See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/5809076

Nurse absenteeism and workload: Negative effect on restraint use, incident

reports and mortality

Article  in  Journal of Advanced Nursing · January 2008

DOI: 10.1111/j.1365-2648.2007.04459.x · Source: PubMed

CITATIONS READS

58 6,585

3 authors, including:

Lynn Unruh M. Lindell Joseph

University of Central Florida University of Iowa
100 PUBLICATIONS   2,056 CITATIONS    25 PUBLICATIONS   204 CITATIONS   

SEE PROFILE SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Workforce Engagement View project

Health and Well-being View project

All content following this page was uploaded by M. Lindell Joseph on 24 November 2015.

The user has requested enhancement of the downloaded file.

JAN ORIGINAL RESEARCH
Nurse absenteeism and workload: negative effect on restraint use,
incident reports and mortality
Lynn Unruh, Lindell Joseph & Margaret Strickland
Accepted for publication 1 August 2007
Correspondence to L. Unruh:
e-mail: lunruh@mail.ucf.edu
Lynn Unruh PhD RN LHRM
Fellow in Nursing Policy and Philanthropy
Robert Wood Johnson Foundation,
Princeton, New Jersey, USA, and
Associate Professor
Health Services Administration, Department
of Health Professions, College of Health and
Public Affairs, University of Central Florida,
Orlando, Florida, USA
Lindell Joseph PhD RN
Candidate and Nurse Researcher
Center for Nursing Research and Innovation,
Florida Hospital Medical Center, Orlando,
Florida, USA
Margaret Strickland BSN RN
Chief Nursing Officer
Florida Hospital Altamonte, Administrative
Offices, Altamonte Springs, Florida, USA
 2007 The Authors. Journal compilation  2007 Blackwell Publishing Ltd
U N R U H L ., J O S E P H L . & S T R I C K L A N D M . ( 2 0 0 7 ) Nurse absenteeism and
workload: negative effect on restraint use, incident reports and mortality. Journal of
Advanced Nursing 60(6), 673–681
doi: 10.1111/j.1365-2648.2007.04459.x
Abstract
Title. Nurse absenteeism and workload: negative effect on restraint use, incident
reports and mortality
Aim. This paper is a report of a study to assess the impact of nurse absenteeism on
the quality of patient care.
Background. Nurse absenteeism is a growing management concern. It can con-
tribute to understaffed units, staffing instability, and other factors that could have a
negative impact on patient care. The impacts of absenteeism on the quality of
nursing care have rarely been studied.
Method. Retrospective monthly data from incident reports and staffing records in
six inpatient units for 2004 were analysed. Dependent variables were the numbers of
restraints, alternatives to restraints, incident reports, deaths, and length of stay.
Explanatory variables were nurse absenteeism hours, patient days per nursing staff,
and interaction between these variables. Controls were patient acuity and unit
characteristics. Fixed effects regressions were analysed as regular or negative
binomial models.
Findings. Neither high Registered Nurse absenteeism nor high patient load was
related to restraint use when taken separately. However, high Registered Nurse
absenteeism was related to restraint use when patient load was high. Registered
Nurse absenteeism was related to a lower use of alternatives to restraints. Incident
reports were increased by high patient load, but not absenteeism, or absenteeism
given patient load. When both patient load and absenteeism were high, deaths were
higher also.
Conclusion. Absenteeism alone may not be a strong factor in lowering quality, but
the combination of high Registered Nurse absenteeism and high patient load could
be a factor. Staffing and absenteeism may be part of a vicious cycle in which low
staffing contributes to unit absenteeism, which contributes to low staffing, and so
on.
Keywords: incident forms, nurse absenteeism, nurse–patient relationships, quality
of care, retrospective analysis, staffing records, work organization
673
L. Unruh et al.
Introduction
Nurse absenteeism is a fact of life of inpatient units but is
problematic because it can contribute to understaffed units,
staffing instability, chaotic working environments, staff
demoralization, and other factors that could negatively
impact on patient care (Adams & Bond 2003, Thomson
2005). In an effort to reduce contributors to the problem,
some of the probable causes of absenteeism, such as
personal and family illnesses, sickness absence policies and
job dissatisfaction, have been identified. However, the
impact of absenteeism on the quality of nursing care and
patient outcomes have rarely been studied, and such
information is important when advocating for changes in
employment that could reduce absenteeism. Concerns over
nurse absenteeism in a 258-bed suburban hospital in the
United States of America (USA) led us to study of its effects
on patient care and patient outcomes using inpatient data
for the year 2004.
Background
Absenteeism, defined simply as unscheduled absences, has
been described as ‘nursing service’s albatross’ (Miller &
Norton 1986, p. 38). The ‘albatross’ of absenteeism has been
hard to overcome, and recently has been on the rise in US
hospitals (Harter 2001). Although present among all levels of
nurses, the extent to which it occurs may differ among types
of nursing staff. In one study, it was found to be highest
among nursing assistants and lowest among Registered
Nurses (RN) (Burton 1992), although in other studies this
differentiation has not been strongly noted (Adams & Bond
2003).
Absenteeism is thought to disrupt the working environ-
ment, producing staffing instability and affecting employee
morale (Taunton et al. 1989, Bauman et al. 2001, Thomson
2005). Nurse staffing instability, in turn, contributes to lower
cohesion among nurses, less collaboration with physicians,
and greater difficulties in handling workload (Adams & Bond
2003). Nursing shortages are accentuated by absenteeism
(Rogers et al. 1990).
Negative work environments in general are related to job
dissatisfaction among nurses (Tumulty et al. 1994). In
particular, inadequate staffing and heavy workload – which
absenteeism may especially affect – have been found to be the
major sources of job dissatisfaction (Aiken et al. 2002, Dunn
et al. 2005, Khowaja et al. 2005, Sheward et al. 2005). Job
dissatisfaction, in turn, can lead to absenteeism, turnover or
professional exit (Parker & Kulik 1995, Shields & Ward
2001, Tzeng 2002, Gardulf et al. 2005). Absenteeism appears
674  2007 The Authors. Journal compilation  2007 Blackwell Publishing Ltd
to be part of a vicious cycle in which it contributes to a
negative work environment, which then leads to more
absenteeism and to other types of staffing instability, such
as increased turnover.
Furthermore, these hypothesized effects of absenteeism
may affect patient care, and therefore patient outcomes
(Taunton et al. 1989, Rogers et al. 1990, Adams & Bond
2003). In a 1995 interview study with nurses, Adams and
Bond (1995) found that nurses put great importance on staff
stability in generating effective working relations and
quality patient care. In a second study, a strong connection
between staff stability and professional nursing practice was
found (Adams & Bond 2003). Inadequate staffing has been
found to lead to poor performance, including increased
errors (Arnetz 1999, Benner et al. 2002, Bridger 1997,
Kalisch, 2006). In contrast, adequate staffing is associated
with the provision of recommended medical care for
patients with selected medical diagnoses (Landon et al.
2006). The negative patient outcomes of inadequate staffing
include: adverse events, such as falls; bloodstream infec-
tions; pneumonia; urinary tract infections; pressure ulcers;
failure to rescue (defined as the death of a patient having
experienced a life-threatening complication); and patient
death (ANA 2000, Aiken et al. 2002, Cho et al. 2003,
Dunton et al. 2004, Elting et al. 2005, Krauss et al. 2005,
Mark et al. 2002, Needleman et al. 2002, Person et al.
2004, Sales et al. 2005, Seago et al. 2006, Unruh 2003,
Whitman et al. 2002).
Nurse dissatisfaction, burnout, exhaustion and stress have
also been associated with lower quality of care and negative
patient outcomes (Dugan et al. 1996, Koivula et al. 1998,
Laschinger & Leiter 2006). Dugan et al. (1996) found a
relationship between nurse stress and patient falls and
medication errors. Koivula et al. (1998) reported that nurse
exhaustion prevented nurses from having necessary prereq-
uisites for quality improvement, while Laschinger and Leiter
(2006) found that emotional exhaustion was related to
greater patient adverse events.
Although a negative relationship between nurse absen-
teeism and the quality of patient care can be hypothesized
through theoretical linkages and studies above, very few
studies have examined this relationship directly. Roszkow-
ski et al. (2005) noted that absenteeism was related to
nurses’ perceptions of poorer performance (both self-rated
and manager-rated). In Taunton et al. (1994) it was found
that absenteeism was related to higher rates of nosocomial
urinary tract and bloodstream infections in hospitalized
patients. We are not aware of other studies of the impact
of absenteeism on patient outcomes or the quality of care.
A review of the literature by Thomson (2005) corroborates
JAN: ORIGINAL RESEARCH Nurse absenteeism a negative effect

