Insights into patient preferences for elective surgery during the COVID-19 pandemic

Study design and data acquisition

All included patients were sent a postal questionnaire on 24 August 2020, five months after the start of the first UK national lockdown. Patients were invited to respond via post, online (Jisc Online Surveys, UK), or via telephone. The closing date for the questionnaire was six weeks following the distribution date (5 October 2020). Patients were informed that their participation and the answers they gave would not influence their care. The questionnaire included a quality of life (QoL) measure (EuroQol five-dimension three-level questionnaire (EQ-5D-3L) and EuroQol visual analogue scale (EQ-VAS))¹⁵ and a joint-specific, patient-reported outcome measure (PROM) (either the Oxford Hip Score (OHS)^16,17 or Oxford Knee Score (OKS)).^18,19 Patient information was collected, including sex, age at the time of survey completion, length of time for which the patient had been on the waiting list, and whether the patient had received a previous joint arthroplasty. Patient-specific information regarding COVID-19 included self-reported COVID-19 risk status, personal history of COVID-19, and family or household history of COVID-19. Patients were asked the “COVID-19 alert” level at which they would be willing to receive arthroplasty surgery in an NHS hospital and in a (COVID-light) private sector hospital during the UK epidemic.

The project was registered as a service evaluation and the questionnaires were approved by the local Trust information governance team (PRN9129). Licencing for EQ-5D-3L (ID 36374), OHS (00OHS-870168), and OKS (00OKS-870166) questionnaires were obtained. EQ-5D-3L index scores were calculated using the UK value set²⁰ as outlined by EuroQol.²¹ A state worse than death was classified as an EQ-5D-3L index score of less than 0.^21,22

Patient and public involvement

Two patient representatives who were on the waiting list for joint arthroplasty at Cambridge University Hospitals were consulted prior to designing the patient survey. They provided insights into their top concerns regarding attending hospital for joint arthroplasty surgery during the COVID-19 epidemic in the UK, their thoughts underlying these concerns, and the information they would wish to know beforehand. These discussions underpinned the design of our questionnaire and informed our analysis.

Statistical analysis

Continuous variables were reported using means and standard deviations (SDs) if normally distributed and medians and interquartile ranges (IQRs) if non-normally distributed. Categorical variables were presented as a percentage. Binary categorical variables were analyzed using either a binomial paired comparison or Fisher’s exact test. Univariate analysis was performed using simple logistic regression and presented as proportional odds ratios (ORs). The differences between COVID-19-alert levels were assessed using the Kruskal-Wallis test for non-parametric continuous variables.

To assess the variables that influence the decision at which COVID-19-alert level patients were willing to attend for surgery, a multivariate ordinal regression model was built.²³ The model was built on the cohort awaiting hip arthroplasty (the larger group) and validated on the cohort awaiting knee arthroplasty. Due to a multimode approach being offered for questionnaire response, a response rate of between 50% and 60% was anticipated (152 to 182 responses).²⁴ With an r² of 0.30, a maximum of seven variables could be incorporated. Independent variables were decided a priori and were based on patient involvement and clinician experience. The variables were joint-specific QoL (OHS or OKS), global QoL (EQ-VAS and EQ-5D-3L index scores), personal or family history of COVID-19 infection, self-reported COVID-19-risk status, patient age, and sex. The EQ-VAS, OHS, and OKS were treated as continuous variables as they represent a continuous range of health states.²⁵ Results of the multivariate analysis were presented as adjusted ORs and probabilities. For all statistical tests, a significance level of p < 0.05 was used. All statistical analyses were performed using R in RStudio (Version 4.0.2, USA)

Patient demographics and survey information

At the time of waiting list screening, 570 patients were waiting for a primary hip or knee arthroplasty and 540 were eligible for inclusion (Figure 1). The final analysis included 400 complete responses (400/540, 74.1%), of which 236 patients were awaiting hip arthroplasty and 164 patients were awaiting knee arthroplasty. Patient information including demographic details, arthroplasty history, COVID-19-related factors, and patient-reported outcome measures are shown in Table II. Patients were asked to self-report COVID-19 risk status as determined by their primary care providers. Of the 400 completed responses, 307 patients (78.7%) reported not knowing their COVID-19 risk status.

Fig. 1

Flow diagram of patient inclusion and exclusion.

