Methods
This is a prospective study with follow-up until 90 days after cardiac
surgery. The protocol for this study has been published. (8)
Study population and data
sources
Patients having undergone cardiac surgery from April 2013 to 2014, aged
18 or above were eligible for inclusion. Information were obtained
through national Danish registers for the total cohort: the Danish
National Patient Register (DNPR) (cardiac surgery procedures, length of
hospitalisation) (9), the Danish Civil Registration System (date of
birth, sex, cohabitation status, migration, vital status) (10), Danish
Education Registers (educational level) (11) and the Danish Register on
Personal Income (disposable income) (12). Information on socio-economic
factors included educational level, income and cohabitation status.
Educational level was categorised as basic school (≤10 years), upper
secondary or vocational education, and higher education. Income was
categorised in three groups according to median; ≤50%,
>50%-150%, and >150%. Cohabitation was
defined as being married or living with a partner. Living alone included
singles, divorced and widowed.
Furthermore, analyses were done for a nested subsample of patients that
participated in the national cross-sectional survey DenHeart (13). All
patients discharged from a Danish Heart Centre were asked to fill out a
questionnaire at hospital discharge to evaluate patient reported
outcomes, reporting information on health-related quality of life
(HRQoL), emotional and cognitive functioning, as well as questions about
health behaviour. (13).
Information on EuroSCORE and length of ICU stay was received from two
clinical databases (14).
Patient-reported outcomes
(DenHeart)
HRQoL was measured using the 12-Item Short-Form Health Survey (SF12) and
the HeartQoL questionnaires.
The (SF-12) is a generic measure of self-rated health constituting a
measure of mental (MCS) and physical (PCS) health. Higher scores (0-100)
indicate better perceived health. (15) As recommended the cut-off was
set as the mean minus one standard deviation, using the Danish normed
score (16).
The HeartQol is a disease-specific tool, scored from 0 (poor) to 3
(best) (17). Scores are summarized in a global, a physical, and an
emotional subscale score. For this study HeartQol quantities were
converted to binary quantities based on the median score. Both the SF-12
and the HeartQoL questionnaires have a 4 week recall period.
Emotional and cognitive perceptions were measured by the Brief Illness
Perception Questionnaire (B-IPQ). Higher scores (0-10) indicates
stronger perceptions. (18). No clear cut-offs for screening have been
determined for B-IPQ. To reduce the degrees of freedom only the summary
score was included for the main analyses in three categories based on
the 25th and 75th quartile in the
studied sample.
Loneliness was assessed by two ancillary questions, which have
previously been used and tested in the Danish National Health Survey
(19). One question concerned whether patients experienced having someone
to talk to if they needed support or were having problems, and the
second question if they were alone, though preferring to be with others.
For health behaviour, patients reported status of current or previous
smoking behaviour and alcohol intake during a typical week, as well as,
current height and weight.
Outcomes
Mortality is a reliable and clinically important outcome in cardiac
surgery; however, duration of hospitalisation and stay in the ICU are
common endpoints in cardiac surgical studies. The ICU stay is a standard
component of the treatment and provides an indication of the patient’s
recovery profile and is in effect a composite measure of the entire
perioperative process. (20). Readmission is frequent, why it is an
outcome with significant health and economic implications. Readmission
rates are about 15% at 30 days after discharge (21,22), but varies
greatly after 30 days from 19 to 56% (22,23). Thus, four outcomes were
included, 1) death within 90 days of cardiac surgery, 2) prolonged stay
in the ICU (≥ 72 hours), 3) prolonged hospital admission (≥ 10 days) and
4) readmission within 90 days from the time of cardiac surgery. Each
outcome was evaluated in separate models.
Death
From the Danish Civil Registration System information on all-cause
mortality within 90 days from cardiac surgery was obtained.
Prolonged length of stay
Length of stay was included as number of days in the ICU (LOS-ICU), as
well as total length of hospital stay (LOS-HOSP). Length of hospital
stay, and ICU stay were dichotomised to designate normal and prolonged
length of stay. There is no consensus on the definition of prolonged
length of stay in hospital following cardiac surgery. Previous studies
have adopted the 75th percentile of the length of stay
distribution, while others have defined prolonged length of stay as
hospitalisation for 10 or more days following cardiac surgery (24,25),
which was used in this study.
In previous studies, prolonged length of stay in the ICU has been
defined as from >24 to as much as >96 hours
(24,26–29). For the present study, based on the existing literature and
clinical framework, prolonged length of stay in the ICU was defined as
>72 hours.
Readmission
Information on rehospitalisation was obtained from the DNPR and was
included as a dichotomous outcome of readmission within 90 days
following cardiac surgery.
Statistical analysis
Baseline characteristics at time of admission were described using means
and standard deviations (SD) for continuous measures and percentages for
categorical measures.
Initially, logistic regression analyses were conducted to investigate
the association between each candidate predictor variable and outcomes
for both the total and DenHeart population. Using logistic regression
models, we estimated odds ratio (OR) for death, readmission, LOS-ICU and
LOS-HOSP adjusting for (1) age (10 years intervals) and sex, and (2)
EuroSCORE I.
The number of missing values in the register-based data was low for
educational level (n = 110 (3%)) and income (n = 28 (<1%)),
however, to determine the best model based on variable selection, data
were imputed, by assigning missing for educational level to basic
education and missing for income to the median value. For the DenHeart
population of 982 patients, 456 patients had missing data in one or more
variables. Thus, single mean imputation for each item was conducted for
continuous variables whilst for categorical variables (smoking and
loneliness), imputations were done by
assigning missing to the category
most frequently occurring, since missingness was <5% (see
Supplementary Table 1).
To determine the incremental value of each candidate predictor variable,
each of the predictor variables were excluded separately in a multiple
regression model by using an automated backwards selection procedure
with a set liberal significance level of 0.10. EuroSCORE was maintained
in the models. The Receiver Operating Characteristic (ROC) curve
including Area Under the Curve (AUC) and Brier score were used to
determine discrimination and calibration, respectively (30,31).
All analyses were conducted using SAS version 9.4.