Uploaded by Adham Moftah

Dissertation Methodology Draft

2. Methodology
2.1 The STAR Care Pathway
2.1.1 Development of the STAR Care Pathway
The need to address the obstacles to long-term pain management after knee replacement, and to
develop a comprehensive care pathway for chronic pain post-TKR, led to the Support and Treatment
After Joint Replacement (STAR) trial. This approach involved an initial assessment clinic at 3 months
post-operatively to identify patients with pain early on, followed by telephone follow-up
appointments. The assessment clinic promotes understanding the underlying reasons for CPSP and
referring patients onwards to existing services. These services aim to target those reasons to
optimise pain management and reduce adverse outcomes due to pain. Evidently, the design of this
care pathway draws on evidence collected from the papers referred to above on the current
management of chronic pain after TKR5,6. The development of the intervention was rigorous,
involving refinement of the intervention and improving on any issues relating to its implementation
before progressing to a randomised controlled trial. Clinicians inputted on the intervention design to
ensure the elements of it are appropriate, and clinicians directly involved in its future delivery were
consulted to further modify the intervention as appropriate. The intervention was trialled on a small
sample of patients with chronic pain after TKR to illuminate any problems in delivery, including how
patients respond to it, and offer a chance to rectify these. Examples of these issues included one of
the neuropathic pain scoring tools, abbreviated as DN-4, missing patients with clinical symptoms of
neuropathic pain. Another clinical neuropathic pain measure, named PainDETECT, was added to the
assessment clinic questionnaire and further questionnaires completed by patients during the follow
up period. These clinical scoring tools are described in further detail below. Another problem
identified was the inconsistency in knee stiffness assessment carried out by the extended scope
practitioner (the healthcare professional carrying out the initial assessment clinic. Their role is
explained further below) and by the orthopaedic surgeon. This was remedied by supplying the
extended scope practitioner (ESP) with a goniometer, a tool which measures range of motion.
Testing the intervention delivery and acceptability to patients also aided in the development of
training materials for the healthcare professionals that will deliver the intervention in the future. The
final stage of development saw health professional stakeholders' views on the implementation of
the STAR trial be evaluated to ensure they understood how this intervention would change their
work, how they would integrate it into their practice and benefit patients11.
2.1.2 Design of the STAR Care Pathway Study
Screening Process:
The STAR care pathway study consisted of a multi-centre randomised controlled trial carried out in 8
NHS hospitals in the UK. It aimed to judge whether this care pathway was clinically effective in
comparison to usual care for people with chronic pain after knee replacement. Patients who had
received a primary TKR for osteoarthritis 2 months prior, at participating sites, were identified and
invited to participate. They were sent a screening study pack with a cover letter, patient information
leaflet, freepost envelope and a complimentary teabag. The patients were asked to complete and
return the screening questionnaire and consent form. The research team scored the Oxford Knee
Score (OKS) data from the patients’ responses to identify patients who scored 0-14, indicating pain
in their knee post-TKR. The OKS is a clinical scoring tool assessing knee pain. In the screening
process, this was adapted into a 7-question score, each question scored from 0-4, yielding a total
score of 0-28. 0 indicates severe pain and 28 indicates no pain12.
Recruitment Process:
Those aged ≥18 years who scored 0-14 on the OKS completed in the screening questionnaire sent to
invitees and consented to further contact were eligible for recruitment into the study. Scoring 0-14
was deemed to be relevant due to previous research suggesting that this range represents pain that
negatively affects patients' health related quality of life13. Those patients were sent a trial
information pack and phoned by a researcher to check if they wished to participate. If so, they
carried out the OKS again via telephone to ensure they still met the inclusion criteria. If patients
wished to participate, they were required to provide written, informed consent. Participants were
further required to complete a detailed baseline questionnaire before being randomly allocates to
the STAR pathway (intervention) group or the usual care (control) group12.
This screening and recruitment process is discussed by Wylde et al. (2020)14, whereby the detailed
questionnaire completed at 13 weeks post-TKR revealed the factors associated with pain and
function. More severe pain was associated with poorer general and physical health, more pain worry
and lower satisfaction with surgical outcome. More severe functional limitation was associated with
higher levels of depression, more pain worry, lower satisfaction with surgical outcome and higher
pain acceptance. These outcomes became the targets for patient-specific interventions aiming to
improve CPSP and patient outcomes.
