Phase 1: investigating at the macro and meso levels of health systems
This phase had three elements: (1) a review of evidence, (2) a scoping
survey74 of key informants at
a national and regional level and (3) a survey of current work deployment of PAs
in UK general practice.
The review of evidence
In this element we undertook three review activities. The first was a review
of empirical evidence using a rapid review methodology76 for empirical evidence of the deployment and
impact of PAs (research questions 1 and 2; see Chapter 1, Study aims and
objectives). The second was a narrative review77 of published commentaries and
opinions of the PA role in the UK setting. The third was a narrative
review77 of national and
international policy and developments on health-care workforce planning in
relation to PAs and primary care. The last two activities addressed the
question as to factors supporting or inhibiting the presence of PAs in the
primary care workforce.
The review of empirical evidence
A review78 was conducted to
investigate the evidence of the contribution of PAs to primary care as
applicable to a European model. The European model of general
practice/family medicine was that given by the European region of the
World Organization of Family Doctors79 (WONCA Europe). Seven databases [British
Nursing Index, Cumulative Index to Nursing and Allied Health Literature
(CINAHL), EMBASE, MEDLINE, PsycINFO, Social Sciences Citation Index
(SSCI) and Scopus] were searched from 1967 (or start date if later) to
September 2010. A broad range of search terms was used singly and in
combination: physician assistant, primary health care, family practice,
general practice, family physician, general practitioner (see
Appendix
3).
The inclusion criteria were defined as follows:
The PA role was one with a recognised PA qualification.
The setting was of general practice/family medicine (including
community paediatrics in the USA) or the data were identified as
pertaining to ‘family medicine’ within other
findings from broader ‘primary care’.
The exclusion criteria were defined as follows:
Settings of secondary care, in-patient care ambulatory or
outpatient care.
Health care provided by primary care specialists in the USA
(obstetrics, gynaecology, internists or primary care physicians)
that would be regarded as secondary care in the UK.
The personnel were student PAs or nurses/others who had received
special training as medical assistants (that is, without a
recognised PA course) for a particular disease condition.
The data presented did not distinguish either between PAs and
other professionals or between the general practice/family
medicine setting from ‘primary care’ overall in
its presentation of results.
Retrieved abstracts were screened and full papers, where available, were
obtained. Each full paper was assessed for inclusion by two researchers
independently and a third (HG, SdeL, VMD, and MH) in cases of indecision
or disagreement. A data extraction framework was developed which
included categories of PA activities, study methods and key findings.
Study outcome measures were categorised using Donabedian’s80 framework of structure,
process and outcome. Owing to the wide range of methods used in the
retrieved studies, a broad quality assessment of the studies was
undertaken using critical appraisal tools81 appropriate to the papers’
methods (e.g. qualitative, cohort, health economic) with additional
questions derived from the British Medical
Journal’s guidance for peer reviewers.82 The heterogeneity of the
retrieved studies precluded any pooling of data and a narrative
synthesis was undertaken using the Donabedian80 framework.
Review of commentaries
A documentary analysis83
was undertaken of published commentaries and opinion pieces about PAs in
the UK health-care system or addressed to UK audiences. The published
material was identified in two ways: first, through the electronic
database search described in the previous section and, second, by an
internet search using the Google™ (Mountain View, CA, USA)
search engine. The internet search used a combination of terms:
physician assistant, physician assistants, National Health Service
(NHS), primary care, secondary care, England, Scotland, Wales, Northern
Ireland, pilots and workforce, and was repeated at intervals by WC from
April 2011 until December 2012. ‘Published’ was defined
as in print and electronic formats including online forums between 1980
and 2012. The lower date was determined by our previous knowledge of the
first paper published on the experience of a PA in UK primary care.58 A data extraction
framework was developed which included a categorisation of the author
(journalists, doctors, nurses, health service managers, etc.), the
intended audience, any catalyst for the item (e.g. another published
item), inclusion of any empirical evidence and from which country,
evidence of support, resistance or neutrality in the publication. The
items were recorded chronologically with the data extraction in a
Microsoft Excel (Microsoft Corporation, Redmond, WA, USA) spreadsheet. A
narrative synthesis was drawn from the data.
