Expert Consensus for the Use of Outpatient Rehabilitation Visits After Total Knee Arthroplasty: A Delphi Study

INTRODUCTION

There is no consensus for how to use rehabilitation visits after total knee arthroplasty (TKA). Recent systematic reviews and clinical practice guidelines have not provided specific visit usage recommendations (e.g., visit timing, visit frequency, care duration) due to lack of supporting evidence—especially in the outpatient setting.^{21, 24} In the absence of guiding evidence, outpatient rehabilitation visits are used quite variably after TKA in the United States. Recent studies have reported standard deviations ranging from 5–23 outpatient visits per episode,^{3, 7, 19, 29, 31, 32} and regional differences in the average timing and duration of outpatient rehabilitation have been reported.⁴² Variable use of visits may undermine the overall quality and cost-effectiveness of outpatient rehabilitation,¹¹ which is especially relevant given payors’ emphasis on reducing TKA costs.^{28, 38}

Given the lack of available evidence, expert consensus could be useful for developing preliminary visit usage recommendations. In 2014, Westby et al. used the Delphi process to build consensus around best practices for TKA rehabilitation.⁴⁸ This study produced many key practice recommendations, but it did not develop consensus around visit usage in post-acute rehabilitation. The lack of consensus may have been partially due to (1) the wide scope of included rehabilitation topics, which might have limited the focus on visit usage and (2) the opinion among participants that rehabilitation should be individualized.⁴⁸

Individuals do have unique needs, goals, and expectations after TKA,^{39, 43} which suggests a preference-sensitive approach to rehabilitation may be ideal.^{10, 14} Preference-sensitive care occurs when well-informed patients make health care decisions in line with their individual preferences.^{10, 45, 47} Some patients with TKA may prefer extensive rehabilitation to help them achieve ambitious goals, while others may prefer to recover mostly independently. However, because no evidence exists to help patients make informed decisions, visit usage is more likely driven by the capacity of local health care systems.^{12, 45–47} This type of supply-sensitive care fosters overuse of health care visits and ignores patients’ individual needs and preferences. Therefore, patients and clinicians could benefit from guidance to anchor decisions about how to use outpatient rehabilitation visits after TKA to patient preferences.

In this study, we used the Delphi method to develop expert consensus for the use of outpatient rehabilitation visits after TKA. We hypothesized our focused objective and approach would help us avoid the challenges experienced in previous studies for developing consensus. Our goal was to develop recommendations that can help patients make decisions about the use of visits in line with their own preferences.

METHODS

We selected the Delphi method for this study because it allowed us to build expert consensus remotely and asynchronously. This format made it feasible to include a diverse panel of experts from across the United States. It also allowed participants to remain anonymous, which may encourage diverse viewpoints.⁴¹

Recommendation Development Phase

We used a reactive modification to the Delphi method, where the research team developed a preliminary list of recommendations for participants to consider in subsequent rounds. ²⁷ This approach ensured subsequent Delphi rounds focused specifically on our research goal—to develop visit usage recommendations for outpatient rehabilitation after TKA. We created the preliminary list to encompass a range of visit usage strategies that are commonly observed in clinical practice and/or TKA-related research. We also used published research to shape the list’s format and contents, which we describe in the subsequent sections of this manuscript.

We piloted and refined the preliminary recommendations and all survey materials using an iterative process involving members of our research team and broader institutional network with relevant expertise in TKA rehabilitation. We used pilot feedback to revise recommendations, generate ideas for new recommendations, and ensure all survey materials were worded clearly and without bias. Using this process, we developed two categories of recommendations: (1) visit frequency recommendations and (2) complementary recommendations, which were designed to be used with the visit frequency recommendations. None of the individuals involved in the development phase participated in subsequent rounds.

