Preoperative education for reducing patient anxiety in elective hip and knee arthroplasty – a systematic review and meta-analysis

Abstract

Background

Preoperative educational interventions are often implicitly regarded as a useful and cost-effective means of improving arthroplasty care. However, previous research in this area is limited – both in terms of age and clinical heterogeneity. This review adds to the existing literature by investigating the relationship between educational interventions and clinically relevant outcome measures, whilst also incorporating a larger sample size, and more stringently defined eligibility criteria. The primary outcome was preoperative anxiety. Secondary outcomes were patient knowledge score, postoperative pain, time-to-discharge, quality-of-life, and surgical complication rate.

Methods

A systematic review and meta-analysis of 19 randomised controlled trials. Inclusion criteria specified an adult population receiving elective hip or knee arthroplasty, an intervention group participating in a structured educational intervention, and a control group receiving no education beyond routine preoperative consenting. Outcome data were independently extracted by the review team, before being pooled for statistical analysis using comparison of mean differences and a random effects model.

Results

Results showed small, but statistically significant, improvements for the outcomes of patient anxiety (SMD -0.37, 95% CI -0.63 to -0.11), knowledge (SMD +0.37, 95% CI +0.05 to +0.69), and pain (SMD -0.31, 95% CI -0.47 to -0.15). However, the results for other outcome measures were not statistically significant.

Conclusions

Findings show a trend in favour of education for reducing anxiety and postoperative pain, though effect sizes were small, and the amount of practical benefit remains questionable due to inter-trial clinical heterogeneity. It was also observed that multi-format interventions may be beneficial for patients with greater anxiety, learning needs, or pain. Further research is necessary due to a lack of trials with reproducibly described interventions.

Supplementary Information

The online version contains supplementary material available at 10.1186/s13741-025-00638-1.

Introduction

Approximately 160,000 patients undergo a hip or knee replacement (arthroplasty) in England and Wales each year (NJR, 2023). These patients come from a range of socioeconomic backgrounds, but most are aged between 60 and 80 at the time of surgery. The leading indication for surgery is osteoarthritis, where joint arthroplasty is used to alleviate pain and restore function. Major surgery to these weightbearing joints can have a considerable physical, as well as psychological, impact on patients – many of whom experience anxiety, pain, and prolonged immobility (Johansson et al. 2005; McDonald et al. 2014). The incidence of these problems, and subsequent costs, is forecast to grow with the UK’s ageing and increasingly comorbid population.

Preoperative patient education can be defined as the processes by which healthcare staff deliver teaching to patients with the aims of improving knowledge, health related behaviours, and clinical outcomes. Historically, education has been delivered in tandem with the routine booking appointment, mainly focusing on the surgical consenting process. However, in recent years there has been a move towards more structured methods of patient education, delivered over a longer time span and employing multiple communication formats. Perhaps the most widely adopted example of a formalised educational intervention is within the orthopaedic ‘enhanced recovery programme’ (Kaye et al. 2019). This is a well-known example of a multi-stage, multi-format, and multi-professional approach to patient communication and rehabilitation – delivered with the aim of hastening recovery and reducing morbidity. Therefore, in the context of growing service pressures, and the need to exercise fiscal restraint in healthcare, similar interventions delivered in the preoperative period offer a safe and appealing means of improving care whilst also increasing efficiency for surgical units.

Previous research by Friedman et al. (2010) and Hounsome et al. (2017) has focused on the teaching and communication strategies that can be employed to optimise information transfer from healthcare staff to patients. However, this review aims to provide a more holistic understanding of the clinical impacts of educational interventions; considering multiple outcome measures and balancing the ‘patient’ and ‘medical’ perspectives. This review’s outcome measures have been carefully chosen to provide insight into how educational interventions might impact the patient experience (e.g. preoperative anxiety, postoperative pain level, recall of clinical information, and quality-of-life score) as well as the demands of the local service (e.g. time-to-discharge and postoperative complication rate). Preoperative anxiety was selected as the primary outcome of interest because, in addition to being stressful for the patient, heightened anxiety has been associated with various clinically important outcomes. For example, a meta-analysis by Shebl et al. (2025) found that anxious patients had increased anaesthetic requirements, increased analgesic requirements, and higher rates of postoperative delirium. Other studies have associated preoperative anxiety with delayed recovery, depression, and morbidity (Andersson et al. 2020; Kassahun et al. 2022).

Previous systematic reviews have assessed how the timing and delivery format of patient education influences clinical outcomes (Johansson et al. 2005; McDonald et al. 2014). Unfortunately, the usefulness of this evidence is limited by its age, comparatively small sample sizes, and the heterogeneity of included trials. This is problematic, as the growing use of increasingly diverse educational models in healthcare seems not to be guided by a clear evidence-based approach, but rather by individual clinicians’ beliefs. Therefore, in the absence of a current ‘gold standard’, this review is needed to provide more robust and up-to-date recommendations that can be applied to a specific and clearly defined patient population.