Review of workforce policy and developments
The purpose of this review was both to identify policy and developments
pertaining to PAs in the primary health-care workforce in the UK and
internationally in order to both help the research team contextualise
the study and to identify supporting or hindering factors in the
development of the PA profession in the English health-care workforce.
The review drew on the methods of the documentary review.83 Publications were
identified through the searches described above. Additional searches
were performed to identify polices and reports on health-care workforce
planning and development relevant to, primarily, the education and
employment of PAs and, secondarily, primary care. The time period
covered was from the mid-1990s to the present day. The searches were
conducted by:
visiting key websites in the UK, for example government health
departments in each of the four countries, the Centre for
Workforce Intelligence (CfWI) and Strategic Health Authority
(SHA) websites
using keyword searches in the Google search engine for open
source materials related to the topics of interest, for example
‘primary care workforce’, ‘physician
assistant/s’ (and other titles to describe this
profession, such as medical care practitioners) and
‘health care workforce’
snowballing techniques of references and website links.
The documents identified in this way were screened to include only those
with relevance to health-care workforce in primary care. Research team
alerts of new policy or reports relevant to the study were generated by
one researcher (WC). A data extraction framework was developed to
classify the document by country, type of author, type of document (e.g.
policy, analysis of primary data, opinion and evaluative reports), any
reference or absence of reference to PAs and key points from or about
the document. The information was saved on an electronic database with
full references, any URLs and electronic copies of the original. A
narrative synthesis was undertaken addressing the questions of
interest.
Scoping survey
A scoping survey74 was
undertaken to address research question 3: what supports or inhibits the use
and development of PAs in the primary care? (See Chapter 1, Study
aims and objectives.) The method is one element of a scoping
review74 and allows for
data to be gathered from those knowledgeable in the issue of interest,
providing a range of perspectives to ensure breadth in the analysis.84 It is an approach that has
been widely used, most recently in exploring European perspectives on human
health resource issues.85 A
purposive sample of key informants at national and regional levels in
England and Wales was approached to participate. These were individuals at a
senior level from central departments of health, professional organisations,
regulatory organisations, patient organisations, and commissioning bodies
who are concerned with the development of primary care services, workforce
development and health professional regulation. At the point of the design
of the study (2009) the Department of Health and Social Services (Wales),
unlike the Departments of Health for England and Scotland, had not explored
the use of PAs. In order to consider diversity in national environments,
individuals were approached to participate in England and Wales. Individuals
were identified for approach through an internet search of public domain
organisational sites, suggestions of the study advisory group and members of
the research team, and snowball techniques from participants. An
aide-memoire was developed of topic areas, based on the issues of interest,
to be used in semi-structured interviews83 which were offered either face to face or by
telephone, dependent on participant preference. The aide-memoire explored
knowledge of mid-level practitioners, PAs and perceptions of the factors
that currently and will in the future support or inhibit their use and
development in the UK settings. Notes were taken during the interview and,
with permission, digitally recorded. These were transcribed, anonymised and
the recordings deleted. The transcriptions and notes were analysed using the
constant comparison method86
by two researchers (WC and VMD) independently, compared and further analysis
confirmed with a third researcher (JG). The analysis of the early interviews
assisted in identifying further stakeholder groups from whom views were
sought. This element of the study was deemed service evaluation by the NHS
ethics query service and as such not for research ethics review. It
conformed to all university requirements.