Visit frequency recommendations

We framed the visit frequency recommendations around efficient visit use, which we defined as the minimum frequency of supervised visits needed to provide adequate care. We used two main strategies to develop these recommendations. First, we anchored each recommendation to a specific timeframe after surgery (i.e., postoperative month 1, 2, or 3). We used this strategy because visit frequency recommendations may depend on postoperative tissue healing.⁴⁸ Also, rehabilitation clinicians typically re-evaluate treatment plans on monthly intervals, and patients often discharge from rehabilitation within three months after surgery.^{3, 32, 42} Second, we anchored each recommendation according to patients’ observed recovery status relative to their expected recovery (i.e., patient is demonstrating a slow, typical, or fast recovery relative to expected). We used this strategy because patients recover differently based on their individual characteristics,²¹ and monitoring recovery against an expected value can be a useful decision making strategy.¹³ Together, these strategies resulted in the development of nine separate combinations of recovery month + recovery status (e.g., postoperative month 1, recovering slower than expected). For each of these combinations, we asked panelists to rate their level of agreement with six different visit frequency options (0x/month, 1x/month, 2x/month, 1x/week, 2x/week, 3x/week) that are commonly used in clinical practice. Overall, panelists considered 54 different visit frequency recommendations (nine combinations of recovery month + recovery status with six frequency options each). See BOX 1 for an illustration of the structure used to create these recommendations.

BOX 1. General structure used to create 54 different preliminary visit frequency recommendations.

Please rate your level of agreement or disagreement with the following visit frequency recommendations for outpatient rehabilitation frequency after TKA:
During the $\frac{\text{first},\text{second,or}\text{third}}{\text{select}\text{one}}$ month after TKA, the most efficient visit frequency of outpatient rehabilitation for patients demonstrating a $\frac{\text{typical},\text{slow},\text{or}\text{fast}}{\text{select}\text{one}}$ recovery relative to expected should be $\frac{0\text{x}/\text{month},1\text{x}/\text{month},2\text{x}/\text{month},1\text{x}/\text{week},2\text{x}/\text{week},\text{or}3\text{x}/\text{week}}{\text{select}\text{one}}$
1 Strongly Disagree	2	3 Somewhat Disagree	4	5 Neutral	6	7 Somewhat Agree	8	9 Strongly Agree

Abbreviations: TKA, Total Knee Arthroplasty

Complementary recommendations

We designed the complementary recommendations to be used with the visit frequency recommendations. These recommendations focused on TKA-related issues that (a) could influence visit usage decisions, (b) have no established consensus, and (c) could be readily used to support the visit frequency recommendations. These issues included (1) the optimal timing for initiating and stopping outpatient rehabilitation,²¹ (2) important range of motion recovery thresholds,³⁴ (3) the proportion of patients who may need more or less rehabilitation than the typical patient,³⁰ and (4) the safety and effectiveness of telerehabilitation after TKA.²¹ We did not intend for the complementary recommendations to be exhaustive because many factors that could influence care (e.g., personal and environmental factors) would be difficult to capture and apply.

Panelist recruitment

We sought to enroll a diverse Delphi panel consisting of physical therapists, orthopedic surgeons, and advanced practice providers (i.e., physician assistants, nurse practitioners) with both clinical and research expertise in TKA recovery. We recruited participants from our own professional networks and through author lists of recently published literature in TKA rehabilitation. We limited our recruitment to individuals based in the United States because practice patterns can vary widely between countries.

Individuals were eligible to participate if they had ≥ 5 years of TKA-related experience and met one of the following volume criteria: (1) physical therapist who sees ≥ 10 patients with TKA/year in the outpatient setting, (2) orthopedic surgeon who performs ≥ 50 TKAs per year, or (3) advanced practice provider who sees ≥ 50 patients with TKA/year. Additionally, clinicians from these professions who did not meet the volume criteria were eligible to participate if they had ≥ 5 years of experience conducting and publishing TKA-related research. We aimed to enroll 30 participants overall with representation from each eligible profession; we chose this sample size as previous Delphi studies have observed stability with as few as 23 participants.¹ During Round 1, we collected baseline information related to participants’ TKA-related experience, geographical location, and practice setting. After enrollment, we sent participants personalized email reminders for each Delphi round to maximize response rates.

Delphi Structure

We compiled the preliminary recommendations into surveys for panelists to consider in Round 1 of the Delphi process. Panelists rated their level of agreement or disagreement with each visit frequency recommendations using a Likert scale ranging from 1 (strongly disagree) to 9 (strongly agree)(see BOX 1). We used the RAND UCLA method to define consensus,⁴ where recommendations with a median response of ≥ 7/9 and less than 30% of responses in the 1–3 range were considered to have reached consensus.^{5, 6} Because panelists separately rated six different visit frequency options for each combination of recovery month + recovery status, more than one frequency within a combination could potentially meet our definition of consensus. In this case, we considered the frequency with the higher mean response to have reached consensus.

