Patient-directed self-management of pain (PaDSMaP) compared to treatment as usual following total knee replacement; a randomised controlled trial

The trial was registered at the Clinical Trials Registry with the International Standard Randomised Controlled Trials Number (ISRCTN10868989) in March 2012. Patient recruitment commenced in August 2011 before registration due to an oversight of the trial team. It had been assumed that it had been registered earlier. The purpose and method of the research were explained to the patients at the pre-admission clinic about 2 weeks prior to their operation when written consent was obtained. The study protocol was published in Trials [12]. The CONSORT checklist is available in Additional file 1.

We conducted a prospective, parallel design, open label, randomised controlled, superiority trial of patient-directed self-management of pain (PaDSMaP) compared with nurse control of oral analgesia (TAU) after a total knee replacement [12]. Participants were recruited from the Norfolk & Norwich University Hospital, a secondary care teaching hospital in the United Kingdom.

Competent adults were recruited from patients waiting to undergo a unilateral total knee replacement. The key inclusion criterion was that these patients were able to manage their medications at home. They also had to be English-speaking and literate. We excluded patients that were; opiate-tolerant, opiate-dependent or using modified-release opiate preparations, or had a history of drug or alcohol abuse within the last 5 years. The reasoning for this was that patients who have an established opiate tolerance might be expected to need to continue on an equivalent opiate regimen to prevent opiate withdrawal, and may also require more opiate to ensure equivalent analgesia, and thus would not be comparable to the rest of the population. Other exclusion criteria included expecting post-operative intensive care, or were not competent by reason of dementia or other cause. Post-operatively patients had to be awake and breathing independently, and able to answer questions and follow commands (i.e. competent). Patients were expected to require standard step 1–3 oral analgesics, the 48 patients that received regional blocks or epidural analgesia, started PaDSMaP or TAU as soon as they began oral analgesia. These inclusion and exclusion criteria were in line with the hospital’s drug self-administration policy [13].

The research nurses checked patient eligibility and consent against a checklist. Patients were allocated a personal 6-digit ‘PIN’ identifier that allowed them access to the independent telephone randomisation service at the Clinical Trials Unit (CTU) at the University of East Anglia (UEA). On entry of a valid PIN, the system generated a unique Study Code and randomly allocated the patients to either the PaDSMaP or TAU arm of the trial. The study code and allocation was sent in an email to the research team and stored in the trial database on the secure CTU server at UEA.

A computer generated randomisation list allocated the patients in a 1:1 ratio to PaDSMaP or TAU groups. Randomisation was unstratified. Patients were allocated in randomly distributed blocks of four and six. The patients were randomised, assessed and trained in the pre-operative clinic or at their homes. Patients, their clinical team and the research team were all aware of their allocation. As the intervention was under the control of the patients and the primary outcome (pain VAS) was a patient-reported measure it was impossible to blind patients or clinical staff as to participant’s allocation. It was originally planned to blind the statistician to allocation but as we accidentally included information on duration of self-medication in the statistician’s database they were unblinded, however the analysis plan was pre-specified and approved in a blinded fashion and therefore any potential bias was minimised.

All participants received the TKR information sheet that is provided as standard to everyone undergoing the operation (Additional file 2). All participants received a study information sheet which informed them that they would be randomised to one of two groups, and in which the two groups’ protocols were described in detail (but this did not include the training pamphlets provided to the self-medicating group once they had been allocated). Care was taken to emphasise the lack of evidence to prefer one protocol over the other. These information sheets are provided in Additional files 3 and 4.

The majority of patients recruited to this study (78%) received the Modified Caledonian Technique (MCT) Norwich Enhanced Recovery Programme (NERP) [14]. This programme aims to enhance the recovery of patients having primary knee replacements by a multi-modal programme, which facilitates early mobility and discharge [15]. The NERP focuses on the provision of safe and effective analgesia with minimal side-effects, which enables early mobility. The analgesia protocol combined a number of classes of analgesics (Table 1). Individual patients were always offered the most suitable protocol for their particular circumstances, so some patients did not follow the NERP protocol. This did not prove an obstacle; as such variations were allowed by our protocol and recorded. Once on the ward TAU patients received their medications as usual from the nurses.