Survey of current use of physician assistants in general practices in
England
An online, anonymous, self-completion survey was developed and piloted with
two PAs who had previously worked in general practice in England. It posed
questions of career history, work environment, work activities and estimates
of volumes of work. PAs known to working in primary care were invited to
participate in 2010 through e-mails sent out by the UK Association of
Physician Assistants (UKAPA), course leaders of PA courses in English
universities and practices identified as employing PAs in previous
research.72 All invitation
e-mails included a link to the survey website and requested forward
distribution to any other PAs thought to be working in primary care. UKAPA
estimated that there were about 25 of their member PAs employed in primary
care. The electronic survey was open for responses for 7 weeks (7 October to
30 November 2010). It addressed three areas: the activities and role of PAs
in the primary care team, their primary care setting and their
qualifications. Those interested in volunteering for the second phase of the
study were given an opportunity to give their contact details at the end of
the survey. The survey did not require NHS ethical review and fulfilled all
university requirements for the ethical conduct of research. Descriptive
analysis of the data was undertaken using Statistical Product and Service
Solutions (SPSS) (IBM Corporation, Armonk, NY, USA).
Phase 2: the micro level
A comparative case study design75
was used to address the three overarching research questions at the micro level
of analysis through examination of 12 case studies: six general practices
currently employing PAs and six matched practices not including PAs in their
staffing. A mixed-methods approach87 allowed for both description and the quantification of the impact
of PAs in the context of general practice. This element of the study addressed
specific additional questions:
How are PAs deployed and supervised in general practice and how do their
roles and responsibilities compare with those of the doctors, NPs and
practice nurses?
How do PA outcomes of care differ from those of GPs, specifically with
respect to prescriptions rates, referrals, investigations, reattendances
for the same problem within 2 weeks, patient safety, and patient
satisfaction?
How do patients understand the role of PAs and what is the experience of
patients when consulting PAs?
How does employment of PAs affect practice organisation, staffing
configurations and costs?
What are the factors that support or inhibit the employment of PAs?
The six practices employing PAs were identified either by the PAs initially
volunteering through the survey (see Survey of current use of physician
assistants in general practices in England) or by an expression of
interest from a practice. The volunteering PA practices achieved diversity in
features such as general practice size, practice population sociodemographic
characteristics, rural and urban coverage, and in the country in which the PAs
had qualified. The comparative that was designed required some similarity
between PA and comparator GP practices. Following the agreement to participate
from PA employing practices, comparison practices were sought in the same health
economy and with similar characteristics and patient population size. After the
adoption of the study by the UK Clinical Research Network (UKCRN) Primary Care
Research Network (PCRN), officers in five local PCRNs made the initial
approaches and requests for comparison general practices. Practices expressing
interest were then contacted and provided with information, and an initial
meeting was set up between the researchers (VMD and MH) and the GP, the practice
manager and, in those employing practices, the PA. A summary description of the
practices is provided at the end of Chapter 2 (see ).
Summary of the practice settings
The general practices and staff likely to be included in the study were provided
with full participant information sheets for all research activities. Written
agreement and consent for the practice to participate was obtained from the
senior partner GP. Procedures for obtaining individual informed consent for each
of the research activities are described below.
Data were gathered through the activities detailed in the following sections in
this chapter. The analysis is described in each section and the economic
analysis in Economic analysis. Analyses were then combined and
synthesised to address the research questions as outlined in .
Summary of data to address research questions
Semi-structured interviews with practice staff
Semi-structured interviews83
were undertaken with PAs, GPs, practice managers, receptionists and practice
nursing staff. A purposive sample in each practice was approached, provided
with information and invited to participate. Aide-memoires for each
professional group were developed addressing the questions (a), (b), (d) and
(e), above, but allowing flexibility to explore new issues or ideas raised
in the interview. GPs, PAs and nurses were also asked to describe their
responsibilities and activities within the practice. In addition, practice
managers and GPs were invited to provide any practice documents that would
provide the research team with information about the practice, its working
practices, services provided, staffing and planning. If documents on
staffing were not available, the practice managers were asked to provide
this information verbally. The interviews were conducted face to face or by
telephone as preferred by the participant, digitally recorded with consent,
transcribed and erased after analysis. Thematic analysis was undertaken88 initially by one researcher
(RL), with further analysis by two other researchers (VMD and MH) and
interpretative discussion with the wider research team. Documentary evidence
on staffing was read and data extracted as relevant to providing contextual
information for each practice.
