Professional Documents
Culture Documents
Measuring Clinical Nurse Performance Dev PDF
Measuring Clinical Nurse Performance Dev PDF
222-230, 1997
0 1997 Elsevier Smnce Ltd. All rights reserved
Pergamon Printed in Great Britam
0020-74X9/97 1617.00+0.00
PII: SOO20-7489(97)00009-6
Department of Nursing Studies, King’s College London, Cornwall House, Waterloo Road, London SE1 8WA, U.K.
Abstract
The development of the King’s Nurse Performance Scale to measure clinical nurse performance is described. Instrument
construction was informed by the Slater Nursing Competencies Rating Scale [Wandelt, M. A. and Stewart, D. S. (1975)
Slater Nursing Competencies Rating Scale. Appleton-Century Crofts, New York] together with key literature and the use of
expert opinion. The instrument was utilised to observe the clinical performance of senior student nurses (n = 99) and data
which were at the ordinal level were statistically analysed using a variety of non-parametric tests. Key findings of students’
observed nursing practice are presented in a separate paper (While et al., unpublished document). Internal consistency testing
of the King’s Nurse Performance Scaleusing Cronbach’s alpha coefficient revealed a promising alpha for the total instrument
(r = 0.93). The subsection alphas indicated that further refinement may enhance the strength of the instrument as a tool for
the measurementof performance in different domains of practice. The possible useof the Scalein the professional development
of newly qualified nurses is suggested.0 1997Elsevier ScienceLtd.
222
J. M. Fitzpatrick et a/./Clinical nurse performance 223
group’ (13 items); ‘physical’ (13 items); ‘general’ (16 instruments in use in the United Kingdom (Redfern
items); ‘communication’ (7 items); and ‘professional et al., 1994).
implications’ (17 items). Reliability testing has been
conducted (using inter-rater reliability, stability and
Enhancing content specificity
internal consistency tests) and construct, content, pre-
dictive and discriminant validity have been examined The King’s Nurse Performance Scale drew sub-
by Ager and Wandelt (1975). stantially upon the Slater Nursing Competencies Rat-
For example, as an index of inter-rater reliability ing Scale (Wandelt and Stewart, 1975). The latter,
interclass correlation coefficients were selected by however, was not without its limitations, for example,
Ager and Wandelt (1975) and were calculated using it was developed for use in the US over twenty years
the scores of pairs of observers who had rated the previously. It was essential therefore to ensure, as far
performance of three student groups (n = 74) sim- as possible, that content was specific to and rep-
ultaneously but independently. Values of 0.72, 0.75 resented a current understanding of nurse per-
and 0.78 were achieved, indicating a modest reliability formance in the United Kingdom (Anastasi, 1976;
when compared to the criterion of 0.80 which Nunally Nunally, 1978; Cronbach, 1984; Linn et al., 1991).
(1978) has specified. As a measure of the instrument’s Indeed, it has been asserted that failure to establish
internal consistency the odd-even split half reliability this requirement: “increases the probability that
and Cronbach’s alpha techniques were employed. The reliability estimates will simply describe the con-
odd&even split half technique produced a reliability sistency of irrelevant measures of subject effec-
coefficient of r = 0.98 which compares favourably tiveness” (Peterson et al., 1985). Thus, while
with Nunally’s (1978) criterion of 0.80. Cronbach’s acknowledging that an index of content validity can-
alpha technique, however, yielded a coefficient of not be computed, it was nevertheless imperative to
r = 0.74 for the total instrument using 71 of the 84 maximise quality of content.