our lack of findings. The lack of information on the

effect of absenteeism on the quality of patient care and
patient outcomes points to the need for more research in
this area.
The study
Aim
The aim of the study was to assess the impact of nurse
absenteeism on the quality of patient care.
The study hypotheses were:
• Higher levels of absenteeism have a negative impact on the
quality of patient care.
• Absenteeism has a more negative effect on quality in units
that also have lower staffing levels.
• Lower staffing levels have a negative impact on the quality
of care.
Design
A retrospective study was carried out using unit-level data
from incident reports and staffing records for 2004.
Data sources
This study was conducted at a 258-bed hospital which is part
of a large hospital system in the southeast USA. Data from
the following six inpatient units were used: 17-bed intensive
care, 24-bed oncology, 39-bed medical–surgical, 39-bed
neuro-medical progressive care, 46-bed pulmonary progres-
sive care, and 46-bed cardiac progressive care.
Available data were recorded as rates or numbers per unit
in monthly intervals. There was a maximum of 12 monthly
observations for the six units, which equalled 72 observations
when all data were present.
Measures
Table 1 presents the measures used in the study, their
definitions, and the completeness of the data. The dependent
variables were nursing process and patient outcomes indica-
tors collected in hospital clinical performance improvement
reports, incident reports or risk management databases for
which data were complete or nearly complete and for which
sufficient variability existed. Despite a large number of
indicators, most data were too incomplete to use or lacked