Consultation record review and linked patient survey
The consultation record review and linked patient survey, which addressed
question (b) above, was of patients presenting in urgent or same-day
surgeries. This was the type of surgery that PAs were predominately working
in as identified from the national survey (p. 9). Each practice identified
time periods in summer and winter periods (known as reference periods)
during 2011 and 2012 in which they were able to assist the research team in
data gathering. The length of time periods were based on the
practice’s estimation as to the number of surgeries required to
reach the sample size. Each practice made staff and patients aware of the
research during the reference periods in keeping with their usual practice,
for example via notices at the reception desk.
The patient surveys
The validated General Practice Assessment Survey (GPAS)89 focuses on patient
outcomes and consultation experience and was used with permission (Paula
Metcalfe, GP Patient Survey Team, 12 October 2010, personal
communication). During a meeting of the study’s public and
patient participation group (see Chapter 1, Study aims and
objectives) the GPAS was scrutinised and amendments
suggested to improve the clarity of the wording and to specifically
address the experience of a consultation with a PA. The amended,
self-completion GPAS (see Appendix 4) was then to be offered to adult
patients in each designated surgery during the reference periods. The
method of distribution and collection varied in each practice, according
to their working practices. In some practices, the GP or PA handed the
patient a sticker to return to the receptionist, following the
consultation, to collect a survey pack (containing a questionnaire,
participant information sheet, and business reply envelope). In other
practices, the survey pack was sent by post either the same day or the
next. All completed questionnaires could be returned in a sealed
envelope to a collection box in the practice reception or by post to the
research team. The questionnaire did not include any information that
could identify the person. Each questionnaire had a unique study
identifier (study ID). Practice staff kept a log of the dates of survey
administration (the index event) and recorded the unique practice
identifier of the patient against the study ID. Practice staff were
asked to send out a research team prepared reminder and duplicate survey
pack to all those who had taken or been sent a questionnaire who had not
responded. The practice staff did not have any access to any of the
completed questionnaires, ensuring that anonymity was protected. Data
were entered onto a SPSS spreadsheet and descriptively analysed. The
data were then also added to the data drawn from the consultation record
review as described below.
In the PA-employing practices, the survey pack included additional
material inviting patients to participate in a face-to-face or telephone
interview. A separate consent form and business reply envelope were
provided to respond to the invitation.
Consultation record review
This review addressed the primary outcome of the study: the rate of
reattendance in general practice or an urgent care facility defined for
the study as GP out-of-hours care, emergency department, walk-in centre,
minor injury unit or other direct entry emergency unit (e.g. early
pregnancy assessment unit) within 2 weeks of the initial consultation
(the index consultation) with a PA or GP, and questions of whether or
not the initial consultation activity by the PA or GP was appropriate in
the light of the subsequent reattendance. The consultation records also
provided the data for the process measures of advice, investigations,
treatments, referrals and prescriptions.