items (13 items had inadequate sample sizes). In previous attempts to design performance evalu-
Coefficients for the six subsections have not been ation tools, a variety of strategies to generate and
reported by Ager and Wandelt. The variation between validate the content domain have been adopted,
the two coefficient measures may be explained by the including: incorporating current literature; drawing
possibility of different reliability estimates being upon programme curricula; seeking expert opinion;
obtained when different combinations of splits have and preliminary observation of people engaged in the
been used (Polit and Hungler, 1987). Further, Ager activity under consideration to identify key indicators
and Wandelt (1975) have suggested that the Cron- (Sommerfield and Accola, 1978; Stecchi et a/., 1983;
bath’s alpha measure in this instance may be under- Cottrell et al., 1986; Gould, 1993). The Critical Inci-
estimated due to unequal sample sizes for the various dent Technique has also been explored as a potentially
intercorrelations (p. 55). To explore the instrument’s useful approach to generate a valid content base (Gor-
underlying dimensions, factor analysis was conduc- ham, 1962; Gorham, 1963; Brumback and Howell,
ted, using 7 1 of the 84 items which had sufficient cases. 1972; Sims, 1976; DeBack and Mentowski, 1986). For
Factors with an eigenvalue > 1 were retained and 12 example, DeBack and Mentowski (1986) interviewed
in total were identified accounting for 83% of the staff nurses and nurse supervisors (n = 83) from three
total variance, with factor 1 accounting for 55%. The health care settings to elicit critical incidents of effec-
authors reported that on the varimax rotation items tive and ineffective nursing practice. A model of nurse
from the six subsections demonstrated some tendency competencies was developed from the interview data
to load on separate factors, which indicates that items and was subsequently used to score perceived differ-
were not exclusive to their subsections. Further details ences in the performance of baccalaureate (n = 37)
regarding where the factors drew their items from associate degree (n = 8) and diploma (n = 30) nurses,
have not been reported by Ager and Wandelt. as well as nurses with a higher degree (n = 5). Once
The original version of the Slater Nursing Com- again, however, the instrument was utilised to explore
petencies Rating Scale (Slater, 1967) was used in a perceived differences rather than actual-situated per-
pilot study by Christman (1971) to examine bacca- formance. Strategies adopted to enhance the content
laureate nurses’ performance (n = 42) using the specificity of the King’s Nurse Performance Scale are
organisation of nursing care as the independent vari- discussed in the forthcoming sections.
able. In another United States (U.S.) study, Petti
(1975) utilised the tool to obtain head nurse and pati-
The construction of scale items
ent ratings of nurse performance. Unfortunately,
neither Christman (1971) nor Petti (1975) reported Taking into consideration the above information,
having carried out further reliability and validity test- the King’s Nurse Performance Scale was constructed
ing of the Scale. Interestingly, the quality assessment with the aim of producing a generic set of observable
instrument Qualpacs (Wandelt and Ager, 1974) nursing actions reflective of nurse performance in the
which was also developed from the Slater Nursing United Kingdom (UK) and amenable to accurate dis-
Competencies Rating Scale, has been found to be the crimination. An analysis of the concept of nurse per-
most valid of three popular generic quality assessment formance identified key facets of the nurse role which
224 J. M. Fitzpatrick et aL/Clinical nurse performance
informed the development of the instrument, together etition and the inclusion of non-specific items. Analy-
with a critical review of the Slater Nursing Com- sis ofitems in the Slater Nursing Competencies Rating
petencies Rating Scale (Wandelt and Stewart, 1975). Scale informed the developmental process in this
Within the King’s Nurse Performance Scale, seven study. For example, items one and two of the ‘psy-
domains of nurse performance were formulated and chosocial: individual’ subsection of the Slater instru-
these are detailed below. Items generated from the ment highlighted the difficulty of generating discrete
literature as well as items drawn from the Slater Nurs- items. Specifically, item 2 refers to the nurse being a
ing Competencies Rating Scale (Wandelt and Stewart, receptive listener, however, this overlaps with item 1,
1975) were assembled to represent each area of prac- the focus of which is the need to give full attention to
tice and the first draft of the instrument consisted of the patient. No similar items were incorporated in the
67 items grouped into the seven domains of nurse King’s Nurse Performance Scale. Further challenges
performance. Items were illustrated with cues, the pur- included ensuring item mutual exclusiveness and logi-
pose of which was to facilitate the observer training cal coherence, as well as achieving discrete subsec-
programme, specifically, enabling the accurate identi- tions. For example, associated with attending to
fication and discrimination of items during obser- clients’ personal hygiene needs (physical domain) is
vation and minimising observer inference. For working in collaboration (use of the nursing process
example, the cues provided for item 4 ‘Ensures patient in planning care domain), and attending to clients’
receives fluid intake as appropriate’ were as follows: sensitivities (psychosocial domain). This challenge is
‘Ensures intravenous fluid is administered according highlighted further when examination of the internal
to regimen’; and ‘Acts upon the evidence of a main- consistency and construct validity of the instrument
tained intake and output chart’. To further facilitate is discussed.