Table 1 Study measures

Variable Definition Source Consistent data available

Dependent variables
Restraints No. episodes of restraint use Incident reports Missing some months in
some units, n = 67
Alternatives to restraints No. alternatives to restraints used Incident reports Missing some months in some
units, n = 67
Incident reports No. incident reports per unit Risk management Yes, n = 72
per no. patient days of care per unit reports
Deaths No. deaths Incident reports Yes, n = 72
Length of stay Length of stay Risk management Yes, n = 72
reports
Explanatory variables and controls
RN, LPN, NA absenteeism hours No. hours of unplanned absences Staffing records Yes, n = 72
per unit per month for
RNs, LPNs, NAs, Total nursing staff
RN, LPN, NA worked hours No. hours worked by each type of Staffing records Yes, n = 72
per patient days nursing staff in the month per the
number of patient days for that unit
Age of RNs on unit Average age of RN per unit Staffing records Yes, but no variation in average
age by unit (most around 43)
RN clinical level Average unit RN clinical Staffing records Yes, but little variation by unit
care level (I–IV) (most around 2)
Average daily census Average no. patients per day Risk management Yes, n = 72
reports
Case mix Average unit case mix for the Staffing records Yes, n = 72
time period
Unit ID ID number for each FHA unit Staffing records Yes, n = 72

RN, Registered Nurses; LPN, Licensed Practical Nurses; NA, nursing assistants.

 2007 The Authors. Journal compilation  2007 Blackwell Publishing Ltd 675
L. Unruh et al.
variability across hospital units. The acceptable data on the
nursing process were the number of restraints and alterna-
tives to restraints (67 observations each), while those for
patients outcomes were the number of incident reports,
deaths and length of stay (a full set of 72 observations each).
The first three of these five measures are direct results of
nursing care and therefore have face and content validity as
indicators of the quality of nursing care. Mortality has been
demonstrated in prior studies as a nursing-sensitive quality
measure (Aiken et al. 2002). Length of stay has likewise been
linked to nursing care (Shamian et al. 1994, Czaplinski &
Diers 1998). As a result, the study was conducted using these
five dependent variables.
The explanatory variables of interest were the number of
absenteeism hours among each type of nursing staff [RN,
Licensed Practical Nurses (LPN), and nursing assistants
(NA)], as listed in Table 1. Another set of explanatory
variables was ratios of the number of worked hours for
each type of nursing staff divided by the number of patient
days. As the study involved six inpatient units, unit-level
controls were needed. The important ones were staff
demographics, such as age, gender, race and ethnicity,
levels of education, and years of experience. However, the
only unit characteristics provided by the hospital records
were RN age and clinical level. Patient characteristics
which could contribute to patient outcomes and which
therefore needed to be controlled were captured in a unit-
level case-mix variable.
Ethical considerations
As this was a retrospective study involving a convenience
sample of archival data, no participant recruitment was
required, and case selection was dependent on the avail-
ability and completeness of the data in the records. There
were no clinical risks to the patients. Confidentiality risks,
because of researcher access to patient-level data, were dealt
with by stripping the data of all patient identifiers prior to
researcher involvement. The study was judged to be exempt
from full Institutional Review Board review by the Boards
at both the University of Central Florida and the partici-
pating hospital.
Statistical analysis
All the data were collected by the hospital’s reporting
systems and were in electronic form. The data existed in
separate data sets. Data in each file were cleaned and
prepared for merging with the other data. Specific measures
of interest were created from the existing raw data through
676  2007 The Authors. Journal compilation  2007 Blackwell Publishing Ltd
manipulations in Excel or recoding in Statistical Analysis
System (SAS) 9.1 (SAS n.d.). For example, nursing staff
absenteeism hours, which were defined as the number of
hours of unplanned absences, were a summation of several
categories of unplanned absences (sick time, no shows, etc.).
Once the data were ready for analysis, SAS was used to
analyse the functional models. Descriptive statistics were
calculated for the absenteeism and staffing variables, controls
and dependent variables (see Table 2). For the associative
analysis, we ran regressions for all types of staff. We were
interested in the impact of each type of nursing staff
absenteeism on each of the dependent variables listed in
Table 1. We were also interested in how absenteeism affected
the dependent variables, given unit staffing. This required the
use of an interaction term between absenteeism hours and
staffing levels, as is indicated in the statistical model below.
Finally, we also examined how staffing impacts the nursing
process and patient outcomes.
The use of an interaction term required that the two
variables in the term had values that went in the same
direction. As increases in absenteeism indicate a negative
change while increases in nurse staffing indicate a positive
change, to be able to interpret the regressions with interaction
between these variables, we used the inverse of the nurse
staffing variables in the regressions. This transformed the
original ‘nursing hours per patient days’ into ‘patient days per
nursing hours,’ analogous to patient load or patients per
nurse. Thus, for the purposes of the regressions, increases in
absenteeism and in patient load ran in the same direction.
Except for length of stay, the dependent variables were
incidences of events (‘count data’) that tend to occur
infrequently. This means that the values were skewed
towards 0 and were not normally distributed. Turning this
count data into rates by weighting them by patient days of
care would result in the same non-normal, skewed distribu-
tion. Our descriptive analysis of the dependent variables
corroborated this fact: the incidences and rates of restraints,
alternatives to restraints, incident reports and deaths were
clustered around zero, skewed and highly dispersed (variance
greater than the mean), whereas the length of stay was fairly
normally distributed (see Table 2). Given these characteristics
of the dependent variables, except for length of stay, ordinary
least squares regressions could not be used. Instead, we used
exponentially based regressions. For the numbers of rest-
raints, alternatives to restraints and incident reports, a
negative binomial regression, which takes into account highly
dispersed distributions, was used. For the number of deaths,
we found that a Poisson regression had better fit. Both the
negative binomial and Poisson regression require that the
dependent variables are in ‘count’ form rather than being a
JAN: ORIGINAL RESEARCH Nurse absenteeism a negative effect