Methods

Search strategy

This research question has been addressed using the quantitative systematic review methodology, as set out in the Cochrane Handbook for Systematic Reviews of Interventions (Higgins et al., 2019). MEDLINE, PUBMED, CT.Gov, ICTRP, CINAHL and EMBASE were searched for relevant randomised controlled trials. Search terms were written in English and developed using a Population, Intervention, Control, Outcome and Study-type (PICOS) framework (Amir-Behghadami and Janati 2020). Full details of the search strategy, completed in April 2021, are presented below (Fig. 1). An accompanying PRISMA diagram can be found in the results section (Fig. 2).

Fig. 1.

Search terms and boolean operators

Fig. 2.

Screening process and application of eligibility criteria

Inclusion and exclusion criteria

Population

Inclusion criteria selected for original randomised controlled trials that focused on adult patients who were in the preoperative period for elective hip or knee arthroplasty. Here, “adult” refers to anyone over the age of 18 years at recruitment. The preoperative period was defined as the 18-week window prior to surgery. The phrase “elective hip and knee replacement” encompassed any non-emergency joint arthroplasty procedure involving the knee, hip, or a combination of both. All participants were required to give fully informed consent to both the surgical procedure, and their participation in the trial.

Intervention

All formats of teaching and information delivery were considered, providing that they adhered to a pre-specified trial protocol and were delivered by professionals. Educational interventions were required to address at least one of the following topics: the patient’s underlying condition, the planned surgical procedure, or optimising postoperative recovery (including instruction on self-care or exercise). Trials involving teaching from multiple healthcare staff, and multi-staged interventions, were also included. Trials with interventions that aimed to optimise preoperative risk factors in isolation (e.g. smoking cessation), or incorporated preoperative procedures (e.g. genicular nerve cryoablation) were excluded to avoid confounding. Pooled analysis of educational interventions with heterogenous content may seem counterintuitive, however it is not without precedent. Anderson et al. (2022) reviewed interventions with heterogenous educational content before knee arthroplasty and found that there was not compelling evidence for a specific curriculum. In fact, the success of educational interventions seemed to be more influenced by delivery format than content – with individualised and multi-format approaches having the most impact for patients. The impact of educational heterogeneity in included trials is further discussed in the limitations section of this review.

Comparison

All included trials featured a comparison group that received no educational input beyond standard preoperative consenting. This was the threshold of education that defined ‘usual care’. Studies were excluded from this review if the comparison group received any supplemental education during the follow-up period.

Types of study

Trials published before 1980 were excluded to help ensure data were reflective of contemporary clinical practices and attitudes towards patient education. The randomised controlled trial methodology was considered the most appropriate for this review’s objectives, to minimise potential researcher bias and confounding clinical factors. This helped to ensure objectivity in assessing the utility of each modality of preoperative education (Yartsev 2021).

Outcome measures

The primary outcome measure for this review was change in patient self-reported preoperative anxiety score. Secondary outcome measures included change in patient knowledge score, postoperative pain level, time-to-discharge, quality of life, and the rate of postoperative complications. These outcome measures were identified as relevant and practical choices following scoping review, though it was not expected or required that included trials would report data for all outcome measures.

Preoperative anxiety was defined by patient’s self-reported anxiety before surgical treatment, measured against an accredited scale (i.e. State Trait-Anxiety Inventory (Julian 2011). Patient knowledge was defined by scores on standardised knowledge tests, delivered before and after each educational intervention. Postoperative pain was defined by patients’ self-reported pain scores following surgical treatment. It was a requirement that pain data must have been recorded within six months using an accredited scale. Preference was given to the first available timepoint wherever serial measurements of pain were reported. Discharge time was defined by the number of days from leaving theatre to the patient being discharged. Quality of life was defined by the patient’s self-reported sense of psychological wellbeing, measured using a recognised and accredited tool (i.e. the EuroQol Group (2021), EQ-5D Instrument). Complication rate was defined by the relative risk of developing a postoperative complication that necessitated a return to theatre or additional medical treatment.

Critical appraisal

The quality of included studies was assessed using Cochrane Risk-of-Bias 2 (Sterne et al. 2019) and inspection of outcome specific funnel plots (Appendix 1) to check for evidence of publication bias. Judgments were made across five domains: random sequence generation, allocation concealment, blinding of participants, incomplete outcome data, blinding of outcome assessment, and selective reporting. Each category was rated as high risk, uncertain risk, or low risk as described in Chap. 8 of the Cochrane systematic review handbook (Higgins et al., 2019). Several of the included studies featured small sample sizes, so critical appraisal placed significance on any bias that may have been introduced through flawed randomisation or missing outcome data. Blinding procedures were also regarded as especially important – given this review’s focus on educational interventions, which are frequently difficult to conceal (Sullivan 2011). Risk of bias assessments were made independently by two reviewers, with discrepancies discussed and addressed in calibration meetings. A set ‘cut-off’ score for high, uncertain, or low risk of bias was not used – as an arbitrary scoring system was deemed unlikely to capture the nuance of each trial. Instead, and in accordance with Cochrane recommendations (Higgins et al., 2019), the overall risk-of-bias for each domain of each trial was critically appraised on a case-by-case basis.

