic obstructive pulmonary disease (ICD 9 code 496)
Proportion of dyspnoeas (ICD 9 code 78609)
Proportion of unplanned return visits within 28 days of previous visit
Constant
.000
.001
.000
.027
.000
.000
.000
.000
1.046
2.363
13.430
4.720
8.807
6.095
14.381
.000
1.036
1.416
7.435
1.189
3.857
3.425
5.487
1.055
3.942
24.258
18.744
20.107
10.845
37.693
Kirby et al. BMC Health Services Research 2010, 10:216
http://www.biomedcentral.com/1472-6963/10/216
Page 5 of 8

Page 6

hospitals on a monthly basis to the NSW Department of
Health. This indicator is considered to be important in
assessing the quality of hospital care [59] on the basis
that the underlying reason for the return visit may be
related to premature discharge. Although the incidence
of unplanned return visits is low, this is a predictor of
frequent readmission and therefore worthy of further
detailed investigation. An early warning system to flag
unplanned return visits would assist in addressing their
needs in a more timely way and reducing the burden on
emergency services. Our findings in relation to
unplanned return visits are in line with those reported
for a US study on the Triage Risk Screening Tool [60].
The finding that frequent readmissions are associated
with diagnoses of COPD, dyspnoeas and chronic heart
failure is consistent with the notion that patients experi-
encing the symptoms associated with these conditions
are more likely to believe their condition is serious
enough for them to choose hospital care rather than pri-
mary care outside the hospital setting.
Seasonal variations in arrivals might have been
expected because of the differences in temperature influ-
encing the incidence of respiratory exacerbations. The
fact that there were no seasonal variations is difficult to
explain.
Patients with ACS chronic conditions were signifi-
cantly more likely to be frequent readmissions. This
finding could mean that access to primary care services
for preventable conditions (immunisation and nutri-
tional interventions) and rapid onset ACS conditions is
adequate and access to chronic disease management ser-
vices is not. Alternatively, the findings could indicate
that patients are choosing hospital services over primary
care services for chronic disease management.
Frequent readmissions for neurosis might be asso-
ciated with reduced access to community based mental
health services in the area. Another interpretation of
this finding is that patients are voting with their feet
expressing a preference for hospital services over com-
munity care.
Does poor access to other services contribute to fre-
quent readmissions? Time and day of arrival had no sig-
nificant impact on frequent readmissions. It might be
expected that the more frequent emergency department
users would opt for after hours or weekend visits when
other community based services are less available. Our
findings tend to negate any notion that access to after-
hours and weekend services is an issue. However, the
finding that ACS chronic conditions were associated
with frequent readmissions suggests there are access
issues at play. Although alternative general or specialist
medical and chronic disease management services exists,
there may be access difficulties other than the time of
availability of services. Access factors such as, availabil-
ity, accessibility, accommodation, affordability and
acceptability, need to be further explored before we
have a definitive answer on the impact on frequent
readmissions. In particular, transport, location of com-
munity-based services and waiting times would act as
barriers to access.
In our analysis, age and type of illness influenced
readmissions. The impact of access issues on frequent
readmissions is less clear. The other factors which
could influence readmissions are related to patient pre-
ferences for the type of services, for example, general
and specialist medical and chronic disease manage-
ment services. If the uptake of these alternative ser-
vices were to be increased, readmissions rates would
be reduced. Further research to explore patient choice
of services would be of benefit to the problem of fre-
quent readmissions.
On the basis of the findings presented in this study,
the question of whether readmissions are preventable
was not directly answered, but the evidence from other
studies shows that hospitalisation of chronic disease
patients can be reduced by a range of targeted interven-
tions to improve chronic disease management [45].
Logistic regression models were developed to identify
the characteristics associated with frequent readmis-
sions. The reliability of the predicting factors identified
is determined by the robustness of the models. Although
the overall fit of the model was significant, the Nagelk-
erk R2values of less than 20% were relatively low. How-
ever, other studies of patient utilisation and clinical data
have quoted similarly low Nagelkerk R2values in logistic
regression models [61,62].