Table 2 Descriptive statistics: average monthly unit values* Statistical models also took into account the fact that
Variable name Mean SD Minimum Maximum
measures were taken on the same six units 12 times (monthly
for the 12 months). This type of ‘repeated measures’ analysis,
RN absenteeism hours 215Æ36 110Æ93 8 520 which causes a heteroskedasticity bias, can be corrected by
RN absenteeism hours 0Æ04 0Æ02 0Æ001 0Æ08
adding a fixed effects variable. This adjustment also allowed
per total worked hours
RN absenteeism hours 0Æ31 0Æ29 0Æ01 1Æ30 us to control for all individual unit characteristics, although
per patient day we did not know what the specific characteristics were. The
RN worked hours 7Æ51 4Æ03 3Æ89 19Æ56 basic statistical model was:
per patient day Nursing care quality = b1 + b2*(RN, LPN and NA absen-
LPN absenteeism hours 41Æ60 48Æ89 0 172Æ00
teeism) + b3*(RN, LPN and NA patient) + b4*(RN, LPN
LPN absenteeism hours 0Æ059 0Æ062 0 0Æ30
per total worked hours
and NA absenteeism* RN, LPN and NA patient
LPN absenteeism hours 0Æ038 0Æ04 0 0Æ14 load) + b5*(CMI) + ui (Fixed Effect = unit) + e.
per patient day Five separate regressions were analysed for each nursing
LPN worked hours 0Æ53 0Æ48 0 2Æ01 staff type corresponding to the five dependent variables
per patient day listed in Table 1. Each type of nursing staff was analysed in
NA absenteeism hours 136Æ80 96Æ85 0 392Æ00
a separate set of regressions because they were highly inter-
NA absenteeism hours 0Æ047 0Æ039 0 0Æ16
per total worked hours correlated in statistical analyses. Therefore, 15 total regres-
NA absenteeism hours 0Æ16 0Æ16 0 0Æ85 sions were performed. For the normally distributed depen-
per patient day dent variable (length of stay), the regressions were analysed
NA worked hours 2Æ97 1Æ51 0 5Æ36 using a regular linear-fixed effects analysis (using the
per patient day
‘mixed’ procedure in SAS) following the basic statistical
RN age 43Æ16 1Æ95 38 46
RN clinical level 2Æ04 0Æ10 1Æ85 2Æ2
model above. For the rest of the dependent variables that
(average of 1–4) were skewed and dispersed, the regressions were analysed as
Patient days, all units 890Æ92 299Æ04 330Æ00 1,329Æ00 log-linear functions of a negative binomial-fixed effects
Patient days, unit 1 571Æ17 122Æ21 463Æ00 933Æ00 model (using the Generalized Linear Model GENMOD
Patient days, unit 2 1,164Æ25 95Æ15 1,022Æ00 1,329Æ00 procedure in SAS). In this log-linear model, the dependent
Patient days, unit 3 980Æ91 45Æ97 912 1,057Æ00
variables are counts of the incidents, and the logs of these
Patient days, unit 4 1,152Æ17 59Æ62 1,079Æ00 1,239Æ00
Patient days, unit 5 1,052Æ58 51Æ74 964Æ00 1,148Æ00 counts are regressed on the log of a weighting factor and the
Patient days, unit 6 424Æ42 51Æ02 330Æ00 520Æ00 explanatory and control variables using a negative binomial
CMI, all units 1Æ97 1Æ36 1Æ16 7Æ27 distribution. For example, in the model below, ‘RNTWHR’
CMI, unit 1 1Æ76 0Æ38 1Æ35 2Æ48 (total number of RN hours worked in the unit in the month)
CMI, unit 2 1Æ29 0Æ09 1Æ16 1Æ41
is the weighting factor, x’i are the explanatory, interaction
CMI, unit 3 1Æ54 0Æ11 1Æ39 1Æ73
CMI, unit 4 1Æ27 0Æ06 1Æ18 1Æ39
and control variables (nurse absenteeism, nurse patient load,
CMI, unit 5 1Æ49 0Æ13 1Æ29 1Æ76 nurse absenteeism*nurse patient load and CMI), and ui are
CMI, unit 6 4Æ46 1Æ83 2Æ15 7Æ27 the fixed effect unit variables:
No. of restraints 11Æ87 9Æ14 1 40
No. of alternatives 6Æ75 5Æ35 0 23 Log{nursing care quality} = log{RNTWHR} + xi¢b + ui + e
to restraints
No. of incident reports 12Æ43 6Æ17 2 31
No. of deaths 4Æ02 3Æ51 0 17 Results
Length of stay 3Æ61 0Æ65 2Æ60 5Æ40
Table 2 reports the results of the descriptive analysis. Average
*Average of all units’ monthly values unless otherwise indicated.
CMI, Case-Mix Index; RN, Registered Nurses; LPN, Licensed
RN absenteeism hours per month were 215. This varied from
Practical Nurses; NA, nursing assistants; SD, standard deviation. only 8 hours in one unit to 520 in another, resulting in a
standard deviation (SD) of 111. As these results were not
rate. A weighting factor in the regression accounted for the weighted by the varied unit staffing levels or patient volume,
unit size, patient volume or staff volume. we also calculated RN absenteeism hours per total scheduled
Descriptive analysis also revealed that there was very little hours and per patient day. These also showed large variance,
variation in RN characteristics (age and clinical level). These as can be seen in Table 2. Mean RN worked hours per patient
variables were therefore not included in the models as they day (RN staffing) were 7Æ5, with a SD of 4. LPN and NA
would not have added anything to the analysis. absenteeism hours were less than those of RN, but only