In the full cohort, 51.7% (207/400) of the included patients expressed a preference to delay their joint arthroplasty until transmission of COVID-19 was either low or nonexistent (COVID-19 alert levels 1 or 2). The remaining 48.3% (193/400) were willing to undergo surgery while “a COVID-19 epidemic is in general circulation” (COVID-19 alert levels 3 to 5), which increased significantly to 54.0% if a private, “Covid-light” hospital was available (p = 0.024, binomial paired comparison).

Preference for timing and setting of surgery in patients awaiting hip arthroplasty

In the cohort awaiting hip arthroplasty, 54.6% (129/236) were willing to attend an NHS hospital for surgery when a “Covid-19 epidemic is in circulation” (COVID-19 alert level 3 and above). Patients were significantly more likely to be willing to attend a private sector (“COVID-light”) hospital compared to an NHS hospital at COVID-19 alert levels 3 (65.3% vs 54.6%, p = 0.024), 4 (35.2% vs 22.0%, p = 0.022), and 5 (18.6% vs 10.2%, p = 0.012, binomial paired comparison) (Figure 2a).

Fig. 2

Percentage of patients on the hip and knee arthroplasty waiting list who would be willing to receive a) hip arthroplasty and b) knee arthroplasty at each of the five COVID-19 alert levels (Table I) in either an NHS hospital (blue bars) or a private sector (COVID-light) hospital (red bars). Exact p-values displayed, binomial paired comparison.

Male sex was significantly associated with a preference for surgery at the more severe COVID-alert levels for hip arthroplasty (OR 1.75, (95% confidence internal (CI) 1.07 to 2.86), p = 0.024, simple ordinal logistic regression). Patient factors that did not significantly influence the COVID-19 alert level that patients were willing to attend for surgery were age at the time of the questionnaire, previous joint arthroplasty surgery, and time on the waiting list (Supplementary Table i). Self-reported COVID-19 risk status and personal or family history of confirmed COVID-19 infection did not significantly influence the COVID-19 alert level that patients were willing to attend for hip arthroplasty surgery (Supplementary Table i).

The median EQ-5D-3L index score of patients awaiting hip arthroplasty was -0.016 (IQR -0.074 to 0.534) and 54.2% (128/236) reported a score “worse than death”. The median EQ-VAS of patients awaiting hip arthroplasty was 45 (IQR 27.5 to 70.0). In univariate analysis, patients with lower EQ-5D-3L index scores and EQ-VAS had a preference for surgery at the more severe COVID-19 alert levels (EQ-5D-3L index score: proportional OR 5.00 (95% CI 2.42 to 11.1), p < 0.001; EQ-VAS: proportional OR 1.02 (95% CI 1.01 to 1.03), p < 0.001).

The median OHS for the cohort awaiting a hip arthroplasty was 11/48 points (IQR 8 to 20). Patients reporting lower OHS had increased odds of being willing to attend for surgery at a more severe COVID-19 alert level (proportional OR 1.10 (95% CI 1.07 to 1.14), p < 0.001, simple ordinal logistic regression). Patients choosing COVID-19 alert level 1 and 2 had significantly higher OHS compared to patients who were willing to have surgery at COVID-19 alert levels 3, 4 and 5 (Supplementary Figure aa).

Multivariate analysis of preferred COVID-19 alert status at which patients are willing to receive hip arthroplasty surgery

Controlling for EQ-5D index score, EQ-VAS, COVID-19 risk status, personal or family history of COVID-19, age, and sex, a lower OHS (adjusted OR 1.10 (95% CI 1.05 to 1.17)) was significantly associated with patients willing to attend for hip arthroplasty surgery at higher COVID-19 alert levels (Table III). As the OHS increased, the probability that a patient would be willing to receive arthroplasty surgery while transmission of COVID-19 was low or nonexistent (COVID-19 alert levels 1 or 2) increased. Conversely, as the OHS worsened, the probability that a patient would be willing to attend while “a COVID-19 epidemic is in general circulation” (COVID-19 alert levels 3, 4, or 5) increased (Figure 3a).

Fig. 3

The probability of patients choosing each individual COVID-19 alert level at each a) Oxford hip or b) Oxford knee score controlling for EuroQol (EQ)-index, EQ-visual analogue scale (EQ-VAS), patient-reported COVID-19 risk status, personal or family history of COVID-19, age, and sex.