Table 4 Odds ratio and confidence limits for the variables having a significant impact on frequent readmissions in
model two ambulatory care sensitive conditions, results from logistic regression backwards elimination
Variable P valueOdds ratio95% Confidence Interval for odds ratio
LowerUpper
Age (years)
Proportion urgent (Triage 1 & 2)
Proportion unplanned return visits within 28 days of previous visit
Proportion ACS Chronic conditions
Constant
.000
.001
.000
.002
.000
1.043
2.320
13.861
3.271
.001
1.035
1.407
5.414
1.519
1.052
3.825
35.485
7.046
Kirby et al. BMC Health Services Research 2010, 10:216
http://www.biomedcentral.com/1472-6963/10/216
Page 6 of 8

Page 7

This study is limited by the fact that it involves data per-
taining to one hospital only. Although the results may not
be generalisable to all hospitals, they are relevant to similar
sized hospitals in non-metropolitan urban areas. Compari-
son with other hospitals with a similar role would lend
weight to these findings. Another limitation is the purpose
of the data collection. The data forms part of the Emer-
gency Department Information System collected by all
NSW hospitals. The accuracy of the recording of primary
diagnosis in the data base is complicated by the fact that
many of the older patients have multiple comorbidities
and the reason for their presentation may not be apparent
until test results are available. It was not possible to make
any estimation of socioeconomic status from the data
available. However, the area which the hospital services,
although it contains some affluent pockets, is generally a
low socioeconomic area [63]. As this study examines
emergency department data only there is an inbuilt selec-
tion bias. We did not examine the use of other primary
care services by people who did not access emergency
department services. An expanded study to include the
patterns of emergency department and community based
primary care, particularly general practitioner services, is
needed to develop population-based solutions to emer-
gency department overcrowding. Another possible limita-
tion is the failure of logistic regression analysis to take into
account the intra-patient variability [51-54]. To address
this issue we have compared the results of logistic and
negative binomial regression models and found similar
results. The relatively wide confidence intervals for some
of the variables reduce the precision of the estimates. The
study is also limited by the lack of a specific measurement
of severity of illness.
Conclusions
This study of routinely collected hospital data identified
the factors associated with frequent readmissions with
the aim of exploring the reasons behind readmissions.
Strategies to identify and improve access to alternative
non-emergency department services for neuroses,
COPD, dyspnoea and chronic heart failure patients,
including chronic disease self management and case
coordination services, would reduce readmissions. An
early warning system for the unplanned return visit
patients would allow for special strategies to be put in
place to address the needs of these patients. Possible
community-based services access barriers such as trans-
port, location of community-based services and waiting
times should be explored and reduced. This study has
opened up the possibility that a patient’s preference for
hospital over non-hospital services might influence read-
missions and warrants further investigation.
Acknowledgements
There are no acknowledgements.
Authors’ contributions
SEK, SMD, UWJ and MFH planned the research study. SMD and MFH
supervised the PhD work. SEK carried out the initial analysis. All authors
reviewed and approved the final manuscript.
Competing interests
The authors declare that they have no competing interests.
Received: 19 November 2009 Accepted: 21 July 2010
Published: 21 July 2010
References
1.Australasian College of Emergency Medicine: Policy on Standard
Terminology. 2009 [http://www.acem.org.au/media/
policies_and_guidelines/P02_-_Standard_Terminology_16.04.09.pdf].
2.Byrne M, et al: Frequent attenders to an emergency department: A study
of primary health care use, medical profile, and psychosocial
characteristics. Annals of Emergency Medicine 2003, 41(3):309-318.
3.Fulde GWO, Duffy M: Emergency department frequent flyers:
Unnecessary load or a lifeline? Medical Journal of Australia 2006,
184(12):595.
4. Moore G, et al: Socio-demographic and clinical characteristics of re-
presentation to an Australian inner-city emergency department:
Implications for service delivery. BMC Public Health 2007, 7.
5. Sprivulis PC, et al: The association between hospital overcrowding and
mortality among patients admitted via Western Australian emergency
departments. Medical Journal of Australia 2006, 184(5):208-212.
6. Cameron PA: Hospital overcrowding: A threat to patient safety? Medical
Journal of Australia 2006, 184(5):203-204.
7.Richardson DB: Increase in patient mortality at 10 days associated with
emergency department overcrowding.[see comment]. Medical Journal of
Australia 2006, 184(5):213-6.
8.Coleman EA: How can we ground the “frequent fliers"? Journal of the
American Geriatrics Society 2007, 55(3):467-8.
9.Locker TE, et al: Defining frequent use of an urban emergency
department. Emergency Medicine Journal 2007, 24(6):398-401.
10. Williams ERL, et al: Psychiatric status, somatisation, and health care
utilization of frequent attenders at the emergency department A
comparison with routine attenders. Journal of Psychosomatic Research
2001, 50(3):161-167.
11.Lynch RM, Greaves I: Regular attenders to the accident and emergency
department. Journal of Accident and Emergency Medicine 2000,
17(5):351-354.
12.Chan BTB, Ovens HJ: Frequent users of emergency departments. Do they
also use family physicians’ services? Canadian Family Physician 2002,
48:1654-60.
13.Pope D, et al: Frequent users of the emergency department: a program
to improve care and reduce visits. Canadian Medical Association Journal
2000, 162(7):1017-20.