 2007 The Authors. Journal compilation  2007 Blackwell Publishing Ltd 677
L. Unruh et al.

because their worked hours were also less. When weighted
for worked hours or patient days, both LPNs and NAs had
higher rates of absenteeism than RN. RN age averaged
43 years, while clinical level was 2 (out of 1–4). As Table 2
indicates, these variables had very little variation (SD of 1Æ95
for age and 0Æ10 for clinical level), justifying their removal
from the statistical model.
Monthly patient days were 891 on average, but two units
(1 and 6) had many fewer on the average (571 and 424
respectively). Average Case-Mix Index (CMI) was 1Æ97, with
a large SD of 1Æ36. The large variation was because of a high
average CMI of 4Æ46 for unit 6. The maximum for unit 6 was
7Æ27, indicating that during 1 month their CMI was
extremely high.
Regarding quality indicators, there was an average of 12
restraint use incidents per unit per month, seven incidents of
the use of alternatives to restraints, 12Æ5 incident reports, and
four deaths. Most indicators had SD nearly as large as the
mean, indicating a large variance and dispersion of the
values, most likely because of the fact that some units, such as
the ICU/PCU had more acutely ill patients than others.
Length of stay was 3Æ61 days on average.
Table 3 reports the results of the regressions for RN staff.
Although regressions were also run separately for LPNs and
NAs, none showed statistically significant results and so these
are not reported. Concerning RN staff, neither high absen-
teeism nor high patient load was related to the use of
restraints when taken separately. However, high RN absen-
teeism in combination with a high patient load was associ-
ated with significantly higher use of restraints.
In contrast, high RN absenteeism was related to lower use
of alternatives to restraints. In this context, alternatives to
restraints are positive actions taken in place of the use of
restraints, so the lower incidence of alternatives to restraints
associated with absenteeism is a negative indicator of nursing
care quality. This result was counterbalanced, however, by
the interaction term, which counter-intuitively was positively
related to the use of restraint alternatives.
The results indicate that the numbers of incident reports
was increased by high patient load, but not absenteeism or
the interaction of absenteeism and patient load. The findings
on number of deaths were partially counter-intuitive: the
greater the RN absenteeism, the lower the number of deaths.
RN patient load was not related to number of deaths.
However, the statistically significantly positive interaction
term indicates that if both patient load and absenteeism are
high, the number of deaths will be high also. This result
would counter-balance the counter-intuitive negative impact
of absenteeism on deaths. Finally, no absenteeism or staffing
variables were statistically significantly related to length of
stay. CMI was statistically significantly related to all quality
indicators except number of incident reports. High case mix
was associated with high restraint use and high length of stay,
but low use of alternatives to restraints and a low number of
deaths.
The fixed effects term (units 1–6) in each of these
regressions yielded coefficients for each unit for each depen-
dent variable. To maintain unit anonymity, in this report we
resorted and relabelled the units 1 through 6 and used unit
number 6 for our regression reference unit. Units 1 through 5

Table 3 Impact of absenteeism on patient outcomes

No. of restraints No. of alternatives No. of incident No. of deaths Length of stay
n = 67 to restraints n = 67 reports n = 72 n = 72 n = 72

Est Est Est Est Est

Explanatory variable:
RN patient load NS NS 5Æ4561* NS NS
RN absenteeism hours NS 0Æ0013* NS 0Æ0050**** NS
RN absenteeism hours 0Æ0103** 0Æ0126** NS 0Æ0330*** NS
given RN patient load
CMI 0Æ0305*** 0Æ0967**** NS 0Æ0302**** 0Æ09779*
Unit 1 1Æ7960**** 1Æ8962**** NS 1Æ4005**** NS
Unit 2 1Æ2323**** 1Æ5294**** NS 2Æ4830**** NS
Unit 3 1Æ4560**** 1Æ7613**** NS 3Æ4998**** NS
Unit 4 0Æ6170* 1Æ1435** NS 2Æ6870**** NS
Unit 5 0Æ8024**** 1Æ2694**** NS 2Æ7689**** NS
Unit 6 – – – – –

RN patient load = the inverse of RN hours per patient days = patient days per RN hours.

Reference unit in relationship to all other units.
NS, not statistically significant; CMI, Case-Mix Index; RN, Registered Nurses; LPN, Licensed Practical Nurses; NA, nursing assistants.
*P < 0Æ05; **P < 0Æ01; ***P < 0Æ001; ****P < 0Æ0001.