Statistical analysis and reporting

Data analysis and reporting were performed using Review Manager (2020) software. Due to the nature of the research question, and focus on patient education as an intervention, it was anticipated that the included studies would unavoidably feature a degree of clinical and methodological heterogeneity (even though this review’s strict inclusion and exclusion criteria had been carefully designed to reduce this). Therefore, assuming a random effects model was deemed to be the most appropriate approach for statistical analysis. As some trials reported comparable results using different scales it was necessary to perform most analysis using standardised mean differences. Two exceptions included the complication rate analysis (calculated as a relative risk ratio) and the time-to-discharge analysis (performed using mean difference, as trial data was reported on a consistent scale). Inter-trial heterogeneity was assessed by calculating the heterogeneity statistic (I²), and by inspection of each outcome measure’s forest plot (Appendix 1). Statistical heterogeneity was deemed low if below the conventionally accepted threshold of I² < 50%, as per Higgins et al. (2019). Following the approach proposed in this review’s protocol, a sensitivity analysis was also performed to evaluate the impact of clinical heterogeneity arising from different patient populations (i.e. hip vs. knee arthroplasty) for each outcome measure.

Results

Search results

19 trials (n = 2116) (Table 1) fulfilled all eligibility criteria, and 16 reported complete outcome data suitable for meta-analysis. 10 trials assessed the primary outcome of preoperative anxiety (n = 1219). However, data relevant to secondary outcomes were less abundant; patient knowledge (4 trials), postoperative pain (9 trials), time to discharge (10 trials), postoperative complication rate (4 trials), and postoperative quality of life (1 trial).

Table 1.

Included study characteristics

Arthroplasty site	Trial	Trial Design	Country	Core educational topic(s) covered	Educational Intervention format	Patients	Outcome(s) assessed
							Anxiety (pre-op)	knowledge	Pain (post-op)	Time-to-discharge	Quality of life	Complications
Hip	Vukomanović et al. 2008	Parallel group	Serbia	Optimising recovery, surgery	Group class	40			X	X
	Berge et al. 2004	Parallel group	UK	Optimising recovery	Group class	44	X		X
	Quintrec et al. 2003	Parallel group	France	Optimising recovery, surgery	Group class	125	X		X	X		X
	Biau et al. 2015	Four arm RCT	France	Optimising recovery	Group class	209			X	X		X
	McGregor et al. 2004	Parallel group	UK	Optimising recovery, surgery	Group class + Written	35			X	X	X
	Cooil and Bithell 1997	Parallel group	UK	Optimising recovery, surgery	Verbal	42		X
	Johansson et al. 2005	Parallel group	Finland	Optimising recovery, surgery	Verbal + Written	123		X		X
	Doering et al. 2000	Parallel group	Austria	Surgery, underlying condition	Video	100	X			X
	Clode-Baker et al. 1997	Parallel group	UK	Surgery, underlying condition	Video + Written	78	X	X
	Butler et al. 1996	Parallel group	Canada	Underlying condition, surgery	Written	123	X			X
Knee	Culliton et al. 2018	Parallel group	Canada	Optimising recovery, surgery	e-Learning	416	X		X
	Atabaki et al. 2019	Parallel group	Iran	Optimising recovery, surgery	Group class + Questions	96			X
	Medina-Garzón 2019	Parallel group	Columbia	Optimising recovery, surgery	Verbal	56	X
	Sjöling et al. 2003	Parallel group	Sweden	Optimising recovery	Verbal + Written	67			X	X
	Huang et al. 2012	Parallel group	Taiwan	Optimising recovery	Verbal + Written	243				X		X
	Eschalier et al., 2017	Parallel group	France	Optimising recovery, surgery	Written	42		X
Both	Cooke et al. 2016	Parallel group	Australia	Optimising recovery, surgery	Video	91	X		X
	O’Connor et al. 2016	Parallel group	USA	Optimising recovery, surgery	Video + Questions	53	X
	Crowe and Henderson 2003	Parallel group	Canada	Optimising recovery	Video + Verbal	133	X			X		X
					Total	2116	10	4	9	10	1	4

Effect sizes

For the primary outcome there was evidence of a small but statistically significant decrease in anxiety that favoured educational intervention (SMD − 0.37, 95% CI −0.63 to −0.11, I² 70%) (Fig. 3). A sensitivity analysis was also completed to assess for differences introduced by trials recruiting hip arthroplasty patients, knee arthroplasty patients, or a mixture of both groups (Fig. 4). The preoperative anxiety outcome funnel plot (Appendix 1) did not show evidence of publication bias.

Fig. 3.

Change in patient reported preoperative anxiety following educational intervention intervention

Fig. 4.