14.Olsson M, Hansagi H: Repeated use of the emergency department:
Qualitative study of the patient’s perspective. Emergency Medicine Journal
2001, 18(6):430-434.
15.Genell Andrén K, Rosenqvist U: Heavy users of an emergency
department–a two year follow-up study. Social Science and Medicine 1987,
25(7):825-831.
16.Hansagi H, et al: Frequent use of the hospital emergency department is
indicative of high use of other health care services. Annals of Emergency
Medicine 2001, 37(6):561-567.
Kirby et al. BMC Health Services Research 2010, 10:216
http://www.biomedcentral.com/1472-6963/10/216
Page 7 of 8

Page 8

17.Ansari Z, Laditka JN, Laditka SB: Access to health care and hospitalization
for ambulatory care sensitive conditions. Medical Care Research and
Review 2006, 63(6):719-741.
Kne T, Young R, Spillane L: Frequent ED users: Patterns of use over time.
Am J Emerg Med 1998, 16:648-652.
Lucas RH, Sanford SM: An analysis of frequent users of emergency care
at an urban university hospital. Annals of Emergency Medicine 1998,
32(5):563-568.
Sun BC, et al: Characteristics and temporal trends of “left before being
seen” visits in US emergency departments, 1995-2002. Journal of
Emergency Medicine 2007, 32(2):211-5.
Zuckerman S, Shen YC: Characteristics of occasional and frequent
emergency department users do insurance coverage and access to care
matter? Medical Care 2004, 42(2):176-182.
Mandelberg JH, Kuhn RE, Kohn MA: Epidemiologic analysis of an urban,
public emergency department’s frequent users. Academic Emergency
Medicine 2000, 7(6):637-646.
Blank FSJ, et al: A descriptive study of heavy emergency department
users at an academic emergency department reveals heavy ED users
have better access to care than average users. Journal of Emergency
Nursing 2005, 31(2):139-144.
Raymond H, Lucas MR, Sandra M, Sanford MSN: An Analysis of Frequent
Users of Emergency Care at an Urban University Hospital. Ann Emerg
Med 1998, 32:563-568.
Braitberg G: Emergency department overcrowding: dying to get in?
Medical Journal of Australia 2007, 187(11/12):624-625.
Bianco A, Pileggi C, Angelillo IF: Non-urgent visits to a hospital
emergency department in Italy. Public Health 2003, 117(4):250-255.
Huang CL, Choi WM: Characteristics of Emergency Services for the
Elderly–A Regional Hospital Example. International Journal of Gerontology
2009, 3(4):217-222.
Huang JA, et al: Analysis of Emergency Department Utilization by Elderly
Patients Under National Health Insurance. The Kaohsiung Journal of
Medical Sciences 2003, 19(3):113-119.
Richardson DB: Elderly patients in the emergency department: a
prospective study of characteristics and outcome.[see comment]. Medical
Journal of Australia 1992, 157(4):234-9.
Dent AW, et al: The heaviest repeat users of an inner city emergency
department are not general practice patients. Emergency Medicine 2003,
15(4):322-329.
UK Department of Health: Patient at Risk of Readmission Tool. 2007
[http://webarchive.nationalarchives.gov.uk/+/www.dh.gov.uk/en/Healthcare/
Longtermconditions/DH_4130655].
Billings J, et al: Case finding for patients at risk of readmission to
hospital: Development of algorithm to identify high risk patients. British
Medical Journal 2006, 333(7563):327-330.
Howell S, et al: Using routine inpatient data to identify patients at risk of
hospital readmission. BMC Health Services Research 2009, 9:96.
Andersen RM: Revisiting the behavioral model and access to medical
care: does it matter? J Health & Soc Behav 1995, 36(1):1-10.
Stein JA, Andersen R, Gelberg L: Applying the Gelberg-Andersen
Behavioral Model for Vulnerable Populations to Health Services
Utilization in Homeless Women. J Health Psychol 2007, 12(5):791-804.
Andersen R, Newman JF: Societal and individual determinants of medical
care utilization in the United States. Milbank Memorial Foundation
Quarterly 1973, 51(1):95-124.
Padgett DK, Brodsky B: Psychosocial factors influencing non-urgent use of
the emergency room: A review of the literature and recommendations
for research and improved service delivery. Social Science and Medicine
1992, 35(9):1189-1197.
Padgett DK, et al: Predictors of emergency room use by homeless adults
in New York City: The influence of predisposing, enabling and need
factors. Social Science and Medicine 1995, 41(4):547-556.
Billings J, Anderson GM, Newman LS: Recent Findings on Preventable
Hospitalizations. Health Affairs 1996, 15(3):239-249.