The PaDSMaP package [12] was developed by members of the research team in conjunction with relevant clinical and lay experts. This resulted in a comprehensive, structured, evidence-based training manual delivered in four pamphlets [3, 16, 17] (Additional file 5). Patients were given these pamphlets in a 20 min training session at the pre-operative clinic outlining why pain control is important, what medications they were likely to be using and how to take them, the common side effects and their management, and non-pharmacological methods for pain relief (e.g. ice packs).

Consultation with lay representatives and healthcare professionals highlighted that it would be inconsistent to allow self-medicating patients to be in control of their oral analgesia but not their oral co-morbid medications. Therefore if patients were prescribed medications for their co-morbidities they were usually in control of these. In general no specific training was provided regarding these medications as patients were expected to have sufficient knowledge to take them appropriately. However patients on anti-hypertensive or anti-coagulant medications were specifically warned to check with nursing staff that it was appropriate for them to take these medications after the operation.

Patients were assessed by a nurse for their competency to manage their own medications, by checking patients were orientated as to location and time, and that they could remember and follow simple instructions [13]. Patients were then given a pack of 14 days-worth of their medications (co-morbid medications and analgesics) previously prepared by the ward pharmacists. All participants received the medications prescribed to them by their clinical team (their consultant, the pain team, and orthopaedic pharmacists).

The UK’s Misuse of Drugs Act (1971) aims to prevent high risk “controlled” medications from being misused, obtained illegally or causing harm. The legislation places legal requirements on the method of prescribing and dispensing of controlled medications. Due to these requirements, any patients prescribed controlled medications (irrespective of their allocation) received them direct from the nurses as they were not approved for self-medication protocols at the hospital [13]. In practice this mostly affected the oxycodone prescribed as part of the standard NERP analgesic protocol. The trial medications were stored in trial-specific lockable bedside cabinets. Patients were given the keys on a lanyard and asked to keep the cabinet locked when not in use and the keys secure. Patients recorded their medication usage on the standard drug chart. This chart had been prepared by the ward pharmacists so that all text was in block capitals, and no abbreviations (e.g. tds) were used. The ward nurses and pharmacists conducted their usual checks of medication usage and record keeping as well as patient competency at the usual drug rounds.

Ward staff (nurses, physiotherapists and the pain team) were trained by the research team about the purpose of the study, the information that the patients would receive and how to assess competency of participants to self-medicate after the operation.

At baseline we measured a number of factors in order to provide a description of the characteristics of our sample. Some of these have been identified by other studies as factors that can affect post-operative pain and recovery, and so were possible confounders (Tables 2, 3, 4) [18‐23]. Baseline measures were recorded in pre-operative outpatient clinics or in patient’s homes.

The primary outcome was pain levels at discharge or after 3 days post-operatively (whichever was the sooner), using a non-graded 100 mm VAS [24]. The pain VAS was anchored by the words “No pain” to “Worst pain” and referred to static pain levels over the last 4 h in the knee that underwent surgery. The primary outcome was measured on the inpatient ward after surgery.

Secondary outcome measures for the patients were:

We planned to investigate medication usage but realised that the high level of variation in the individualised analgesic protocols after surgery (see Table 1) meant that direct comparison between the two groups was impossible.

Secondary measures were measured in the inpatient ward for the 3 days after surgery, and at an outpatient clinic for the 6 week data.

A qualitative evaluation and cost-effectiveness analysis were also undertaken and are reported in detail elsewhere [30].

The qualitative evaluation found that patients had a positive experience of self-medicating even for those patients who reported an initial poor experience of pain control. Some patients reported feeling “isolated” and “a bit concerned I was taking their (nurses) jobs away”. Every patient when asked if they had the option to self-medicate for a similar operation responded positively. Healthcare professionals expressed positive views about the empowering nature of self-medication but had realistic concerns about capacity, documentation and accountability. The paternalistic nature of healthcare was evident (Balls P, Darrah C, Deane KHO, Gray R, Swift L, Barton G, et al: Patient self-management of pain (PaDSMaP); a qualitative investigation of the acceptability of self-medicating with oral analgesics in the context of a randomised controlled trial following total knee replacement, Unpublished qualitative data from this trial).