Sensitivity analysis: comparison of reported anxiety scores between trials recruiting hip vs. knee. vs. mixed arthroplasty patients

Secondary outcomes assessing patient knowledge scores (SMD + 0.37, 95% CI + 0.05 to + 0.69, I² 0%) and self-reported postoperative pain (SMD − 0.31, 95% CI −0.47 to −0.15, I² 16%) also indicated small clinical improvements that favoured education. However, the outcomes of time-to-discharge (MD −1.12, 95% CI −2.45 to + 0.21, I² 89%), postoperative quality of life (SMD + 0.17, 95% CI −0.52 to + 0.85) and postoperative complication rate (RR 0.92, 95% CI + 0.41 to + 2.06, I² 70%) did not provide compelling evidence of an effect in either direction.

Discussion

Preoperative patient anxiety

Ten trials (n = 1219) assessed the impact of education on anxiety, and eight provided data suitable for quantitative synthesis. Overall risk of bias was deemed as low in two of the trials assessing anxiety (Berge et al. 2004; Medina-Garzón 2019) uncertain in two (Doering et al. 2000; Culliton et al. 2018), and high in four (Quintrec et al. 2003; Butler et al. 1996; Cooke et al. 2016; Crowe and Henderson 2003) (Fig. 4). Notably, Butler et al. (1996), which showed the greatest individual effect size of any trial, was deemed at high risk of attrition bias because 11% of participants were discharged early for undisclosed reasons and were therefore missing outcome data.

There was a statistically significant effect size suggesting that educational interventions help reduce preoperative anxiety (SMD − 0.37, 95% CI −0.63 to −0.11) (Fig. 3). However, while any reduction in anxiety at the population level might be considered an ethically relevant benefit, it is important to consider this effect size in terms of the minimally clinically important difference (MCID). As some trials used different anxiety scales, and a combined analysis was conducted using standardised mean difference, it is not possible to draw direct clinically relatable comparisons (i.e. directly comparing the pooled overall effect size to a single scoring scale). However, some inferences can still be made from the available effect size data. For example, the State-Trait Anxiety Index (STAI) was the most used scale for assessing anxiety in included trials (Butler et al. 1996; Cooke et al. 2016; Medina-Garzón 2019; Quintrec et al. 2003). Calculating effect size using data from these four trials alone produced a standardised mean difference of −0.26 (95% CI −0.54 to 0.02, I² 35%). For context, the STAI uses a scale going up to 80 points where higher scores represent greater anxiety (Spielberger et al. 1983). Therefore, this observed effect size is comparatively small when set against the broad range of the scale being used. Therefore, the amount of any observable clinical benefit at the individual patient level seems likely to be small.

Sensitivity analysis (Fig. 4) was performed to identify any significant differences introduced by the different hip and knee patient subgroups, and to allow for comparison with joint specific data from previous systematic reviews (McDonald et al. 2014; Friedman et al. 2010; Hounsome et al. 2017). Ultimately, only the result for the hip arthroplasty subgroup was statistically significant (SMD − 0.31, 95% CI − 0.53 to −0.08). Results for the knee arthroplasty subgroup (SMD − 0.26, 95% CI −0.66 to + 0.14) and ‘combined’ hip and knee subgroup (SMD − 0.53, 95% CI −1.53 to + 0.47) were not significant. This would seem to corroborate the findings of McDonald et al. (2014) which reported that “… education was found to reduce preoperative anxiety in people undergoing hip replacement” but that it was “no better or worse than usual care…” for patients receiving knee replacement.

Interestingly, McDonald et al. (2014) reported a larger overall effect size despite also including several of the trials included in this review (calculated SMD − 0.53, 95% CI −0.64 to −0.42). This is explained by several key differences in the trials included in either review’s ‘hip-specific’ analysis. Firstly, McDonald et al. (2014) excluded trial data from Berge et al. (2004), because patients received an intervention starting more than six weeks before surgery. And secondly, outcome data from Crowe and Henderson (2003), which was incorporated into McDonald’s (2014) review, was excluded from this review’s ‘hip-specific’ analysis because participants received a mixture of hip and knee arthroplasties and it was not possible to distinguish between the two groups.

Since both trials (Berge et al. 2004; Crowe and Henderson 2003) provided approximately equivalently weighted data for meta-analysis, the observed disparity between this review’s finding and McDonald’s (2014) appears to be because of the relative effect sizes of the two trials. While Berge et al. (2004) demonstrated a marginal effect size (SMD − 0.23), data from Crowe and Henderson (2003) indicated a substantially larger benefit following educational intervention (SMD − 1.03). An unusually large effect size set Crowe and Henderson (2003) apart from other included trials (Fig. 3). There are two plausible explanations for this. Firstly, the fact that participants were unnaturally selected from a local population “…who scored high on the Oxford Questionnaire (i.e. more disabled at baseline), and who were identified on the functional data collection form as having coexisting medical conditions, sub-optimal social support, and requiring home alteration…” (Crowe and Henderson 2003). In other words, this may have introduced a selection bias for vulnerable participants with potentially more to gain from educational intervention.