The sample
The sample size required for the primary outcome was based upon
randomised controlled trial data which compared the rate of
reattendance in general practice within 2 weeks and outcomes of care
delivered by NPs and GPs in the UK for patients attending
‘same-day appointments’. These studies reported
figures ranging from a self-reported 18.2%90 through to extracted medical records
28.4%91 and
29%92 for patients
seeing GPs and 20.4%, 31% and 37.2% for NPs, respectively. Taking
Venning et al.’s91 analysis as the most robustly
adjusted of these, the reattendance rate within 2 weeks of the
consultation for this sample size calculation was taken as 28.4% for
GPs and 37.2% for NPs. Assuming an OR of 1.5 and adjusting for
slight multicollinearity as reported by Venning et
al.,91 we
estimated that a sample size of 205 in each group (consultations
with a PA or a GP) was needed to give 80% power at a significance
level of 5% for a logistic regression using the covariates of age,
gender and general practice. We estimated that anonymised clinical
records data might be required for up to approximately 600 patients
each attended by PAs and GPs in the 12 practices in the data
collection periods in order to meet the required sample size,
assuming a response rate of 30% to the patient survey. However, in
conducting the study, two problems arose. The first was that a
condition of the research ethics approval was that child
consultations were excluded from the survey element but could be
included in the review of clinical records. We therefore included
all child consultations in the designated surgeries in the reference
periods as our review and survey had shown that PAs were consulted
by children and their parents. The second problem arose through
inaccurate research team and practice staff estimates of patient
survey return rates in their practice (i.e. lower than our
assumptions) resulting in more surgeries being designated for survey
distribution. In some practices, clinical records were passed to the
research team some months after the designated surgeries and only
then did it become apparent to the research team that these had much
greater numbers than anticipated. In addition, in some practices,
there was a lack of fidelity to the protocol in that not all
eligible patients were offered the survey (in some instances only
30% were offered). This was apparent to the research team staff only
when they returned to discuss the participant numbers and check that
the reference periods were similar to those of the preceding weeks
for the practice. As the possibility of selection bias could not be
eliminated, the research team decided to include all of the clinical
records sent by the practices. This resulted in a larger sample than
in the study design.
The data
Data from the electronic records of those patients who consulted in
the designated surgeries during the reference periods were extracted
and all patient identifiers removed in the practice. These data were
then passed to the research team using the unique study ID number
only. The data extraction occurred at a time point more than 2 weeks
after the end of the reference period. The data extracted by the
practice staff for the index event included the consultation date,
person consulted (PA or GP) and length of consultation; the
patient’s age, gender, ethnicity and postcode; the
patient’s presenting conditions; number of prescriptions
from the consultation; the number and type of Quality and Outcome
Framework93
chronic conditions registers the patient was entered on; the number
of repeat prescriptions; and the number of times the patient had
contact with the practice in the three months preceding the index
event. For any consultations in the following 2-week period, the
location of consultation, the professional consulted, presenting
condition(s) and number of prescriptions issued were recorded. The
research team used the postcode on each record to assign the Index
of Multiple Deprivation (IMD)94 and then deleted the postcode. The practice records
varied in the extent to which Read codes95 were used for the presenting problem
or diagnosis. The research team subsequently reviewed each record
and assigned Read codes95 to presenting problems that were documented but not
coded by practice staff and assigned study codes for groups of
diagnostic tests, referrals, procedures, certificates, medicine
management and advice given (see Appendix 5). Each consultation
record was coded by one member of the research team and then checked
by a second. Uncertainties were discussed and resolved with a third
member of the team. A categorisation of the presenting problems in
the index consultation was also assigned to each patient record in
order to examine the case mix of the consulting patients. This
classification of acute, chronic, minor/symptoms, prevention and
processes was an adaptation of that used by de Jong et
al.,96
developed by the research team in which the GP in the research team
(SdeL) made the final decisions on categorisation and assignment.
The method of assignment of a condition category to each patient is
presented in Appendix
6.
Each index consultation that had subsequent consultations was
identified. Each subsequent consultation was coded by two research
team members to identify whether it was for exactly the same, a
linked or a different problem to that of the index consultation. The
method for designation was developed and refined with the GP
research team member (SdeL).
All data were entered into a SPSS spreadsheet, to which the data from
the patient survey (where available) could be matched by the study
unique identifying number.
The data were then analysed descriptively, involving summary measures
of location (e.g. means /medians/proportions) and dispersion
[standard deviations (SDs)/percentiles] appropriate for the type and
distribution of the individual variables. Differences in processes
and outcome between those practices with and without PAs were
assessed by generalised estimating equation models.97 The resulting ORs or
rate ratios were adjusted for the case mix and IMD variables where
they were significant predictors or notably confounded the
relationship between practice type and the dependent variable. The
correlation matrix was set to be exchangeable, that is to say
patients were assumed to have some shared characteristics within
each practice, whether from demographics or organisation of the
practice.