A few of the complementary recommendations were scored using the Likert scale and method described above, but most of them required numeric responses. We did not evaluate the numerically-scored recommendations for consensus. Instead, we reported the mean response and standard deviation for these recommendations from the final Delphi round. In addition to Likert and numeric responses, the Delphi survey also included open-ended text boxes to record panelists’ comments after each recommendation. Panelists were encouraged to provide the rationale behind their response, their opinions on specific recommendations, and suggestions for new/revised recommendations.

We conducted additional rounds as needed to develop consensus. During each subsequent round, we included all previous recommendations that had not reached consensus, and we revised and added new recommendations based on panelist feedback. We also provided panelists with (1) a list of recommendations that previously reached consensus, (2) a comparison between the individual’s response and the group’s response for each recommendation, and (3) a qualitative summary of the group’s comments for recommendations from the previous round. Supplementary File 1 describes the complete list of recommendations included in each Delphi round.

To determine whether additional rounds were needed, we examined the response stability for each Likert-scored recommendation that had not reached consensus. Specifically, we compared the absolute difference in the coefficient of variation (CV, standard deviation / mean) between rounds and considered values < 0.2 to be indicative of stability.^{23, 37} We also monitored the number and content of comments between rounds; we considered fewer comments with no major change in content as further evidence of response stability.¹⁸ We discontinued the Delphi process once all Likert-scored recommendations had either reached consensus or demonstrated response stability. See FIGURE 1 below for a summary of the Delphi round structure.

FIGURE 1.

Visual summary of study structure

All study procedures were approved by the Colorado Multiple Institutional Review Board. We administered all surveys electronically via REDCap electronic data capture tools hosted at the University of Colorado.^{16, 17} We maintained panelists’ anonymity throughout the entirety of the study.

RESULTS

We invited 49 individuals to participate, and 30 panelists enrolled with 29 completing two Delphi rounds. The panel had a median of 17 years of TKA-related experience and included individuals from 11 US states and 22 unique zip codes. The panel included representation from the Department of Veterans Affairs (n=6), university settings (n=6), non-profit organizations (n=9), private practice (n=5), and other practice settings (n=3). Twelve out of the 18 participating physical therapists held either a PhD or board-certified specialization. Additional details regarding the panel’s experience and participation by round are available in FIGURE 2 and TABLE 1. We did not conduct additional rounds beyond Round 2 because all Likert-scored recommendations had either reached consensus or met our definition of response stability.

FIGURE 2.

Flow chart depicting panelist recruitment and participation

TABLE 1.

Summary of panel’s TKA-related experience. All values presented as median (range) unless otherwise noted.

Characteristic	Orthopedic surgeons (n=8)	Physical therapists (n=18)	Advanced practice providers (n=3)	Total (N=29)
Years of TKA experience	15.5 (11–36)	15 (6–48)	6 (6–18)	15 (6–48)
TKA experience Primarily clinical Equally clinical + research Primarily research	n = 5 n = 3 n = 0	n = 12 n = 4 n = 2	n = 3 n = 0 n = 0	n = 19 n = 8 n = 2
Patients with TKA treated or operated on per year *	250 (120–400)	45 (12–180)	240 (144–240)	120 (12–400)

Abbreviations: TKA, Total Knee Arthroplasty

^*Panelists with primarily research experience were not included in this calculation

Visit frequency recommendations

The panel reached consensus in Round 1 for visit frequency recommendations in eight of the nine combinations of recovery month + recovery status. The panel did not reach consensus on a recommendation for patients recovering faster than expected in postoperative month 3 during either round. In both rounds, some panelists commented that fast-recovering patients should already be discharged by postoperative month 3. Other panelists commented that these patients’ need for rehabilitation depends on their individual goals. Most surgeon panelists fell within the first group; non-surgeon panelists were more equally divided.