Another plausible explanation might be the unique format of the intervention delivered by Crowe and Henderson (2003). In contrast to most of the other trials in this review, it utilised a multi-format approach to preoperative education. Patients in the intervention group received an outpatient educational package, consisting of a 50-minute video and a booklet. These outpatient resources included comprehensive information on the planned procedure, length of stay, discharge criteria, diet, and use of equipment. The intervention group also received extended individualised counselling. An outpatient phone number and physiotherapy programme were also provided to participants in the intervention group at discharge, allowing for a longer follow-up period and more sustained support than interventions featured in other trials. This individualised and multi-format approach to patient education was not attempted by any other trials. Therefore, the unique format of the intervention may have contributed to the unusually large reduction in preoperative anxiety. One mechanism that might explain this would be the educational reinforcement from the multi-staged intervention, and patients’ anticipation of extended outpatient support. This hypothesis would be congruent with the findings of Friedman et al. (2010), which support the use of multiple teaching strategies and multi-staged information delivery in preoperative education. The theoretical benefits of a multi-stage approach to preoperative education are also supported by adult learning theory – where the quality of learning is believed to be enhanced by the opportunity for periods of self-directed study, reflection, and subsequent expert feedback (Taylor and Hamdy 2013). In a clinical setting this might be characterised by a patient receiving an educational intervention, reflecting on it, and then using their newfound understanding to have a more informed discussion with their clinician (allowing for a more individualised approach to the consultation and an efficient use of time).

Patient knowledge

Four trials (Clode-Baker et al. 1997; Cooil and Bithell 1997; Eschalier et al. 2016; Johansson et al. 2005) evaluated the impact of education on patient knowledge. Complete outcome data was only available from Eschalier et al. (2016) and Johansson et al. (2005) - indicating a small but statistically significant increase in test scores (SMD + 0.37, 95% CI 0.05 to 0.69). However, as the design of knowledge tests was not clearly described in either study, it is not possible to exclude heterogeneity in the precise timing, format, and content of assessments. As a result, findings may only be treated as weak evidence of an advantage over standard care.

Interestingly, a recurring theme of ‘educational reinforcement’ was raised in three separate trials (Clode-Baker et al. 1997; Eschalier et al. 2016; Johansson et al. 2005). Patients who were given reading materials before admission, and then went on to receive further educational input, appeared the most satisfied – “…staff felt that they were able to offer reinforcement to patients on admission rather than presenting them with new information. Nurses reported that this also saved a lot of their time, which is often pressured.” (Clode-Baker et al. 1997, p113). A similar idea was also discussed elsewhere – “Our study showed that admission discussions are shorter when the patient has received both written educational materials and [verbal] education…” (Johansson et al. 2005, p89). Eschalier et al. (2016) did not feature a verbal intervention, instead participants in the intervention group did receive written course materials that they could study in the leadup to admission. This meant that ‘reinforcement’ was achieved by repeated exposure to the curriculum. Resulting knowledge scores showed “significantly greater increases over time, versus baseline, in the intervention group than in the control group” (Eschalier et al. 2016). This finding would also seem to support ‘reinforcement’ as an important aspect of communicating patient education information.

Postoperative pain

Nine trials assessed the impact of educational interventions on pain. Of these, seven reported outcome data suitable for meta-analysis (Berge et al. 2004; Biau et al. 2015; Cooke et al. 2016; Culliton et al., 2018; McGregor et al. 2004; Quintrec et al. 2003; Vucomanović et al., 2008). Overall, the standardised mean difference in pain scores was calculated at −0.31 (95% CI −0.47 to −0.15) - indicating a statistically significant result in favour of educational intervention. Given that reported pain scores were commonly assessed on a zero to ten scale, this effect size would appear to represent a clinically relevant reduction in pain. However, this result may have been subject to performance bias that created an overestimation of the true effect size. It is possible that the content of educational interventions may have led participants in some trials (Berge et al. 2004; Biau et al. 2015; Culliton et al. 2018; Sjöling et al. 2003; Quintrec et al., 2003) to better anticipate pain, and therefore request analgesics more readily than their counterparts in the control groups. This is because most trials featured education that related to preventative pain management – perhaps best represented by the quote “the most important issue was that they [participants] were strongly encouraged to be active in their own treatment, i.e. to let staff know at an early stage when their pain returned, in order to receive treatment and prevent peaks of pain.” (Sjöling et al. 2003, p170).

Baseline imbalances may also have affected the result for postoperative pain. Quintrec et al. (2003) featured a control group with more women than men (60% vs. 50%), and Atabaki et al. (2019) featured significantly more women than men in both the intervention (95.8% female) and control (85.4% female) groups. This is relevant because gender seems to influence the extent to which patients report pain – with men underreporting overall pain despite achieving similar absolute reductions to women following analgesia (Hussain et al. 2013).