The clinical review of reconsultations records
All records in which a patient had reconsulted within 2 weeks for the
same or a linked problem were assigned to clinical record review. The
review of medical records, or ‘chart review’, is a
longstanding method of assessing the consultation98 used in assessing overall quality99 and considered reliable
in detecting prescribing errors.100
An anonymised printout of the computerised medical record system was
reviewed to assess the quality of the record and to assess the
appropriateness and safety. The data on whether it had been a GP or PA
who had conducted the index consultation were removed, as well as any
other references that might identify the practice or its geographical
area (e.g. hospital name). A clinical review process, which addressed
the question as to whether or not, in the light of the subsequent
consultation(s), the first consultation had been appropriate, was
developed by the GP member of the research team (SdeL) in consultation
with the clinical review team of four GPs. The four GPs all had
experience of assessing the quality of the consultation and, to a more
limited extent, of the medical record. They were all active in
education: one was a trainer while the other three took medical students
in their practices; two were research active; two were male, two were
female. They were trained in two 2-hour training sessions led by
SdeL.
Based on the problem-orientated medical record format,101 the experienced GP
reviewers were asked to judge each element of the consultation, the
overall appropriateness and whether a GP or a PA had been consulted.
Weed’s taxonomy,101,102 for
the elements of a problem-orientated medical record, was used as an
measure of quality of the record, and the opinion of experienced GPs
about appropriateness of the content of each part was used for the
problem presented. The GP raters also gave an overall assessment.
Weed’s101,102
approach was to suggest that there were four key elements of the record,
described by the acronym SOAP:
Subjective: what the patient complains of, the problem from the
patient’s perspective.
Objective findings from the consultation (e.g. blood pressure
measurement).
Analysis by the clinician of the problem.
Plan for future action, including follow-up.
We modified this by making it explicit that the plan could have different
elements: investigations, prescriptions and other aspects of the plan,
such as referral or follow-up ().
Modification of Weed’s SOAP to judge the appropriateness of
the medical record
Each record was assessed by a GP and 20% by two GPs independently using a
study-designed form (see Appendix 7). Data were entered into a SPSS
spreadsheet, analysed descriptively and inter-rater reliability was
assessed. Inter-rater reliability is a measure of the extent to which
the reviewers agreed with each other in their responses.
Fleiss’s kappa103
is a simple measure of this inter-rater reliability. A kappa score of 0
indicates no more agreement than one would expect to occur simply by
chance, and a score of 1 indicates complete agreement on all the data
that were rated. Negative scores indicate agreement that is worse than
chance. It is often said that a kappa score of 0.0–0.2 is
‘poor’, one of 0.21–0.4 is
‘fair’, 0.41–0.6 is ‘moderate’,
0.61–0.8 is ‘good’, and 0.81–1.0 is
‘very good’.104 The raters assessed one case/medical record as a group.
Next, they analysed three cases on their own and checked back
afterwards. They then did a further 10 cases, which were assessed for
inter-rater reliability.
Patient interviews
Patients who had consulted PAs and received a survey were also invited to
participate in telephone or face-to-face interviews. This was designed to
address question (c): how do patients understand the role of PAs and what is
the experience of patients when consulting PAs? An aide-memoire was
developed to explore issues not captured by the patient survey such as
patient choice, understanding of the role of the PA, perceptions of the
consultation with regard to communication and perceived expertise. A
researcher (WC) contacted those who responded to the invitation as soon
after the index consultation as possible. With consent, the interviews were
digitally recorded, transcribed and analysed thematically88 by three researchers (LJ, MH,
WC) using the NVivo software programme (QSR International Pty Ltd,
Doncaster, Australia) (LJ, MH).
Observation of physician assistants and general practitioners
Patient consultations were video recorded, with consent of the practitioner
and the patients, during one surgery per PA and GP. This element addressed
questions (a), (b) and (e). We recorded GPs and PAs using two video
channels, and assessed competence at consulting using the validated
Leicester Assessment Package (LAP) rating scale.105 Video recording is an established method of
assessing the consultation106
and the use of two channels was a compromise between the multichannel
recording we planned106 and
what we could gain the NHS REC permission for.