STAR Trial Procedure
363 patients were recruited. These participants completed a baseline questionnaire at 13 weeks
post-operatively, then were randomly allocated to the STAR pathway group or usual care and
informed of their allocation. 242 were allocated to the intervention plus usual care group, while 121
were allocated to usual care alone. Usual care entails entailed a 6-week follow up postoperatively,
although one site offered an additional appointment at 3 months12.
Those in the STAR intervention group also received usual care in addition. The intervention was
carried out by ESPs, which are registered allied healthcare professionals with a specialist training in
orthopaedics. Furthermore, all ESPs attended a 1-day training session and received a training manual
relating to the assessment clinic. The intervention involved a 1-hour assessment clinic appointment
with an Extended Scope Practitioner (ESP) where a clinical history was taken, patient-reported
outcome measures were reviewed (Brief Pain Inventory (BPI), Hospital Anxiety and Depression Scale
(HADS), painDETECT and Douleur Neuropathique 4 (DN4)18), knee examination is carried out, and
radiographs and blood tests are reviewed or carried out12. The BPI is a measurement tool for
assessing clinical pain as reported by patients: patients rate the severity of their pain via a pain
severity scale and how much their pain affects their daily activities and feelings via a pain
interference scale. Both scales are scored from 0-10 (best to worst)15. The HADS is a simple clinical
measure of anxiety and depression in patients, comprising 7 questions on anxiety and 7 questions on
depression and both scales are scored separately from 0-21 (no to severe depression/anxiety).
PainDETECT and DN4 are neuropathic screening questionnaires. PainDETECT contains 9 items: 7 on
sensory sytmptoms, one on pain course pattern and another on pain radiation17. DN4 is a 10-item
questionnaire split into questions on symptoms and physical exam signs18.
The outcomes of the assessment clinic were recorded on a standardised proforma and based on this,
participants were referred to appropriate existing services for further treatment. This included
referral to a surgeon, physiotherapy, GP for treatment of depression or anxiety and/or pain
specialists for neuropathic pain or Complex Regional Pain Syndrome. The assessment clinic was held
3-4 months after surgery if possible. Participants then received up to a maximum of six telephone
follow-up consultations over 12 months from the ESP to monitor how referrals are progressing and
to make other referrals if necessary. These consultations were also recorded on a standardised
proforma. Outcome measurements were self-reported by patients when they completed the
baseline questionnaire at 3 months after their surgery, then also at 9 months after surgery and 15
months after surgery (12 months after randomisation). The co-primary outcomes were pain severity
and interference, assessed using the BPI14. The mean BPI pain severity and frequency scores at 12
months after randomisation will be compared between the usual care and intervention groups12.
2.2 Comparing Outcomes Within the Intervention Group
2.2.1 Reviewing and Coding the STAR Intervention Data
In order to examine the impact of treatments (from referrals) on patient reported outcomes, namely
BPI scores, the data collected on patients’ referrals from the assessment clinic and subsequent
follow-up telephone calls with ESPs will be analysed. The data was made available in a Microsoft
Excel spreadsheet presenting which referrals, out of all those available to be made from the
assessment clinic, were selected for all participants (for whom data was available and complete) in
one worksheet, and in another worksheet the data recorded on participants’ follow-up calls. The
worksheet presenting the assessment clinic data was filtered to show only which referrals were
made for each participant and exclude all other data recorded from the clinic. Similarly, the
worksheet containing the data on follow-up calls was reviewed to remove data identifying
participants and healthcare professionals involved in their care to ensure the data is anonymous,
and to ensure that, where information was available to do so, dates recorded were consistent. To
process the data pertinent to my research project, a new worksheet was created, titled “Project
Data”, where the follow-up call data was coded according to participant ID, what referrals they
received, a variable reflecting whether the referral was received or not (Yes/No/Unknown/N/A),
treatment received for the referral and the outcome of this. The range of number of referrals
participants received was 1-9. In addition to this, a variable reflecting the overall outcome
subjectively (Improved/Stayed the same/Worse/Unknown) was included. The status of the referral
was judged from the follow up call data – where the ESP recorded that a participant received all
their referrals, then ‘Yes’ was selected for each referral. Where an ESP recorded that some or all of
the referrals were not received then ‘No’ was selected for all/the relevant referrals for that
participant. Where data from the follow-up calls was incomplete and the status of referrals for that
participant was unclear, then ‘Unknown’ was selected. ‘N/A’ was applied to some of the ‘referrals’
that fell under the category of “Other” where this did not actually represent a referral to further
healthcare. For reference, the below flow-diagram illustrates all possible referrals that could be
made from the assessment clinic.