Laditka JN, Laditka SB, Mastanduno MP: Hospital utilization for ambulatory
care sensitive conditions: Health outcome disparities associated with
race and ethnicity. Social Science and Medicine 2003, 57(8):1429-1441.
Zook CJ, Savickis SF, Moore FD: Repeated Hospitalization for the Same
Disease: A Multiplier of National Health Costs. The Milbank Memorial Fund
Quarterly. Health and Society 1980, 58(3):454-471.
18.
19.
20.
21.
22.
23.
24.
25.
26.
27.
28.
29.
30.
31.
32.
33.
34.
35.
36.
37.
38.
39.
40.
41.
42. Calver J, et al: High-cost users of hospital beds in Western Australia: A
population-based record linkage study. Medical Journal of Australia 2006,
184(8):393-397.
Malone RE: Heavy users of emergency services: Social construction of a
policy problem. Social Science and Medicine 1995, 40(4):469-477.
Richardson DB: 2009 Access Block Point Survey. The Road Trauma and
Emergency Medicine Unit Australian National University 2009.
O’Connell T: Redesign for Hospital Avoidance, in Hospital Avoidance.
2008 [http://www.changechampions.com.au/resource/Tony_OConnell.pdf].
Siminski P, et al: Primary care presentations at emergency departments:
Rates and reasons by age and sex. Australian Health Review 2008,
32(4):700-709.
Masso M, et al: Why patients attend emergency departments for
conditions potentially appropriate for primary care: Reasons given by
patients and clinicians differ. Emergency Medicine Australasia 2007,
19:333-340.
Howard MS, et al: Patients’ perspective on choosing the emergency
department for nonurgent medical care: A qualitative study exploring
one reason for overcrowding. Journal of Emergency Nursing 2005,
31(5):429-435.
Hong R, Baumann BM, Boudreaux ED: The emergency department for
routine healthcare: race/ethnicity, socioeconomic status, and perceptual
factors. Journal of Emergency Medicine 2007, 32(2):149-58.
Australasian College of Emergency Medicine: Australian Triage Scale. 2000
[http://www.acem.org.au/media/policies_and_guidelines/
P06_Aust_Triage_Scale_-_Nov_2000.pdf].
Coxe SS, West GL, Aiken S: The Analysis of Count Data: A Gentle
Introduction to Poisson Regression and Its Alternatives. Journal of
Personality Assessment 2009, 91(2):121-136.
Elhai JD, Calhoun PS, Ford JD: Statistical procedures for analyzing mental
health services data. Psychiatry Research 2008, 160(2):129-136.
Hunt KA, et al: Characteristics of Frequent Users of Emergency
Departments. Annals of Emergency Medicine 2006, 48(1):1-8.
Tsia : Appropriate Statistical Treatment of Frequent Emergency
Department Visits in Health Services Research. Annals of Emergency
Medicine 2007, 49(3):385.
Pallant J: SPSS Survival Manual. Sydney: Allen & Unwin 2007.
Tabachnick BG, Fidell LS: Using Multivariate Statistics. Boston: Allyn and
Bacon, 5 2007.
Nagelkerke NJD: A note on the general definition of the coefficient of
determination. Biometrika 1991, 78(3):691-692.
Norusis MJ: SPSS professional statistics 6.1. Chicago, Illinois: SPSS Inc 1994.
Government N: NSW State Plan. Premier’s Department: Sydney 2006.
Meldon SW, et al: A brief risk-stratification tool to predict repeat
emergency department visits and hospitalizations in older patients
discharged from the emergency department.[see comment]. Academic
Emergency Medicine 2003, 10(3):224-32.
Bernstein GA, et al: Somatic symptoms in anxious-depressed school
refusers. Journal of the American Academy of Child and Adolescent Psychiatry
1997, 36(5):661-668.
Fleishman JA, et al: Using the SF-12 health status measure to improve
predictions of medical expenditures. Medical Care 2006, 44(5 SUPPL).
IRIS: Shellharbour 2009 [http://www.iris.org.au/index.pl?page=222].
43.
44.
45.
46.
47.
48.
49.
50.
51.
52.
53.
54.
55.
56.
57.
58.
59.
60.
61.
62.
63.
Pre-publication history
The pre-publication history for this paper can be accessed here:
http://www.biomedcentral.com/1472-6963/10/216/prepub
doi:10.1186/1472-6963-10-216
Cite this article as: Kirby et al.: Patient related factors in frequent
readmissions: the influence of condition, access to services and patient
choice. BMC Health Services Research 2010 10:216.
Kirby et al. BMC Health Services Research 2010, 10:216
http://www.biomedcentral.com/1472-6963/10/216
Page 8 of 8