Patients in the designated surgeries were provided with information and
consent was sought in writing by a researcher in the waiting room before
recording commenced. Recording was stopped at the end of each consultation.
Patients were then asked whether they continued to give consent or would
prefer that it was withdrawn and the recording deleted. Practitioners also
reconsented to the inclusion of each recording or could ask for a recording
to be withdrawn at the end of the surgery. The recordings were made using
ALFA: a computer-mediated consultation observation technique developed to
provide an analysable overview of the consultation.107 Each consultation was assessed by two GPs
independently using the LAP105 for assessing competence in general practice on dimensions
of interviewing and history taking, patient management, problem solving and
behaviour and relationship with patients.
The LAP for video consultation is a tool for measuring competence by direct
observation. It was initially developed in the 1990s and demonstrated to
have validity108–110 and reliability across
levels of clinical competence.107 Testing of reliability found a range of mean scores between
51.3% and 70.2% in one study, depending on levels of experience,110 and between 45% and 75% in
another looking at the effect of interruption.109 Although it has been criticised, no better
package has been developed.111,112 The LAP
has also been used in a range of clinical contexts – assessment of
cancer patients,112 the needs
of caregivers113 and in the
assessment of psychosexual problems.114 This package has also been used in an international setting,
where a 50% pass mark was included,115 for assessing practice nurses116 and medical students.117
Grades were given by the four raters, using the seven main categories of the
assessment package, and the results were then analysed. The seventh element
of the package, the record, could not be rated as it could not be properly
observed from the two-camera set-up used. Record keeping referred to how the
health-care professional (HCP) interacted with the computer. It was not
possible to view the detail of the individual record being collected by the
HCP owing to the fact that only two channels of video were being recorded,
and an additional channel is required for recording video directly from the
HCP’s computer screen. Three of the seven categories –
interview/history taking, patient management and problem solving –
are given a double weighting, and count for 20% of overall marks
(). When combined, this means that the LAP scale total is
100, allowing final results to be expressed as a mean score (out of 5 for
each category) or be expressed as a total percentage score.
Components of the LAP (adapted from Fraser et
al.)
A workshop was held to discuss the findings with the raters (unblinded) which
was video recorded and transcribed.
Observation was also undertaken in a sample of practice clinical meetings in
which PAs and GPs were both present to understand more broadly how the PAs
interact with the GPs in clinical decision-making in the practice.
Permission and invitations was sought from the practice manager, the GPs and
the PA. Permission and consent to observe was sought at each meeting that
the research team was invited to. Field notes were made of only the activity
and interaction of the team members. Analysis of these notes was used to
complement data from the staff interviews (see Semistructured
interviews with practice staff) and consultation video
observations.
Work activities of physician assistants and general practitioners
Work activity diaries, using the 2006–7 general practice workload
survey,118 were offered
to PAs to complete for 1 week in the summer and winter. This addressed
question (a), above. Early discussions with GPs made it evident that these
diaries were unlikely to be completed by practice professionals other than
the PAs and the data were therefore sought through interviews (see
Semi-structured interviews with practice staff) and
administrative systems. Data were entered onto a SPSS spreadsheet and
analysed descriptively.
Economic analysis
The economic analysis was conducted at two levels:
The resource implications and costs of different team configurations
were explored using practice-level data on staffing. Skill mixes,
and the relative contribution of different practitioners, were
compared within and between practices that do and do not employ PAs,
after consideration of case mix differences. Nationally validated
scales
119 were
used to calculate total human resource costs and costs per
patient.
The impact of PAs, compared with GPs, on consultation processes (e.g.
prescribing, referral and investigations), consultation outcomes
(e.g. reattending, consultation satisfaction) and length, at
individual patient level, as analysed in the clinical record element
were incorporated in a high-level analysis of costs and
effectiveness.