Figure 1 - Flow diagram demonstrating all referrals made from STAR intervention assessment clinic and the number of
participants who received each referral.
Similarly, the follow up call free-text data for each participant was read and where details on
treatment received and outcome of this for each referral was available, it was recorded. Where this
information was not available, this was recorded as “Not recorded”. The overall outcome
(Improved/Stayed the Same/Worse/Unknown) was subjectively judged from the comments that
ESPs made in the follow up call data. “Improved” was noted when there was clear mention of the
participant’s symptoms and function being better than their pre-operative state in the follow-up call
comments. “Stayed the Same” was recorded when follow-up call comments suggested that
compared to their pre-operative state, a patient’s symptoms largely remained the same or similar.
“Worse” was only noted when a patient reported that their symptoms, function and ability had
deteriorated in comparison to their pre-operative state. For some participants, the follow-up call
data was incomplete or lacked the necessary detail to record whether a patient improved, stayed
the same or became worse, so “Unknown” was recorded under “Overall Outcome”.
2.2.2 Descriptive Analysis of Referrals Made in the STAR Intervention Group
All referrals made for each participant were listed in a new Excel worksheet alongside whether the
referral was received or not using the aforementioned variable for this. The “Sort and Filter” Excel
function was used to separately select each individual referral type to count the total number of the
individual referrals and the number of “Yes”, “No”, “Unknown” and “N/A” options for whether the
referral was received, which builds on the data in Figure 1. From the “Project Data” worksheet, the
total number of participants who received all their referrals, and those who did not, was counted.
Within the group of participants who received all their referrals, the number of each “Overall
Outcome” variable (Improved/Stayed the same/Worse/Unknown) was counted. The same was done
for the participants who did not receive all of their referrals, so that the two groups can be
2.2.3 Statistical Analysis of Outcomes for Participants in the Intervention Group
To perform statistical analysis of outcomes for participants in the STAR trial who received referrals,
and compare outcomes between participants who received all their referrals and those who did not,
BPI pain severity and pain interference scores, collected from the 12-month trial questionnaire, were
used. BPI scores were available for 216 participants of the 233 who attended the assessment clinic
as not all patients who attended the intervention completed a 12-month questionnaire.
To determine what statistical test to apply to the BPI scores, a histogram was first derived for BPI
pain interference scores and for pain severity scores to determine if the data was normally
distributed or not. Based on this, a parametric (T-test) or non-parametric test (Mann-Whitney U test)
would be performed on the data.
BPI scores were formatted by referral group as depicted in Figure 1 (i.e. surgeon, GP, pain clinic,
physiotherapy, follow-up calls to monitor pain and “other” referrals). For each referral type, a table
was made with two columns: referral received and referral not received. The BPI scores of the
participants who received the individual types of referrals within each referral group, and those who
did not (i.e. that referral was selected for them at the intervention assessment clinic but they did not
receive it), were listed in their respective columns. For each referral group, a table with BPI pain
severity and a table for pain interference scores was produced, on which a Mann-Whitney U test
was carried out, yielding a P value which represented whether, for that referral group, there was a
statistically significant difference between the BPI scores of those who received that referral and
those who did not.
Looking more closely at the outcomes within each individual referral type, the above process of
formatting BPI scores on which to perform a Mann Whitney U test was repeated for the individual
referrals within the referral groups (as in Figure 1) to establish whether any individual referrals
yielded a statistically significant difference in BPI scores between participants that received the
referral and those who did not.
Descriptive statistics were also carried out on the tables of BPI scores as mentioned above.