quick and accurate recording in the field setting, each
item was identified as one whose rating will usually be
Validation of content
direct, one whose rating will usually be indirect (e.g.
from a written record) or one whose rating may be Key criteria for maximising content validity include
direct or indirect. The first subsection ‘physical an adequate collection of items which represent the
domain’ (14 items) focused upon nursing actions to domain of investigation (i.e. nurse performance) and
address the physical needs of clients. The ‘psycho- appropriate methods of test construction (Nunally,
social domain’ (6 items) referred to nursing actions 1978; Messick, 1989; Streiner and Norman, 1989).
which addressed the psychosocial needs of clients. The From the outset of instrument development, the
‘professional domain’ (9 items) centred upon actions objective was to select items which provided the most
directed towards fulfilling the professional role. Meet- accurate and representative description of effective
ing the knowledge needs of clients, self and others was nurse performance and this process was informed by
the focus of the ‘promotion of health and teaching an extensive review of the literature. The use of experts
skills domain’ which consisted of four items. Man- is also an accepted strategy to validate the content
agement of self and others was the focus of the ‘care domain of instruments (Anastasi, 1976; Cronbach,
management skills and organisation of workload 1984; Streiner and Norman, 1989) and a panel of nine
domain’ and was represented by six items. Com- experts drawn from clinical, educational and research
munication with clients and others was reflected in settings, together with the Steering Group of the larger
subsection six which consisted of five items. The final study, were involved in the process of instrument
subsection encompassed the use of the nursing process review.
approach to the planning and delivery of care (21 The King’s Nurse Performance Scale items and cues
items). The subsections were designed to reflect the were sent to each member of the expert panel and the
training regulations of pre-registration courses of Steering Group of the larger study who were asked to
nurse preparation (Statutory Instrument No. 1456, respond independently. Experts were asked for their
1989) the Code of Professional Conduct (UKCC, comments upon the Scale as a whole, and the appro-
1992) and criteria for effective nurse performance priateness of subsections and individual items. The
which were derived from a wide literature base. For experts were asked to review the instrument for clarity,
example, it is expected that the newly qualified nurse comprehensiveness, mutual exclusiveness and to sug-
should be able to devise, implement and evaluate a gest any additional items for inclusion. They were also
plan of care (Statutory Instrument No. 1456, 1989; asked to score each item as an observable indicator of
UKCC, 1992). The conceptual framework under- effective nurse performance and to assign a rating on
pinning this, the nursing process, was therefore a scale of O-5 (0 being totally irrelevant and 5 being
reflected in the subsection: ‘use of the nursing process absolutely essential). It emerged, however, that the
in planning care’ and the work of several key authors scoring system did not always correspond with com-
contributed to item generation (Mayers, 1978; Yura ments regarding the technical quality of items. Thus
and Walsh, 1978; Kratz, 1979; Brooking, 1986; Hunt while an item may have been considered an important
and Marks-Maran, 1986). indicator of effective nurse performance its con-
Some of the challenges associated with the process struction required modification. Further, in some
of instrument development included avoiding rep- cases experts chose to note comments rather than
J. M. Fitzpatrick et uI/Clinical nurse performance 225
items and Bondy’s (1983) rating criteria were modified Cronbach’s alpha technique. The latter provides a
(Fitzpatrick et al., 1996). No item reduction occurred. good estimate of reliability in most situations since
Internal consistency testing of the tool and principal the major source of measurement error focuses upon
components analysis were not conducted at this stage sampling of content (Nunally, 1978). The total Scale
due to the small data set. Examination of inter- yielded an alpha of Y = 0.93 using the means of the
observer reliability and observer drift are discussed in subsection scores (n = 98, 1 missing case due to
a separate paper which details operationalisation of insufficient data for analysis). Since the Scale com-
the instrument (Fitzpatrick et al., 1996). prised of seven subsections it was also important to
examine the internal consistency of each (Waltz et al.,
1991). Using the item means it was only possible to
Reliability and validity testing of the instrument
calculate an alpha coefficient for 3 of the 7 subsections
As highlighted previously, the target population for (where >50 cases were included). The coefficient
this research were students completing their nursing alphas were: ‘physical domain’ Y = 0.74 (n = 51); ‘pro-
programmes for Part 1 (RGN) and Part 12 (Adult fessional domain’ Y = 0.70 (n = 60); and ‘promotion
Nursing) of the Register. Non-participant observation of health and teaching skills domain’ r = 0.71
of senior students representing the three programmes (n = 61). The subsection alphas did not reach
(n = 99) took place in the hospital setting and each Nunally’s (1978) criterion of 0.80 which suggests
participant was observed on three separate occasions independent components within the instrument.