Time-to-discharge

This review included 10 trials that assessed time-to-discharge, five of which included data suitable for meta-analysis. (Butler et al. 1996; Crowe and Henderson 2003; Huang et al. 2012; Quintrec et al. 2003; Vukomanović et al. 2008). Overall, the mean difference in time-to-discharge was calculated at −1.12 days (95% CI −2.45 to + 0.21) in favour of education. While the overall effect size would appear to be significant in the context of a typical inpatient stay of 4–5 days after arthroplasty (Mundi et al. 2020), as this confidence interval spans zero it cannot be regarded as a statistically significant result. It should be noted that this effect size is likely an overestimation, owing to the selection bias from Crowe and Henderson (2003) as previously discussed.

One theme amongst the trials reporting the greatest benefit (Crowe and Henderson 2003; Huang et al. 2012; Johansson et al. 2005; Sjöling et al. 2003) was an educational intervention featuring an individualised component. Here, structure in the educational interventions was achieved by the format (e.g. written information, decision aid or group presentation), while individualisation came from clinical staff exploring patients’ concerns and responding to questions raised on an opportunistic basis. The success of this approach may be linked to theories of adult learning, which state that adult learners benefit from more self-directed, individualised, and goal-oriented interventions (Niksadat et al. 2022).

Postoperative quality of life

Only one trial, McGregor et al. (2004), reported data on postoperative quality of life, so meta- analysis was not possible. This individual result was not significant (SMD + 0.17, 95% CI −0.52 to + 0.85) but does serve to highlight the lack of patient reported outcomes in currently available trials.

Postoperative complication rate

Four trials (Biau et al. 2015; Crowe and Henderson 2003; Huang et al. 2012; Quintrec et al. 2003) assessed the impact of educational interventions on postoperative complication rates. Overall patient data (n = 676) indicated a relative risk ratio of 0.92 (95% CI + 0.41 to + 2.06) in favour of preoperative education. This result was congruent with the result of an earlier systematic review, which also found that “adverse events such as infection and deep vein thrombosis were lower [less frequent] in participants receiving preoperative education for hip or knee replacement compared to usual care, though differences were not statistically significant” (McDonald et al. 2014, p18).

Crowe and Henderson (2003) was the only trial to find a significant reduction in postoperative complications. Here, participants in the intervention group received an informational video and individualised teaching from a physiotherapist. Content “focused on the [patient’s] responsibilities during the post-operative phase, and [their] use of equipment” (Crowe and Henderson 2003, p90). After follow-up, patients in the intervention group (65 patients out of a total of 133) experienced 37% of all hip dislocations, 28% of all cardiac complications, 0% of all infections, and 12.5% of all other complications. The authors theorised that this pattern might be due to “comprehensive education…which placed emphasis on positioning post-operatively and prevention of complications” (Crowe and Henderson 2003, p95). This seems like a rational explanation for the observed difference, since educated patients would be more equipped to prevent, recognise, and report postoperative complications to staff. However, it should also be remembered that the result from this trial has already been discussed as an outlier in the dataset and remains at risk of bias due to the unnaturally selected population.

Quality of the evidence

Multiple trials reported on the primary outcome of preoperative anxiety and the secondary outcomes of postoperative pain and time-to-discharge. However, evidence for the patient knowledge, complication rate, and quality of life outcomes was more limited. Therefore, the confidence placed in each finding should reflect the availability and quality of evidence (Figs. 3 and 4). A GRADE appraisal framework has been applied to assist interpretation (Table 2). Inspection of funnel plots (Appendix 1) did not suggest publication bias.

Table 2.

GRADE assessment

							№ of patients		Effect		Certainty	Importance
№ of studies	Study design	Risk of bias	Inconsistency	Indirectness	Imprecision	Other considerations	Education	No education	Relative (95% CI)	Absolute (95% CI)
Anxiety
8	RCTs	seriousa, b	seriousc	seriousc	not serious	none	445	477	-	SMD 0.37 lower (0.63 lower to 0.11 lower)	⨁⨁◯◯ Lowa, b,c	CRITICAL
Patient Knowledge
2	RCTs	very seriousa, d	not serious	seriousc	not serious	none	80	73	-	SMD 0.37 higher (0.05 higher to 0.69 higher)	⨁◯◯◯ Very lowa, c,d	IMPORTANT
Pain score
7	RCTs	seriouse, f	not serious	seriousc	not serious	plausible residual confounding would suggest spurious effect, while no effect was observed	400	411	-	SMD 0.31 lower (0.47 lower to 0.15 lower)	⨁⨁⨁◯ Moderatec, e,f	CRITICAL
Time-to-discharge
5	RCTs	not serious	not serious	seriousc	seriousg	none	289	297	-	MD 1.12 lower (2.45 lower to 0.21 higher)	⨁⨁⨁◯ Moderatec, g	IMPORTANT
Quality of Life
1	RCTs	not serious	very serioush	seriousc	not serious	none	14	20	-	SMD 0.17 higher (0.52 lower to 0.85 higher)	⨁⨁◯◯ Lowc, h	IMPORTANT
Complications
4	RCTs	seriousb, d	not serious	seriousc	not serious	none	42/343 (12.2%)	46/333 (13.8%)	RR 0.89 (0.41 to 2.06)	14 fewer per 1,000 (from 87 fewer to 133 more)	⨁⨁⨁◯ Moderateb, c,d	CRITICAL