Based on this, the patient’s selected COVID-19 alert level was categorized into a binary outcome of low alert level (levels 1 and 2) versus high alert level (levels 3, 4, and 5). When controlling for the same variables, patients with lower OHS had higher odds of being willing to attend while “a COVID-19 epidemic is in general circulation” (COVID-19 alert levels 3, 4, or 5) (adjusted OR 1.54 (95% CI 1.45 to 1.63)). Furthermore, patients with lower OHS were more likely to choose a private, “COVID-light” hospital while “a COVID-19 epidemic is in general circulation” (COVID-19 alert levels 3, 4, or 5) (adjusted OR 3.50 (95% CI 3.26 to 3.71)) (Figure 4a).

Fig. 4

The probability of patients choosing COVID-19 alert levels 3, 4, or 5 at each a) Oxford Hip or b) Oxford Knee Score controlling for EuroQol (EQ)-index, EQ-visual analogue scale (VAS), patient-reported COVID-19 risk status, personal or family history of COVID-19, age, and sex.

Preferences for elective surgery among patients awaiting knee arthroplasty

Patients awaiting elective knee arthroplasty were significantly less willing to attend for surgery at an NHS hospital at COVID-19 alert level 3 and above compared to those awaiting elective hip arthroplasty (64/164, 39.0%; p < 0.001, Fisher’s exact test). Patients awaiting knee arthroplasty were significantly more willing to attend a private sector hospital than an NHS hospital for their surgery at COVID-19 alert levels 4 (20.7% vs 11.6%, p < 0.001) and COVID-19 alert level 5 (10.4% vs 4.9%, p = 0.003) but not at COVID-19 alert level 3 (54.3% vs 39.0%, p = 0.092, all binomial paired comparison) (Figure 2b).

In patients waiting for knee arthroplasty, male sex did not have a significant association with the COVID-19 alert level that patients would be willing to attend for their surgery in an NHS hospital (proportional OR 1.28 (95% CI 0.71 to 2.33), p = 0.405, simple ordinal logistic regression). Similar to the hip cohort, factors not significantly associated with the COVID-19 alert level at which patients were willing to attend for surgery were age, previous joint arthroplasty surgery, time on the waiting list, self-reported COVID-19 risk status, and personal or family history of COVID-19 (Supplementary Table i).

The EQ-VAS and EQ-index scores in the knee arthroplasty group were significantly higher when compared to the hip arthroplasty group (median EQ-VAS: 60.0, median EQ-5D-3L index score: 0.159; both p < 0.001; Mann-Whitney U test). There was also a lower proportion of patients waiting for knee arthroplasty reporting an EQ-5D-3L index score that was worse than death (34.8%, 57/164). Patients awaiting knee arthroplasty with a lower EQ-5D-3L index score and EQ-VAS had a preference in attending for surgery at higher COVID-19 alert levels (EQ-5D-3L index score: proportional OR 14.9 (95% CI 5.70 to 38.5), p < 0.001; EQ-VAS: proportional OR 1.03 (95% CI 1.01 to 1.04), p = 0.002, all simple ordinal logistic regression).

The median OKS for the cohort awaiting a knee arthroplasty was 15 out of 48 (IQR 10 to 22). Patients reporting lower OKS had increased odds of being willing to attend at a higher COVID-19 alert level (proportional OR 1.16 (95% CI 1.11 to 1.21), p < 0.001). Patients choosing COVID-19 alert level 1 and 2 had significantly higher OKS compared to patients who were prepared to have surgery at COVID-19 alert levels 3, 4, and 5 (Supplementary Figure ab).

Using a multivariate analysis, controlling for factors hypothesized to play a role in influencing a patient’s decision for surgery (age, sex, EQ-VAS, EQ-5D-3L index score, personal or family history of COVID-19 infection, and reported COVID-19 risk status) demonstrated that lower OKS was significantly associated with an increased odds of patients being willing to have their surgery in an NHS hospital at a higher COVID-19 alert level (adjusted OR 1.14 (95% CI 1.06 to 1.22), p < 0.001) (Table III). As OKS increased, the probability that patients would prefer to attend for surgery at a lower COVID-19 alert level (level 1 or 2) increased. Conversely, as the OKS decreased, the probability that a patient would be willing to attend for surgery at a higher COVID-19 alert level (level 3, 4, or 5) increased (Figure 3b). When using a binary outcome (COVID-19 alert levels 1 and 2 vs levels 3, 4, and 5) and controlling for the same variables, patients reporting lower OKS had an increased odds of being willing to attend at COVID-19 alert levels 3, 4, or 5 in an NHS hospital (adjusted OR 1.16 (95% CI 1.07 to 1.26)). These odds increased further if surgery in a private (COVID-light) hospital was available (adjusted OR 1.41 (95% CI 1.29 to 1.53)) (Figure 4b).