at different times of the day, totalling 742.5 hr of Interestingly, insufficient score data in the ‘psycho-
observation. Data analysis, which commenced on social: group’ section for Qualpacs (Wandelt and
completion of data collection, involved calculating a Ager, 1974) ratings was a significant problem in the
total mean performance score and a mean score for research of both Carr-Hill et al. (1992) and Redfern
each of the seven domains for each participant. The et al. (1994).
formula used was: the sum of weighted totals divided An alternative method was also used to compute
by the total number of ratings. The sum of the total coefficient alphas for the subsections, that is, using the
weighted scores reflected the performance levels: inde- mean score for each observation period. Using this
pendent (4) assisted (3) marginal (2) and dependent method it was possible to compute coefficient alphas
(1). The omitted column was not assigned a numerical for the seven subsections, however, once again they
weighting but was reflected in the total number of did not reach the criterion of 0.80 (Nunally, 1978):
ratings for each observation period. Non-parametric ‘physical domain’ Y = 0.58 (n = 99); ‘psychosocial
tests were applied to explore relationships between domain’ Y = 0.53 (n = 62); ‘professional domain’
the different participant groups and key findings are r = 0.60 (n = 99); ‘promotion of health and teaching
presented elsewhere (While et al., 1996). The data were skills domain’ Y = 0.46 (n = 93); ‘care management
used to conduct reliability and validity testing of the skills and organisation of workload domain’ r = 0.60
instrument, details of which are presented below. (n = 99); ‘communication skills domain’ r = 0.65
(n = 99); and ‘use of the nursing process in planning
care domain’ r = 0.72 (n = 99).
Internal consistency testing of the instrument
In summary, internal consistency testing suggested
The data were used to examine the internal con- some independence within the subsections of the
sistency of the King’s Nurse Performance Scale using instrument. There was evidence nonetheless of overall
Table1
Content and percentage of variance for each component
I Provides for the psychological, social and physical needs of patients using a 43.5
multidisciplinary approach to planning and delivery of care
2 Self-directing care management and organisation of workload 9.3
3 Attention to patient safety with adherence to regulations. policy directives and 1.4
research findings
4 Attention to patients’ physical care needsfor hygiene with sensitivity and 5.2
encouraging patient participation in care
5 Attention to patients’ activity in accordance with their current and potential 4.6
health status
6 Safe administration of intravenous/parenteral fluids with patient consultation 4.2
and teaching
I Effective communication about patient supported by patient participation in the 3.x
evaluative process
8 Attention to patients’ dietary intake appropriate to associated actual/potential 3.2
problems
228 J. M. Fitzpatrick et &./Clinical nurse performance
Table 2 ency to load on separate factors (Table 2). The 14
Number of items within each component items which loaded on component 1 are set out in
Table 3 and it was noteworthy that the items were
Component Number of items
drawn from six of the seven subsections. Component
1 15 (11. 14, 16, 21-23, 27-29, 32-35, 38, 41) 1 appears to focus upon two inter-related areas of
2 14 (4, 7, 11-12, 20-21,23,27-28, 32, 37-38, nursing: (i) an individual and holistic plan of care
41,51) including skilled communication, performing effec-
3 11 (&9, 12, 16, 19, 27,40,4748) tively and responsibly in the planning and delivery of
4 9 (24, 16, 22, 38, 41, 47, 51) care with a multidisciplinary approach (derived from
5 8 (1, 11-12, 20, 24, 29, 40, 51 10 items); and (ii) the psychosocial component of nurs-
6 7 (6, 21-22, 24, 28, 32, 47) ing care and in particular developing and maintaining
7 4 (16, 29, 37, 51) a therapeutic nurseslient relationship (derived from
8 3 (5,20,40) four items).
Table 3
Items loading on first principal component