CI Confidence interval, MD Mean difference, RR Relative risk, SMD Standardised mean difference

Explanations

^aSelection bias introduced by recruitment of more vulnerable participants/tertiary setting

^bInadequate/unclear description of allocation concealment

^cHeterogeneity in clinical environment and/or educational content

^dInadequate blinding - patients and staff aware of intervention

^eProbable confounding introduced by educational content (e.g. preemptively requesting analgesia)

^fSignificant gender imbalance

^gHigh or otherwise unreported attrition rate

^hInadequate available data for comparison, small sample size

The generalisability of this review’s findings must be critically considered in context. Most trials were conducted within research-active tertiary units that specialise in hip and knee arthroplasty. Furthermore, in some cases trials recruited participants that were not representative of the general population – for example those with complex needs (Crowe and Henderson 2003) or a significant gender imbalance (Atabaki et al. 2019).

The usefulness of many of the included trials was also hampered by inadequate descriptions of their methods. Very few of the included trials provided a truly replicable description of their educational intervention. Some trials also provided incomplete descriptions of their intervention and control groups at baseline (Clode-Baker et al. 1997; Butler et al. 1996).

Overall, the most prevalent reason for downgrading the quality of evidence was risk of bias introduced by inadequate blinding of trial participants (Fig. 5). This was a concern because it may have introduced a source of performance bias into six of the studies included in this review – where patients with knowledge of the intervention, and intended benefits, may have been influenced by social desirability bias, feeling that they ‘owed’ their teachers results in return.

Fig. 5.

Risk of bias domains as percentages of all included studies

The forest plot for the anxiety outcome showed moderate evidence of heterogeneity (I² = 70%). Interestingly, subgroup analysis (Fig. 3) to compare the primary formats of education delivery showed that Crowe and Henderson (2003) were a prominent outlier, while other educational formats were broadly similar. As already mentioned, from a methodological perspective this trial was the only one in the anxiety analysis to feature a multi-stage multi-format educational intervention. Therefore, it seems plausible that the difference in effect size could have been introduced by patients receiving a dual benefit from educational reinforcement and a more individualised approach to care. Furthermore, the possibility of confounding also needs to be remembered since this trial took place in a specialist unit where participants were recruited from a more disabled at baseline – perhaps resulting in them having more to gain from the intervention.

Limitations

The most apparent limitation of this review is the clinical and methodological heterogeneity of included trials. Although this problem was anticipated during scoping review of the literature, it could not be wholly addressed by tightening the inclusion and exclusion criteria. In practice, it was challenging to develop criteria that clearly defined the scope of ‘educational interventions’, and the target patient population, whilst still ensuring enough relevant search results. With hindsight, the search strategy (Fig. 1) could have been improved by the including additional search terms which referenced specific intervention formats (e.g. schools, programme, workshop etc.). This would likely have increased the chances of detecting relevant trials, at the expense of additional time and resources spent abstract screening.

A key decision in the analysis was pooling two similar patient subgroups (hip and knee arthroplasty patients). This inevitably introduced clinical heterogeneity; however it was necessary due to the surprising lack of contemporary high-quality trials addressing either group of patients. Sensitivity analyses were conducted to explore the impact of this decision. Nevertheless, further primary research is needed before a robust meta-analysis can tackle either the hip or knee arthroplasty patient groups individually.

Another methodological constraint to consider was the choice to conduct a merged analysis of different formats of educational intervention (e.g. written, individual, group-class, video etc.). Although taking this approach limits the certainty that can be placed in findings, it was considered a pragmatic decision made for several reasons. Firstly, and most importantly, broader grouping of education formats was necessary due to the scarcity of trials featuring a single common format. Secondly, the conceptual similarity of educational interventions (e.g. the end goal of reducing anxiety and improving care) was thought more important than the specific mode of delivery. And lastly, it was acknowledged a priori that many existing trials did not describe their educational interventions in sufficient detail to be replicable. Therefore, in the absence of a single rigidly defined gold standard educational protocol, a merged statistical analysis using a random effects model was considered more representative of real-world clinical heterogeneity.

Another hinderance to this review was the lack of detailed reporting of individual trial protocols - specifically regarding the content, timing, and attendance rates of educational interventions. Most of the included trials featured core content relating to optimising recovery (Table 1) through improving patients’ understanding of the healing process, planning, or self-directed rehabilitative physiotherapy. However, methods were often incompletely described, and educational resources (e.g. videos, eLearning, or leaflets) were generally unavailable for comparison. This makes it difficult to comment confidently on the efficacy of individual approaches and make generalisable recommendations. Furthermore, only one trial included a patient reported outcome (quality of life score), and no trials attempted to report patient satisfaction with the educational intervention itself. Each of these deficits should be carefully addressed by future trials that wish to investigate the benefits of preoperative patient education in a replicable and holistic fashion. Interestingly, Anderson et al. (2020) looked at methodologically varied approaches to preoperative knee arthroplasty intervention and found no definitive evidence for an optimal educational format or content – ultimately recommending an individually tailored and multi-format approach as the most impactful for patients. This would seem to corroborate this review’s findings and, to some extent, help validate the decision to pool randomised controlled trials with differing educational content.