678  2007 The Authors. Journal compilation  2007 Blackwell Publishing Ltd
JAN: ORIGINAL RESEARCH
What is already known about this topic
• Nurse absenteeism may disrupt the working environ-
ment, contributing to greater difficulties in handling
workload and affecting employee morale.
• Negative work environments, particularly insufficient
staffing, are related to nurse dissatisfaction, absenteeism
and turnover; poorer nurse performance; and worse
patient outcomes.
• Nurse absenteeism may be related to higher rates of
nosocomial urinary tract and bloodstream infections in
hospitalized patients.
What this paper adds
• High Registered Nurse absenteeism was related to
higher restraint use and more patient deaths when
patient load is also high.
• High Registered Nurse absenteeism was independently
associated with fewer uses of alternatives to restraints.
• High patient load was related to greater numbers of
incident reports.
had lower restraint use, lower alternatives to restraint use,
and fewer deaths than unit 6, the reference unit. No one unit
stood out compared with another in terms of incident reports
or length of stay.
Discussion
Of the five quality indicators in this study, nurse absenteeism
was related to only one: lower use of alternatives to
restraints. However, the combination of high RN absentee-
ism and high patient load was related to greater use of
restraints and a higher number of deaths. This could mean
that it is not absenteeism alone that contributes to lower
quality of nursing care, but absenteeism on units that have
poor staffing. Further, as mentioned in the introduction,
staffing and absenteeism may be part of a vicious cycle in
which low staffing contributes to unit absenteeism, which
then contributes to low staffing, and so forth. Therefore,
despite the lack of robust findings, this study supports others
that point to RN staffing as important in maintaining quality
of care, and it gives tentative indications that absenteeism
also plays a role.
In addition, some results were counter-intuitive: a higher
use of alternatives to restraints was found on units with both
higher RN absenteeism and higher RN patient load, while
RN absenteeism (unrelated to patient load) was associated
 2007 The Authors. Journal compilation  2007 Blackwell Publishing Ltd
Nurse absenteeism a negative effect
with fewer patient deaths. The lower mortality associated
with higher absenteeism may be an indication of the lack of
an independent association between absenteeism and mor-
tality. Although it may be tempting to hypothesize that
uncontrolled unit characteristics contributed to confounding
results, the effects of unit characteristics should have been
controlled through the CMI variable and the fixed effects unit
variable.
It is possible that the results for absenteeism were not more
robust because we worked with monthly, instead of daily or
shift, associations. This was necessary because our absentee-
ism and staffing measures were reported on a pay-period
(every 2 weeks) basis, while the patient care and patient
outcomes measures were reported on a monthly basis. This
necessitated the use of monthly aggregations of absenteeism
and staffing data to match the monthly quality data. Yet the
effects of absenteeism and staffing are much more directly
connected in reality. The higher use of restraints, the lower
use of alternatives to restraints, and increases in incident
reports occurred in the same shift as the absenteeism and/or
staffing problem, while mortality and length of stay changes
showed up within hours or days of the problems. Aggregating
the explanatory and response variables to monthly values
diluted this connection, probably weakening the association
in the process. This data specification problem was a likely
primary reason why we did not see more statistically
significant results. What is really needed to test the relation-
ship between absenteeism and patient outcomes is shift (or
daily) absenteeism and staffing data linked to patient
outcomes flagged by the time of day. For example, such a
study would look at whether the higher number of absentee
hours in a shift or a day on a unit was related to a certain type
of patient outcome (code blue, medication error, fall, etc.)
during that shift or day on that unit.
Another probable reason why we did not have statistically
significant results was that the monthly aggregation of data
resulted in a smaller number of observations. As we could not
study shift, daily or weekly measures, the monthly measures
we used meant that the number of observations was reduced
to 72 at the maximum (six units · 12 months). Then, for
some patient outcomes, there were missing values that
reduced the number of observations to 67. Out of concern
for the low number of observations, we conducted a
retrospective power analysis of the study. Software for
conducting a power analysis of an exponential distribution
analysed through a negative binomial regression does not
exist. However, we used the GLMPower analysis in SAS,
which is a power analysis of generalized linear models. The
negative binomial regression is in the family of generalized
linear models. The power analysis indicated that, with
679
L. Unruh et al.
observations of 67 and 72 and with one categorical variable
(fixed effects unit variable) and four covariates, the power
was adequate for hypothesis-testing at the 5% level.
A final limitation of our study was that restraint use and
alternatives to restraints are quality indicators that may be
more appropriate for some units than others. For example,