Practice level
Data
Each of the 12 practices (six with and six without a PA) was asked in
November 2012 to provide information about their staffing
configurations – headcounts and whole-time equivalents
(WTEs) of:
clinical personnel
– GPs (partners and salaried/other)
– mid-level professionals – PAs and
NPs
– nursing and other staff engaged in direct
patient care – practice nurses (PNs), HCAs,
phlebotomists
support staff
– managers and professionals [e.g. for
information technology (IT), finance]
– secretarial, reception, clerical.
The number of patients on the list of each practice and the
proportions aged > 65 years and < 15
years were obtained from the Department of Health NHS Commissioning
Board.120 IMD
2010 for each practice was obtained from the NHS Information
Centre.121
Costs
Unit costs of clinical personnel were obtained from validated
national sources.119
These unit costs provide a value, per annum, that covers
remuneration, on-costs, qualifications (where relevant), training,
capital and other overheads associated with each role (see
Appendix
8). PAs were costed as Band 7 nurses (secretary
of UK Association of PAs, 2012, personal correspondence). The same
costs were applied to GP partners and salaried GPs because evidence
on this was contradictory. One study suggested that partners cost
less than salaried GPs122 and others argued partners cost slightly more123 or the same.124
The remuneration levels of management and other support staff were
obtained by looking at several job vacancies within the NHS for each
role (outside London rates) in January 2013, and inferring an
average salary. The on-costs and overheads associated with each role
were assumed to be equivalent to those of clinical NHS staff working
in general practice with the same remuneration. For example, the
salary of a NP was £38,800 and job advertisements for
practice managers were around the same level, so the on-costs and
overheads of the NP were used to estimate the costs of a practice
manager.
Analysis
The numbers of WTE staff in each category, and numbers of patients
per GP, were calculated and compared within and between practices
with and without a PA. Reported vacancies were included in the
calculations.
Staffing costs were calculated for each category in each practice on
an annual basis by multiplying the WTE by the unit cost. Total staff
costs for each practice per annum were obtained by summing across
all staff categories. An average cost per patient per practice was
calculated, and compared within and between practices with and
without a PA.
The sample of practices included in the study in both the PA and
no-PA groups were compared with national staff headcount data from
the NHS Information Centre 2012125 with respect to the proportions of
GPs who were salaried (vs. partners), and the numbers of
administrative/clerical staff and nurses/direct patient care staff
per GP.
Patient level
Analysis
The patient-level analysis used findings from the analysis of
consultation records and the patient survey to compare the costs and
effectiveness of consultations by PAs and GPs. The mean length of
consultations in minutes for PAs and GPs was calculated from
available data recorded by the ‘time enter’ and
‘time leave’ functions on practice management
systems. Nationally validated unit costs119 were applied to mean consultation
lengths to explore differences between PA and GP practices in the
cost of consultations. Other ways in which the use of PAs may affect
practice costs were also considered, such as GP time spent in
supervision and training, but could not be quantified as data were
not available. The relative effectiveness of PAs and GPs was judged
by rates of reconsultation (primary outcome), patient satisfaction,
and rates of referring, prescribing, testing, fitness to work
certification, procedures and advice giving (process outcomes).
Adjusted values of variables were used, after controlling for
covariates.
Ethics and research governance
Phase 2 of the study required a number of types of reviews and agreements.
The research ethical review was undertaken by NHS Research Ethics Committee
(REC) South East Coast – Surrey who gave approval but required that
child patients were not included in some elements (REC reference number:
10/H1109/28). Local NHS primary care research governance approvals were
obtained using different local processes from five different bodies. The
procedures for local agreement of the UKCRN NHS service support costs to
research in general practice varied in each area. Agreements were obtained
for each participating general practice.
The case study settings
Twelve practices participated in the study in the south-west, east and
south-east of England. They were diverse in the sociodemographic features of
their locale, the number of practice partners and the size of their
registered patient population. These data are presented within ranges, to
ensure anonymity for the practices, in .
The report now turns to presenting the findings.