All Likert-scored complementary recommendations reached consensus. The panel agreed that outpatient rehabilitation should be initiated within 1 week following TKA, and that telerehabilitation can be recommended for patients demonstrating a typical or fast recovery, but not for patients demonstrating a slow recovery. Many panelists voiced concern that telerehabilitation would deprive slow-recovering patients of manual therapy treatments or in-person assessment, which they viewed as important for this population. However, no safety concerns were raised for telerehabilitation, and several panelists commented it could be appropriate for slow-recovering patients with care access challenges.

The panelists also provided numeric responses for (1) the optimal discharge timing from outpatient rehabilitation, (2) the proportion of patients who demonstrate a slower or faster recovery than expected, which may indicate they need more or less care than anticipated, and (3) important knee flexion and extension range of motion thresholds. The complementary recommendations are displayed in Table 3 along with the format (Likert or numerical) used to score them.

TABLE 3.

Summary of complementary recommendations and their potential application the visit frequency recommendations

Complementary recommendation	Scoring method (score)	Potential application
Timing of outpatient rehabilitation
Assuming a patient does not receive sub-acute or home health rehabilitation, outpatient rehabilitation should be initiated within the first week after TKA	Likert (Consensus reached in Round 2)	Clinicians may recommend patients to schedule their first visit within one week after surgery
On average, patients who are demonstrating a slower recovery than expected are anticipated to discharge from outpatient rehabilitation around 15 weeks after surgery	Numerical (Mean = 15 weeks, SD = 3 weeks)	Clinicians may initiate discussions about discharge around these timeframes. Patients can use them to inform their preferences for when to discontinue supervised rehabilitation.
On average, patients who are demonstrating a typical recovery compared to expected are anticipated to discharge from outpatient rehabilitation around 10 weeks after surgery	Numerical (Mean = 10 weeks, SD = 3 weeks)
On average, patients who are demonstrating a faster recovery than expected are anticipated to discharge from outpatient rehabilitation around 8 weeks after surgery	Numerical (Mean = 8 weeks, SD = 2 weeks)
Proportion of patients who demonstrate a slow, typical, or fast recovery
Approximately 22% of patients demonstrate a slower recovery than expected after TKA, which may indicate they need more rehabilitation than anticipated	Numerical (Mean = 22%, SD = 10%)	Categorizing a patient’s recovery requires clinical judgment. Clinicians can use these values to compare how frequently they are recommending more/less care than typical compared to expert opinion.
Approximately 54% of patients demonstrate a typical recovery compared to expected after TKA	Numerical (Mean = 54%, SD = 14%)
Approximately 24% of patients demonstrate a faster recovery than expected after TKA, which may indicate they need less rehabilitation than anticipated	Numerical (Mean = 24%, SD = 10%)
Important range of motion thresholds
Most patients need approximately 114 degrees of knee flexion range of motion to be functional with most daily activities after TKA	Numerical (Mean = 114 degrees, SD = 6 degrees)	Certain tasks that most patients perform (e.g., walking) require a minimum range of motion. Patients who are not tracking towards these thresholds could be considered as slow recovering regardless of their recovery in other domains.
Most patients need to be within approximately 2 degrees of knee extension from neutral to be functional with most daily activities after TKA	Numerical (Mean = 2 degrees, SD = 2 degrees)
Telerehabilitation
Telerehabilitation should not be recommended for patients demonstrating a slower recovery than expected after TKA	Likert (Consensus reached in Round 1)	Some patients may prefer telerehabilitation or have access to care challenges (transportation, rural area), which may influence their preferences for rehabilitation care. Clinicians may choose to recommend telerehabilitation to patients if their recovery is progressing fast or typically.
Telerehabilitation can be recommended for patients demonstrating a typical recovery compared to expected after TKA	Likert (Consensus reached in Round 1)
Telerehabilitation can be recommended for patients demonstrating a faster recovery than expected after TKA	Likert (Consensus reached in Round 1)

Abbreviations: TKA, Total Knee Arthroplasty

DISCUSSION

The panelists reached consensus on visit usage recommendations for outpatient TKA rehabilitation that can be used during the first 3 months after surgery. We envision that patients and clinicians can use these expert recommendations as the starting point for preference-sensitive care decisions. Their clinical utility may best be illustrated by example (see BOX 2).