one of the units in our study was a medical–surgical ICU, in
which restraint use may be a more acceptable practice and
alternatives to restraints less commonly practised. Therefore,
in this unit these may not be good indicators of negative and
positive quality respectively.
Given our tentative results, and the limitations of the study
which could have led to the weakness of those results, we
believe that the issue of absenteeism and quality of care
remains open. Additional research needs to be conducted
using more types of nursing processes and patient outcomes
indicators, with a larger number of observations, and more
complete and more detailed data.
Conclusion
As this was a case study of one hospital, and the results were
not robust, they cannot be readily generalized and policy
recommendations would be premature. However, it is
important to understand the issue of absenteeism and its
impact on patient outcomes, and this study has begun to
investigate the issue. We see this study as an exploration that
gives initial results and identifies data needs and requirements
for future studies.
Several recommendations for further research can be made.
Research on the relationship between nurse absenteeism and
the quality of care should be conducted with more detailed
data on absenteeism, staffing and several different types of
nursing-sensitive quality indicators. The indicators could be
those used by the Joint Commission on the Accreditation
of Healthcare Organizations (Joint Commission 2007),
National Quality Forum (NQF 2003), or other quality
improvement programmes in which the hospital may partici-
pate. Indicators should have face, content, or criterion
validity as nursing-sensitive quality measures. The additional
requirement for these studies, however, is that the unit, hour,
and date of the incidents are flagged so that they can be linked
with unit-level daily absenteeism and staffing data. In
addition to a single hospital setting, it is important that these
studies also be conducted in multiple hospital settings to be
able to better generalize research results. The research should
also be conducted in other institutional settings, such as
nursing homes.
In the introduction, we mentioned the lack of information
on other possible absenteeism effects: on the work environ-
680  2007 The Authors. Journal compilation  2007 Blackwell Publishing Ltd
ment and nursing staff, and on hospital financial perfor-
mance. These are other areas also in need of further research.
The same methodologies would apply: detailed data, nursing-
sensitive quality indicators, and single, system or multi-
hospital samples.
There is a dearth of information on the effect of absentee-
ism on the workplace, the workforce, patient care, patient
outcomes and financial performance. Future research on any
one of these would contribute immensely to our understand-
ing of the impacts of absenteeism.
Author contributions
LU, LJ and MS were responsible for the study conception
and design and the drafting of the manuscript. LJ
performed the data collection and LU performed the data
analysis. LJ and MS obtained funding and LJ provided
administrative support. LU, LJ and MS made critical
revisions to the paper. LU provided statistical expertise.
LU supervised the study.
References
Adams A. & Bond S. (1995) Nursing Organisation in Acute Care: the
Development of Scales. Report no. 71. Centre for Health Services
Research, University of Newcastle upon Typne, Newcastle upon
Tyne.
Adams A. & Bond S. (2003) Staffing in acute hospital wards: part 2.
Relationships between grade mix, staff stability and features of
ward organizational environment. Journal of Nursing Manage-
ment 11, 293–298.
Aiken L., Clarke S., Sloane D., Sochalski J. & Silber J. (2002) Hos-
pital nurse staffing and patient mortality, nurse burnout, and job
dissatisfaction. Journal of the American Medical Association
288(16), 1987–1993.
American Nurses Association (ANA) (2000) Nurse Staffing and
Patient Outcomes in the Patient Hospital Setting. American Nurses
Publishing, Washington, DC.
Arnetz B.B. (1999) Staff perception of the impact of health care
transformation on quality of care. International Journal for
Quality in Health Care 11(4), 345–351.
Benner P., Sheets V., Uris P., Malloch K., Schwed K. & Jamison D.
(2002) Individual, practice, and system causes of errors in nursing:
a taxonomy. Journal of Nursing Administration 32(10), 209–523.
Bridger J.C. (1997) A study of nurses’ views about the prevention of
nosocomial urinary tract infections. Journal of Clinical Nursing 6,
379–387.
Burton R. (1992) Tackling staff absenteeism. Nursing Standard 7(3),
37–40.
Cho S., Ketefian S., Barkauskas V.H. & Smith D.G. (2003) The
effects of nurse staffing on adverse events, morbidity, mortality,
and medical costs. Nursing Research 52(2), 71–79.
Czaplinski C. & Diers D. (1998) The effect of staff nursing on length
of stay and mortality. Medical Care 36(12), 1626–1638.
JAN: ORIGINAL RESEARCH
Dugan J., Lauer E., Bouquot Z., Dutro B., Smith M. & Widmeyer L.
(1996) Stressful nurses: the effect on patient outcomes. Journal of
Nursing Care Quality 10(3), 46–58.
Dunn S., Wilson B. & Esterman A. (2005) Perceptions of working as
a nurse in an acute care setting. Journal of Nursing management
13, 22–31.
Dunton N., Gajewski B., Taunton R.L. & Moore J. (2004) Nurse
staffing and patient falls on acute care hospital units. Nurse Out-
look 52, 53–59.
Elting L.S., Pettaway C., Bekele B.N., Grossman H.B., Cooksley C.,