The cost effectiveness analysis [30] showed that the self-medicating intervention cost £307 to deliver. However most of the costs (£243) were due to set-up costs such as the purchase of bedside cabinets and printing costs. The mean incremental cost for the self-medication group was calculated to be £774 (95%CI £174 to £1374) (when adjusted for age (ns), gender (ns), education (ns) and baseline cost (ns) based on 133 observations, Adj R-squared = 0.0983). However, much of this incremental cost was accounted for by four self-medicating patients who needed intensive care for extended periods of time (for reasons unrelated to the self-mediation protocol). The mean incremental QALY gain was small, 0.002 (95%CI –0.002 to 0.006) (adjusted for age (ns), gender (ns), education (ns) and baseline EQ-5D (p < 0.001) based on 133 observations, Adj R-squared = 0.7225). This leads to a mean cost per quality of life adjusted year (QALY) of > £390,000 [30].

Seventy-two patients in each arm was calculated to be sufficient to detect a between-group difference of 0.5 standard deviations in the VAS pain intensity with 80% power using an independent samples t-test. Dahlen [5] reported a standard deviation in VAS pain intensity score of 24 mm 2–5 days after total knee arthroplasty so assuming similar variability this represents a difference of about 12 mm which is within the range of minimum clinically important differences (MCID) that have been reported for pain VAS measures [31, 32]. However, the study was formally powered to detect a difference of 0.5 standard deviations which is therefore the definition of MCID formally used. Calculations used a significance level of 0.05. A drop-out rate of 10% was expected and accounted for.

Linear regression models were used to compare primary and secondary outcomes between the intervention and control. There is no single ‘right’ way of selecting covariates for adjustment. In particular Pocock [33] says that ‘pre-specification of all covariates for adjustment’ while ‘desirable in principle, is often unachievable in practice’ due to ‘inadequate prior knowledge as to which baseline factors are related to prognosis’. This was the case in this study as many factors have been proposed to affect post-operative pain levels [18, 19, 22]. Thus Pocock [33] advocates ‘defining objective variables selection algorithms’ a priori to ‘overcome any suspicions that post hoc selection of covariates might be based on subjective criteria’. Therefore as planned in our original protocol [15], we finalised our analysis plan 6 months before the final data being available, identifying baseline variables which differed between the groups (i) and a set of candidate covariates supplied by the research team (ii) in line with current research [18, 19, 22]. Therefore estimates were also obtained, (i) adjusted by baseline VAS, Hospital Anxiety and Depression Scale (HADS) anxiety and depression scores and (ii) by further potential covariates. (Variables which gave P < 0.10 when examined individually as a predictor of the primary outcome, adjusting for variables in (i), were included in a multiple predictor model and then eliminated one by one until all potential predictors had P < 0.10); this resulted in further adjustment by gender and number of prescribed medications. The assumptions of the models were checked using residual analysis which found no outliers or heavily influential observations.

The main analysis was conducted on an ITT basis. For a per protocol analysis successful self-medication was defined as those who started self-medicating and were continuing at 72 h/discharge, even if they stopped for a period during the hospital stay.

Linear regression models were used to investigate, for those randomised to the self-medication group, which baseline variables individually predicted successful self-medication. In order to account for repeated measures, mixed linear regression models were used to compare series of VAS scores measured at breakfast, lunch and supper over the 72 h follow up period, between groups.

In the light of the results, unplanned analyses used the same methods to additionally adjust by day 1 breakfast static VAS scores. SPSS v18 was used for the analysis. The number of co-morbidities was compared between individual included and excluded from the per-protocol analysis using a Mann-Whitney test.