A final weakness of this review was the reliance placed upon patient self-reported outcome measures, as these tools are known to be subject to social desirability bias - especially when associated with educational interventions in healthcare (Miller 2012).

Conclusions

In summary, this review finds mixed evidence for the use of structured educational interventions in patients awaiting hip and knee arthroplasty. Results indicate that statistically significant benefits for the outcomes of anxiety (SMD − 0.37, 95% CI −0.63 to −0.11), patient knowledge (SMD + 0.37, 95% CI + 0.05 to + 0.69), and reported pain (SMD − 0.31, 95% CI −0.47 to −0.15). However, there was no strong evidence of an effect in either direction for the time-to-discharge, quality of life, or complication rate outcomes.

Interestingly there was more evidence of a benefit for the patient centric ‘soft’ outcomes of reported anxiety and postoperative pain, whereas the more clinician centric ‘hard’ outcomes of time-to-discharge and complication rate showed no significant difference following educational intervention. This might be explained by the fact that educational interventions do not occur in isolation; they are typically used as an adjunctive measure within broader systems spanning primary and secondary care which aim to pre-habilitate the patient, minimise surgical risk, and streamline discharge planning. As a result, these ‘hard’ outcome measures should be considered in context - as the cumulative effect of multiple interventions where patient education may not be the ‘rate limiting step’ in the process. Further research to compare the impact of educational interventions on patient reported outcomes and clinician centric outcomes is needed to corroborate this theory.

Effect sizes calculated by meta-analysis need to be carefully considered in context, and with relation to the likelihood of an observable clinical impact at the individual level. Based on this review the clinical impact of preoperative education on anxiety remains uncertain. Firstly, because the observed effect size was relatively small. And secondly, because despite carefully planned methodology, there was still evidence of statistical heterogeneity - a perennial challenge for systematic reviews involving educational interventions (Friedman et al. 2010; Hounsome et al. 2017; McDonald et al. 2014). Furthermore, there was also a potentially high risk for bias in seven of the 19 included trials (Fig. 6), mostly due to problems with blinding or incompletely reported outcome data. Any conclusions and recommendations should therefore be regarded with critical consideration and on an individual basis.

Fig. 6.

Risk of bias domains by indivudal trial

However, it is still possible to draw a number of interesting conclusions. Firstly, that the teaching strategy recommendations of Friedman et al. (2010) appear to be generalisable to patients awaiting hip and knee arthroplasty; namely the advantages of employing multiple communication methods, individualisation of teaching resources, and targeting interventions towards more vulnerable groups of patients. And secondly, this review did not observe compelling evidence to suggest the existence of a single rigidly defined ‘gold standard’ delivery format or timeframe for preoperative patient education. Confidence in these observations is increased by the fact that a scoping review, by Anderson et al. (2022), arrived at similar conclusions independently. Here, authors evaluated preoperative interventions for knee arthroplasty and found no definitive evidence of a single optimal approach but did identify the value of employing multiple delivery formats and an individualised approach to educational content.

Recommendations for future clinical practice and research

This review has highlighted several important aspects of current preoperative patient education and communication strategies that ought to be considered. They can be summarised as follows:

1. Preoperative education can represent a safe and cost-effective addition to standard care for those with heightened anxiety or greater risk of difficult to manage pain.

2. Certain patient groups may stand to gain more benefit from targeted educational interventions – for example those with complex care needs, learning difficulties, or a lower socio-economic status.

3. In the absence of clear evidence supporting one ‘gold standard’ method, educators may wish to consider using multiple communication techniques and teaching formats, tailored to their local population, available resources and individual patient preferences.

4. There is a notable lack of high-quality randomised controlled trials investigating preoperative education for hip or knee arthroplasty patients. More primary research is needed; where emphasis should be placed on clearly describing the population demographics, truly replicable methodology, and incorporating patient reported outcomes such as satisfaction and longitudinal quality of life measures.

Authors’ contributions

Conceptualisation and design: A.L.Data collection, abstract screening, and analysis: A.L. and L.C. and S.V. Administrative: L.C. and S.V. Writing original draft: A.L.Review of original draft: S.V. and L.C.Revisions and editing of original draft: A.L. and L.C. and S.V.

Funding

No external funding, donations, or research grants were received.

Data availability

No datasets were generated or analysed during the current study.

Declarations

Ethics approval and consent to participate

Not applicable.

Consent for publication

Participants - Not applicable. Authors – freely given and unanimous informed consent.

Competing interests

The authors declare no competing interests.