BOX 2. Case examples of using recommendations to promote preference sensitive care for patients demonstrating typical, fast, and slow recoveries.

Typical Recovery Patient A is pleased with their recovery at week 8 (start of month 3) after TKA. Their pain has improved, and they have met their goal of returning to a walking program. Patient A’s physical therapist assures them their recovery has progressed typically thus far. Their physical therapist suggests that most patients are recommended to be seen 1x/week at this point after surgery with a goal of discharging around week 10. Patient A considers this information and decides to return for one additional visit at week 10. They would prefer to rehabilitate independently until then because they feel confident with their home exercise program and have already achieved their primary goal. Fast Recovery Patient B is discussing with their physical therapist how often to attend treatment next week (postoperative week 5). Patient B feels like their recovery is going very well, but they assume they must continue at 2x/week to keep their positive momentum. Their physical therapist informs them their recovery is progressing much faster than expected and that 1 visit per week is usually recommended for patients this far ahead of schedule. Patient B feels reassured and agrees to reduce their visit frequency. Now they can spend more time next week with their family who is visiting from out of town. Slow Recovery Patient C feels overwhelmed 6 weeks after their TKA. Their recovery has been slower and more painful than expected, and they are struggling to attend rehabilitation 2x/week while caring for an aging parent. Patient C asks their physical therapist about using telerehabilitation. Their physical therapist tells them that in-person and telerehabilitation may lead to similar outcomes on average, but experts have recommended in-person treatment for patients experiencing a slow or challenging recovery. Their therapist also believe that in-person sessions will be critical for managing their pain and monitoring their progress moving forward. After further discussion, they both agree to trial a hybrid approach. Patient C will attend one in-person and one telerehabilitation appointment per week, and they will re-evaluate the plan together in two weeks.

The patient cases above demonstrate simplified examples of how these recommendations could facilitate preference-sensitive care. Each patient used the best available evidence (expert opinion) to help make care decisions in line with their own preferences.¹² This type of preference-sensitive approach requires supporting evidence, to ensure patients are well informed in their preferences. Conversely, supply-sensitive care occurs in the absence of medical theory or evidence, and care usage is driven primarily by the capacity of the local healthcare system.^{12, 45–47} Current post-TKA rehabilitation practice exemplifies supply-sensitive care, where visit usage varies considerably by clinician, facility, and region in the absence of guiding evidence.^{3, 19, 29, 31, 32, 42}

The clinician’s role in promoting preference-sensitive care using this study’s recommendations—or any medical evidence—should not be overlooked. To provide preference-sensitive care, clinicians must engage patients in the decision-making process,⁴⁷ attempt to maintain equipoise,¹⁰ and avoid rigidly applying evidence.¹⁵ Clinicians may face barriers to consistently practicing this way in outpatient rehabilitation after TKA.¹⁴ Therefore, the recommendations from this study may be most effective when combined with clinician training,²⁵ patient decision aids,³³ or patient-facing support programs^{40, 44} to promote preference-sensitive decisions.

Preference-sensitive care aligns closely with the principles of value-based care—models that promote guideline-adherent, patient-centered, and cost effective treatments.⁹ They also reward treatment quality over quantity, which contrasts with the fee-for-service models that frequently guide rehabilitation.²⁶ There is an urgent need for value-based models of rehabilitation after joint replacement, because payors are incentivizing streamlined treatment pathways after surgery.³⁸ Rehabilitation organizations could use this study’s results within a value-based care model for joint replacement. Specifically, the recommendations could be used to promote efficient use of rehabilitation visits, guided by patient preferences. When combined with guideline-adherent treatment,²¹ value-based care models could meaningfully improve patient outcomes; patients with knee osteoarthritis who make informed, preference-sensitive care decisions have reported higher quality of life, function, and satisfaction compared to patients who do not.^{8, 35, 36} Value-based models may also reduce overall costs,^{8, 35, 36} which could help ensure that outpatient rehabilitation remains a valued and reimbursable service after TKA.