Between July 2011 and March 2013, 144 competent adults about to undergo a primary unilateral TKR were recruited (Fig. 1). Patients without co-morbidities (American Society of Anesthesiologists (ASA) grade 1) [34] were generally sent to a private healthcare provider not involved in this study (n = 448). Of the 665 patients screened for eligibility for the RCT 23% were excluded because they were not managing their own medications at home (usually due to dementia). A further 28% refused to participate, often because they felt unsure of their ability or desire to self-medicate after a major operation.

Seventy-one patients were randomly allocated to TAU, and 73 to the PaDSMaP group. However seven patients (five PaDSMaP, two TAU) did not undergo surgery for reasons unrelated to the study and were excluded them from the ITT analysis. The remaining patients (n = 137) are characterised in Tables 2, 3, 4. They had a median of three co-morbidities in addition to their osteoarthritis. The HADS assessment determined that 13% of patients scored above the cut off for depression, and 17% scored above the cut off for anxiety. Twenty-four patients (18%) had musculoskeletal diagnoses other than osteoarthritis, of which the most common were rheumatoid arthritis (22%) or fibromyalgia (15%).

Half of patients were not taking any opiates prior to their surgery (Table 5). Of those participants prescribed opiates for after their operation, the majority (57 of 68, 84%) were prescribed a single opiate usually to be taken as required (50 of 57, 88%). The maximum prescribed Equivalent Oral Morphine Dose (EOMD) [35] was 96, but most patients were taking substantially lower dosages. It was not possible to calculate the EOMD accurately for each patient due to the variability in the usage of “as required” medication.

The Beliefs about Medicines Questionnaire (BMQ) assesses commonly held beliefs about medicine [21]. The baseline scores (Table 3) show that overall our population agreed that medicines were generally not harmful or overused. The differences between the specific necessity and concerns scores were positive, which indicated that these patients perceive that the benefits of taking medications outweigh the risks.

Of the 68 patients analysed in the PaDSMaP group; 47 started self-medicating (six of whom later stopped (due to losing competency, n = 7, or anxiety n = 1), and four who started, stopped and then restarted before the end of day 3 (due to losing competency n = 3 or anxiety n = 1)), 21 never started (including two who withdrew from the study; due to losing competency up to day 3 n = 16 or anxiety n = 5), and 41 were classed as having successfully completed the self-medicating protocol (Fig. 1). None of the patients reported any difficulties in opening the medication lockers, and all keys were kept secure for the duration of the study.

Self-medicating patients were also in control of their co-morbid medications. They took a median of 5 (range 0–11) prescriptions. Although at the initiation of the project we were assured that no patients self-medicated on the ward, we found that one diabetic patient allocated to the TAU group did self-medicate with their injectable insulin. All of their remaining medications were delivered by the nurses as usual.

Little difference was observed between the two groups’ baseline characteristics (Tables 2, 3, 4). More of the PaDSMaP group were in employment (28% vs 16%) or had musculoskeletal problems (27% vs 9%), and fewer scored above the cut off for depression (a HADS depression score of ≥8) (7% vs 17%) [20].

No evidence was found for a between group difference in ITT mean pain score (static VAS for the previous 4 h) at 72 h/discharge (Table 6). Similar results were obtained after (i) adjustment by baseline VAS, HADS anxiety, HADS depression as specified in the statistical analysis plan and (ii) further adjustment by gender and number of prescribed medications, identified by the covariate selection procedure (Table 6).

Using the per protocol definition, patients who were able to self-medicate had significantly lower static VAS (effect − 9.9 mm, 95% CI -18.7 to − 1.1, T-test) and mobilised VAS (effect − 13.1 mm, 95% CI-23.8 to − 2.4, T-test) at 72 h/discharge compared to TAU patients.

The population selected were elderly, had co-morbidities, and the operation required a high level of opiates in order to control pain levels suitably. All of these factors led to a high incidence of adverse events such as confusion, dizziness or vomiting post-operatively, which were the usual reasons for patients not self-medicating. As 40% of patients were unable to complete the self-medicating protocol, the study was effectively under-powered.