Implications for clinicians and policymakers

Our results demonstrate how patient perceptions around elective orthopaedic surgery have changed in the context of the pandemic. Understanding how factors such as joint-specific QoL influence the decision to undergo surgery is of great importance to surgeons, waiting list management teams, and commissioners. Informed consent for a surgical procedure is only valid if a patient understands and accepts the risks involved.²⁷ The present data has made it clear that COVID-19 alert levels play an important role for patients in their assessment of the risks of surgery during the pandemic and therefore should be accounted for during the consenting process.

Recent documents published throughout Europe emphasize the importance of prioritizing patient preferences in elective surgery^8,9 and elective orthopaedic surgery.^5,7 A high response rate to our survey was achieved at 74.1% (400/540). Postal surveys without financial incentives have been documented as typically achieving a response rate of around 25%,²⁸ with the rate increasing if financial incentives are offered or if the survey is of high interest to the population being surveyed.²⁹ Factors contributing to our high response rate may include the multimodal methods for return of the questionnaire (online, mail, telephone), and the topic being a high priority to patients.²⁴ Our strong survey response indicates the importance in engagement of patients in the planning of surgical waiting lists, and that patients want to be involved in this process.

The OHS, OKS, and EQ-5D-3L are widely used patient-reported outcome measures (PROMs) in research, audits, and clinical practice.³⁰ Indicators of patient experience are prioritized at policy, commissioning, and management levels and are widely used within NHS accountability frameworks. Conversely, PROM questionnaires are rarely included in such frameworks nor do they routinely form part of the patient-surgeon consultation. We recommend surgeons consider prospectively performing QoL questionnaires with patients at the time of being placed on the waiting list. This will help patients weigh up the risks versus benefits of surgery based on their individual QoL and joint symptoms against a backdrop of the pandemic.

Study strengths and limitations

One strength of our study is the high patient response rate achieved. We consulted with patient representatives to help design our survey and plan our analysis a priori. Patients were offered a range of options for returning their questionnaires, supported by existing evidence that different methods of questionnaire return do not significantly influence the reliability of completion of QoL questionnaires.³¹ Self-reported measures can be vulnerable to ‘response bias’ where a person’s response to a survey can be influenced by a number of factors, including perceived desirability of a given response (for example, some individuals may select extreme responses in order to expedite surgery). To avoid this, the cover letter accompanying the survey stated that ‘response would not impact clinical care’. While we offered multimodal methods of questionnaire return, there may be a proportion of patients who experience barriers to completing such forms, owing to literacy levels or language barriers. Further, such services could include alternative language options. One factor that we hypothesized would play a role in a patient’s decision for surgery was the personal COVID-19 risk status. However, only 21.3% knew their COVID-19 risk status and therefore only limited data were available to assess whether this factor effects the COVID-19 alert level that patients are willing to receive surgery.

Strengths and limitations in the context of a changing COVID-19 landscape

One challenge to translating the findings of this study into clinical practice is posed by the changing landscape of the UK’s response to managing COVID-19. On 12 October 2020, one week after our study period closed, the UK’s categorization of the COVID-19 alert level changed such that each area of England was categorized as medium, high, or very high-alert according to the risk of infection in that area. On 23 October the introduction of a new five-tier system was announced in Scotland. This regional approach means that some neighbouring areas may be assigned different tiers. Therefore, a patient can be resident in an area in a different tier to that of the hospital at which they are awaiting surgery. This is important, given how significantly COVID-19 alert level impacts patients’ decisions on proceeding with surgery.

In conclusion, while the magnitude of delayed elective surgical procedures poses serious challenges for healthcare systems, patients are also facing challenges with regards to understanding the proportionality of risk associated with surgery during the ongoing pandemic. We have reported on the use of patient-reported QoL to aid shared decision-making and guide prioritization on elective waiting lists. We recommend that QoL measures form a component of prioritization frameworks, ensuring patient preferences are carefully considered against the background of a fluctuating COVID-19 landscape.