Avritscher E.B.C., Saldin K. & Dinney C.P.N. (2005) Correlation
between annual volume of cystectomy, professional staffing, and
outcomes. Cancer 104(5), 975–984.
Gardulf A., Soderstrom I.-L., Orton M.-L., Eriksson L.E., Arnetz B. &
Nordstorm G. (2005) Why do nurses at a university hospital want
to quit their jobs? Journal of Nursing Management 13(4), 329–337.
Harter T.W. (2001) Minimizing absenteeism in the workplace:
strategies for nurse managers. Nursing Economics 19(2), 53–5.
Joint Commission (2007). Performance Measurement. Retrieved
from http://www.jointcommission.org/PerformanceMeasurement/
on 10 September 2007.
Kalisch B.J. (2006) Missed nursing care: a qualitative study. Journal
of Nursing Care Quality 21(4), 306–313.
Khowaja K., Merchant R.J. & Hirani D. (2005) Registered nurses
perception of work satisfaction at a tertiary care university hos-
pital. Journal of Nursing management 13, 32–39.
Koivula M., Paunonen K.M. & Laippala P. (1998) Prerequisites for
quality improvement in nursing. Journal of Nursing Management
6(6), 333–342.
Krauss M.J., Evanoff B., Hitcho E., Ngugi K.E., Dunagan W.C.,
Fischer I., Birge S., Johnson S., Costantinou E. & Fraser V.J.
(2005) A case control study of patient, medication, and care-
related risk factors for inpatient falls. Journal of General Internal
Medicine 20, 116–122.
Landon B.E., Normand S.T., Lessler A., O’Malley A.J., Schmaltz S.
et al. (2006) Quality of care for the treatment of acute medical
conditions in US hospitals. Archives of Internal Medicine 166,
2511–2517.
Laschinger H.K.S. & Leiter M.P. (2006) The impact of nursing work
environment on patient safety outcomes. Journal of Nursing
Administration 36(5), 259–267.
Mark B.A., Salyer J. & Harless D.W. (2002) What explains nurses’
perceptions of staffing adequacy? Journal of Nursing Administra-
tion 32(5), 234–242.
Miller D.S. & Norton V.M. (1986) Absenteeism: nursing services
albatross. Journal of Nursing Administration 16(3), 38–42.
Needleman J., Buerhaus P., Mattke S., Stewart M. & Zelevinsky K.
(2002) Nurse-staffing levels and the quality of care in hospitals.
New England Journal of Medicine 346(22), 1715–1722.
NQF (2003) National Voluntary Consensus Standards for Hospital
Care: an Initial Performance Measure Set. Retrived from http://
www.qualityforum.org/publications/reports/hospital_measures.asp
on 10 September 2007.
Parker P.A. & Kulik J.A. (1995) Burnout, self- and supervisor-rated
job performance, and absenteeism among nurses. Journal of
Behavioral Medicine 18(6), 581–99.
 2007 The Authors. Journal compilation  2007 Blackwell Publishing Ltd
View publication stats
Nurse absenteeism a negative effect
Person S.D., Allison J.J., Kiefe C.I., Weaver M.T., Williams O.D.,
Centor R.M. & Weissman N.W. (2004) Nurse staffing and mor-
tality for medicare patients with acute myocardial infarction.
Medical Care 42(1), 4–12.
Rogers J.E., Hutchins S.G. & Johnson B.J. (1990) Nonpunitive dis-
cipline. A method of reducing absenteeism. Journal of Nursing
Administration 20(7–8), 41–3.
Roszkowski M.J., Spreat S., Conroy J., Garrow J., Delaney M.M. &
Davis T. (2005) Better late than never? The relationship between
ratings of attendance, punctuality, and overall job performance
among nursing home employees. International Journal of Selection
and Assessment 13(3), 213–219.
Sales A.E., Sharp N.D., Li Y., Greiner G.T., Lowy E., Mitchell P.,
Sochalski J.A. & Cournoyer P. (2005) Nurse staffing and patient
outcomes in Veteran Affairs hospitals. Journal of Nursing
Administration 35(10), 459–466.
SAS (n.d.). Statistical Analysis Software, Version 9.1. Retrieved from
http://www.sas.com/technologies/analytics/statistics/ on 10 Sep-
tember 2007.
Seago J.A., Williamson A. & Atwood C. (2006) Longitudinal anal-
yses of nurse staffing and patient outcomes. Journal of Nursing
Administration 36(1), 13–21.
Shamian J., Hagen B., Hu T.W. & Fogarty T.E. (1994) The rela-
tionship between length of stay and required nursing care hours.
Journal of Nursing Administration 24(7/8), 52–57.
Sheward L., Hunt J., Hagen S., Macleod M. & Ball J. (2005) The
relationship between UK hospital nurse staffing and emotional
exhaustion and job dissatisfaction. Journal of Nursing Manage-
ment 13, 51–60.
Shields M.A. & Ward M. (2001) Improving nurse retention in the
National Health Service in England: the impact of job satisfaction
on intentions to quit. Journal of Health Economics 20(5), 677–
701.
Taunton R.L., Krampitz S.D. & Woods C.Q. (1989) Absenteeism
– retention links. Journal of Nursing Administration 19(6),
13–20.
Taunton R.L., Kleinbeck S.V.M., Stafford R., Woods C.Q. & Bott
M. (1994) Patient outcomes: are they linked to registered nurse
absenteeism, separation, or work load? Journal of Nursing
Administration 24(4S), 48–55.
Thomson D. (2005) Absenteeism. In Quality Work Environments for
Nurses and Patient Safety (McGillis-Hall L., ed.), Jones and
Bartlett, Boston, pp. 229–357.
Tumulty G., Jernigan I.E. & Kohurt G.F. (1994) The impact of
perceived work environment of job satisfaction of hospital staff
nurses. Applied Nursing Research 7(2), 84–90.
Tzeng H.M. (2002) The influence of nurses’ working motivation
and job satisfaction on intention to quit: an empirical investiga-
tion in Taiwan. International Journal of Nursing Studies 39(8),
867–878.
Unruh L. (2003) Licensed nurse staffing and adverse events in hos-
pitals. Medical Care 41(1), 142–152.
Whitman G.R., Kim Y., Davidson L.J., Wolf G.A. & Wang S.L.
(2002) The impact of staffing on patient outcomes across
specialty units. Journal of Nursing Administration 32(12), 633–
639.
681