Clinicians must use their clinical judgement when applying the recommendations from this study in practice. In particular, they need to determine whether a patient is demonstrating a slow, typical, or fast recovery relative to expected. The complementary recommendations were designed to help clinicians make these determinations (TABLE 3). For example, a clinician may determine a patient’s recovery is slower than expected if they are not on pace to achieve an important range of motion threshold (e.g., around 114 degrees of knee flexion based on this study’s results). However, clinicians must consider numerous individual factors when assessing recovery, such as the patient’s preoperative prognosis,²¹ pain, wound healing status,⁴⁸ and personal goals/expectations.^{39, 43} Clinicians and patients should also discuss external factors that may influence the patient’s care preferences such as familial support, access/transportation to rehabilitation, and insurance.^{14, 48}

Clinicians should also consider published research when using these recommendations with patients. For example, clinicians should inform patients of the evidence surrounding telerehabilitation if they recommend against using it for patients perceived to be recovering slowly (Box 2, slow recovery). The panelists in our study agreed that telerehabilitation should not be recommended for patients who are recovering more slowly than expected, and many commented that in-person treatment and assessment was especially important for these individuals. This recommendation appears to conflict with recent TKA evidence, which suggests that telerehabilitation and in-person rehabilitation lead to similar outcomes on average.²⁰ However, most clinical studies investigate the average treatment response of a population, so it is possible the slow-recovering subgroup could benefit more from in-person treatment. Regardless, clinicians should discuss the evidence and rationale underpinning their recommendations to help patients make preference-sensitive decisions.

The panelists did not reach consensus on a visit frequency recommendation for individuals recovering faster than expected in postoperative month 3. It appeared unlikely that additional rounds would lead to consensus because panelists’ responses and comments were stable between rounds. The panelists appeared to be split between two groups based on their comments. One group felt that fast-recovering patients should be recommended to discharge by month 3, while the other group stated their recommendation would depend on the ambitiousness of the patient’s goals. Most surgeon panelists fell within the first group, while panelists from other professions were more equally divided. These findings may reflect differences in how orthopedic surgeons view recovery after TKA compared to physical therapists or advanced practice providers.

Limitations

Patients with TKA were not included on the Delphi panel. We chose not to include patients because we felt it was important for panelists to have experience with a wide range of cases given the heterogeneity of the TKA population. We also envisioned that an individual patient’s input on care decisions would be most valuable when applied to their own care (i.e., preference-sensitive care). Regardless, future work should examine patients’ perceptions of the acceptability and usefulness of the recommendations developed in this study. The recommendations from this study are also most generalizable to the United States because practice patterns (e.g., length of hospital stay, use of different rehabilitation settings) may differ substantially in other countries.

This study also has notable strengths. We enrolled an experienced panel with considerable diversity in terms of profession, geography, and practice setting. To strengthen the validity of our findings, we used an iterative recommendation development phase, established predefined criteria for consensus and for stopping the Delphi process, provided participants with both quantitative and qualitative feedback between rounds, and incorporated panelist feedback into subsequent rounds.^{2, 22}

CONCLUSION

We used the Delphi method to develop recommendations for the use of outpatient rehabilitation visits after TKA. These recommendations could be used to promote preference-sensitive care. Future research should explore patients’ perceptions of these recommendations and examine whether they can be used within a value-based model to improve care quality and efficiency.

TABLE 2.

Visit frequency recommendations by patient’s postoperative month and recovery status

Recovery status	Visit frequency recommendation	Round consensus reached
Recovery month 1
Slow	2x/week	Round 1
Typical	2x/week	Round 1
Fast	2x/week	Round 1
Recovery month 2
Slow	2x/week	Round 1
Typical	1x/week	Round 1
Fast	1x/week	Round 1
Recovery month 3
Slow	2x/week	Round 1
Typical	1x/week	Round 1
Fast	No consensus	No consensus

Abbreviations: TKA, Total Knee Arthroplasty

KEY POINTS.

Findings:

Using the Delphi method, we developed expert consensus for the use of outpatient rehabilitation visits after TKA.

Implications:

There is no consensus for how to use outpatient rehabilitation visits after TKA. We envision the recommendations from this study could help patients make visit usage decisions in line with their own preferences, which could help improve the quality and efficiency of outpatient rehabilitation.

Caution:

The findings from this study are expert opinion only; future work should examine the acceptability and clinical impact of this study’s recommendations compared to alternative approaches.