The high rate of co-morbidities in our population was due to ASA grade 1 patients (being healthy with a well-controlled co-morbidity [34]) generally being referred to a private healthcare provider which was not involved in the study. This high rate of co-morbidities limits the generalisability of our results to populations with low or no co-morbidities. However our data suggest that these populations would be more likely to manage a self-medicating protocol.

Obviously only patients willing to participate in the trial were recruited. Many of those who refused stated that they were concerned about their ability to manage the perceived complexity of self-medication after an operation. Overall our participants had more positive beliefs regarding the benefits of taking medications when compared to other chronic conditions [21]. Therefore if self-medication was rolled out for all eligible patients, further time might be needed to ensure concerns about the medicines and self-medicating were addressed to patient’s satisfaction.

There was rarely more than one self-medicating patient on the ward at any one time. Sawhney [45] showed that professionals’ knowledge and confidence in self-medicating improved with time and numbers of patients treated. Therefore it is likely that having a greater number of self-medicating patients would have improved understanding and embedded the protocols with ward staff.

There were a few missing pain scores, mostly in the intermediate time points. We know that at least some of these scores were missing due to patients’ unwillingness to complete the research outcomes when they were in severe pain i.e. they were not missing at random. Therefore the intermediate time point’s pain scores may have been under-estimated.

Due to the trial allowing patients to be prescribed individualised analgesia, it is impossible to compare the analgesic “dose” between the groups. Patients were prescribed a range of analgesics. Although the daily Equivalent Oral Morphine Dose (EOMD) could be estimated [35], it would not fairly represent the analgesia provided. Patients usually received additional non-opioid analgesics which have been shown to improve quality of analgesia and reduce opioid use [9‐11]. There are currently no methods that allow for the calculation of an overall analgesic dose that includes both opioid and non-opioid analgesics, and even the EOMD estimates are not without controversy and variations in how they are calculated [46].

As noted in our introduction the evidence base is limited and not consistent. Striebel’s two studies [38, 39] claimed PCOA increased patient satisfaction and pain control compared to patient controlled intravenous analgesia, and Kastanias’ study [37] showed no difference between PCOA and nurse-controlled analgesia on day 2 after a TKR.

Currently elective surgical patients have all their inpatient medications controlled by healthcare professionals, yet patients are expected to manage to self-medicate once they are discharged home. For patients capable of self-medication there are potential benefits in terms of improved pain control. However this study demonstrated that self-medication was not suitable for all and a patient’s ability to self-medicate needs to be assessed on an on-going basis. The factors which may limit the ability of adults to self-medicate include having high pain levels on day 1, having higher numbers of co-morbidities, and suffering from adverse events of the surgery and anaesthesia such as confusion and nausea.

Our cost effectiveness analysis [30] showed that the additional costs of self-medication were not large, particularly once set-up costs are accounted for. The calculation of the self-medicating group’s health related costs was complicated by four self-medicating patients requiring periods of time in intensive care (for reasons unrelated to their self-medicating). A longer period of evaluation with a larger cohort would give a clearer picture on the costs of a self-medicating protocol.

We are aware that there are significant political pressures on hospitals to implement self-medicating policies in order to improve patients’ autonomy and dignity [36, 47‐50]. The hospital in which this research was conducted now has a target of having all suitable inpatients self-medicating. This study supports this target but highlights that there are significant training needs for both staff and patients. Additionally organisational aspects, such as the availability of Specialist Ward Pharmacists and secure accessible patient lockers, are required to ensure the safe and successful implementation of this policy.

Further research is needed to identify how important this reduction in pain and improvement in autonomy is to patients; whether it is worth the additional training time needed for staff and patients; or if the implementation of self-medication should be for all eligible patients in inpatient settings, and finally, and the evaluation of the health economic implications. The risks of the protocol need to be quantified and management protocols developed that include the development of predictive tools to identify patients unlikely to manage self-medication. Finally the development of a method to calculate the overall analgesic dose for patients prescribed combinations of opioid and non-opioid medications would be helpful.